Social Work Environment and Mental Health

Transcription

1 A C U L T Y O F H E A L T H A N D M E D I C A L S C I E N C E S U N I V E R S I T Y O F C O P E N H A G E N Social Work Environment and Mental Health An epidemiological study of interpersonal work environment factors as predictors of psychotropic treatment Ida Elisabeth Huitfeldt Madsen PhD thesis

2

3 Social Work Environment and Mental Health An epidemiological study of interpersonal work environment factors as predictors of psychotropic treatment Ida Elisabeth Huitfeldt Madsen PhD thesis Faculty of Health and Medical Sciences University of Copenhagen and National Research Centre for the Working Environment 2012

4 Social Work Environment and Mental Health An epidemiological study of interpersonal work environment factors as predictors of psychotropic treatment Ida Elisabeth Huitfeldt Madsen PhD thesis Submitted June 2012 Defended September 2012 ISBN Academic Advisors Professor Reiner Rugulies Professor Finn Diderichsen National Research Centre for the Working Environment, University of Copenhagen, Dept. of Public Health University of Copenhagen, Dept. of Public Health Assessment Committee Professor Jens Peter Bonde Professor Tarani Chandola Dr. Maria Melchior University of Copenhagen, Denmark (Chair) University of Manchester, United Kingdom National Institutes of Health and Medical Research (INSERM), France

5 PREFACE This thesis concludes the PhD project Social Work Environment Factors and Psychopharmateuticals which was conducted at the National Research Centre for the Working Environment in Copenhagen during July June The project was funded by the Danish Work Environment Research Fund (grant ) and examined whether social factors in the workplace are associated with risk of developing common mental disorder, as indicated by treatment with psychotropic medications (antidepressants, anxiolytics, hypnotics). The project was approved by the Danish data protection agency (reference number ). Several individuals have been instrumental in conducting this PhD and thus deserve my gratitude. First, I would like to thank my supervisors Reiner Rugulies and Finn Diderichsen for their invaluable and insightful guidance and support throughout the project. Second, I wish to thank my co-authors on the journal articles for their constructive criticism, and especially Linda L Magnusson Hanson who was in charge of the Swedish analyses. Third, I would like to thank my colleagues at the National Research Centre for the Working Environment, in particular the members of the PIVA research group; amongst them, a special thanks goes to Maj Britt Dahl Nielsen for her ever supportive companionship in our office. Finally, the project was based on existing data that were collected for previous work environment studies and I would like to thank all researchers and respondents who helped make these data available through their work or participation in the studies PUMA, COPSOQ, DWECS, DHCWC-2004, and SLOSH. Ida E. H. Madsen, June 29, 2012

6 1. INTRODUCTION COMMON MENTAL DISORDER CAUSES OF COMMON MENTAL DISORDERS WORK ENVIRONMENT AND COMMON MENTAL DISORDER SOCIAL WORK ENVIRONMENT FACTORS HYPOTHESES METHODS DATA SOURCES OPERATIONALIZATIONS OF EXPOSURES OPERATIONALIZATION OF OUTCOME COVARIATES STUDY POPULATIONS STATISTICAL ANALYSES STATISTICAL MODELLING CONFOUNDER SELECTION SUMMARY OF FINDINGS ARTICLE 1: WORK-RELATED VIOLENCE AND PSYCHOTROPICS ARTICLE 2: PERSON-RELATED WORK AND PSYCHOTROPICS ARTICLE 3: ELDERCARE WORK AND ANTIDEPRESSANT TREATMENT ARTICLE 4: WORK ENVIRONMENT AND ANTIDEPRESSANTS IN DENMARK AND SWEDEN ARTICLE 5: EMOTIONAL DEMANDS AND ANTIDEPRESSANTS: BUFFERING BY LEADERSHIP DISCUSSION METHODOLOGICAL DISCUSSION RELATIONS TO PREVIOUS FINDINGS PRACTICAL IMPLICATIONS CONCLUSION LIST OF APPENDICES 7. LIST OF ENCLOSURES Madsen 2012 Social Work Environment and Mental Health

7 1. INTRODUCTION Common mental disorders are prevalent and incur substantial human and societal costs (1;2). These disorders are thought to result from an interplay of biological, psychological and social factors (3-7), and research shows that common mental disorders may be predicted by work environment exposures including high psychological demands and low social support (8-11). Knowledge is scarce (10), however, concerning the role of work environment factors outside the established theories of Job Strain (12) and Effort Reward Imbalance (13). This thesis examines the associations between social work environment factors and psychotropic treatment, as an indicator of clinically significant common mental disorder. The thesis starts with an overview of the existing knowledge in the field, including the epidemiology and costs of common mental disorders, the causes of common mental disorders, and a presentation of the examined work environment exposures. This introduction ends by stating the specific hypotheses tested in the thesis. Following the introduction is a review of the methodologies applied in the thesis, including the data sources, operationalizations of key constructs, study populations, and the statistical analyses. After this methodological presentation, the results are summarized. The results encompass those of five journal articles, three of which are published (articles 1-3). Also, a supplementary analysis is included of the hypotheses examined in article 2 as this article was the first written and it was based on a subsample of data which were available early in the project. The journal articles are included as appendices 1-5. The results are followed by a discussion of key methodological concerns and a relation of the findings to previous literature. This discussion is concluded by an outline of the practical implications of the results. Finally, the thesis ends with an overall conclusion to the findings. Madsen 2012 Social Work Environment and Mental Health 1

8 1.1. COMMON MENTAL DISORDER In this thesis, the term Common Mental Disorder refers collectively to depressive and anxiety disorders, and clinically significant insomnia. These conditions are interrelated as they have high comorbidity (5-7;14;15) and insomnia may be a symptom of depression or anxiety (16;17). The symptoms also overlap between depressive and anxiety disorders as anxiety symptoms are prevalent in individuals with depression (5;6). The main symptom difference between these disorders is that individuals with anxiety disorder rarely experience certain depressive symptoms, particularly anhedonia and cognitive and motor retardation (5;6). Furthermore, there may be a causal effect of insomnia and anxiety disorders on depression (3-5;7;14). Common mental disorders are prevalent and incur high human and societal costs. European data estimate 12-month prevalences of 9% for depressive disorders and 12% for anxiety disorders (1), and a review found primary insomnia prevalences of 2%-4% (16). Common mental disorders are associated with substantially reduced quality of life (2;7) and The World Bank estimates that unipolar depressive disorders are the third greatest cause of disease burden in high-income countries and the greatest world-wide cause of years lived with disability (18;19). Societal costs of common mental disorders include not only treatment expenses, but also productivity loss (2;7); the European Study on Epidemiology of Mental Disorders (ESEMed) (2) found that individuals with major depression lost on average 9% of work days during the past month. For anxiety disorders, the most disabling subtypes were panic disorder, specific phobia and posttraumatic stress syndrome each associated with 11% lost work days (2). Furthermore, common mental disorders are associated with exclusion from the labour market (20;21) and they are the most frequent reason for disability retirement in Denmark (22). Madsen 2012 Social Work Environment and Mental Health 2

9 1.2. CAUSES OF COMMON MENTAL DISORDERS Common mental disorders are multifactorially caused in an interplay of biological, social and psychological factors (3-7). The psychological theories of depression include psychodynamic, behavioural and cognitive perspectives (23). Representing the psychodynamic view, Freudian theory suggests that depression may result from unprocessed grief when an individual loses an object (e.g. person, group, ideal) with which they identify and feel ambivalent towards (23). Other psychodynamic theorists propose that depression stems from a loss of self-esteem when ideals cannot be fulfilled, or that depression is aggression turned against the self due to anger expression issues (23). Seligman found, based on animal experimentation, that repeated exposure to uncontrollable circumstances may cause a state of learned helplessness and eventually depression (23). Behavioural theories build on this idea and propose that depression results from lacking adaptive social behaviours (e.g. assertiveness, positive response to challenge, seeking affection and caretaking) caused by a loss of social reinforcement (23). According to the cognitive perspective, early experiences may cause global negative assumptions (schemata) and schemata involving all-or-nothing thinking are a central cause of depression (23). Anxiety disorders, according to Freud, stem from intrapsychic conflict between instinctual drives, super-ego prohibitions and external reality demands (24). Anxiety signals the ego to mobilize defence mechanisms (e.g. repression) to restore intrapsychic balance (24). Behavioural theories, on the other hand, suggest that anxiety is the conditioned response to a specific environmental stimulus that becomes generalised, or that anxiety may be learned from social modelling of anxious behaviours (24). The cognitive theories of anxiety are disorder-specific; as an example, generalised anxiety disorder (GAD) is thought to stem from insecure and anxious early attachment relations, or cognitive schemas of the world as a dangerous place (24). In GAD, worrying becomes a strategy to avoid intense negative affect, and worrying about unlikely future Madsen 2012 Social Work Environment and Mental Health 3

10 threat removes the need to deal with present and more realistic threats and conflicts (24). Also, a degree of magical thinking occurs, and the individual may find that worrying about future threats helps prevent them (24). Despite the distinct psychological theories of depressive and anxiety disorders, some have argued that similar factors may affect common mental disorders collectively (5-7;14). Studies have indicated that depressive and anxiety disorders have shared risk factors, possibly indicating common underlying biological and psychological vulnerabilities (5;7;14). Following this perspective, common mental disorders develop in a stress-diathesis framework and stressful situations may trigger existing vulnerabilities (3-7). Vulnerability factors may be genetic, biological, psychological or social (3-7), and the stressful situations may be singular events (e.g. assault) or ongoing difficulties (e.g. marital discord) (3-7). Genetics may affect vulnerability to environmental stressors, and so might early environmental factors such as the relationship with caregivers and parenting quality (5;6). In addition to acting as vulnerability factors, these early influences may moderate genetic effects (5;6). Psychological vulnerabilities to common mental disorder include neuroticism, low self-worth and low sense of control over external events (5;6). Effects of stressful situations may furthermore be influenced by attributions about their causes as internal or external, global or specific, and stable or unstable (5;6;25). Not only early life social factors such as the quality of care and parenting (5;6), but also social factors in adulthood have shown important for the risk of common mental disorders. Marriage has been found protective for common mental disorder (6;26), and problematic social relations such as low social support, relationship difficulties, marital discord, lack of an intimate confidant and discontentment with partner status have been associated with increased risk (3;4;6;27). A causal effect of interpersonal factors on common mental disorders is plausible because negative interpersonal interactions and social rejection may lower selfesteem and self-worth (28;29), which may in turn increase risk of disorder (3;4;14;29;30). These factors have been studied extensively in the private life domain (29), but less so within a workplace context. Madsen 2012 Social Work Environment and Mental Health 4

11 1.3. WORK ENVIRONMENT AND COMMON MENTAL DISORDER Research on the relations between the working environment and common mental disorder has mainly focused on the theoretical models of Job Strain (12) and Effort-Reward-Imbalance (13). According to a systematic literature review from 2006 (8), robust evidence exists that high demands and low decision latitude, and high efforts and low rewards are prospective risk factors for common mental disorders. This review, however, did not distinguish between different types of common mental disorder (10), and the clinical significance of the outcome-measures, which were largely self-reported, was uncertain (9). Consequently, three systematic reviews were subsequently published in 2008 by Bonde (9), Netterstrøm et al. (10) and Siegrist (11), focusing on the effects of the psychosocial work environment on depression. All three reviews found evidence of a prospective association between self-reported high demands and low social support at work and depression, whereas the findings for control at work were less consistent. The reviews did, however, point to several limitations of the literature. Bonde and Siegrist pointed out that most studies had examined self-reported exposures that may have been affected by baseline mental health (9;11). Studies using independent measures of exposure and outcome were lacking, and self-report on both factors could cause common method bias (9). Also, according to Bonde, studies examining the risk of clinically significant depression were lacking, as many studies used an outcome that did not necessarily correspond to a clinical diagnosis (9). Based on these limitations, Bonde concluded that there was no sound evidence for a causal effect of work environment on depression (9). The Netterstrøm review found indications of publication bias in the published findings and stated a need for further research with more precise assessment of exposure duration and intensity (10). Furthermore, the review found a need for more studies examining common mental disorders other than depression using a clinically significant outcome measure (10). Finally, the authors pointed to a lack of studies examining exposures outside the established job strain and effort reward imbalance models (10). Madsen 2012 Social Work Environment and Mental Health 5

12 1.4. SOCIAL WORK ENVIRONMENT FACTORS The focus of this thesis is on the effects of social, i.e. interpersonal, work environment factors on common mental disorder. More specifically, the work environment factors examined are: violence at work, personrelated work, emotional demands and emotional labour. The following sections present these exposures, and argue why they might be related to the development of common mental disorder VIOLENCE AT WORK The World Health Organization defines violence as The intentional use of physical force or power, threatened or actual, against oneself, another person, or against a group or community, that either results in or has a high likelihood of resulting in injury, death, psychological harm, maldevelopment or deprivation. (31, p. 5). One of the exposures examined in this thesis is work-related violence, i.e. violence that occurs in the work-place and is directed toward staff. Work-related violence has been associated with self-reported mental health outcomes such as emotional exhaustion (32), sleep problems (33), anxiety (34), and depression (35). Research shows that workrelated violence is associated with feelings of helplessness (36), and a causal effect on common mental disorders is therefore plausible, as helplessness may be causally related to these disorders (25). There is a paucity of evidence, however, linking work-related violence with clinically significant common mental disorder. Also, most existing studies have been occupation-specific (32-35;37;38), raising questions regarding the generalizability of the findings. One large register-based study (39) showed that employees in occupations with high levels of work-related violence were at increased risk of hospitalization with depression or anxiety, but the lacking individual level exposure measures restricted causal inference due to the ecological fallacy (40). Madsen 2012 Social Work Environment and Mental Health 6

13 PERSON-RELATED WORK Person-related work is an umbrella term for jobs that require face-to-face or voice-to-voice interactions with individuals not employed at the work-place (clients) (41). This term covers both customer service work, and the occupations sometimes referred to as human service work, e.g. nursing, care work, teaching and social work. Human service work, in this thesis, is work in human service organizations. The term human service organization was coined by Hasenfeld & English (42), and names organizations where the main purpose of the organization is to increase or maintain the functioning and well-being of the client (42). This could be in a hospital, where the purpose is to make ill patients healthy, or in a school, where the aim is to increase the pupils educational level. Human service work has been argued to be particularly emotionally demanding because it involves confrontation with the clients problems and suffering (43) and because the interactions with clients require empathy and sensitivity to their emotions (41). Also, the relationship between human service workers and their clients may become imbalanced if the worker does not receive the rewards they expect for their work, in terms of for instance gratitude, or that the client improves or attempts to do so (44). Some researchers have even argued that imbalance is a fundamental premise in the relationship between human service workers and clients as the employee provides emotional resources that the client receives as illustrated by the terminology of caregiver and care-recipient (45). Because reciprocity in interpersonal relationships is pivotal to human well-being, such imbalance is thought to drain the employee s emotional resources and lead to adverse emotional consequences (13;44). Jenkins & Maslach (46) argue that human service occupations such as health care, educational and social work place high interpersonal demands on the employees by placing them in a helping role or requiring that they deal with people in stressful situations (46). They further argue that customer service work involves intermediate interpersonal demands, as they are associated with high social requirements but lower levels of social-emotional stress in the daily work routine (46). Madsen 2012 Social Work Environment and Mental Health 7

14 Human service work has been related to poor mental health outcomes such as suicide (47) and hospitalization with depression and anxiety (48-50), and the potential mental health effects of human service work was the focus of early burnout research (51). Burnout is a condition characterized by exhaustion, cynicism and inefficacy (51). Burnout is correlated with both depression and anxiety, but whereas depression affects all life-domains, burnout is specific to the work context (51). The clinical diagnosis closest resembling the phenomenon of burnout is neurasthenia (51). Burnout is thought to result from chronic emotional and interpersonal job stressors (51) and some research suggests that burnout may be a mediating factor in the relationship between work environment and depression (52) POTENTIAL MEDIATORS BETWEEN PERSON-RELATED WORK AND COMMON MENTAL DISORDER Person-related work may affect the risk of common mental disorder through work environment factors disproportionately prevalent within these occupations, namely the emotional demands of the work, the emotional labour, and the exposure to work-related violence. Regarding violence, a potential effect on common mental disorder has been argued in a previous section (see section 1.4.1), and it is known that violence is more prevalent in person-related work (39). The following section of this thesis therefore focuses on emotional demands and demands for hiding emotions, and why they might mediate an effect of personrelated work on common mental disorder. Emotional demands at work may be defined as aspects of the work that require sustained emotional effort (53), following the de Jonge and Dormann definition of demands as those aspects of the job which require sustained behavioural, physical, cognitive and/or emotional effort (54, p. 49). Emotional demands are elevated in human service occupations (55), and they may be related to special characteristics of these occupations, such as the imbalanced interpersonal relationship with clients (45). Emotional demands may also Madsen 2012 Social Work Environment and Mental Health 8

15 be related to the confrontation with the problems and suffering of the clients (53;56), and in health care, critical incidents such as treating victims of terror or critically ill patients, may cause emotional responses (57). Also, person-related work requires empathy and sensitivity to the feelings of clients (58), and empathy might elicit sympathy (59) which could be experienced as emotionally demanding. Empirically, emotional demands at work have been associated with indicators of poor mental health including psychological distress (60;61) and depressive and anxiety symptoms (62). Also, a register based study found that women employed in occupations with higher levels of emotional demands were at increased risk of psychiatric treatment for depression (63). The mechanism by which emotional demands might affect mental health is not clear, but they may result in prolonged expenditure of emotional resources or energy as they require sustained emotional effort (53). A causal effect of emotional demands on common mental disorder might then be explained by the conservation of resources theory (64) which states that negative functional and emotional outcomes may occur when resources that an individual values are threatened or lost. Emotional Labour is a term coined by Hochschild (65). She suggested that customer service work involves a special form of emotional work because employees must manage their feelings and emotions in the provision of customer service, to adhere to situation specific feeling rules. One example is the suppression of anger towards a customer, because the expression of anger would be inappropriate according to organizational rules. According to Hochshild (65), employees performing emotional labour may be at increased risk of mental health problems due to an alienation from their true feelings and selves. Zapf (41) uses the term emotion work rather than emotional labour, and argues that such work is a significant part not only of customer service work, but of all types of person-related work. According to this perspective, emotion work has three central characteristics (41): 1) it occurs in face-to-face or voice-to-voice interactions with clients, 2) emotions are displayed to influence other people s emotions, attitudes and behaviours, and 3) the display of emotions Madsen 2012 Social Work Environment and Mental Health 9

16 has to follow certain rules. Emotional labour, or emotion work, has been associated empirically with mental health outcomes such as emotional exhaustion (41;66;67), irritation (41) and psychosomatic complaints (41). This thesis does not examine emotional labour directly, but demands for hiding emotions at work. These demands, though, will likely result in the performance of emotional labour, at least if the employee attempts to meet the demands. The choice to measure demands for hiding emotions was based on the data available, but is also reasonable as emotional labour may be performed without substantial conscious effort (41). Hence, the specific construct of emotion work might be difficult to measure self-reportedly. Demands for hiding emotions could possibly be conceptualized as a type of emotional demands, because the performance of emotion work might involve substantial emotional effort as indicated by its association with emotional exhaustion (41). In the context of this thesis, however, emotional labour and emotional demands are treated as separate constructs for three reasons: 1) emotional labour may not always require sustained emotional effort (41), 2) the two concepts developed as separate entities in different strands of scientific literature, 3) the empirical distribution indicates that the constructs are related but not identical as some occupations have high levels of emotional demands but low demands for hiding emotions (55) EFFECT-MODIFICATION BY GOOD LEADERSHIP If human service work is particularly emotionally demanding because of particular characteristics of the work, such as the confrontation with the clients problems and suffering (43), these types of work may be inherently emotionally demanding and their emotional demands may not be amenable to change. Consequently, it is important to examine if any effects of emotional demands on common mental disorder may be buffered by other factors, which are modifiable. Madsen 2012 Social Work Environment and Mental Health 10

17 According to the conservation of resources theory, negative functional and emotional outcomes may occur when resources that an individual values are threatened or lost (64). Emotional demands at work may result in prolonged expenditure of emotional resources or energy, as they are aspects of the job that require sustained emotional effort (53). Following the conservation of resources theory, the availability of resources will determine the impact of resource loss (64). The theory further states that acknowledgement of accomplishments and tasks, and understanding from superiors are important resources in a Western context (64). Hence, leadership quality may be an important work environment resource, which could have the potential to buffer detrimental effects of emotional demands at work on employee mental health. Good and supportive leadership could also be construed as a type of social support, and social support could buffer effects of stressors (68;69). According to Thoits (68), this buffering may occur because supportive others help the individual cope with the situation (active coping assistance) or provide emotional support leading to an increased sense of self-esteem and mattering (emotional sustenance). In a work context, the leader likely has previous experience with the work demands the employee is facing, and may thus be considered an experientially similar other. Such individuals may offer emotional sustenance in terms of empathic understanding, acceptance of ventilation and validation of feelings and concerns (68). They may also provide active coping assistance in form of assistance in re-appraising the threat, and offering information and advice (68). Finally, they may be a source of social comparison, offering a role model and inspiring hope that one can deal with the situation (68). According to both the social support perspective and the conservation of resources theory, then, good leadership is a work environment factor that might have the potential to buffer any effects of emotional demands at work on common mental disorder. Madsen 2012 Social Work Environment and Mental Health 11

18 1.5. HYPOTHESES Based on the literature presented, the following hypotheses are stated and tested: 1. Employees exposed to work-related violence are at increased risk of developing common mental disorder as indicated by entering psychotropic treatment. (Article 1) 2. Employees in person-related work are at increased risk of developing common mental disorder as indicated by entering antidepressant treatment. (Article 2) 3. This increased risk of common mental disorder is mediated by work environment factors elevated in person-related work, namely emotional demands at work, demands for hiding emotions, and/or workrelated violence. More specifically, there is no direct effect of person-related work on psychotropic treatment when these work environment factors are fixed. (Article 2) 4. The odds of antidepressant treatment in care workers increase more after the employees enter their profession than it did before they entered care work, indicating a causal effect of this work on common mental disorder. (Article 3) 5. Associations between work environment factors and common mental disorder, as indicated by antidepressant treatment, are similar in Denmark and Sweden, substantiating the generalizability of the findings. (Articles 4, 5) 6. The effects of emotional demands at work on common mental disorder, as indicated by antidepressant treatment, is buffered by good leadership. (Article 5) Madsen 2012 Social Work Environment and Mental Health 12

19 2. METHODS The following section presents the methods used in this thesis, including the applied data sources, operationalizations and study populations and the analyses conducted DATA SOURCES This thesis is based on data from the project Social work environment factors And PsychophArmaceuticals (SAPA) which combines self-reported data on work environment exposures with register data on psychotropics purchases and sociodemographic factors. The following section describes the data sources applied WORK ENVIRONMENT DATA The work environment data used in this thesis originate from four Danish and one Swedish work environment studies, namely: 1) the Project on Burnout, Motivation and Job Satisfaction (PUMA), 2) the Copenhagen Psychosocial Questionnaire Study (COPSOQ), 3) the Danish Work Environment Cohort Study (DWECS), 4) the Danish Health Care Worker Cohort - Class of 2004 (DHCWC-2004), and 5) the Swedish Longitudinal Occupational Survey of Health (SLOSH). Table 1 summarizes key information about these studies. PUMA was a three wave open cohort study during designed to examine burnout. PUMA recruited employees from seven human service organizations, and the PUMA participants worked in social work, health care work, elder care work and care work dealing with handicapped persons. Details of PUMA were published by Borritz et al. (70). Madsen 2012 Social Work Environment and Mental Health 13

20 Madsen 2012 Social Work Environment and Mental Health 14 Table 1. Overview of work environment data Study Wave Period for datacollection Method for datacollection Population No. Respondents PUMA Questionnaire Employees in 7 Human service organizations (n= 2,391) Questionnaire Respondents PUMA 1 + new employees in the organizations (n= 2,335) 2005 Questionnaire Respondents PUMA 2 + new employees in the organizations (n= 2,568) COPSOQ a) Questionnaire (2/3, random assignment) b) Telephonic Interview (1/3, random assignment) Representative sample, ages (n=4,000) Questionnaire a) Follow up wave 1 (n=1,698) b) Representative sample (n= 8,000) DWECS 2000 Oct January 2001 (% responserate) DHCWC Oct May 2006 Telephonic Interview Representative sample (supplemented follow-up of DWECS 1995, n= 11,955) a) Questionnaire (9/10, random assignment) b) Telephonic Interview (1/10, random assignment) a) Follow-up of DWECS 2000 (n= 10,131) b) Supplemental random sample, ages (n=943) c) Supplemental random sample, immigrants (n=236) d) New random sample, ages (n=8,545) Social- and health care helpers and assistants (respondents from DHCWC-2004 wave 1, n=5,696) SLOSH 2006 Respondents to SWES (representative Swedish sample, n=9,154 ) Selection criteria for SAPA No. included respondents 1,914 (80.1) a) PNR valid b) No previous response included in SAPA 1,759 (75.3) a) PNR valid b) No previous response included in SAPA 1,747 (68.0) a) PNR valid b) No previous response included in SAPA 2,454 (61.4) a) PNR valid b) No previous response included in SAPA a) 1,281 (75.4) b) 4,732 (60.4) a) PNR valid b) No previous response included in SAPA 8,583 (75.0) a) PNR valid b) No previous response included in SAPA 12,413 (62.5) a) PNR valid b) No previous response included in SAPA 3,708 (65.1) a) PNR valid b) Employed in eldercare in 2005 (self-reported) c) Female 5,985 (65.4) a) Gainfully employed b) Aged years For study details see 1,856 Borritz et al (70) 747 Borritz et al (70) 644 Borritz et al (70) 2,445 Kristensen et al (55) 4,719 Pejtersen et al (71) 8,551 Burr et al (72) 6,743 Feveile et al (73) 2,582 Carneiro et al (74) 4,351 Kinsten et al (75); Magnusson Hanson et al (76) Abbreviations: No.: Number; PUMA: Project burnout, job satisfaction and motivation (Danish Acronym); COPSOQ: Copenhagen psychosocial questionnaire; DWECS: Danish work environment cohort study; DHCWC-2004: Danish health care worker cohort-class of 2004; SLOSH: Swedish longitudinal survey on health; SAPA: Social work environment factors and psychopharmaceuticals.

21 COPSOQ had two waves of data collection in 1997 (COPSOQ I) and (COPSOQ II). COPSOQ I invited participants from a random sample, and COPSOQ II included both a follow up of COPSOQ I participants and a new random sample. Individual dates for questionnaire response were unavailable for COPSOQ I, and all response dates were set to December first Details of COPSOQ were published by Kristensen et al. (55) and Pejtersen et al. (71). DWECS is an on-going cohort study of the Danish working environment, with follow up every five years. The cohort was started in 1990, but the data included in this thesis were collected in 2000 (DWECS 2000) and 2005 (DWECS 2005). At each follow up the population is supplemented with age and migration panels, to remain representative of the Danish working population. Details of DWECS were published by Burr et al. (72) and Feveile et al. (73). DHCWC-2004 was a three-wave study of individuals graduating as social and health care helpers or assistants in Denmark in Social and health care helpers and assistants are primarily trained to work with elder care (77). All 28 Danish schools providing such training were invited to participate, and 27 schools accepted. Details of the DHCWC-2004 were published in Danish by Carneiro et al. (74). SLOSH is an ongoing cohort study which started as a follow up of individuals from the Swedish Work Environment Survey (SWES) from SWES was based on a random sample of gainfully employed Swedish citizens aged in Respondents to SWES were invited to participate in SLOSH. Details of SLOSH were published by Kinsten et al. (75) and Magnusson Hanson et al. (76) REGISTER DATA This thesis uses Danish data from the following registers: the Civil Registration System (78) (sex, date of birth, migration), the Population Education Register (79) (education), the Income Statistics Register (80) (income), Madsen 2012 Social Work Environment and Mental Health 15

22 the Danish Register of Causes of Death (81) (date of death) and the Register of Medicinal Product Statistics (82;83) (purchases of psychotropic medication). Also, data were included from the Swedish Register of Causes of Death (date of death) (84) and the Swedish Prescribed Drug Register (purchases of psychotropic medications) (83) OPERATIONALIZATIONS OF EXPOSURES The operationalizations of the key constructs from articles 1-5 are presented in the articles (see appendices 1-5). The following section of the thesis gives the operationalizations used for the supplementary analyses for article TYPE OF WORK Person-related work was defined to encompass all participants reporting substantial client-contact. The items measuring client-contact are presented in table 2, and substantial client-contact was defined in PUMA as over half the time, in DWECS as three quarters of the time or more, and in COPSOQ by indicating that one s primary work tasks were related to customers or clients. The type of person-related work (health care, social, educational, customer service, other) was defined in PUMA by the organization through which the respondent was recruited. In COPSOQ and DWECS type of work was defined using the Danish Industrial Classification 1993 (85) which is compatible with Eurostat s Statistical classification of economic activities in the European Community (NACE) (86). Table 3 gives an overview of the coding and categorization applied. The industrial classification was chosen for consistency across the datasets, as the classification of PUMA participants was organization based. Madsen 2012 Social Work Environment and Mental Health 16

23 Madsen 2012 Social Work Environment and Mental Health 17 Table 2. Supplementary analysis: measures of work environment exposures CONSTRUCT ITEMS RESPONSE OPTIONS Client contact Emotional demands Demands for hiding emotions PUMA I-III STUDIES COPSOQ II STUDY DWECS 2000 STUDY How much direct contact with clients do you have during 1. All the working hours 2. Working with clients, patients, pupils, children, citizens 4. Working with machines, tools, facilities/plants, an average workweek? a 2. More than half of the working hours 3. Less than half of the working hours 4. Never/almost never What are your primary daily work tasks? 1. Working with contact to customers etc. 3. Working with language, figures, writing, communication transportation 5. Working with animals, plants 6. Other, write: Are you in contact with people who are not employed at your workplace when conducting your work? (e.g. customers, clients, passengers, pupils?) Does your work put you in emotionally disturbing situations? 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever Is your work emotionally demanding? 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. to a very small extent Do you get emotionally involved in your work? 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent Does your work require that you hide your feelings? 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent Violence Have you been exposed to threats of violence at your workplace during the last 12 months? Table 2 Continues Have you been exposed to physical violence at your workplace during the last 12 months? 1. No 2. yes 1. No 2. Yes 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. Yes, daily 2. Yes, weekly 3. Yes, monthly 4. Yes, a few times 5. No 1. Yes, daily 2. Yes, weekly 3. Yes, monthly 4. Yes, a few times 5. No 1. Almost all my working hours 2. Approx. ¾ of the working hours 3. Approx.½ of the working hours 4. Approx. ¼ of the working hours 5. Only few working hours 6. None at all 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. To a very large extent 2. To a large extent 3. Some-what 4. To a small extent 5. To a very small extent 1. No 2. Yes, from a colleague 3. Yes, from a manager 4. Yes, from subordinates 6. Yes, from clients* 1. No 2. Yes, from a colleague 3. Yes, from a manager 4. Yes, from subordinates 5. Yes, from clients*

24 Madsen 2012 Social Work Environment and Mental Health 18 Table 2 Continued Quantitative demands Quantitative demands Do you have to work very fast? 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever Is your work unevenly distributed so it piles up? 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever How often do you not have time to complete all your work tasks? Influence Do you have a large degree of influence concerning your work? Social support 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever Do you have any influence on what you do at work? 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever Can you influence the amount of work assigned to you? 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever How often do you get help and support from your colleagues? How often do you get help and support from your immediate superior? 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 1. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 6. Not relevant 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 6. Not relevant 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 6. Have no colleagues 1. Always 2. Often 3. Sometimes 4. Seldom 5. Never/hardly ever 6. Have no superior Abbreviations: PUMA: Projekt Udbrændthed, Motivation, og Arbejdsglæde (Danish Acronym); COPSOQ: Copenhagen Psychosocial Questionnaire; DWECS: Danish Work Environment Cohort Study. adepending on the organization and the types of clients dealt with, the term client was exchanged with resident, inmate or patient.

25 Table 3. Supplementary analysis: definition of type of person-related work by industry used in COPSOQ and DWECS TYPE OF PERSON-RELATED WORK Health care INDUSTRY ACCORDING TO THE DANISH INDUSTRIAL CLASSIFICATION Hospital activities General medical practice activities Dental practice activities Other human health activities Residential institutions for physically or mentally handicapped adults Residential nursing homes and sheltered homes Home help Day institutions for elderly people Rehabilitation institutions Refugee centres Associations combatting diseases and performing activities aimed at social work, etc Special schools for handicapped children Fire service activities Education Primary education (except Special schools for handicapped children ) General secondary education Technical and vocational secondary education Higher education Driving school activities Other adult education Child day-care in private homes Day nurseries Kindergartens After school centres Age-integrated institutions Recreation centres for young people Social work Customer service Justice and judicial activities, law and order activities Police Compulsory social security activities Social work activities with accommodation (except Residential care activities for disabled adults and Residential care activities for the elderly) Other social work activities without accommodation Employment exchange 45 Construction 50 Sale and rep. of motor vehicles, sale of auto. fuel 51 Wholesale except of motor vehicles 52 Retail trade and repair work except of motor vehicles 55 Hotels and restaurants 60 Land transport and transport via pipelines 61 Water transport 62 Air transport 63 Supporting transport activities 64 Post and telecommunications 65 Finance 66 Insurance 67 Activities auxiliary to finance 70 Real estate activities 71 Renting of transport equipment and machinery 72 Computer and related activities 73 Research and development 74.2 Consulting architectural activities 74.4 Advertising Temporary employment agencies Security and surveillance activities Cleaning activities Photographers Packaging activities, etc Envelope addressing activities Activities of interior designer Charitable trusts and foundations Madsen 2012 Social Work Environment and Mental Health 19

26 EMOTIONAL DEMANDS Emotional demands at work were measured by the three items presented in table 2. A scale was constructed as the mean of the three items scored with equally spaced response categories from For presentational purposes the continuous scale was divided by 10, yielding a scale from High versus low emotional demands were defined dichotomously by the median value from DWECS 2000 (41.67) DEMANDS FOR HIDING EMOTIONS Demands for hiding emotions were measured by a single item as presented in table 2. A continuous score was constructed scoring each response category equally spaced from 0-10 with higher scores indicating higher demands. High versus low demands for hiding emotions were defined by collapsing to a very large extent and to a large extent as high and the remaining categories were defined as low WORK-RELATED VIOLENCE Exposure to work-related violence was defined as reporting being exposed to violence or threats in the workplace during the past 12 months, measured by the items presented in table 2. Threats were included in the definition to conform with the World Health Organization s definition of violence (31). Madsen 2012 Social Work Environment and Mental Health 20

27 2.3. OPERATIONALIZATION OF OUTCOME Psychotropic (antidepressant, anxiolytic, hypnotic) treatment was used in this thesis as an indicator for clinically significant common mental disorder. This choice was based on a number of considerations, which are presented in the following section. In clinical psychiatry, the psychiatric interview by a trained mental health professional is the most important method of diagnosing mental disorders (87) and this method is considered the most valid for diagnosing mental disorders (88). Applying this measure, however, may not be feasible in large-scale epidemiological studies due to logistic and financial challenges (89). Furthermore, this measure was unavailable for this thesis which was based on existing data. Other measures for common mental disorder include self-reported data such as asking respondents if they have been diagnosed with a common mental disorder or to fill out symptom scales, such as the general health questionnaire (90). Self-report measures, however, may be problematic because of reporting issues such as recall and social desirability bias (91;92) and their correspondence with clinically significant mental disorder is sometimes unclear (93). A further challenge for the present analyses, was a lack of consistent selfreported measures across the included studies, and the self-reported nature of the measures; as the examined exposures were measured by self-report, a non-self-reported outcome was preferred to avoid common method bias (94). In Denmark, all psychiatric treatment, both in- and outpatient, is registered since 1995 (95). From the Danish Psychiatric Central Research Register it is thus possible to obtain records of psychiatric treatment and diagnoses for research purposes (95). Most cases of common mental disorder are, however, not treated (1;96), and amongst those who are treated, individuals in specialist treatment likely suffer from more severe disorder (97). For the present analyses, examining associations in the working population, this register may therefore leave out many relevant cases of common mental disorder. Madsen 2012 Social Work Environment and Mental Health 21

28 Purchases of prescription medications at Danish pharmacies are also registered at the individual level since 1995 in the Register of Medicinal Product Statistics (82) and purchases of psychotropic medication were chosen in this thesis as an indicator of common mental disorder. The main advantages of this outcome measure compared to diagnostic interviews or self-reported outcome measures are the cost-efficiency, the minimal loss to follow up and the day-to-day measurement. When applying other outcome measures which require respondent participation, caseness of ill-health may be underestimated because individuals with poor health might be less likely to participate at follow up (98) and because the follow up is carried out at a specific time. Compared to self-reported measures, psychotropic treatment further has the advantage that it is nonself-reported which may be preferable when the examined exposures are self-reported. Also, the measure likely has substantial validity for clinically significant conditions as it reflects a clinical decision that psychotropic treatment is warranted. Compared to the psychiatric register, the main advantage of psychotropic treatment is the inclusion of less severe cases of common mental disorder which may not be specialist-treated. It must be noted, though, that despite these advantages of using psychotropic treatment as an indicator for common mental disorder, several important methodological concerns exist regarding the chosen outcome. Most notably, although most psychotropics are used to treat common mental disorder (99-101), the various types of psychotropics are used to treat different disorders, and antidepressants for instance are also given for anxiety disorders (100). Hence, the effects reported in this thesis cannot be interpreted as effects on any specific common mental disorder. The methodological issues concerning the outcome measure are discussed more fully in a later section of the thesis (see section 4.1.2) DEFINING PSYCHOTROPIC TREATMENT The Register of Medicinal Product Statistics (82) classifies all medications according to their main indications following the World Health Organization s Anatomical Therapeutic Chemical Classification System (ATC) (102). Madsen 2012 Social Work Environment and Mental Health 22

29 For this thesis, psychotropics were defined as medications N06A (antidepressants), N05B (anxiolytics), or N05C (hypnotics) and data were included from 1995 onwards. The medication data were updated during the project, and the available follow up period expanded from the analyses for articles 1, 2, and 3 ( ), to the supplementary analysis for article 2, as medication data for 2009 and 2010 became available. The specific outcome definitions varied between the articles and are described in the separate articles (appendices 1-5). For the supplementary analysis for article 2, the outcome was defined as follows: Antidepressant treatment was defined by at least one purchase of antidepressants, as indicated by ATC-code N06A. Anxiolytic or hypnotic treatment was defined by at least one purchase of anxiolytics or hypnotics, as indicated by ATC-codes N05B and N05C, in individuals without antidepressant treatment. Hence, individuals purchasing both antidepressants and either anxiolytics or hypnotics were classified as in antidepressant treatment. This definition was based on the findings in article 1 (see appendix 1) that patterns for treatment with both antidepressants and anxiolytics and antidepressants only were similar COVARIATES The covariates included in the analyses varied, and the covariates used in the articles are described in the separate appendices (appendices 1-5). For the supplementary analysis for article 2, the following covariates were included: Data on sex and date of birth were included from the Civil Registration System (78). Cohabitation (living with a partner or spouse, yes/no) was measured by self-report. Education was obtained through the Population Education Register (79) and defined as the norm length of the respondents highest attained education. Income was obtained through the Income Statistics Register (80), and defined as income after tax and interests in the year of responding to the questionnaire. Migration data were obtained through the Civil Registration System and death through the Danish Register of Causes of Death (81). Madsen 2012 Social Work Environment and Mental Health 23

30 2.5. STUDY POPULATIONS The study populations differed between the analyses and were chosen for availability of key measures. The study populations for articles 1-5 are presented in the separate articles (see appendices 1-5). Briefly, article 1 uses data from PUMA I-III, COPSOQ I-II, and DWECS , and article 2 applies data from DWECS Article 3 is based on the DHCWC-2004 and DWECS Articles 4 and 5 use data from DWECS 2005 and SLOSH. This section focuses on study population for the supplementary analyses for article 2. The supplementary analysis for article 2 used data from PUMA I-III, COPSOQ-II, and DWECS 2000, with a total of 12,766 employed first-time respondents. I excluded 1,916 individuals with purchases of any psychotropic during 1,494 days before baseline to examine incident treatment. The length of exclusion period used to define prevalent use (1,494 days) was chosen because it was the maximum available for all studies. Further, 334 individuals with poor mental health at baseline and 544 individuals with missing data on any covariates were excluded, yielding a final study population of 9,972 individuals. The characteristics of the study population are shown in table 4. Madsen 2012 Social Work Environment and Mental Health 24

31 Table 4. Supplementary analysis: characteristics of participants and incident psychotropic treatment Participants Antidepressants treatment, No. (%) Anxiolytics/hypnotics treatment, No. (%) No. Mean (SD) No. (%) No. (%) Total 9, (7.2) 681 (6.8) Type of work Health care 1, (9.1) 146 (8.4) Education (7.1) 44 (7.1) Social (10.3) 59 (7.3) Customer service 1, (6.4) 63 (6.0) Other person-related 1, (6.5) 78 (6.0) Non-person-related 4, (6.3) 291 (6.5) Emotional demands Score, (2.49) high 4, (8.2) 308 (7.6) low 5, (6.5) 373 (6.3) Demands for hiding emotions Score, (2.70) to a very large extent (10.9) 26 (8.3) to a large extent (8.3) 74 (7.8) some-what 3, (7.7) 253 (7.3) to a small extent 2, (7.0) 152 (6.2) to a very small extent 2, (5.9) 176 (6.3) Work-related violence Yes 1, (9.5) 76 (7.5) No 8, (6.9) 605 (6.8) Sex Female 5, (8.2) 442 (8.0) Male 4, (6.0) 239 (5.4) Cohabitation No 2, (7.2) 127 (6.2) Yes 7, (7.2) 554 (7.0) Age, years 41.5 (10.5) Education, years 13.1 (2.46) Income, Dkk per year 181,819 (88,266.0) Abbreviations: No.: Number; SD: Standard deviation. Madsen 2012 Social Work Environment and Mental Health 25

32 2.6. STATISTICAL ANALYSES This section gives an overview of the main statistical methods applied in this thesis. More specifically, the section introduces the Cox proportional hazards regression analysis, the competing risks framework, mediation analysis, individual participant meta-analysis, and the principals applied for confounder selection COX PROPORTIONAL HAZARDS REGRESSION Articles 1, 4, 5 and the supplementary analyses for article 2 applied Cox proportional hazards regression analyses. The following presentation of Cox regression analysis is based on the text by Kleinbaum & Klein (103). The Cox proportional hazards model is a regression model for examining time-to-event outcomes. The model specifies that the hazard of an event at any time t, h(t), is a function of the baseline hazard h0(t) and the exponential expression e to the power of the sum of the included covariates X1, X2,..., Xi, multiplied by their respective regression parameters β1, β2,..., βi. The hazard ratio (HR) estimates the relative effect of a covariate in the model, and is the ratio of the hazard rate among exposed individuals divided by that of the unexposed group. The baseline hazard is included in both numerator and denominator and thus need not be estimated. The Cox proportional hazards model is a robust semiparametric model, and yields regression coefficients that are close to the underlying parametric model, without having to fully specify this model. Hence, Cox regression is a good choice when the underlying parametric model is unknown. The model assumes, however, that the effects of the covariates remain constant over time the proportional hazards assumption. This assumption may be tested empirically, for instance by visual inspection of the log-log survival plots. If the log-log curves for the examined exposure groups are non-parallel the assumption is violated. In this case, data may be analysed by dividing follow-up into periods with proportional hazards (extended Cox modelling). Madsen 2012 Social Work Environment and Mental Health 26

33 COMPETING RISKS In time-to-event analysis with multiple outcomes, the occurrence of one outcome may prevent the occurrence of another. If examining death from cancer, for instance, this outcome cannot occur if an individual dies from a heart attack first. Such outcomes are referred to competing risks (104) and were applied in article 1 and the supplementary analysis for article 2 as the outcomes of different types of psychotropics were defined as mutually exclusive. When analysing competing risks data, Cox regression may be employed to calculate cause-specific hazard ratios (104). These ratios assess the relative hazard rate for experiencing the corresponding outcome at time t, in individuals who have experienced none of the examined outcomes at time t (104). Technically, the cause-specific hazard ratio is estimated by treating the competing outcomes as censored observations when constructing the model (104). This approach is valid for interpreting the relative associations between an exposure and the outcomes (105), as is the purpose of this thesis. It should be noted, though, that causespecific hazard ratios do not correspond to cumulative incidence functions unless the examined outcomes are independent, an issue which complicates the estimation of absolute risk of the outcomes in competing risks analysis (105) MEDIATION A mediating factor is a step in the causal pathway between an exposure and an outcome (40). Article 2 and its supplementary analyses deal with mediation, and the potential mediating relations examined there are presented in figure 1. Madsen 2012 Social Work Environment and Mental Health 27

34 One traditional approach to examining the extent of mediation by a factor (Z) in the relation between an exposure (X) and outcome (Y), is to compare the regression parameter for X in a model of the effect of X on Y (βx), to the regression parameter for X when Z is controlled (βx_z) (106). βx then estimates the total effect of X on Y, βx_z estimates the direct effect of X on Y, and βx-βx_z estimates the indirect effect of X on Y through Z. This method has been used in linear regression models, but may be problematic when these models cannot be applied, e.g. due to a dichotomous outcome (106;107). Petersen et al. (108) distinguish between controlled and natural direct effects. The controlled direct effect is defined as...the difference in counterfactual outcome if the individual was unexposed and her intermediate variable was controlled (or set) at level Z=z versus the counterfactual outcome if she was exposed and her intermediate variable set at the same level Z=z. (108, p. 277) In contrast, the natural direct effect direct effect is defined as:...the difference in counterfactual outcome if the individual was unexposed Figure 1. Mediating relations between person-related work and psychotropic treatment Madsen 2012 Social Work Environment and Mental Health 28

35 versus the counterfactual outcome if she was exposed, but her intermediate remained at its counterfactual level under no exposure. (108, p. 277) Notably, the difference between these two types of direct effects is whether Z is fixed or allowed to vary between individuals. When X and Z interact in their effects on Y, there will be as many controlled direct effects as there are levels of Z (108). The natural direct effect, however, is a population average of these controlled direct effects, and when X and Z do not interact, the controlled direct effect and the natural direct effect are identical (108). In this thesis, mediation is assessed using controlled direct effects. For the three examined potential mediators emotional demands, demands for hiding emotions, and work-related violence, the respective controlled direct effects correspond to the effect of person-related work on psychotropics treatment if the levels of emotional demands or demands for hiding emotions were fixed at the population mean, or if no employees were exposed to work-related violence. These effects are estimated by controlling the effect of person-related work for the potential mediator as a continuous variable (fixed at mean level) or as a dichotomous variable with absent as the reference-group (fixed at no exposure) (108). The absence of a controlled direct effect of person-related work on psychotropics indicates full mediation by the examined mediator (108) INDIVIDUAL PARTICIPANT DATA META-ANALYSIS The data used for this thesis were collected in five different studies, each with multiple waves. The data may be clustered within each study and wave, for instance due to differences in study populations and timing of the study (the data collection took place during the years ). The data analysis must account for such clustering (109). Combining data from different studies may be done by meta-analytic techniques. When individual level data is available, individual participant data (IPD) meta-analysis is possible (109). Such analyses are Madsen 2012 Social Work Environment and Mental Health 29

36 either one-step or a two-step. The one-step IPD meta-analysis combines the raw data of the studies and analyses them simultaneously, whilst accounting for clustering within the separate studies (109). In a time-toevent framework, this may be achieved in Cox regression through stratifying the model by an indicator of the study (110). This method was employed for article 1 and the supplementary analysis for article 2. In the two-step IPD meta-analysis, separate risk estimates are calculated for each study, using harmonized definitions of exposures, outcomes, and covariates. These study-specific risk estimates are then combined using meta-analytic approaches (109). The two-step approach is suitable when raw data cannot be combined, and was applied in articles 4 and 5, which used data from both Denmark and Sweden STATISTICAL MODELLING The statistical models used in the analyses for articles 1-5 are presented in the separate articles (appendices 1-5). This section presents the statistical modelling for the supplementary analyses for article 2 on personrelated work and psychotropics. The associations between person-related work and the potential mediators emotional demands, demands for hiding emotions, and work-related violence, were examined by Poisson regression, adjusting for sex, age, cohabitation, education, income and method for data-collection. Study and wave specific relative risks were calculated in SAS using proc genmod, as explained by Spiegelman and Hertzmark (111). Poisson regression with robust standard errors was used as the log-binomial model failed to converge. The relative risks were pooled by fixed effect meta analysis in R version ( using the meta package (112). The analyses with psychotropics use as outcome estimated cause-specific hazard ratios for antidepressants-use or anxiolytics/hypnotics-use, treating the competing outcome as censoring. Participants Madsen 2012 Social Work Environment and Mental Health 30

37 were followed from questionnaire response until first psychotropics purchase, migration, death, or end of follow up (after 5.3 years/1,938 days), whichever came first. End of follow up was fixed at the maximum available for all studies to decrease between study variation. The hazard ratios were estimated using Cox regression adjusting for sex, age, cohabitation, education, income and method for data collection. The models were stratified by an indicator of study and wave to account for clustering within studies. Mediation was assessed by the magnitude of the controlled direct effect in relation to the total effect. The absence of a controlled direct effect indicated full mediation. The robustness of the results was examined in sensitivity analyses further adjusting for quantitative demands, influence, social support at work, and mental health at baseline as continuous covariates. Also, post hoc analyses tested for interactions (departure from multiplicativity) between the potential mediators and type of work, to assess whether the controlled direct effect could be interpreted as a natural direct effect. Given statistically significant interaction, the controlled direct effects within each level of the mediator were estimated CONFOUNDER SELECTION For all analyses in this thesis, potential confounders to be statistically controlled were chosen a priori based on existing literature. All analyses adjusted for sex, age, cohabitation, and one or more indicators of socioeconomic position (education, income, occupational position) as these are known risk factors for common mental disorder (26; ) that could be related to the examined work environment exposures. Analyses in article 4 further adjusted for baseline health indicators such as mental health, self-rated health and sickness absence, and analyses in article 5 adjusted for baseline mental health in a sensitivity analysis. These adjustments were included in some but not all analyses, as baseline health might be a consequence of the working environment, and these indicators could be considered mediating factors. Madsen 2012 Social Work Environment and Mental Health 31

38 3. SUMMARY OF FINDINGS This section summarizes the findings from the project. The full articles are included as appendices 1-5. This section presents a summary of the findings in the articles, and the supplementary analyses for article 2 regarding the association between person-related work and psychotropic treatment ARTICLE 1: WORK-RELATED VIOLENCE AND PSYCHOTROPICS Article 1 (see appendix 1) examined whether employees exposed to work-related violence are more likely to enter treatment with psychotropic medication MAIN FINDINGS The article s table 2 shows the cause-specific hazard ratios (HRs) for entering treatment with antidepressants, antidepressant and anxiolytics, anxiolytics, and hypnotics only. Employees exposed to work-related violence were at increased risk of entering treatment with antidepressants alone (HR=1.38, 95% CI: ) or in combination with anxiolytics (HR=1.74, 95% CI: ). There was no association, however, between work-related violence and entering treatment with anxiolytics alone (HR=1.05, 95% CI: ) or hypnotics only (HR=1.05, 95% CI: ). As a time-dependent effect of work-related violence on anxiolytics treatment was indicated, I also fitted an extended analysis with a time-dependent effect on anxiolytics, with differing effects before and after 900 days. This analysis also found no substantial association between workrelated violence and anxiolytics. The HRs before and after 900 days were 1.11 (95% CI: ) and 0.91 (95% CI: ), respectively. Madsen 2012 Social Work Environment and Mental Health 32

39 SENSITIVITY ANALYSES To examine the robustness of the results I conducted four sensitivity analyses, none of which substantially altered the findings. The results were robust to adjustment for occupational group, use of pain medications, and applying the survey date as baseline for follow up. Adjusted for occupational group the hazard ratio for antidepressant treatment was 1.36 (95% CI: ). When adjusting antidepressant treatment for regular use of pain medications, the hazard ratio for work-related violence was 1.51 (95% CI: ). When applying the survey-date as baseline, the hazard ratios for antidepressant treatment alone and combined with anxiolytics were 1.22 (95% CI: ) and 2.21 (95% CI: ), respectively. Also, results were largely similar when examining the separate effects of threats of violence and physical assault; the hazard ratios for antidepressant treatment were 1.45 (95% CI: ) for threats and 1.29 (95% CI: ) for physical assault. For antidepressants and anxiolytics combined, the hazard ratios were 1.62 (95% CI: ) for threats and 1.90 (95% CI: ) for physical assault NUMBER OF VIOLENT EPISODES In a sub-sample of the data, information on number of violent episodes was also recorded. The article s table 3 shows the cause-specific hazard ratios for psychotropics treatment in relation to number of violent episodes. There was no trend of increasing risk of psychotropic treatment with increasing number of violent episodes. Although the risk estimates for anxiolytics tended to increase with increasing numbers of episodes, the logrank test for trend was statistically nonsignificant (P=0.35). Madsen 2012 Social Work Environment and Mental Health 33

40 3.2. ARTICLE 2: PERSON-RELATED WORK AND PSYCHOTROPICS Article 2 (see appendix 2) examined the association between person-related work and antidepressants, and whether any effects may be mediated by emotional demands at work, demands for hiding emotions, and workrelated violence. This article was the first written, and it was based on the data available at an early stage in the project (DWECS 2000). Due to the potential limitations of this data, I have re-examined the association between person-related work and antidepressants, using data from all relevant studies (COPSOQ II, PUMA I- III, DWECS 2000). This supplementary analysis is also expanded to include associations between personrelated work and anxiolytics/hypnotics. This section summarises the findings from article 2, and presents the supplementary results SUMMARY OF ARTICLE 2 Employees in person-related work (health care, educational, social, customer service, or other person-related) were more likely to experience the potentially mediating work environment factors emotional demands, demands for hiding emotions and work-related threats and violence as shown in the article s table 2. The only exception was customer service work which was associated with demands for hiding emotions and threats, but not emotional demands or violence. The article s table 1 shows that health care workers were at increased risk of antidepressant treatment with an odds ratio (OR) of 1.70 (95% CI: ). Also, educational workers had an elevated risk, though not statistically significantly (OR= 1.37, 95% CI: ). No increased risk was observed for social workers, customer service workers, or employees in other person-related work. Madsen 2012 Social Work Environment and Mental Health 34

41 High emotional demands and demands for hiding emotions were associated with antidepressant treatment with respective ORs of 1.51 (95% CI: ) and 1.26 (95% CI: ). Work-related threats or violence were not associated with antidepressant treatment in this population. The risk estimate for health care work was reduced to from 1.70 to 1.47 and became statistically insignificant when adjusting for emotional demands. As seen from the article s table 4, this adjustment reduced the risk estimate for educational work from 1.37 to The risk attenuations obtained by adjustment for demands for hiding emotions were smaller than those for emotional demands SUPPLEMENTARY ANALYSIS The supplementary analysis confirmed that employees in person-related work are at increased risk of high emotional demands, demands for hiding emotions, and work-related threats or violence (see table 5). Also, the supplementary analysis confirmed an increased risk of entering antidepressant treatment for health care workers (HR=1.31, 95% CI: ) as shown in table 6. The supplementary analyses further showed an increased risk for social workers (HR=1.48, 95% CI: ), and reiterated a statistically non-significant tendency of increased risk for educational workers (HR=1.18, 95% CI: ). Emotional demands and demands for hiding emotions were associated with antidepressant treatment, with similar risk estimates of 1.05 (95% CI: ) and 1.05 (95% CI: ) per unit increase. Work-related violence was associated with antidepressants with a hazard ratio of 1.31 (95% CI: ). Madsen 2012 Social Work Environment and Mental Health 35

42 Table 5. Supplementary analysis: relative risks of high emotional demands, high demands for hiding emotions and violence, in relation to type of work Emotional demands Demands for hiding emotions Work-related Violence RR a 95% CI RR a 95% CI RR a 95% CI Type of work Health care Education Social Customer service Other personrelated Non-personrelated 1 (ref) - 1 (ref) - 1 (ref) - Abbreviations: RR: Relative Risk; CI: Confidence Interval. astatistical model includes: sex, age, cohabitation, education. Separate risk estimates calculated by Poisson regression using proc genmod for each study/wave, which were merged by fixed effect meta analysis. Adjustment for the potential mediators resulted in weak risk reductions as shown in table 6. The largest reduction was found for emotional demands, which changed the HR for health care work from 1.31 to 1.21, for educational work from 1.18 to 1.07 and for social work from 1.48 to Table 7 shows the HRs for entering treatment with anxiolytics or hypnotics in relation to type of work and the potential mediators. There were no associations between person-related work, or any of the potential mediators, and anxiolytic or hypnotic treatment SENSITIVITY ANALYSES Adjustment for quantitative demands, influence, social support at work, and mental health at baseline did not change results (data not shown). Post hoc analyses testing for interaction (departure from multiplicativity) between type of work and the potential mediators on antidepressant treatment found interaction between emotional demands and type of work. Stratified analyses, revealed that this interaction pertained to social Madsen 2012 Social Work Environment and Mental Health 36

43 Madsen 2012 Social Work Environment and Mental Health 37 Table 6. Supplementary analysis: hazard ratios for antidepressant treatment in relation to type of work and potential mediators Type of work Adjusted for sociodemographics (Model 1) Model 1 adjusted for emotional demands Model 1 adjusted for demands for hiding emotions Model 1 adjusted for workrelated violence HR a 95% CI HR a 95% CI HR a 95% CI HR a 95% CI Health care Education Social Customer service Other person-related Non-person-related 1 (ref) - 1 (ref) - 1 (ref) - 1 (ref) Emotional demands Score 0-10, per 1 unit increase High vs. low Demands for hiding emotions Score 0-10, per 1 unit increase High vs. low Work-related violence Yes vs. no Abbreviations: HR: Hazard ratio; CI: Confidence interval. a Adjusted for sex, age (continuous), cohabitation, education (continuous), income (continuous), method, study/wave (strata variable) badjusted for model 1+mediator (separately)

44 Table 7. Supplementary analysis: cause-specific hazard ratios for anxiolytics/hypnotics treatment in relation to type of work and mediators HR a 95% CI Type of work Health care Education Social Customer service Other person-related Non-person-related 1 (ref) - Emotional demands score 1-10, per 1 unit increase High vs. low Demands for hiding emotions score 1-10, per 1 unit increase High vs. low Work-related violence Yes vs. no Abbreviations: HR: Hazard ratio; CI: Confidence interval. aadjusted for sex, age, cohabitation, education, income, method, data (as strata variable) Madsen 2012 Social Work Environment and Mental Health 38

45 workers and employees in other person-related work. These two types of work showed a direct effect on antidepressants only in employees with high emotional demands, with respective HRs of 1.78 (95% CI: ) and 1.39 (95% CI: ). The respective HRs in employees with low emotional demands were 0.92 (95% CI: ) and 0.93 (95% CI: ) COMPARING FULL SAMPLE ANALYSIS TO ARTICLE 2 Overall, both analyses showed increased antidepressant treatment amongst health care workers, and some indications of increased treatment amongst educational and social workers. Also, the increased risk of the potential mediators amongst employees in person-related work was consistent. Both analyses showed that emotional demands and demands for hiding emotions were associated with antidepressants, and the supplementary analysis showed the association between violence and antidepressants also found in article 1. Regarding mediation, article 2 indicated part mediation by emotional demands, but this finding was less clearly supported by the supplementary analyses. Madsen 2012 Social Work Environment and Mental Health 39

46 3.3. ARTICLE 3: ELDERCARE WORK AND ANTIDEPRESSANT TREATMENT Article 3 (see appendix 3) examined the association between care work, as a type of human service work, and antidepressants longitudinally. The analysis examined the prevalence of antidepressant treatment in a cohort of female eldercare workers before and after entering their profession, and compared these prevalences to those of the general female working population (selected from DWECS 2005) MAIN FINDINGS The article s figure 1 presents the prevalences of antidepressant treatment in the eldercare worker cohort (CWC) during the years Also, the figure shows the age standardised prevalences of antidepressant treatment in a representative sample of the female population gainfully employed in The prevalence of antidepressant treatment was consistently elevated amongst the eldercare workers compared to the representative sample. In the article s figure 2, the prevalences in the two cohorts were compared by agestandardized prevalence ratios (SPRs), which showed statistically significantly increased prevalence of antidepressant treatment amongst the care workers in all but two years (1997, 1998). Also, the generalized estimating equation (GEE) analysis showed that the odds of antidepressant treatment for the care workers increased log-linearly throughout the period, and that there was no added effect of entering care work in To assess whether the observed differences reflected treatment seeking behaviours, I also compared the risk of suffering from depressive symptoms in 2005, between the two cohorts. The care workers were also more likely to suffer from depressive symptoms (SPR=1.28, 95% CI: ). Madsen 2012 Social Work Environment and Mental Health 40

47 SENSITIVITY ANALYSES To examine the robustness of these findings I conducted four sensitivity analyses. The findings were substantially unchanged when restricting the reference population to individuals with shorter vocational training, to control for differences in socioeconomic position. Also, results did not change when excluding CWCparticipants graduating as eldercare assistants or CWC-participants who reported that previous experience in carework was a motivation for them to undertake the eldercare training. The SPR for depressive symptoms was slightly decreased when restricting DWECS to questionnaire-data (SPR=1.17, 95% CI: ). Also, the age-stratified analysis revealed that the difference in antidepressant prevalence increased with age (see the article s figure 3) and was largely unpresent in the youngest agegroup (<30 years) ARTICLE 4: WORK ENVIRONMENT AND ANTIDEPRESSANTS IN DENMARK AND SWEDEN Article 4 (see appendix 4) examined whether the associations between work environment factors and antidepressant treatment were similar in representative samples of the Danish and Swedish working populations. More specifically, the work environment factors included were: quantitative demands, conflicting demands, emotional demands, and learning opportunities and influence at work MAIN FINDINGS The article s table 4 presents the country-specific and pooled hazard ratios for entering antidepressant treatment in relation to the examined work environment factors. High quantitative and conflicting demands were associated with antidepressant treatment after adjustment for sociodemographics with pooled hazard ratios of 1.25 (95% CI: ) and 1.38 (95% CI: ), respectively. These associations, however, Madsen 2012 Social Work Environment and Mental Health 41

48 attenuated in the final model when adjusting for depressive symptoms, sickness absence and self-rated health. Emotional demands at work were associated with antidepressant treatment in all models, with a fully adjusted pooled risk estimate of 1.45 (95% CI: ). High learning opportunities and influence at work were not associated with antidepressants. Overall, the associations between work environment factors were similar in Denmark and Sweden, as indicated by small or statistically non-significant values of I 2. There was some indication, though, that the crude effect of emotional demands was stronger in Sweden (HR=2.14, 95% CI: ) compared to Denmark (HR=1.69, 95% CI: ) SENSITIVITY ANALYSES Sensitivity analyses were conducted regarding the effects of 1) changing the dichotomizations of conflicting demands, emotional demands, and influence, and 2) modelling exposures as linear predictors. The results were similar in the sensitivity analyses, though most exposures did not show clear linear trends. Working fast, conflicting demands and emotional demands seemed to have a u-shaped relation with antidepressants, with higher risk for individuals reporting these demands never/almost never compared to seldom. Madsen 2012 Social Work Environment and Mental Health 42

49 3.5. ARTICLE 5: EMOTIONAL DEMANDS AND ANTIDEPRESSANTS: BUFFERING BY LEADERSHIP Article 5 (see appendix 5) examined whether the association between emotional demands and antidepressants is buffered by good leadership. More specifically, I hypothesized that the effects of high emotional demands and good leadership were less than additive MAIN FINDINGS The article s table 4 shows the pooled hazard ratios for incident antidepressant treatment in relation to combinations of emotional demands and quality of leadership. The group with low emotional demands and poor quality of leadership is used as reference. There was no clear indication of effect-modification. Although the risk estimate for high emotional demands and good leadership was slightly lower (HR=1.70, 95% CI: ) than that for high emotional demands and poor leadership (HR=1.84, 95% CI: ), this difference was not statistically significant, as indicated by the confidence interval for the synergy index SENSITIVITY ANALYSES The sensitivity analyses assessed whether the associations were affected by adjustment for baseline mental health, adjustment for occupational group, stratification by country, or moving the cut off point for good leadership to the best quartile of the scale. None of the sensitivity analyses substantially changed the results, although some variation was seen in the country-specific risk estimates, with opposite synergy indices in the two countries. This variation was, however, not statistically significant, although this lack of statistical significance could be due to lacking statistical power. Madsen 2012 Social Work Environment and Mental Health 43

50 4. DISCUSSION The following section starts by discussing key issues concerning the methods applied, followed by relating the results to previous findings in the field, and concludes by outlining some practical implications of the findings METHODOLOGICAL DISCUSSION This methodological discussion contains first a section on issues related to the exposure assessment, second a section on issues related to the outcome assessment, third a section on residual confounding and finally a section on recent statistical developments in the field. The discussion is not exhaustive of important methodological issues but focuses on issues pertaining particularly to this thesis, namely the use of selfreported exposure data, psychotropic treatment as an outcome measure and time-to-event analyses EXPOSURE ASSESSMENT This thesis applied self-reported exposure measures. As a non-self-reported outcome assessment was applied the common method bias was averted (94). However, the self-reported exposure assessments may have been affected by mental health at baseline, which may also be associated with future mental health (3;4;23), possibly biasing risk estimates away from the null (9). Self-reported exposure assessments have been commonly employed in work environment research, and the potential bias when examining mental health outcomes is a serious concern in the literature according to the 2008 review by Bonde (9). Since 2008, several studies have, however, examined the effects of non-selfreported work environment exposures on mental health. A US study found increased depression in jobs rated Madsen 2012 Social Work Environment and Mental Health 44

51 more demanding by a safety and hygiene manager (116). There was no effect of job control after adjusting for sociodemographics. A UK study found that observed high skill utilization decreased risk of depression and anxiety, and observed hindrances causing more than 2 hours of extra work per week were associated with increased risk, although not statistically significantly (117). A Finnish study of hospital employees showed that overcrowding in the hospital wards was associated with increased risk of entering antidepressant treatment with a dose-response association (118). In addition to observed work environment exposures, some studies have assessed work environment factors by aggregating self-reported data. These studies have mostly not found consistent associations between work environment and common mental disorder ( ), although one study showed an association with poor work climate satisfaction (122). This lack of associations may be related to methodological issues concerning the data aggregation, although it should be noted that two analyses (119;122) with the same data and exposure definitions found associations for hospitalization with common mental disorder but not for antidepressant treatment. Taken together, the observational studies do, however, indicate that the associations between demands and common mental disorder may not fully be explained by bias due to self-reported exposure measurement. Some work environment factors may, however, be less amenable to non-self-reported measurement and this thesis used self-reported exposures. According to Frese & Zapf (123) the degree of subjectivity of a self-reported measure depends on the level of cognitive and emotional processing involved. It is plausible that the effects of individuals states or traits are limited on reports such as occupation, degree of client contact, and incidences of violence. Self-reported experiences of emotional demands at work, conversely, may be more affected by the emotional processing of the respondent, and self-report of this exposure may be more vulnerable to bias by baseline mental health. Although the findings for emotional demands and antidepressants were robust to adjustment for baseline mental health indicators, residual confounding of this measure cannot be ruled out. It should be noted, though, that because emotional demands as conceptualized Madsen 2012 Social Work Environment and Mental Health 45

52 in this thesis per definition require sustained emotional effort, this concept may be complicated to measure non-self-reportedly, as the emotional efforts may be difficult to establish. Also, the effort-reward-imbalance model (13) proposes that the efforts spent at work depend not only on the demands of the external environment but also individual characteristics. Consequently, the phenomenon of emotional demands, if per definition linked to emotional efforts, may not be possible to measure separate from the individual. Further research exploring the concept of emotional demands, and the potentially causal links with mental health, is warranted OUTCOME ASSESSMENT This thesis examined risk of entering psychotropic treatment. The advantages of using this register-based outcome measure were the minimal loss to follow up, the day-to-day measurement, and that the measure is non-self-reported so the common method bias is avoided. Also, as the measure is based on a prescription from a medical professional, it is likely to validly measure clinically significant health conditions. The risk of entering treatment with psychotropics does not, however, correspond directly to the risk of developing mental disorder. This discrepancy exists for three reasons: First, most cases of mental disorder are untreated (1;96). Second, treatment options are not limited to pharmaceuticals, but also include psychotherapy (1;124). Third, the indications for psychotropic treatment include physical problems such as neuropathic pain (125;126). Hence, the results of this thesis cannot directly be interpreted in terms of effects on risk of common mental disorder, due to misclassification of the outcome. To the extent that such misclassification is nondifferential in relation to the exposure it may cause an underestimation of the examined associations (40). Finnish analyses have in deed suggested that effects of social support on mental health were less clear when Madsen 2012 Social Work Environment and Mental Health 46

53 examining antidepressant treatment compared to diagnosed depression or anxiety (127). Similarly, a Danish study found that associations between antidepressant treatment and work environment factors (quantitative demands, social support from colleagues) were strengthened when applying statistical methods accounting for misclassification (128). In case of differential misclassification, the examined associations may, however, be overestimated (40). The following sections of the thesis examine two main sources of misclassification when using psychotropic treatment as a proxy for mental disorder, namely 1) factors influencing the probability of entering psychotropic treatment given mental disorder (affecting the sensitivity of the measure), and 2) the multiple indications for treatment with psychotropics (affecting the specificity of the measure). Determinants of Health Service Utilization. Psychotropic treatment is - not surprisingly - more frequent in individuals suffering from mental disorder (96;124; ). Many factors, however, affect if individuals with mental disorder enter psychotropic treatment (96;124; ). According to the behavioural model (133;134), health service utilisation is influenced by both the health services system and various social and individual determinants, as illustrated in figure 2. As the health services system and the social determinants were relatively fixed in the data for this thesis, the individual determinants of health service use are of particular concern. According to the model, the most proximal individual determinant of health services use is the illness level or need, which is affected by enabling resources and predisposing characteristics (134). Illness level is composed of the individual s perceived illness level and the level evaluated by the medical practitioner (134). Before using health services, the individual must perceive illness or the probability of illness, but the clinical judgment of the health care practitioner also affects the type and extent of health care provided (134). Examples of perceived illness level are self-reported symptoms and level of disability, whereas evaluated illness level is ideally reflected by a clinical diagnosis (134). Enabling resources are means that Madsen 2012 Social Work Environment and Mental Health 47

54 Figure 2. The behavioral model of health service utilization. Adapted from Andersen & Newman (133,134) enable the individual to use health services (134). Examples of such resources are income, health insurance and availability and accessibility of health services (134). Predisposing factors affecting health service use are associated with a person s propensity to service use and pre-exist the specific illness episode (134). Examples of predisposing factors are sex, education, health beliefs and previous illness episodes (134). Predictors of mental health service use. Empirically, mental health service use in individuals with mental disorder has been found increased with indicators of higher illness level such as more severe disorder (97), higher disability (97), mental comorbidity (97), suicide thoughts (135), chronic health conditions (135), high daily stress (135) and parental mental problems (97). Enabling (or rather impeding) factors reported to hinder mental health service use include low social support (135), waiting times (135;136), costs (135;136), no available professional help (135;136), and problems with child care and transportation (135;136). Predisposing factors associated with decreased mental health service use in individuals with mental disorder include low education (135) and single parenthood to an adult child (135). Increased use has been found in individuals living alone (97), with previous depressive episodes (135) and positive attitudes to mental Madsen 2012 Social Work Environment and Mental Health 48

55 health help-seeking (137). Similarly, non-use of mental health services may be related to attitudinal acceptability barriers such as preferring to manage oneself, not thinking anything more can help, and fear of what others might think (136). Three studies found no effects of sex on mental health service use (97;135;138). In a large European study, positive attitudes towards mental health treatment seeking were associated with being female, under 65 years, and having higher income, but there was no independent association with years of education or employment status (137). Also, having a mood disorder and previous use of mental health care were associated with more positive attitudes, whereas having an anxiety disorder, previous mental disorder, emotional role impairments and a parent with a psychiatric history were not (137). A Canadian study (136) found that acceptability barriers to mental health service use were less likely in individuals with higher education, income and age. Conversely, increased acceptability barriers were found in employed individuals, with higher self-reported distress and parents with a partner (136). Predictors of psychotropic treatment. Psychotropic treatment is a specific type of mental health service use, and factors affecting the type of service entered may differ from factors affecting use or non-use of health services ( ). Psychotropic treatment has also been associated with illness level indicators such as more severe disorder (124), greater disability (124), physical comorbidity (130;132), no physical comorbidity (131), mental comorbidity (124;132), more medical consulations (132), and help-seeking for emotional or mental problems (130). In relation to enabling factors, psychotropic treatment has - perhaps counterintuitively - been found increased with lower income (124) and unemployment (139). Predisposing factors for psychotropic treatment include being female (96;124;129;130), older (124;130), previous psychotropic use (140), and having higher education (antidepressants) (129), or lower education (benzodiazepines) (130). Antidepressant treatment has also been found increased in single individuals with depression (124). Madsen 2012 Social Work Environment and Mental Health 49

56 Health services utilization may also be affected by medical practices of the health care provider (134). Correspondingly, a Danish study of general practitioners (GPs) (141) found substantial variation in antidepressant prescription; antidepressants were more frequently prescribed by GPs with more years of practice, who worked in a solo-pratice, with higher use of counselling sessions, and who were more prone to prescribe medications in general (general prescribing attitude). Finnish data (124) suggest that psychiatrists are more likely to treat depressed patients psychologically or with antidepressants. Primary care providers, conversely, more frequently prescribed one of either anxiolytics, hypnotics or antipsychotics (124). Although recent Danish data on providers of psychotropic treatment are lacking, figures from the early 90 s suggest that at this time, 94% of purchased antidepressants were prescribed by a GP (142). A more recent European study also found that the GP was the most frequent prescriber of psychotropics; the GP was the prescriber in 81.5% of hypnotics treated individuals, 69.8% of anxiolytics treated individuals and 55.8% of those treated with antidepressants (132). The type of health care provider consulted may also be affected by a range of factors, and Finnish data suggest that specialist treatment is more common in cases with greater severity and comorbidity (97). Also, a Canadian study (135) found that treated women were less likely than men to be in specialist care. Amongst women, specialist treatment was decreased in middle-aged (30-59 years) women and those who had lived in Canada less than 10 years, whereas it was increased for women with low income adequacy, poor selfrated health, high stress-levels, chronic health conditions, low social support, and multiple depressive episodes (135). Due to few men in the sample, predictors of type of treatment could not be examined among men. Madsen 2012 Social Work Environment and Mental Health 50

57 Occupation and treatment seeking. A particular concern for this thesis are occupational patterns in treatment seeking. Different norms, health beliefs and insurance coverage might affect health services utilization (134), and could be occupationally patterned; if so, such patterns might explain the observed increased risks of antidepressant treatment found in health care and social workers. There is no firm evidence on treatment seeking for mental disorder in relation to occupation, but a Norwegian study of ambulance personnel (143) found that when controlling for symptoms of depression and anxiety, these employees were less likely than the general working population to seek treatment from a GP, other physician, physiotherapist, or occupational health practitioner. They were, though, more likely to seek a chiropractor. Unfortunately, the probabilities of seeking a psychologist or psychiatrist could not be compared (143). Although ambulance personnel were classified as health care workers in the present analyses, these findings may, however, not be generalizable to other health care workers as ambulance personnel are predominantly male, and the professional culture is relatively masculine (143). Contrastingly, other care professions, such as nursing, are predominantly female (144). A Canadian survey of physicians reported very high levels of depressive symptoms, and 14.5% of participants did not know of treatment resources they would feel comfortable using (145). The American Medical Associations reports that physicians may be particularly hesitant to seek treatment for mental disorder due to concerns about confidentiality, and the effects on medical licensing and career prospects (146). Danish data also show increased suicide risk amongst doctors and nurses, and this increased risk is not explained by differences in psychiatric morbidity (47). Such increased suicide risk might indicate inadequate treatment (147), although greater access to lethal medications may also be involved (47). A US study (148) examined helpseeking for misuse of alcohol and other drugs amongst social workers and reported that 24% of participants with problematic use of alcohol or other drugs had sought help. Reasons for not seeking help included confidentiality concerns, lack of acceptable providers nearby and feeling uncomfortable because of knowing the provider personally or professionally (148). Madsen 2012 Social Work Environment and Mental Health 51

58 Together, these previous findings do not suggest increased treatment seeking amongst human service professionals. If employees in human service work do, however, have more positive attitudes towards or greater accessibility to psychotropic treatment, this may have biased the results of this thesis upwards. Article 3 assessed whether treatment seeking behaviours explained the increased antidepressant treatment in female eldercare workers by also comparing self-reported mental health between the cohorts. These findings indicated that the eldercare workers also had poorer self-reported mental health, suggesting that occupational patterns in treatment seeking do not explain results. Nevertheless, treatment seeking behaviours are a substantial concern in interpreting the results of this thesis. Indications for psychotropic treatment. In addition to treatment seeking behaviours, which may cause differential sensitivity of psychotropic treatment as a measure of common mental disorder, the specificity of the outcome measure may be problematic. The examined psychotropics are used for a range of disorders and one type of treatment does not correspond to one type of disorder (99-101). Antidepressants are used to treat depression but also generalized anxiety disorder, social phobia, anxiety or panic disorder (100), bulimia nervosa, obsessive-compulsive disorder (100), posttraumatic stress disorder, premenstrual dysphoric disorder (126), neuropathic pain (100), headache (100;125), psychotic disorders (125), dementia (125), sleep disorders (100) and incontinence (100). A Danish study from 1995 (99) found that amongst Danish GPs, the most frequent indication for antidepressant prescription was depression (73%), followed by neuropathic pain (7%) and anxiety/nervousness (4%). It should be noted, though that the types of antidepressants prescribed have changed somewhat since the mid 1990 s, as the selective serotonin reuptake inhibitors (SSRIs) have grown from 80% to 88% of first line antidepressants prescribed during the years (149). A more recent study from the US (101) found that the three most frequent reasons for prescribing antidepressants amongst physicians were depression (65.3%), anxiety disorders (16.4%) and Madsen 2012 Social Work Environment and Mental Health 52

59 attention-deficit or conduct disorder (2.8%). In total, the indication was a mental disorder in 92.7% of the antidepressants prescribed (101). An Italian study from 2007 found that the main indications for prescribed antidepressants was depressive disorders (50.1% for tricyclic antidepressants (TCAs); 67.2% for SSRIs), anxious disturbances (13.8%; 10.6%), and bipolar disorders (8.5%; 9.2%) (125). A Dutch study from 2007 found that the three most frequent indications for antidepressant prescription amongst general practitioners were depression (45.5%), anxiety disorders (17.2%) and sleep disorders (9.3%) (100). A European study from 2002 (132) examined self-reported reasons for psychotropic treatment and found that the most commonly stated reasons for antidepressant treatment were depression (30.7%), sleep problems (25.5%), anxiety (24.5%), or sleep problems and anxiety (13.4%) (132). When antidepressants were taken in combination with anxiolytics, the most commonly stated treatment reasons were sleep problems, anxiety and depression (45.1%) and sleep problems and anxiety (33.8%) (132). The first line treatments for anxiety disorders in Denmark are antidepressants and/or psychotherapy (150), but anxiolytic treatment may be indicated for short term treatment in cases of acute crisis (151;152). Similarly, Danish guidelines (152) state that hypnotic treatment should only be initiated in patients with substantial impairment, who have not responded to other treatments, and the treatment should be reassessed after 1-2 weeks. A US study (101) found that the three most frequent reasons for prescribing anxiolytics were anxiety disorders (39.6%), mood disorders (18.9%) and medical examination or evaluation (6.0%). A mental disorder was the indication for 67.7% of the anxiolytic prescriptions (101). A European study (132) found that, amongst individuals treated with anxiolytics, the most commonly stated reasons for treatment were sleep problems (52.9%), anxiety (24.5%), and sleep problems and anxiety (15.8%). For hypnotics, the most commonly stated reasons were sleep problems (89.1%), sleep problems and anxiety (4.1%), and sleep problems and depression (3.1%). In summary, the examined psychotropics are mainly used for common mental disorders but there is a lack of specificity in relation to type of disorder, as each psychotropic has multiple indications. This lack of Madsen 2012 Social Work Environment and Mental Health 53

60 specificity may result in an underestimation of the examined effects, if non-differential in relation to examined exposures (40). Given previous findings showing non-differential specificity of incident antidepressant treatment as a measure of depression in relation to sex, age and socioeconomic position (129) such nondifferentiality is plausible. The cited studies on indications for psychotropic treatment do, however, emphasize that any type of psychotropic treatment does not directly correspond to any type of common mental disorder RESIDUAL CONFOUNDING The risk estimates found in this thesis were generally modest in size, and their magnitude raise the concern of bias by residual confounding. The analyses adjusted for several risk factors for common mental disorder, including sociodemographics (sex, age, cohabitation), socioeconomic indicators (education, income, occupation), and other work environment factors (quantitative demands, influence, social support). A number of factors, however, could not be accounted for, including genetics, personality, self-esteem, previous life experiences, and private life factors such as marital problems and low private life social support. These factors have all been associated with common mental disorder (3;4), and could have confounded effects if they were associated with the examined work environment exposures. The analyses could not account for genetic factors which may be related to vulnerability to common mental disorders (5;6). If these factors are associated with the examined work environment exposures, they may have biased risk estimates upwards. Such association is possible as genetics have been related to personality factors and self-employment (153). The direction and magnitude of any confounding by genetics is, however, difficult to gauge, given lacking knowledge of specific genetic components affecting work characteristics, and how, or if, these factors are associated with common mental disorder. Madsen 2012 Social Work Environment and Mental Health 54

61 The analyses also could not account for previous life experiences and private life factors such as marital problems and low social support. However, many of these factors cluster with social disadvantage (114), and the adjustment for socioeconomic indicators may have provided some, albeit crude, account for these potential confounders. Work environment studies occasionally control for indicators of self-esteem or personality, for instance neuroticism or negative affectivity; previous analyses have for instance showed that the association between elements of the job strain model and common mental disorder was robust to adjustment for negative affectivity in the Whitehall II study (154). It is, however, controversial whether such adjustment is warranted (155;156); baseline measures of these factors may be affected by the social environment of the individual (28;156;157) and consequently these factors may be considered mediating factors rather than confounders. Accordingly, it is preferable to use measurements prior to the exposure assessment, when adjusting for these factors. Such adjustment was unfortunately not possible given the available data. Hence, the potential for residual confounding by personality factors should be considered when interpreting the findings, although adjustment for baseline mental health, when present, likely reduced such bias. Equally, the analyses did not adjust for lifestyle indicators such as physical activity and alcohol use. Although these indicators have been associated with common mental disorder (158;159) they may be affected by the working environment (160). Consequently, adjustment for lifestyle before exposure to work environment was preferable, but impossible given the data available. Several meta-analytic analyses, however, show only minor associations between adverse work environment (job strain) and lifestyle (160;161), suggesting that residual confounding due to these factors may be modest. Madsen 2012 Social Work Environment and Mental Health 55

62 RECENT STATISTICAL DEVELOPMENTS Article 2 examined whether the effects of person-related work on antidepressant treatment were mediated by work environment factors. This mediation was operationalized as the absence of a controlled direct effect of person-related work on antidepressants. It should be noted, however, that in the context of time-to-event analysis, this method has been criticised (162). A central issue is that in the presence of interaction, the controlled direct effect is not equivalent to the natural direct effect (108). Because the natural direct effect corresponds to an indirect effect, the mediated proportion may be estimated from the natural direct effect. Although the indirect effect, or the mediated proportion, may not be quantified from the controlled direct effect with time-to-event outcomes (108), controlled direct effects can verify the presence or absence of an indirect effect (108). Methods for the estimation of mediated proportions outside linear regression analysis are currently developing, and one example is using the Aalen additive hazard model rather than the Cox proportional hazards model as the underlying statistical model (162). Within this framework, it is possible to estimate the number of events attributable to the examined mediator, and thus to quantify the mediated proportion using an additive statistical model rather than a multiplicative. Similarly, it should be noted that the examination of effect-modification as departure from additivity in article 5 might also have benefited from using an additive statistical model and testing for statistical interaction, rather than calculating the synergy index based on risk estimates from the multiplicative Cox proportional hazards model. Although Rothman s synergy index is an established method for examining departure from additivity, it is possible that the use of an additive statistical model such as the Aalen additive hazard model, may have been more statistically efficient and resulted in narrower confidence limits for the interaction effect. Also, such estimation would not have required dichotomising the two exposures. Hence, this procedure may have revealed some statistically significant effect-modification, as the pooled risk estimate in employees with high emotional demands was slightly lower in those with good leadership (HR=1.70, 95% CI: ) Madsen 2012 Social Work Environment and Mental Health 56

63 compared to those with poor leadership (HR=1.84, 95% CI: ). However, there was still substantially increased antidepressant treatment in employees with high emotional demands and good leadership. Hence, the results do not support the hypothesis that good leadership is a substantial buffer and it is unlikely that this overall conclusion would be changed considerably by using a more statistically efficient additive model RELATIONS TO PREVIOUS FINDINGS This thesis found that work-related violence, employment in health care or social work, and high emotional demands at work predicted antidepressant treatment. There was also some indication of an effect of educational work and high quantitative or conflicting demands, although the latter association attenuated with adjustment for baseline depressive symptoms. Previous studies examining work environment factors as predictors of antidepressant treatment have found increased antidepressant treatment in individuals with high job insecurity (163), low social support at work (127), poor team climate (164), high job dissatisfaction (165), high quantitative demands and low social support from colleagues in men only (128), high job demands or job strain in men (166), high job demands and low job control in women (165), poor workplace climate in men (165), and high psychological demands and overtime work in blue collar workers, and low psychological demands and shift work in white collar workers (167). The association between work-related violence and antidepressants is consistent with previous smaller occupation-specific studies showing associations between work-related violence and self-reported mental health outcomes including general well-being (37), emotional exhaustion (32), feelings of sadness, helplessness, lack of control, depression (33;34;38); sleep problems (33), anxiety (34), and depression (35). This finding is also consistent with a large register based study showing increased hospitalization with depression or anxiety in employees in occupations with high levels of work-related violence (39). The Madsen 2012 Social Work Environment and Mental Health 57

64 association between work-related violence and antidepressants was robust to adjustment for a range of factors including occupation and, together with previous findings, suggests that there may be an effect of workrelated violence on clinically significant common mental disorder. It is, however, unclear whether the findings indicate a specific effect of work-related violence on depression or anxiety, or a generalized effect on disorders treated with antidepressant medication. Given the increased hospitalization with both depression and anxiety in occupations with high risk of violence (39), the effects of work-related violence on common mental disorders may be general rather than specific. The increased antidepressant treatment in health care, educational and social workers is consistent with previous literature theorizing that human service work may be particularly straining for employee mental health (44;51). If this increased risk was caused by the work, however, it was expected to be mediated by some of the examined work environment factors which had particularly high prevalence in these professions. Such mediation was not consistently found. This thesis did, though, not examine all potential mediators; other possible mediators include for instance shift work (168), although findings regarding this exposure and common mental disorder are inconsistent (165). Alternatively, the increased risk may not be mediated by the working environment but related to factors outside the workplace. Although the longitudinal analyses of antidepressant treatment in eldercare workers could only examine shorter-term effects of eldercare work, the analyses suggested that selection might be involved. Such selection is consistent with a British study (169) which found that internalising problems in childhood predicted lower status jobs at 42 years, including being a personal service worker. Internalising problems in childhood also predicted lower job demands and lower decision latitude at 45 years, whereas psychological distress in young adulthood predicted higher job demands, lower decision latitude, lower social support and higher job insecurity at 45 years. It should be noted, though, that the present selection of eldercare workers seemed unexplained by general socioeconomic status, suggesting that specific mechanisms concerning care work might be involved. The psychological concept of parentification (170), Madsen 2012 Social Work Environment and Mental Health 58

65 which might lead individuals to seek employment in care work professions ( ) and increase risk of depression (175), offers one possible explanation. Previous research on mental health and selection into human service occupations partially contrast the findings of this thesis. A US study from 1994 (46) examined psychological health in adolescence amongst employees in occupations with high interpersonal demands (equivalent to health care, educational, and social work in the present analyses), moderate interpersonal demands (equivalent to customer service work) and low interpersonal demands (equivalent to non-person-related work). The employees were recruited for the study in early childhood, and there were two separate cohorts of individuals born in (n=73) and (n= 100), respectively. The authors hypothesized that occupations with high interpersonal demands were particularly psychologically straining, and that employees in these occupations might have had better mental health in adolescence than employees in occupations with lower interpersonal demands, due to both selection and attrition. Notably, this hypothesis is opposite those of this thesis which expected poorer mental health in human service workers. The authors hypothesis was partially confirmed by their analyses, but only for the oldest cohort. For the youngest cohort there were no statistically significant differences in adolescent mental health in relation to interpersonal demands in adulthood occupation. These previous findings are opposite those of this thesis which found indications of reduced mental health in employees in health care, educational work and social work. This difference is possibly explained by cohort specific patterns in selection as the US study found associations only in the eldest cohort. Given these cohort specific patterns it is likely that selection patterns may have changed from the 1950 s to when the data for this thesis were collected. The association between working fast and conflicting demands and antidepressants might support an association between psychological demands and depression, as reported previously (9-11). As the association was, however, attenuated by adjustment for baseline health indicators, this finding underlines the problems associated with self-reported exposure measures (9). The association between emotional demands and antidepressants was robust to adjustment for baseline health indicators, and is consistent with previous Madsen 2012 Social Work Environment and Mental Health 59

66 findings relating emotional demands to self-reported psychological distress (60;61), anxiety symptoms (62) and depressive symptoms (62). The findings are also in line with a large register-based study which found that women employed in occupations with higher levels of emotional demands were at increased risk of psychiatric treatment for depression (63). Although the association between emotional demands and antidepressants was robust to adjustment for baseline mental health indicators, it is still unclear whether the association reflects a causal effect. Given the present findings regarding selection into eldercare, an occupation with high levels of emotional demands, it is possible that selection might cause employees in occupations with high emotional demands to have different trajectories in odds of antidepressant treatment over time compared to other employees. In that case, even adjustment for baseline mental health indicators and occupation might not fully account for non-workrelated mental health differences. This thesis did not find indications that the effects of emotional demands on antidepressants were substantially buffered by good leadership. This is in contrast to the conservation of resources theory (176), which states that high availability of valued resources reduces the risk that resource expenditure results in adverse health outcomes. The measure used for leadership quality, however, was partly based on the availability of similar items in the two studies. Hence, it was not a validated scale, and may not have captured the concept of leadership quality satisfactorily. If emotional demands at work are caused by particular characteristics of the work, for instance the confrontation with clients problems and suffering, the prevention of emotional demands may not be feasible. In case of a causal effect of emotional demands on employee mental health, the identification of other potential buffers which are modifiable is an important target for future research. Madsen 2012 Social Work Environment and Mental Health 60

67 This thesis found no association between work environment factors and anxiolytics or hypnotics. A recent Finnish study reported increased anxiolytic or hypnotic treatment in men working overtime, women with desk top work, and individuals with highly mentally strenuous work and high job dissatisfaction (165). Previous studies have also found associations between social phobia and job strain or job insecurity (177); anxiety disorders and low social support from colleagues or supervisors, work/family imbalance, high psychological demands and low control (178); generalised anxiety disorder and high psychological job demands (179); and subclinical anxiety and high psychological demands, low social support, conflicts with supervisors and high emotional demands (180). Sleep problems have been linked with organizational injustice (181), and insomnia symptoms have been associated with control over days off, and control over daily working time in men (182). This apparent inconsistency may be related to the indications for anxiolytic and hypnotic treatment. According to Danish guidelines, the first line pharmacological treatment for anxiety disorders is not anxiolytics but antidepressants (150). Anxiolytic treatment is recommended only in cases of acute crisis (151). Hence, any effects of the examined exposures on anxiety disorders may not manifest in effects on anxiolytics but in effects on antidepressants. Regarding the discrepancy to the Finnish study which found associations between several work environment factors and treatment with anxiolytics or hypnotics, it should be noted that the Finnish study did not differentiate anxiolytics cases according to whether individuals were also in antidepressant treatment, as the present analyses did. Hence, it is possible that the Finnish findings were largely driven by an effect on depression, given the substantial comorbidity (183). Regarding the previous findings on sleep disorder, Danish guidelines state that hypnotic treatment is only indicated in sleep disorders associated with substantial impairment that have not responded to other treatments (152). Hence, it is possible that the discrepancy between current and previous findings may be due to differences between predictors of sleep problems generally and predictors of clinically significant sleep disorder requiring pharmaceutical treatment. Madsen 2012 Social Work Environment and Mental Health 61

68 4.3. PRACTICAL IMPLICATIONS The association between work-related violence and antidepressant treatment found in this thesis, together with previous findings on the association of work-related violence and poor mental health, indicates that the prevention of work-related violence may help prevent the development of common mental disorder amongst employees. Interventions that may reduce the risk of workplace violence include training staff to better manage aggressive clients and to recognize risk factors for violence, and providing workplace security staff or surveillance (184). It should be noted though, that the evidence of the effectiveness of these strategies in preventing workplace violence has been mixed (184). The consistent association between emotional demands and antidepressant treatment found in this thesis warrants further inquiry. The experience of emotional demands at work may be influenced by baseline mental health, and the association may be affected by confounding, despite adjustment for a baseline mental health indicator. Also, given the findings regarding selection into eldercare work, the association may be biased by pre-existing differences between individuals who experience their work as emotionally demanding and those who do not. Clarifying the concept of emotional demands at work, and the factors underlying this phenomenon, is an important target for future research. Equally, further research should aim to identify potential buffers of any causal effects of emotional demands on common mental disorder. The findings indicate that health care, social, and to some extent educational workers may be at increased risk of common mental disorders that are treated with antidepressant medication. Regardless of whether this increased risk is causally related to the working environment, these employees appear to be a high-risk group for common mental disorder. Consequently, secondary prevention, ensuring the detection and treatment of common mental disorder in these employees, is a public health concern that may help prevent common mental disorder in the employed population. Madsen 2012 Social Work Environment and Mental Health 62

69 5. CONCLUSION To conclude, this thesis finds that employees exposed to work-related violence, with high emotional demands, or working in health care, social and to some extent educational work, have increased risk of entering antidepressant treatment. There were no associations between work environment factors and treatment with anxiolytics or hypnotics. It is, however, unclear whether the associations between health care, educational and social work and antidepressant treatment reflect a causal effect of the work, as selection of vulnerable individuals into these professions may be involved. The patterns of associations between work environment factors and antidepressant treatment were largely similar in Denmark and Sweden, substantiating the generalizability of the findings. There was no indication that the effects of emotional demands on incident antidepressant treatment were substantially buffered by good leadership. Unfortunately, the multiple indications for psychotropic treatment and issues regarding treatment seeking bias impede firm conclusions regarding the associations between the examined work environment exposures and common mental disorder. Nonetheless, the results indicate, in light of previous studies, that preventing work-related violence may help prevent clinically significant common mental disorders amongst employees. Further research should clarify factors underlying the experience of work as emotionally demanding. Health care, social, and to some extent educational workers, appear to be at high risk of common mental disorder. Secondary preventive measures in these occupational groups may reduce levels of psychiatric morbidity in the employed population. As common mental disorders are associated with substantial societal and human costs, such prevention is a significant public health concern. Madsen 2012 Social Work Environment and Mental Health 63

70 REFERENCES (1) Wittchen HU, Jacobi F. Size and burden of mental disorders in Europe--a critical review and appraisal of 27 studies. Eur Neuropsychopharmacol 2005 Aug;15(4): (2) Alonso J, Angermeyer MC, Bernert S, Bruffaerts R, Brugha TS, Bryson H, et al. Disability and quality of life impact of mental disorders in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatr Scand 2004;109: (3) Kendler KS, Gardner CO, Prescott CA. Toward a comprehensive developmental model for major depression in women. Am J Psychiatry 2002 Jul 1;159(7): (4) Kendler KS, Gardner CO, Prescott CA. Toward a comprehensive developmental model for major depression in men. Am J Psychiatry 2006 Jan 1;163(1): (5) Barlow DH. Anxiety and its disorder: the nature and treatment of anxiety and panic. 2. paperback ed. New York: Guilford; (6) Goldberg DP, Goodyer IM. The origins and course of common mental disorders. Hove, East Sussex: Routledge; (7) Drake CL, Roehrs T, Roth T. Insomnia causes consequences, and therapeutics: An overview. Depress Anxiety 2003;18(4): (8) Stansfeld S, Candy B. Psychosocial work environment and mental health--a meta-analytic review. Scand J Work Environ Health 2006;32(6): (9) Bonde JPE. Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med 2008 Jul;65(7): (10) Netterstrøm B, Conrad N, Bech P, Fink P, Olsen O, Rugulies R, et al. The relation between workrelated psychosocial factors and the development of depression. Epidemiol Rev 2008 Nov 1;30(1): (11) Siegrist J. Chronic psychosocial stress at work and risk of depression: evidence from prospective studies. Eur Arch Psychiatry Clin Neurosci 2008 Nov;258: (12) Karasek R, Theorell T. Healthy work: stress, productivity, and the reconstruction of working life. New York: Basic Books; (13) Siegrist J. Adverse health effects of high-effort/low-reward conditions. J Occup Health Psychol 1996 Jan;1(1): (14) Mathew AR, Pettit JW, Lewinsohn PM, Seeley JR, Roberts RE. Co-morbidity between major depressive disorder and anxiety disorders: shared etiology or direct causation? Psychol Med 2011 Oct;41(10): (15) Wittchen HU, Beesdo K, Bittner A, Goodwin RD. Depressive episodes-evidence for a causal role of primary anxiety disorders? Eur psychiat 2003 Dec;18(8): Madsen 2012 Social Work Environment and Mental Health 64

71 (16) Ohayon MM. Epidemiology of insomnia: what we know and what we still need to learn. Sleep Med Rev 2002 Apr;6(2): (17) World Health Organization. International statistical classification of diseases and related health problems - 10th Revision. World Health Organization 2007 [cited 2011 Jan 7]; Available from: (18) Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJL. Global and regional burden of disease and risk factors, 2001: systematic analysis of population health data. Lancet 2006 May 27;367(9524): (19) Mathers CD, Lopez AD, Murray CJL. The burden of disease and mortality by condition: data, methods, and results for In: Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJL, editors. Global burden of disease and risk Factors.Washington (DC): World Bank; (20) Øverland S, Glozier N, Sivertsen B, Stewart R, Neckelmann D, Krokstad S, et al. A comparison of insomnia and depression as predictors of disability pension: The HUNT Study. Sleep 2008 Jun 1;31(6): (21) Knudsen AK, Øverland S, Aakvaag HF, Harvey SB, Hotopf M, Mykletun A. Common mental disorders and disability pension award: Seven year follow-up of the HUSK study. J Psychosom Res 2010 Jul;69(1): (22) Ankestyrelsen [The National Social Appeals Board]. Førtidspensioner: Årsstatistik 2009 [Disability retirements: Annual Statistics 2009]. København: Ankestyrelsen; 2010 Jul 1. (23) Dubovsky SL, Davies R, Dubovsky AN. Mood disorders. In: Hales RE, Yudofsky SC, editors. Textbook of clinical psychiatry. 4. ed. ed. Washington, DC: American Psychiatric Publishing; p (24) Hollander E, Simeon D. Anxiety disorders. In: Hales RE, Yudofsky SC, editors. Textbook of clinical psychiatry. 4. ed. ed. Washington, DC: American Psychiatric Publishing; p (25) Abramson LY, Seligman ME, Teasdale JD. Learned helplessness in humans: critique and reformulation. J Abnorm Psychol 1978;87(1): (26) Simon RW. Revisiting the relationships among gender, marital status, and mental health. AJS 2002 Jan;107(4): (27) Flensborg-Madsen T, Tolstrup J, Sørensen H, Mortensen E. Social and psychological predictors of onset of anxiety disorders: results from a large prospective cohort study. Soc Psychiatry Psychiatr Epidemiol 2011;Online first. DOI: (28) Semmer NK, Jacobshagen N, Meier LL, Ellaway A. Occupational stress research: the "stress-asoffense-to-self" perspective. In: Houdmont J, McIntyre S, editors. Occupational health psychology - European perspectives on research, education and practice.nottingham: ISMAI publishers; p (29) Sacco WP, Vaughan CA. Depression and the response of others: a social-cognitive interpersonal process model. In: Joiner TE, Brown JS, Kistner J, editors. The interpersonal, cognitive, and social Madsen 2012 Social Work Environment and Mental Health 65

72 nature of depression. Mahwah, New Jersey: Lawrence Erlbaum Associates, Publishers; p (30) Brown GW, Bifulco A, Andrews B. Self-esteem and depression - III. Aetiological issues. Soc Psychiatry Psychiatr Epidemiol 1990 Sep 1;25(5): (31) World Health Organization. World report on violence and health. Geneva: World Health Organization; (32) Korkeila JA, Toyry S, Kumpulainen K, Toivola JM, Rasanen K, Kalimo R. Burnout and self-perceived health among Finnish psychiatrists and child psychiatrists: a national survey. Scand J Public Health 2003;31(2): (33) Gerberich SG, Church TR, McGovern PM, Hansen HE, Nachreiner NM, Geisser MS, et al. An epidemiological study of the magnitude and consequences of work related violence: the Minnesota Nurses' Study. Occup Environ Med 2004 Jun 1;61(6): (34) Ryan EP, Aaron J, Burnette ML, Warren J, Burket R, Aaron T. Emotional responses of staff to assault in a pediatric state hospital. Journal of the American Academy of Psychiatry and the Law 2008;36(3): (35) Geiger-Brown J, Muntaner C, McPhaul K, Lipscomb J, Trinkoff A. Abuse and violence during home care work as predictor of worker depression. Home Health Care Serv Q 2007;26(1): (36) Brewin CR, Andrews B, Rose S. Fear, helplessness, and horror in posttraumatic stress disorder: investigating DSM-IV criterion A2 in victims of violent crime. J Trauma Stress 2000 Jul;13(3): (37) Leather P, Beale D, Lawrence C, Dickson R. Effects of exposure to occupational violence and the mediating impact of fear. Work Stress 1997 Oct;11(4): (38) Menckel E, Viitasara E. Threats and violence in Swedish care and welfare - magnitude of the problem and impact on municipal personnel. Scand J Caring Sci 2002 Dec;16(4): (39) Wieclaw J, Agerbo E, Mortensen PB, Burr H, Tüchsen F, Bonde JP. Work related violence and threats and the risk of depression and stress disorders. J Epidemiol Community Health 2006 Sep;60(9): (40) Rothman KJ, Greenland S, Lash TL. Modern epidemiology. 3. edition ed. Philadelphia, Pa.: Wolters Kluwer; (41) Zapf D. Emotion work and psychological well-being: A review of the literature and some conceptual considerations. Human Resource Management Review 2002;12(2): (42) Hasenfeld Y, English RA. Human service organizations: a conceptual overview. In: Hasenfeld Y, English RA, editors. Human service organizations: a book of readings.ann Arbor: University of Michigan Press; p (43) Söderfeldt B, Söderfeldt M, Muntaner C, O'Campo P, Warg LE, Ohlson CG. Psychosocial work environment in human service organizations: a conceptual analysis and development of the demandcontrol model. Soc Sci Med 1996 May;42(9): Madsen 2012 Social Work Environment and Mental Health 66

73 (44) Schaufeli WB. The balance of give and take: Toward a social exchange model of burnout. Revue Internationale de Psychologie Sociale-International Review of Social Psychology 2006;19(1): (45) Schaufeli WB, Maslach C, Marek T. Interpersonal approaches. In: Schaufeli WB, Maslach C, Marek T, editors. Professional burnout: recent developments in theory and research.washington, D.C.: Taylor & Francis; p (46) Jenkins SR, Maslach C. Psychological health and involvement in interpersonally demanding occupations - a longitudinal perspective. J Organ Behav 1994 Mar;15(2): (47) Agerbo E, Gunnell D, Bonde JP, Mortensen PB, Nordentoft M. Suicide and occupation: the impact of socio-economic, demographic and psychiatric differences. Psychol Med 2007 Aug;37(8): (48) Wieclaw J, Agerbo E, Mortensen PB, Bonde JP. Risk of affective and stress related disorders among employees in human service professions. Occup Environ Med 2006 May;63(5): (49) Hannerz H, Tüchsen F, Pedersen BH, Dyreborg J, Rugulies R, Albertsen K. Work-relatedness of mood disorders in Denmark. Scand J Work Environ Health 2009 Jul;35(4): (50) Wieclaw J, Agerbo E, Mortensen PB, Bonde JP. Occupational risk of affective and stress-related disorders in the Danish workforce. Scand J Work Environ Health 2005;31(5): (51) Maslach C, Schaufeli WB, Leiter MP. Job burnout. Annu Rev Psychol 2001;52: (52) Ahola K, Hakanen J. Job strain, burnout, and depressive symptoms: A prospective study among dentists. J Affect Disord 2007 Dec;104(1-3): (53) van Vegchel N, de Jonge J, Söderfeldt M, Dormann C, Schaufeli W. Quantitative versus emotional demands among swedish human service employees: moderating effects of job control and social support. Int J Stress Manag 2004;11(1): (54) de Jonge J, Dormann C. The DISC model: Demand-Induced Strain Compensation mechanisms in job stress. In: Dollard MF, Winefield HR, Winefield AH, editors. Occupational Stress in the Service Professions.London: Taylor & Francis Ltd; p (55) Kristensen TS, Hannerz H, Høgh A, Borg V. The Copenhagen psychosocial questionnaire--a tool for the assessment and improvement of the psychosocial work environment. Scand J Work Environ Health 2005 Dec;31(6): (56) de Jonge J, Mulder MJ, Nijhuis FJ. The incorporation of different demand concepts in the job demandcontrol model: effects on health care professionals. Soc Sci Med 1999;48(9): (57) de Boer J, Lok A, van't Verlaat E, Duivenvoorden HJ, Bakker AB, Smit BJ. Work-related critical incidents in hospital-based health care providers and the risk of post-traumatic stress symptoms, anxiety, and depression: a meta-analysis. Soc Sci Med 2011 Jul;73(2): (58) Zapf D, Seifert C, Schmutte B, Mertini H, Holz M. Emotion work and job stressors and their effects on burnout. Psychol Health 2001;16(5): Madsen 2012 Social Work Environment and Mental Health 67

74 (59) Neumann M, Bensing J, Mercer S, Ernstmann N, Ommen O, Pfaff H. Analyzing the "nature" and "specific effectiveness" of clinical empathy: A theoretical overview and contribution towards a theorybased research agenda. Patient Educ Couns 2009 Mar;74(3): (60) Burr H, Albertsen K, Rugulies R, Hannerz H. Do dimensions from the Copenhagen Psychosocial Questionnaire predict vitality and mental health over and above the job strain and effort-reward imbalance models? Scand J Public Health 2010 Feb;38: (61) Bültmann U, Kant IJ, Schröer CAP, Kasl SV. The relationship between psychosocial work characteristics and fatigue and psychological distress. Int Arch Occup Environ Health 2002 May 1;75(4): (62) Andrea H, Bültmann U, Beurskens AJHM, Swaen GMH, van Schayck CP, Kant IJ. Anxiety and depression in the working population using the HAD Scale. Soc Psychiatry Psychiatr Epidemiol 2004 Aug 1;39(8): (63) Wieclaw J, Agerbo E, Mortensen PB, Burr H, Tüchsen F, Bonde JP. Psychosocial working conditions and the risk of depression and anxiety disorders in the Danish workforce. BMC Public Health 2008 Aug 7;8:280. (64) Hobfoll S. The influence of culture, community, and the nested-self in the stress process: advancing conservation of resources theory. Applied Psychology: An international review 2001;50(3): (65) Hochschild A. The managed heart : commercialization of human feeling. 1st ed. London: University of California Press; (66) Grandey AA. When "the show must go on": surface acting and deep acting as determinants of emotional exhaustion and peer-rated service delivery. Acad Manage J 2003 Feb;46(1): (67) Brotheridge CM, Grandey AA. Emotional labor and burnout: comparing two perspectives of "people work". J Vocat Behav 2002 Feb;60(1): (68) Thoits PA. Mechanisms linking social ties and support to physical and mental health. J Health Soc Behav 2011 Jun;52(2): (69) Cohen S. Social relationships and health. Am Psychol 2004;59(8): (70) Borritz M, Rugulies R, Bjorner JB, Villadsen E, Mikkelsen OA, Kristensen TS. Burnout among employees in human service work: design and baseline findings of the PUMA study. Scand J Public Health 2006;34(1): (71) Pejtersen JH, Kristensen TS, Borg V, Bjorner JB. The second version of the Copenhagen Psychosocial Questionnaire. Scand J Public Health 2010 Feb;38:8-24. (72) Burr H, Bjorner JB, Kristensen TS, Tüchsen F, Bach E. Trends in the Danish work environment in and their associations with labor-force changes. Scand J Work Environ Health 2003;29(4): (73) Feveile H, Olsen O, Burr H, Bach E. Danish work environment cohort study 2005: from idea to sampling design. Statistics in Transition 2007;8(3): Madsen 2012 Social Work Environment and Mental Health 68

75 (74) Carneiro IG, Faber A, Flyvholm M, Giver H, Jensen JN, Nabe-Nielsen K, et al. SOSU'er årgang Det første år [Social and Health care workers year The first year]. Copenhagen: National Research Centre for the Working Environment; Report No.: 5. (75) Kinsten A, Magnusson Hanson L, Hyde M, Oxenstierna G, Westerlund H, Theorell T. Swedish Longitudinal Occupational Survey of Health - a nationally representative psychosocial survey of the Swedish working population. Stockholm University: Stress Research Institute; Report No.: 321. (76) Magnusson Hanson L, Theorell T, Oxenstierna G, Hyde M, Westerlund H. Demand, control and social climate as predictors of emotional exhaustion symptoms in working Swedish men and women. Scand J Public Health 2008 Sep;36(7): (77) Giver H, Faber A, Hannerz H, Stroyer J, Rugulies R. Psychological well-being as a predictor of dropout among recently qualified Danish eldercare workers. Scand J Public Health 2010 May;38(3): (78) Pedersen CB. The Danish civil registration system. Scand J Public Health 2011 Jul;39:22-5. (79) Jensen VM, Rasmussen AW. Danish education registers. Scand J Public Health 2011 Jul;39:91-4. (80) Baadsgaard M, Quitzau J. Danish registers on personal income and transfer payments. Scand J Public Health 2011 Jul;39: (81) Helweg-Larsen K. The Danish Register of Causes of Death. Scand J Public Health 2011 Jul;39:26-9. (82) Kildemoes HW, Sørensen HT, Hallas J. The Danish national prescription registry. Scand J Public Health 2011 Jul;39: (83) Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sørensen HT. The Nordic countries as a cohort for pharmacoepidemiological research. Basic Clin Pharmacol Toxicol 2009;106(2): (84) Socialstyrelsen [The National Board of Health and Welfare]. Dödsorsaksregistret [Register of Causes of Death]. Socialstyrelsen [The National Board of Health and Welfare] 2012 [cited 2012 May 30]; Available from: (85) Statistics Denmark. Statistical Yearbook Statistics Denmark; (86) Eurostat. NACE Rev. 2 - Statistical classification of economic activities. Luxembourg: Office for Official Publications of the European Communities; (87) Scheiber SC. The psychiatric interview, psychiatric history, and mental status examination. In: Hales RE, Yudofsky SC, editors. Textbook of clinical psychiatry. 4. ed. ed. Washington, DC: American Psychiatric Publishing; p (88) Comstock GW, Helsing KJ. Symptoms of Depression in 2 Communities. Psychol Med 1976;6(4): Madsen 2012 Social Work Environment and Mental Health 69

76 (89) Kessler R, Wittchen H-U, Abelson J, McGonagle K, Schwarz N, Kendler K, et al. Methodological studies of the composite international diagnostic interview (CIDI) in the US national comorbidity survey. International Journal of Psychiatric Methods 1998;7: (90) Goldberg DP, Hillier VF. A scaled version of the General Health Questionnaire. Psychol Med 1979 Feb;9(1): (91) Tourangeau R, Yan T. Sensitive questions in surveys. Psychol Bull 2007 Sep;133(5): (92) Bromet EJ, Dunn LO, Connell MM, Dew MA, Schulberg HC. Long-term reliability of diagnosing lifetime major depression in a community sample. Arch Gen Psychiatry 1986 May;43(5): (93) Boyd JH, Weissman MM, Thompson WD, Myers JK. Screening for Depression in A Community Sample - Understanding the Discrepancies Between Depression Symptom and Diagnostic Scales. Arch Gen Psychiatry 1982;39(10): (94) Podsakoff PM, MacKenzie SB, Jeong-Yeon L, Podsakoff NP. Common method biases in behavioral research: a critical review of the literature and recommended remedies. J Appl Psychol 2003 Oct;88(5): (95) Centre for Psychiatric Research. Danish Psychiatric Central Research Register. Centre for Psychiatric Research 2012 [cited 2012 Jun 18]; Available from: (96) Alonso J, Angermeyer MC, Bernert S, Bruffaerts R, Brugha IS, Bryson H, et al. Psychotropic drug utilization in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatr Scand 2004;109: (97) Hämäläinen J, Isometsa E, Sihvo S, Pirkola S, Kiviruusu O. Use of health services for major depressive and anxiety disorders in Finland. Depress Anxiety 2008;25(1): (98) Goldberg M, Chastang JF, Zins M, Niedhammer I, Leclerc A. Health problems were the strongest predictors of attrition during follow-up of the GAZEL cohort. J Clin Epidemiol 2006 Nov;59(11): (99) Rosholm JU, Gram LF, Damsbo N, Hallas J. Antidepressant treatment in general-practice - an interview study. Scand J Prim Health Care 1995 Dec;13(4): (100) Gardarsdottir H, Heerdink ER, van Dijk L, Egberts ACG. Indications for antidepressant drug prescribing in general practice in the Netherlands. J Affect Disord 2007 Feb;98(1-2): (101) Mark TL. For what diagnoses are psychotropic medications being prescribed? A nationally representative survey of physicians. Cns Drugs 2010;24(4): (102) World Health Organization. The Anatomical Therapeutic Chemical classification system with Defined Daily Doses (ATC/DDD). World Health Organization 2009 [cited 2009 Sep 7]; Available from: (103) Kleinbaum DG, Klein M. Survival analysis: a self-learning text. 2. ed. ed. New York: Springer-Verlag; Madsen 2012 Social Work Environment and Mental Health 70

77 (104) Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol 2009 Jul 15;170(2): (105) Varadhan R, Weiss CO, Segal JB, Wu AW, Scharfstein D, Boyd C. Evaluating health outcomes in the presence of competing risks: a review of statistical methods and clinical applications. Med Care 2010 Jun;48(6 Suppl):S (106) MacKinnon DP, Fairchild AJ, Fritz MS. Mediation analysis. Annu Rev Psychol 2007 Feb;58(1): (107) MacKinnon DP, Lockwood CM, Brown CH, Wang W, Hoffman JM. The intermediate endpoint effect in logistic and probit regression. Clin Trials 2007 Oct 1;4(5): (108) Petersen ML, Sinisi SE, van der Laan MJ. Estimation of direct causal effects. Epidemiology May 2006;17(3): (109) Riley RD, Lambert PC, bo-zaid G. Meta-analysis of individual participant data: rationale, conduct, and reporting. BMJ 2010;340. (110) Tudur Smith C, Williamson PR. A comparison of methods for fixed effects meta-analysis of individual patient data with time to event outcomes. Clin Trials 2007 Dec 1;4(6): (111) Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol 2005 Aug 1;162(3): (112) Schwarzer G. Package 'meta'. The R Foundation for Statistical Computing 2012 [cited 2012 May 9]; Available from: (113) Alonso J, Angermeyer MC, Bernert S, Bruffaerts R, Brugha IS, Bryson H, et al. Prevalence of mental disorders in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatr Scand 2004;109:21-7. (114) Harris T. Recent developments in understanding the psychosocial aspects of depression. Br Med Bull 2001 Mar 1;57(1): (115) Olsen LR, Mortensen EL, Bech P. Prevalence of major depression and stress indicators in the Danish general population. Acta Psychiatr Scand 2004 Feb;109(2): (116) DeSanto Iennaco J, Cullen MR, Cantley L, Slade MD, Fiellin M, Kasl SV. Effects of externally rated job demand and control on depression diagnosis claims in an industrial cohort. Am J Epidemiol 2010 Feb 1;171(3): (117) Griffin JM, Greiner BA, Stansfeld SA, Marmot M. The effect of self-reported and observed job conditions on depression and anxiety symptoms: A comparison of theoretical models. J Occup Health Psychol 2007 Oct;12(4): (118) Virtanen M, Pentti J, Vahtera J, Ferrie JE, Stansfeld SA, Helenius H, et al. Overcrowding in hospital wards as a predictor of antidepressant treatment among hospital staff. Am J Psychiatry 2008 Nov;165(11): Madsen 2012 Social Work Environment and Mental Health 71

78 (119) Bonde JPE, Munch-Hansen T, Wieclaw J, Westergaard-Nielsen N, Agerbo E. Psychosocial work environment and antidepressant medication: a prospective cohort study. BMC Public Health 2009 Jul 27;9. (120) Kolstad HA, Hansen Å, Kærgaard A, Thomsen JF, Kaerlev L, Mikkelsen S, et al. Job strain and the risk of depression: is reporting biased? Am J Epidemiol 2011 Jan 1;173(1): (121) Kouvonen A, Oksanen T, Vahtera J, Stafford M, Wilkinson R, Schneider J, et al. Low workplace social capital as a predictor of depression: the Finnish public sector study. Am J Epidemiol 2008 May 15;167(10): (122) Jensen HK, Wieclaw J, Munch-Hansen T, Thulstrup AM, Bonde JP. Does dissatisfaction with psychosocial work climate predict depressive, anxiety and substance abuse disorders? A prospective study of Danish public service employees. J Epidemiol Community Health 2010 Sep;64(9): (123) Frese M, Zapf D. Methodological issues in the study of work stress: Objective vs subjective measurement of work stress and the question of longitudinal studies. In: Cooper CL, Payne R, editors. Causes, coping and consequences of stress at work. Wiley series on studies in occupational stress. John Wiley & Sons, Chichester, England; p (124) Hämäläinen J, Isometsä E, Sihvo S, Kiviruusu O, Pirkola S, Lönnqvist J. Treatment of major depressive disorder in the Finnish general population. Depress Anxiety 2009 Nov;26(11): (125) Trifirò G, Barbui C, Spina E, Moretti S, Tari M, Alacqua M, et al. Antidepressant drugs: prevalence, incidence and indication of use in general practice of Southern Italy during the years Pharmacoepidemiol Drug Saf 2007;16(5): (126) Schatzberg AF. New indications for antidepressants. J Clin Psychiatry 2000;61(Suppl 11):9-17. (127) Sinokki M, Hinkka K, Ahola K, Koskinen S, Kivimäki M, Honkonen T, et al. The association of social support at work and in private life with mental health and antidepressant use: The Health 2000 Study. J Affect Disord 2009 May;115(1-2): (128) Thielen K, Nygaard E, Rugulies R, Diderichsen F. Job stress and the use of antidepressant medicine: a 3.5-year follow-up study among Danish employees. Occup Environ Med 2011 Mar 1;68(3): (129) Thielen K, Nygaard E, Andersen I, Rugulies R, Heinesen E, Bech P, et al. Misclassification and the use of register-based indicators for depression. Acta Psychiatr Scand 2009;119(4): (130) Demyttenaere K, Bonnewyn A, Bruffaerts R, de Girolamo G, Gasquet I, Kovess V, et al. Clinical factors influencing the prescription of antidepressants and benzodiazepines: Results from the European study of the epidemiology of mental disorders (ESEMeD). J Affect Disord 2008 Sep;110(1-2): (131) Kendrick T, Dowrick C, McBride A, Howe A, Clarke P, Maisey S, et al. Management of depression in UK general practice in relation to scores on depression severity questionnaires: analysis of medical record data. BMJ 2009;338. (132) Ohayon MM, Lader MH. Use of psychotropic medication in the general population of France, Germany, Italy, and the United Kingdom. J Clin Psychiatry 2002 Sep;63(9): Madsen 2012 Social Work Environment and Mental Health 72

79 (133) Andersen RM. Revisiting the behavioral-model and access to medical-care - does it matter? J Health Soc Behav 1995 Mar;36(1):1-10. (134) Andersen R, Newman JF. Societal and individual determinants of medical care utilization in the United States. Milbank Quarterly 2005;83(4):Online Only. DOI: (135) Gadalla TM. Comparison of users and non-users of mental health services among depressed women: a national study. Women Health 2008;47(1):1-19. (136) Steele L, Dewa C, Lee K. Socioeconomic status and self-reported barriers to mental health service use. Can J Psychiatry 2007 Mar;52(3): (137) ten Have M, de Graaf R, Ormel J, Vilagut G, Kovess V, Alonso J. Are attitudes towards mental health help-seeking associated with service use? Results from the European Study of Epidemiology of Mental Disorders. Soc Psychiatry Psychiatr Epidemiol 2010 Feb;45(2): (138) Koopmans G, Lamers L. Gender and health care utilization: The role of mental distress and help seeking propensity. Soc Sci Med 2007;64: (139) Colman I, Croudace TJ, Wadsworth MEJ, Jones PB. Factors associated with antidepressant, anxiolytic and hypnotic use over 17 years in a national cohort. J Affect Disord 2008 Oct;110(3): (140) Colman I, Wadsworth MEJ, Croudace TJ, Jones PB. Three decades of antidepressant, anxiolytic and hypnotic use in a national population birth cohort. Br J Psychiatry 2006 Aug;189: (141) Gilså Hansen D, Søndergaard J, Vach W, Freng Gram L, Rosholm JU, Kragstrup J. Antidepressant drug use in general practice: inter-practice variation and association with practice characteristics. Eur J Clin Pharmacol 2003 Jun 1;59(2): (142) Rosholm JU, Hallas J, Gram LF. Outpatient utilization of antidepressants - a prescription database analysis. J Affect Disord 1993 Jan;27(1):21-8. (143) Sterud T, Hem E, Ekeberg O, Lau B. Health problems and help-seeking in a nationwide sample of operational Norwegian ambulance personnel. BMC Public Health 2008;8(1):3. (144) England P, Budig M, Folbre N. Wages of virtue: The relative pay of care work. Soc Probl 2002 Nov;49(4): (145) Compton MT, Frank E. Mental health concerns among Canadian physicians: results from the Canadian Physician Health Study. Compr Psychiatry 2011 Sep;52(5): (146) Center C, Davis M, Detre T, Ford DE, Hansbrough W, Hendin H, et al. Confronting depression and suicide in physicians - A consensus statement. Jama 2003 Jun 18;289(23): (147) Kivimäki M, Gunnell D, Lawlor DA, Davey Smith G, Pentti J, Virtanen M, et al. Social inequalities in antidepressant treatment and mortality: a longitudinal register study. Psychol Med 2007;37(03): (148) Siebert DC. Help seeking for AOD misuse among social workers: Patterns, barriers, and implications. Social Work 2005 Jan;50(1): Madsen 2012 Social Work Environment and Mental Health 73

80 (149) Lægemiddelstyrelsen [Danish Medicines Agency]. Forbruget af antidepressiva i Danmark i den primære sundhedssektor i perioden [The use of antidepressants in Denmark in the primary healthcare sector during ]. Lægemiddelstyrelsen; (150) Sundhedsstyrelsen [National Board of Health]. Refererenceprogram for angstlidelser hos voksne [Reference program for anxiety disorders in adults]. Copenhagen: Sundhedsstyrelsen [National Board of Health]; (151) Institut for Rationel Farmakoterapi [Institute for Rational Pharmacotherapy]. National Rekommendationsliste- ATC-gruppe N05B og N05C - Angstdæmpende midler og sovemidler [National Recommendationlist - ATC-groups N05B and N05C - Anxiolytics and hypnotics]. Institut for Rationel Farmakoterapi [Institute for Rational Pharmacotherapy] 2010 June 16 [cited 2011 Feb 11]; Available from: aka/atc-gruppe_n05b_og_n05c_-_angstdaempende_midler_og_sovemidler.htm (152) Sundhedsstyrelsen [National Board of Health]. Vejledning om ordination af afhængighedsskabende lægemidler og om substitutionsbehandling af personer med opioidafhængighed [Guidelines on ordination of dependency-creating medications and substitute treatment for persons with opioidaddiction]. Copenhagen: Sundhedsstyrelsen [National Board of Health]; (153) Shane S, Nicolaou N, Cherkas L, Spector TD. Genetics, the big five, and the tendency to be selfemployed. J Appl Psychol 2010 Nov;95(6): (154) Stansfeld SA, Fuhrer R, Shipley MJ, Marmot MG. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II Study. Occup Environ Med 1999;56(5): (155) Payne RL. Comments on 'Why negative affectivity should not be controlled in job stress research: don't throw out the baby with the bath waters'. J Organ Behav 2000;21(1):97-9. (156) Spector PE, Zapf D, Chen PY, Frese M. Why negative affectivity should not be controlled in job stress research: don't throw out the baby with the bath water. J Organ Behav 2000;21(1): (157) Caspi A, Roberts BW, Shiner RL. Personality development: stability and change. Annu Rev Psychol 2005;56: (158) Teychenne M, Ball K, Salmo J. Sedentary behavior and depression among adults: a review. Int J Behav Med 2010 Dec;17(4): (159) Rehm J, Room R, Graham K, Monteiro M, GmeI G, Sempos CT. The relationship of average volume of alcohol consumption and patterns of drinking to burden of disease: an overview. Addiction 2003 Sep;98(9):1209. (160) Siegrist J, Rodel A. Work stress and health risk behavior. Scand J Work Environ Health 2006 Dec;32(6): (161) Nyberg ST, Heikkilä K, Fransson EI, Alfredsson L, De Bacquer D, Bjorner JB, et al. Job strain in relation to body mass index: pooled analysis of adults from 13 cohort studies. J Intern Med 2012 Jul;272(1): Madsen 2012 Social Work Environment and Mental Health 74

81 (162) Lange T, Hansen JV. Direct and indirect effects in a survival context. Epidemiology 2011 Jul;22(4): (163) Rugulies R, Thielen K, Nygaard E, Diderichsen F. Job insecurity and the use of antidepressant medication among Danish employees with and without a history of prolonged unemployment: a 3.5- year follow-up study. J Epidemiol Community Health 2010 Jan 1;64(01): (164) Sinokki M, Hinkka K, Ahola K, Koskinen S, Klaukka T, Kivimaki M, et al. The association between team climate at work and mental health in the Finnish Health 2000 Study. Occup Environ Med 2009 Aug 1;66(8): (165) Laaksonen M, Lallukka T, Lahelma E, Partonen T. Working conditions and psychotropic medication: a prospective cohort study. Soc Psychiatry Psychiatr Epidemiol 2011;47(4): (166) Virtanen M, Honkonen T, Kivimäki M, Ahola K, Vahtera J, Aromaa A, et al. Work stress, mental health and antidepressant medication findings from the Health 2000 Study. J Affect Disord 2007;98(3): (167) d'errico A, Cardano M, Landriscina T, Marinacci C, Pasian S, Petrelli A, et al. Workplace stress and prescription of antidepressant medications: a prospective study on a sample of Italian workers. Int Arch Occup Environ Health 2011 Apr;84(4): (168) Driesen K, Jansen NWH, van Amelsvoort LGPM, Kant I. The mutual relationship between shift work and depressive complaints - a prospective cohort study. Scand J Work Environ Health 2011 Sep;37(5): (169) Stansfeld SA, Clark C, Caldwell T, Rodgers B, Power C. Psychosocial work characteristics and anxiety and depressive disorders in midlife: the effects of prior psychological distress. Occup Environ Med 2008 Sep;65(9): (170) Hooper LM, Doehler K, Wallace SA, Hannah NJ. The parentification inventory: development, validation, and cross-validation. Am J Fam Ther 2011 Apr 12;39(3): (171) Lackie B. The families of origin of social-workers. Clin Soc Work J 1983;11(4): (172) Fussell FW, Bonney WC. A comparative study of childhood experiences of psychotherapists and physicists: Implications for clinical practice. Psychotherapy: Theory, Research, Practice, Training 1990;27(4): (173) Nikcevic AV, Kramolisova-Advani J, Spada MM. Early childhood experiences and current emotional distress: what do they tell us about aspiring psychologists? J Psychol 2007 Jan;141(1): (174) Vincent J. Why ever do we do it? Unconscious motivation in choosing social work as a career. Journal of Social Work Practice 1996 May;10(1):63-9. (175) Jacobvitz DB, Bush NF. Reconstructions of family relationships: Parent child alliances, personal distress, and self-esteem. Dev Psychol 1996;32(4): (176) Hobfoll S. The Influence of Culture, Community, and the Nested-Self in the Stress Process: Advancing Conservation of Resources Theory. Applied Psychology: An international review 2001;50(3): Madsen 2012 Social Work Environment and Mental Health 75

82 (177) Stansfeld SA, Blackmore ER, Zagorski BM, Munce S, Stewart DE, Weller I. Work characteristics and social phobia in a nationally representative employed sample. Can J Psychiatry 2008 Jun;53(6): (178) Wang JL, Lesage A, Schmitz N, Drapeau A. The relationship between work stress and mental disorders in men and women: findings from a population-based study. J Epidemiol Community Health 2008 Jan;62(1):42-7. (179) Melchior M, Caspl A, Milne BJ, Danese A, Poulton R, Moffitt TE. Work stress precipitates depression and anxiety in young, working women and men. Psychol Med 2007 Aug;37(8): (180) Andrea H, Bültmann U, van Amelsvoort LGPM, Kant Y. The incidence of anxiety and depression among employees-the role of psychosocial work characteristics. Depress Anxiety 2009 Nov;26(11): (181) Elovainio M, Ferrie JE, Gimeno D, De Vogli R, Shipley M, Brunner EJ, et al. Organizational justice and sleeping problems: the Whitehall II study. Psychosom Med 2009 Apr;71(3): (182) Takahashi M, Iwasaki K, Sasaki T, Kubo T, Mori I, Otsuka Y. Worktime control-dependent reductions in fatigue, sleep problems, and depression. Applied Ergonomics 2011 Jan;42(2): (183) Alonso J, Angermeyer MC, Bernert S, Bruffaerts R, Brugha TS, Bryson H, et al. 12-Month comorbidity patterns and associated factors in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatr Scand 2004;109: (184) Gillespie GL, Gates DM, Miller M, Howard PK. Workplace violence in healthcare settings: risk factors and protective strategies. Rehabilitation Nursing 2010 Sep;35(5): Madsen 2012 Social Work Environment and Mental Health 76

83 6. LIST OF APPENDICES Appendix 1 Article 1: Work-related violence and incident use of psychotropics Appendix 2 Article 2: Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study Appendix 3 Article 3: Paid care work and depression: a longitudinal study of antidepressant treatment in female eldercare workers before and after entering their profession Appendix 4 Article 4: Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish gainfully employed Appendix 5 Article 5: Does good leadership buffer effects of high emotional demands at work on risk of antidepressant treatment? A prospective study from two Nordic countries Madsen 2012 Social Work Environment and Mental Health

84 APPENDIX 1. ARTICLE 1 Madsen IEH, Burr H, Diderichsen F, Pejtersen J, Borritz M, Bjorner JB, Rugulies R. Work-related violence and incident use of psychotropics. American Journal of Epidemiology 2011;174(12): Madsen 2012 Social Work Environment and Mental Health

85 American Journal of Epidemiology ª The Author Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please Vol. 174, No. 12 DOI: /aje/kwr259 Advance Access publication: October 29, 2011 Original Contribution Work-related Violence and Incident Use of Psychotropics Ida E. H. Madsen*, Hermann Burr, Finn Diderichsen, Jan H. Pejtersen, Marianne Borritz, Jakob B. Bjorner, and Reiner Rugulies * Correspondence to Ida E. H. Madsen, National Research Centre for the Working Environment, Lersø Parkallé 105, DK-2100 Copenhagen, Denmark ( [email protected]). Initially submitted February 18, 2011; accepted for publication July 7, Although the mental health consequences of domestic violence are well documented, empirical evidence is scarce regarding the mental health effects of violence in the workplace. Most studies have used data from small occupationspecific samples, limiting their generalizability. This article examines whether direct exposure to work-related violence is associated with clinically pertinent mental health problems, measured by purchases of psychotropics (antidepressants, anxiolytics, hypnotics), in a cross-occupational sample of 15,246 Danish employees free from using psychotropics at baseline. Self-reported data on work-related violence were merged with other data on purchases of medications through a national registry to estimate cause-specific hazard ratios during 3.6 years (1,325 days) of follow-up in the years Outcomes were examined as competing risks, and analyses were adjusted for gender, age, cohabitation, education, income, social support from colleagues, social support from supervisor, and influence and quantitative demands at work. Work-related violence was associated with purchasing antidepressants alone (hazard ratio ¼ 1.38, 95% confidence interval: 1.09, 1.75) or in combination with anxiolytics (hazard ratio ¼ 1.74, 95% confidence interval: 1.13, 2.70) but not with purchasing anxiolytics or hypnotics only. The frequency of violent episodes and risk of caseness were unrelated. Work-related violence is associated with increased risk of clinically pertinent mental health problems. Reducing levels of work-related violence may help to prevent mental disorders in the working population. mental disorders; psychotropic drugs; violence; work Abbreviations: CI, confidence interval; COPSOQ, Copenhagen Psychosocial Questionnaire Study; DWECS, Danish Work Environment Cohort Study; HR, hazard ratio; PUMA, Project on Burnout, Motivation, and Job Satisfaction. The severe mental health consequences of domestic violence are well established (1, 2). There is little evidence, however, concerning the effects of work-related violence, which does not involve an intimate relationship between victim and perpetrator. With an estimated 1.7 million annual episodes of violence in US workplaces alone (3), it is of substantial public health concern to document the consequences of this work environment exposure. Although small occupation-specific studies show adverse emotional consequences including depressed mood and sleep disturbance (4, 5), there is a paucity of larger cross-occupational studies examining clinically pertinent mental health outcomes. One study showed increased hospitalization rates for depression and anxiety disorders in occupations with high levels of work-related violence (6). This study, though, assessed exposure only ecologically, limiting individual-level inference (7). To the best of our knowledge, this article is the first to examine whether direct exposure to work-related violence is associated with purchases of 3 types of psychotropics (antidepressants, anxiolytics, hypnotics), by use of a unique cross-occupational sample of Danish employees (n ¼ 15,246) with individual-level exposure data. In this article, we examine the effects of work-related violence defined dichotomously and whether more frequent exposure is associated with greater risk. MATERIALS AND METHODS Study design and population This article presents results from a Danish project synthesizing data from 3 previous studies and merging these data with the Danish Register of Medicinal Product Statistics. The Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, Am J Epidemiol. 2011;174(12):

86 Work-related Violence and Use of Psychotropics 1355 PUMA I n = 1,856 PUMA II n = 747 PUMA III n = 644 COPSOQ I n = 2,445 COPSOQ II n = 4,719 DWECS 2000 n = 8,551 DWECS 2005 n = 6,743 Total n = 25,705 Self-employed n = 1,429 Not working n = 6,452 3 original studies are 1) the Project on Burnout, Motivation, and Job Satisfaction (3 waves: PUMA I, PUMA II, PUMA III) (8); 2) the Copenhagen Psychosocial Questionnaire Study (2 waves: COPSOQ I, COPSOQ II) (9, 10); and 3) the Danish Work Environment Cohort Study (2 waves: DWECS 2000, DWECS 2005) (11, 12). All studies were designed to examine the associations between work environment and health, and they contain comprehensive self-reported work environment data. Whereas populations for COPSOQ and DWECS were drawn randomly from the working-aged Danish population, PUMA was workplace based, recruiting employees from 7 human service organizations. The studies were open cohort studies conducted during the years , and they had response rates between 60% (COPSOQ II) and 80% (PUMA I). Details of the studies are published elsewhere (8 12). This article uses all first-time responses from these studies to increase statistical power and merges them with data from the Register of Medicinal Products Statistics a national Danish registry containing data on all purchases of prescription medications at Danish pharmacies since January 1, 1995 (13). Violence Missing data n = 751 Following the World Health Organization s definition of violence, which includes actual and threatened use of physical force (14), we measured work-related violence with 2 items: Have you been exposed to threats of violence at your workplace during the last 12 months? and Have you been exposed to physical violence at your workplace during the last 12 months?. The items measured direct exposure (i.e., being the target of the act), and the response options were yes and no. We defined exposure to work-related violence as responding yes to either question. Furthermore, some studies recorded the number of violent episodes (available from Included participants n = 15,246 PUMA I III, COPSOQ I, and DWECS 2000), and we categorized exposure frequency as 0, 1 2, and 3 or more times. Covariates Prior psychotropics use n = 1,827 Figure 1. Selection of study population and exclusions from PUMA I III, COPSOQ I and II, and DWECS 2000 and 2005, Denmark, COPSOQ, Copenhagen Psychosocial Questionnaire Study; DWECS, Danish Work Environment Cohort Study; PUMA, Project on Burnout, Motivation, and Job Satisfaction. Gender, age, education, occupational group, and disposable income (individual income after tax) were assessed through registry data, as were migration and death during follow-up. The occupational group was coded according to the Danish version of the International Standard Classification of Occupations approved in 1988 (ISCO-88), which classifies jobs on the basis of the skills needed to perform them (15). For the present article, we combined agricultural and fishery workers with craft and related trades workers because of the few respondents in agricultural and fishery occupations. The resulting groups were as follows: armed forces, legislators, senior officials, and managers, professionals, clerks, service workers and shop and market sales workers, skilled agricultural and fishery workers and craft and related trades workers, plant and machine operators and assemblers, and elementary occupations. Furthermore, we included self-reported information on the following covariates: cohabitation (living with a partner or spouse, yes/no), social support from colleagues, social support from supervisor, work-related influence (3-item scale), and quantitative demands (3-item scale). We selected these covariates because they have been related to mental disorders (16 20) and may be associated with work-related violence. All self-reported covariates were measured by items from the validated COPSOQ questionnaire (9). The specific items were chosen for commonality among the original studies. Social support from colleagues was measured by the item: How often do you get help and support from your colleagues?. Social support from the supervisor was measured by the item: How often do you get help and support from Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012 Am J Epidemiol. 2011;174(12):

87 1356 Madsen et al. Table 1. Characteristics of Participants and Incident Use of Psychotropics, Denmark, No. Participants Mean (SD) your immediate superior?. The response options for the social support items were always, often, sometimes, seldom, never/hardly ever, and not applicable. Two studies, however, applied no not applicable option and showed high levels of missing values in these items. To retain respondents in the analyses, we categorized missing values in these studies as not applicable and collapsed this category with seldom and never/hardly ever as the lowest level of support. Influenceatworkwasmeasuredbythe3items: Doyou have a large degree of influence concerning your work?, Can you influence the amount of work assigned to you?, and Do you have any influence on what you do at work?. Quantitative demands at work were measured by the 3 items: Do you have to work very fast?, Is your work unevenly distributed so it piles up?, and How often do you not have time to complete all your work tasks?. Because antidepressants are prescribed as pain management (21), we also included self-reported information on regular use of pain medications (monthly or more frequently) as an indicator of chronic pain (not available for DWECS). Antidepressants No. of Cases % Follow-up Antidepressants and Anxiolytics No. of Cases % Anxiolytics No. of Cases Hypnotics No. 15, Work-related violence Yes 1, No 13, Sex Men 6, Women 8, Living arrangements Noncohabiting 3, Cohabiting 12, Age, years 41.3 (10.6) <30 2, , , , Education, years (2.48) , , , , Disposable income, 32,319.8 (16,103.3) US dollars/year by quartile Following the World Health Organization-developed Anatomical Therapeutic Chemical (ATC) Classification System (22), psychotropics were identified in the registry as medications with codes N06a (antidepressants), N05b (anxiolytics), and N05c (hypnotics). The baseline for follow-up was set at 365 days before the survey date, as exposure to violence was assessed retrospectively within the past 12 months. We assessed prior use of psychotropics during 1.9 years (700 days) before baseline and followed purchases of psychotropics for 3.6 years (1,325 days) after baseline. These time periods were chosen because they were the longest available for all studies. Figure 1 shows the selection of participants. Of the 25,705 first-time respondents, we excluded individuals not working (n ¼ 6,452) and individuals self-employed (n ¼ 1,429), as one study did not obtain exposure data for the self-employed. Additionally, we excluded 1,827 respondents who had purchased any psychotropic medication within 1.9 years before % No. of Cases 1(<24,895) 3, (24,895 30,285) 3, (30,286 37,163) 3, (>37,163) 3, % Table continues Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012 Am J Epidemiol. 2011;174(12):

88 Work-related Violence and Use of Psychotropics 1357 Table 1. Continued No. Participants Mean (SD) baseline. Finally, we excluded 751 respondents with missing data, yielding a study population of 15,246 employees. For the included participants, we followed purchases of psychotropics from baseline (365 days before the survey date) until migration, death, or end of follow-up, whichever came first. We examined 4 mutually exclusive types of use of psychotropics (caseness), defined a priori as purchasing 1) antidepressants, 2) antidepressants and anxiolytics, 3) anxiolytics, or 4) hypnotics only. Because sleep disturbance is a symptom of depressive and anxiety disorders (23), we did not define separate types of caseness for use of antidepressants and use of anxiolytics with and without use of hypnotics. Hence, antidepressant cases are participants who purchased antidepressant medications, regardless of any purchases of hypnotics, if they did not also purchase anxiolytics. Anxiolytic caseness was defined similarly, whereas hypnotics-only cases were defined by purchasing only hypnotics. We applied full follow-up information in defining caseness; that is, individuals were classified as combined antidepressant-anxiolytic cases if they purchased both types of medication during follow up, although Antidepressants No. of Cases % Antidepressants and Anxiolytics No. of Cases not necessarily simultaneously. The first purchase of any psychotropic medication was used as the caseness date. Statistical analysis % Anxiolytics No. of Cases Hypnotics Social support from colleagues 4 (always) 3, (often) 5, (sometimes) 4, (seldom, never/ hardly ever, not applicable) 2, Social support from manager Using Cox regression analysis, we examined the rates of the 4 types of caseness for participants exposed to workrelated violence compared with unexposed participants. Because the types of cases were mutually exclusive, they were competing risk outcomes (24). As we aimed to study etiology, we calculated cause-specific hazard ratios for each type of caseness, treating the other types as censorings, as recommended by Lau et al. (24). Further, we tested whether the effects of work-related violence were similar across outcomes, as described by Putter et al. (25). Hazard ratios were adjusted for gender, age, cohabitation, education, income, social support from colleagues, social support from supervisor, influence at work, and quantitative demands at work. We adjusted for differences between the original studies by stratifying models for the original study and method of data collection (questionnaire, telephonic interview, face-to-face interview). This approach % No. of Cases 4 (always) 3, (often) 4, (sometimes) 4, (seldom, never/ hardly ever, not applicable) 3, Quantitative demands at work by quartile 48.0 (19.8) 4 (highest) 5, , , (lowest) 2, Influence at work by 54.2 (24.2) quartile 4 (highest) 2, , , (lowest) 3, Abbreviation: SD, standard deviation. % Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012 Am J Epidemiol. 2011;174(12):

89 1358 Madsen et al. also deals with clustering within the original studies, assuming a fixed effect of exposure across studies (26). We did not stratify analyses by gender, as an interaction term between gender and violence was statistically nonsignificant, indicating no effect modification by gender. The proportional hazards assumption was assessed visually by inspecting the log-log hazard plots. Because this inspection revealed possible time dependency for anxiolytic caseness, with hazards crossing at 900 days, we fitted an additional extended model for this outcome, allowing for different effects before and after 900 days. The functional form of covariates was tested by log likelihood, comparing models with covariates entered as categorical, ordinal, or continuous variables. Further, we conducted 4 types of sensitivity analyses: 1) adjusting for occupational group, 2) adjusting for regular use of pain medication, 3) examining the separate effects of physical assault and threats, and 4) applying the survey date as baseline. Adjustment for occupational group was as sensitivity analysis only, as work-related violence is strongly related to occupational group (6, 27), and occupational group has been used as a proxy measure for work-related violence (6). Hence, adjusting for occupational group could be problematic because of collinearity between the factors. The adjustment for regular use of pain medication was to examine whether the association with antidepressant use was independent of chronic pain. This analysis excluded data from DWECS, as this information was unavailable. The separate effects of physical assault and threats were examined by defining 3 categories of exposure: 1) unexposed to threats and physical assault (referent), 2) exposed to threats only, and 3) exposed to physical assault. This analysiswasdonetoreducepossibleresponsebiascausedby depressive symptoms affecting the experience of others behavior as threatening, which is plausible as depression affects recognition of emotional expressions (28). We also examined whether the risk of using psychotropics increased with the frequency of violent episodes, excluding data from COPSOQ II and DWECS 2005 where these data were unavailable. On the basis of the distribution of participants, we categorized the number of episodes into 0, 1 2, and 3 ormoreandusedthelog-ranktest for trend to test if risk increased with frequency. The study was approved by the Danish Data Protection Agency (29). All statistical tests were 2 sided, used a statistical significance level of P < 0.05, and were conducted with SAS, version 9.1, software (SAS Institute, Inc., Cary, North Carolina). RESULTS Table 1 shows the characteristics of participants and numbers of cases. The majority of the population were female (55%), and most were cohabiting (79%). The mean age was 41 years. Exposure to work-related violence was reported by 1,447 participants (9%). Of the exposed participants, 841 (58%) were exposed to threats but not to physical assault (data not shown in table). The most frequent type of caseness was antidepressant use (617 participants, 4%), and the least frequent type was combined use of antidepressants and anxiolytics (156 participants, 1%). Just over 400 participants (3%) entered treatment with either anxiolytics or hypnotics only. The crude analyses showed higher risk of entering treatment with antidepressants alone or combined with anxiolytics for participants exposed to work-related violence (6% and 2%, respectively) compared with nonexposed participants (4% and 1%, respectively). For use of anxiolytics and hypnotics,the risks were similar for exposed and nonexposed participants. Table 2 shows the modeled cause-specific hazard ratios for use of psychotropics in relation to work-related violence and the covariates gender, age, cohabitation, education, income, social support from colleagues, social support from supervisor, influence at work, and quantitative demands at work. In this multivariable model, work-related violence was associated with increased risk of entering use of antidepressants alone (hazard ratio (HR) ¼ 1.38, 95% confidence interval (CI): 1.09, 1.75) or in combination with anxiolytics (HR ¼ 1.74, 95% CI: 1.13, 2.70). However, there was no association for anxiolytics alone (HR ¼ 1.05, 95% CI: 0.76, 1.45) or hypnotics only (HR ¼ 1.05, 95% CI: 0.75, 1.46). The effects of work-related violence were statistically significantly different across the outcomes (P ¼ 0.02, data not shown in tables). The extended analysis with a time-dependent effect of work-related violence on use of anxiolytic medications also showed no association. Respective hazard ratios before and after 900 days were 1.11 (95% CI: 0.76, 1.63) and 0.91 (95% CI: 0.49, 1.70) (data not shown in tables). Sensitivity analyses None of the sensitivity analyses substantially changed results (data not shown in tables). When adjusted for occupational group, the hazard ratio for antidepressant use was 1.36 (95% CI: 1.07, 1.73). When adjusting antidepressant use for regular use of pain medications, we found that the use of pain medications was associated with antidepressants (HR ¼ 1.75, 95% CI: 1.38, 2.22) but that the hazard ratio for work-related violence was unchanged (HR ¼ 1.51, 95% CI: 1.12, 2.04). Regarding the separate effects of physical assault and threats, the risk estimates were largely similar for the 2 types of exposure. The hazard ratios for antidepressant use were 1.45 (95% CI: 1.08, 1.95) for threats and 1.29 (95% CI: 0.91, 1.83) for physical assault, and for antidepressants and anxiolytics, the hazard ratios were 1.62 (95% CI: 0.91, 2.89) for threats and 1.90 (95% CI: 1.04, 3.46) for physical assault. Findings were also largely unchanged by using the survey date as the baseline: The hazard ratios for antidepressant use alone and combined with anxiolytics were 1.22 (95% CI: 0.92, 1.62) and 2.21 (95% CI: 1.32, 3.68), respectively. Frequency of violent episodes Table 3 shows use of psychotropic medications in relation to frequency of violent episodes in the subsample with these data (n ¼ 8,593). One or 2 episodes were reported by 480 participants (6%), and 352 participants (4%) experienced 3 or more violent episodes. Neither the crude nor the multivariable analysis indicated a relation between number of episodes and use of psychotropics. For anxiolytic medications, though, Am J Epidemiol. 2011;174(12): Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012

90 Work-related Violence and Use of Psychotropics 1359 Table 2. Cause-specific Hazard Ratios for Use of Psychotropics in Association With Work-related Violence and Covariates, Denmark, the risk estimates tended to increase with episodes, but the log-rank test for trend was statistically nonsignificant (P ¼ 0.35) (data not shown in tables). DISCUSSION Antidepressants With a large cross-occupational sample of Danish employees, this article shows that exposure to work-related violence is associated with increased risk of entering treatment with antidepressants but not treatment with anxiolytics or hypnotics only. This association was independent from occupational group and use of pain medications. More frequent exposure was not associated with greater risk of use of psychotropics. Our results are in line with those of smaller occupationspecific studies linking work-related violence with selfreported mental health outcomes, such as depressed mood and sleep problems (4, 5). Our findings also correspond to an ecologic study (6) that showed that employment in occupations highly exposed to work-related violence was associated with hospitalization for depression and anxiety. Our findings contrast, though, with a previous null finding on work-related violence and antidepressant use (17). However, because this previous study used a subsample of the present study, with only 4,958 participants, we believe that the null finding was due to lack of statistical power. A previous study among home-care workers showed increasing risk of depression with greater severity of the violent episode (5). Our results, however, do not indicate that more episodes of work-related violence are associated with greater risk of mental health problems. This seeming discrepancy may be because severity better measures the emotional impact of the episode, compared with frequency of exposure. Furthermore, qualitative research indicates that frequent workrelated exposure to violence may cause habituation (30), possibly counteracting a relation between number of episodes and use of psychotropics. Alternatively, the lack of association Antidepressants and Anxiolytics Anxiolytics Hypnotics Only HR a 95% CI HR a 95% CI HR a 95% CI HR a 95% CI Work-related violence (yes vs. no) , , , , 1.46 Women vs. men , , , , 1.79 Age per 5-year increase , , , , 1.27 Cohabitation (yes vs. no) , , , , 1.03 Education per SD increase , , , , 1.15 Income per quartile increase , , , , 1.18 Social support from colleagues per unit increase , , , , 1.12 Social support from supervisor per unit increase , , , , 1.07 Influence per unit increase , , , , 1.04 Quantitative demands per SD increase , , , , 1.10 Abbreviations: CI, confidence interval; HR, hazard ratio; SD, standard deviation. a Statistical model includes the following: work-related violence, gender, age, cohabitation, education, income, social support from colleagues, social support from supervisor, influence at work, and quantitative demands at work. could be explained by healthy worker effects, as employees may leave jobs with frequent exposure to violence if they are experiencing distress (31). A causal effect of work-related violence on mental disorder is plausible, given substantial evidence concerning the effects of domestic violence (1, 2). The mechanism by which workrelated violence may affect mental health possibly involves the emotional reactions of helplessness, which can be triggered by violent acts (32) and have been related to the etiology of mental disorder (33). A number of strengths and weaknesses of this study should be noted. Because the outcome was assessed through registry data, the loss to follow-up was limited to individuals who died or migrated, who were censored at date. This comprehensive follow-up strengthens the study as it precludes differential attrition of participants potentially biasing results toward the null (7). Moreover, because the purchase of psychotropic medications requires a prescription, this outcome measure is based on a medical assessment that treatment with psychotropics is needed. Consequently, it likely has high validity for measuring clinically pertinent mental health problems. However, the use of purchases of medications as an outcome measure leaves a problem of what specific mental health problem is measured, as antidepressants are treatments for a range of conditions, including depressive disorders (34), anxiety disorders (35), and post-traumatic stress disorder (36). Hence, this study cannot determine whether work-related violence is associated with several of these disorders or is specific to one. In light of a previous study (6) that found associations with hospitalization for both depressive and anxiety disorders, the exposure likely affects these disorders broadly rather than specifically. Notably, though, we found no association with anxiolytics-only and hypnotics-only treatment. This lack of association may suggest that work-related violence is associated with mental health problems requiring longer-term treatment rather than acute psychological crises the only situation for which treatment by anxiolytics is Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012 Am J Epidemiol. 2011;174(12):

91 1360 Madsen et al. Table 3. Frequency of Work-related Violence and Use of Psychotropics in the Subpopulation, Number of Cases, and Cause-specific Hazard Ratios, Denmark, Anxiolytics Hypnotics Only Antidepressants and Anxiolytics Antidepressants No. % HR a 95% CI No. of Cases % HR a 95% CI No. of Cases % HR a 95% CI No. of Cases % HR a 95% CI No. of Cases No. 8, Violent episodes 0 7, Referent Referent Referent Referent , , , , , , , , 3.08 Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusted for gender, age, cohabitation, education, income, social support from colleagues, social support from supervisor, influence at work, and quantitative demands at work. recommended in Denmark (37). However, as we did not examine the length of treatment course and as the indication for treatment was unavailable, further research is needed to establish the specific mental health consequences of workrelated violence. Because the outcome measure was based on pharmaceutical treatment, this study is limited to recognized and pharmaceutically treated mental disorders. This is of methodological concern, because most mental disorders are untreated (38), and treatments also include nonpharmaceutical options such as psychotherapy (38). Furthermore, psychotropics can be used to treat physical health problems (21). Results could be biased if exposure was related to treatment-seeking behaviors and the type of treatment entered, possibly through occupational group. The robustness of the results to adjustment for occupational group and the specificity of the associations to antidepressant treatment suggest that bias by treatment-seeking behaviors may be minor. Nonetheless, when interpreting the results, one should keep in mind that use of psychotropics does not measure mental disorder but pharmaceutical treatment. Our study population was cross-occupational. It was, however, not representative of the Danish work force, as one study (PUMA) encompassed only human service professionals. As these professions are highly exposed to work-related violence (6), the prevalence estimates in this paper are not representative of Danish employees. The nonrepresentative study population may also have biased our results toward the null if employees become habituated to work-related violence, as suggested previously (30). The overrepresentation of human service professionals was more pronounced in the sample with frequency data, and potential underestimation may particularly pertain to the results from this subsample. Another methodological issue is the retrospective assessment of exposure, which could result in response bias if cases were more likely than noncases to report exposure. The exposure operationalization is likely to be relatively unbiased, as it probably does not require extensive cognitive and emotional processing of the respondent. However, as depression reduces recognition of emotional facial expressions (28), individuals with depressive symptoms may experience others as more threatening, possibly causing response bias. As we consider the reporting of physical assault less sensitive to such bias, we did sensitivity analyses examining the effect of physical assault only. These analyses showed similar results, suggesting that response bias does not substantially influence our results. Nonetheless, future studies should use more frequent exposure assessment to minimize length of recall and assess timing of exposure more accurately. Further, the study is limited by possible residual confounding by unmeasured variables related to mental health, including exposure to violence outside the workplace. Were these factors unequally distributed between exposed and unexposed participants, they could have biased results, despite adjustment for confounders including gender, age, cohabitation, education, income, and work-related social support, influence, and quantitative demands. Following this line, it should be noted that work-related violence is highly related to occupation. The consequences of this association are 2-fold. First, respondents exposed to work-related violence are more likely to witness work-related violence against others, possibly affecting mental Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012 Am J Epidemiol. 2011;174(12):

92 Work-related Violence and Use of Psychotropics 1361 health negatively (39). Hence, our results may be biased away from the null by contextual effects of witnessing violence. Second, selection into occupation could cause preexisting differences between employees exposed and unexposed to work-related violence. Although we tried to account for such differences by adjusting for occupational group, residual confounding may remain, as we could apply only the major groupings of ISCO-88. Exclusion of respondents purchasing psychotropic medications within 700 days before baseline also aimed to reduce potential reverse causality, but future studies should include more comprehensive measures of mental health status before exposure to substantiate a causal effect of work-related violence. In summary, this paper finds that work-related violence is associated with increased risk of entering treatment with antidepressants. These findings extend existing occupationspecific research and substantiate an effect of work-related violence on clinically pertinent mental health problems. Reducing levels of workplace violence may thus help to prevent clinically pertinent mental health problems, a source of substantial public health costs and disability worldwide (40). ACKNOWLEDGMENTS Author affiliations: National Research Centre for the Working Environment, Copenhagen, Denmark (Ida E. H. Madsen, Jan H. Pejtersen, Jakob B. Bjorner, Reiner Rugulies); Department of Public Health, University of Copenhagen, Copenhagen, Denmark (Finn Diderichsen, Reiner Rugulies); Department of Psychology, University of Copenhagen, Copenhagen, Denmark (Reiner Rugulies); Federal Institute of Occupational Safety and Health, Berlin, Germany (Hermann Burr); and Department of Occupational and Environmental Medicine, Bispebjerg University Hospital, Copenhagen, Denmark (Marianne Borritz). This work was supported by a grant from the Danish Working Environment Research Fund (grant ). The funding source had no role in the study design, data collection, analysis, interpretation of data, or the decision to submit the paper to publication. Conflict of interest: none declared. REFERENCES 1. Weich S, Patterson J, Shaw R, et al. Family relationships in childhood and common psychiatric disorders in later life: systematic review of prospective studies. Br J Psychiatry. 2009;194(5): Campbell JC. Health consequences of intimate partner violence. Lancet. 2002;359(9314): Duhart DT. Violence in the Workplace, Washington, DC: Bureau of Justice Statistics, US Department of Justice; Gerberich SG, Church TR, McGovern PM, et al. An epidemiological study of the magnitude and consequences of work related violence: the Minnesota Nurses Study. Occup Environ Med. 2004;61(6): Geiger-Brown J, Muntaner C, McPhaul K, et al. Abuse and violence during home care work as predictor of worker depression. Home Health Care Serv Q. 2007;26(1): Am J Epidemiol. 2011;174(12): Wieclaw J, Agerbo E, Mortensen PB, et al. Work related violence and threats and the risk of depression and stress disorders. J Epidemiol Community Health. 2006;60(9): Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. Philadelphia, PA: Wolters Kluwer; Borritz M, Rugulies R, Bjorner JB, et al. Burnout among employees in human service work: design and baseline findings of the PUMA study. Scand J Public Health. 2006;34(1): Kristensen TS, Hannerz H, Høgh A, et al. The Copenhagen Psychosocial Questionnaire a tool for the assessment and improvement of the psychosocial work environment. Scand J Work Environ Health. 2005;31(6): Pejtersen JH, Kristensen TS, Borg V, et al. The second version of the Copenhagen Psychosocial Questionnaire. Scand J Public Health. 2010;38(suppl 3): Feveile H, Olsen O, Burr H, et al. The Danish Work Environment Cohort Study 2005: from idea to sampling design. Stat Transit. 2007;8(3): Burr H, Bjorner JB, Kristensen TS, et al. Trends in the Danish work environment in and their associations with labor-force changes. Scand J Work Environ Health. 2003; 29(4): Danish Medicines Agency. About the register of medicinal product statistics. (In Danish). Copenhagen, Denmark: Danish Medicines Agency; ( 1024/visUKLSArtikel.asp?artikelID¼10895). (Accessed December 1, 2009). 14. World Health Organization. World Report on Violence and Health. Geneva, Switzerland: World Health Organization; International Labour Organization. ISCO-88: major, sub-major, minor and unit group titles. Geneva, Switzerland: International Labour Organization; ( bureau/stat/isco/isco88/major.htm). (Accessed January 25, 2011). 16. Simon RW. Revisiting the relationships among gender, marital status, and mental health. AJS. 2002;107(4): Madsen IE, Diderichsen F, Burr H, et al. Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study. Scand J Work Environ Health. 2010;36(6): Harris T. Recent developments in understanding the psychosocial aspects of depression. Br Med Bull. 2001;57(1): Netterstrøm B, Conrad N, Bech P, et al. The relation between work-related psychosocial factors and the development of depression. Epidemiol Rev. 2008;30(1): Bonde JP. Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med. 2008;65(7): Trifirò G, Barbui C, Spina E, et al. Antidepressant drugs: prevalence, incidence and indication of use in general practice of southern Italy during the years Pharmacoepidemiol Drug Saf. 2007;16(5): World Health Organization. The anatomical therapeutic chemical classification system with defined daily doses (ATC/ DDD). Geneva, Switzerland: World Health Organization; ( (Accessed September 7, 2009). 23. World Health Organization. International Classification of Diseases. Geneva, Switzerland: World Health Organization; ( (Accessed May 2, 2005). 24. Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol. 2009;170(2): Putter H, Fiocco M, Geskus RB. Tutorial in biostatistics: competing risks and multi-state models. Stat Med. 2007; 26(11): Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012

93 1362 Madsen et al. 26. Katsahian S, Latouche A, Mary JY, et al. Practical methodology of meta-analysis of individual patient data using a survival outcome. Contemp Clin Trials. 2008;29(2): Salminen S. Violence in the workplaces in Finland. J Safety Res. 1997;28(3): Bourke C, Douglas K, Porter R. Processing of facial emotion expression in major depression: a review. Aust N Z J Psychiatry. 2010;44(8): The Danish Data Protection Agency. Introduction to the Danish Data Protection Agency. Copenhagen, Denmark: Danish Data Protection Agency; ( english/). (Accessed December 1, 2009). 30. Chapman R, Styles I, Perry L, et al. Nurses experience of adjusting to workplace violence: a theory of adaptation. Int J Ment Health Nurs. 2010;19(3): Farrell GA, Bobrowski C, Bobrowski P. Scoping workplace aggression in nursing: findings from an Australian study. J Adv Nurs. 2006;55(6): Brewin CR, Andrews B, Rose S. Fear, helplessness, and horror in posttraumatic stress disorder: investigating DSM-IV criterion A2 in victims of violent crime. JTraumaStress. 2000;13(3): Abramson LY, Seligman ME, Teasdale JD. Learned helplessness in humans: critique and reformulation. J Abnorm Psychol. 1978; 87(1): National Board of Health. Reference program for unipolar depression in adults. (In Danish). Copenhagen, Denmark: National Board of Health; ( aspx). (Accessed February 11, 2011). 35. National Board of Health. Reference program for anxiety disorders in adults. (In Danish). Copenhagen, Denmark: National Board of Health; ( aspx). (Accessed February 11, 2011). 36. National Collaborating Centre for Mental Health. Post-Traumatic Stress Disorder The Management of PTSD in Adults and Children in Primary and Secondary Care. London, United Kingdom: Gaskell and the British Psychological Society; Institute for Rational Pharmacotherapy. ATC-groups N05B and N05C anxiolytics and hypnotics. National recommendation list. (In Danish). Copenhagen, Denmark: Institute for Rational Pharmacotherapy; ( rekommandationsliste/baggrundsnotater/nervesystemet_ analgetika_og_psykofarmaka/atc-gruppe_n05b_og_ n05c_-_angstdaempende_midler_og_sovemidler.htm). (Accessed February 11, 2011). 38. Wittchen HU, Jacobi F. Size and burden of mental disorders in Europe a critical review and appraisal of 27 studies. Eur Neuropsychopharmacol. 2005;15(4): Scarpa A. Community violence exposure in a young adult sample lifetime prevalence and socioemotional effects. J Interpers Violence. 2001;16(1): Lopez AD, Mathers CD, Ezzati M, et al. Global and regional burden of disease and risk factors, 2001: systematic analysis of population health data. Lancet. 2001;367(9524): Downloaded from at Det Nationale Forskningscenter for Arbejdsmiljà (NFA) on June 14, 2012 Am J Epidemiol. 2011;174(12):

94 APPENDIX 2. ARTICLE 2 Madsen IEH, Diderichsen F, Burr H, Rugulies R. Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study. Scandinavian Journal of Work, Environment & Health 2010 June 2;36(6): Madsen 2012 Social Work Environment and Mental Health

95 Original article Scand J Work Environ Health 2010;36(6): Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study by Ida EH Madsen, MSc, 1 Finn Diderichsen, PhD, 2 Hermann Burr, PhD, 1 Reiner Rugulies, PhD 1, 2, 3 Madsen IEH, Diderichsen F, Burr H, Rugulies R. Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study. Scand J Work Environ Health. 2010;36(6): Objectives Previous Danish studies have shown that employees who work with people (ie, do person-related work) are at increased risk of hospitalization with a diagnosis of depression. However, these studies were purely register-based and consequently unable to point to factors underlying this elevated risk. This paper examines whether person-related work is associated with incident use of antidepressants, and whether this association is mediated by several work environment exposures. Methods Self-reported data from the Danish work environment cohort study in 2000 were linked with the use of antidepressants between We included 4958 respondents in our study after excluding those with severe depressive symptoms or use of antidepressants at baseline. Results Compared to employees doing non-person-related work, the use of antidepressants was increased statistically significantly for healthcare workers and statistically non-significantly for educational workers. The use of antidepressants was not elevated for social or customer service workers, or those doing other types of person-related work. The increased risks of antidepressant-use for healthcare and educational workers were attenuated when adjusted for emotional demands at work. Conclusions The results imply that healthcare and educational workers in Denmark are at increased risk of depression and that this risk is partly mediated by the high emotional demands of the work. Key terms depressive disorder; emotional demand; emotional labor; human service work; occupational health; threat; violence. Depressive disorders are costly, both for those afflicted and society at large (1, 2). The causes of depression are, as yet, poorly understood but thought to involve an interplay of biological, psychological and social factors (3). Work environment factors such as high demands, low control and low social support have been associated with depression in longitudinal studies; however, knowledge is scarce concerning the potential role of other work environment factors (4, 5). Previous Danish studies (6 8) have found increased hospitalization with depression for employees doing person-related work [ie, work that requires face-to-face or voice-to voice interactions with clients (9)]. However, as these studies were register-based, exposure was defined solely by job group and not the actual amount of time spent with clients. Furthermore, the studies could not point to specific work environment exposures mediating the risk. Several characteristics of person-related work could be involved in an increased risk of depression. Firstly, it has been argued that such work is particularly emotionally demanding (6, 10), for instance due to confrontation with clients problems and suffering (9, 10). Previous research has related emotional demands at work [ie, aspects of the job that require sustained emotional effort (10)] to depression (11), depressive- and anxiety-symptoms (12), fatigue (13 15), psychological distress (13, 15), emotional well-being (14), and burnout (10, 16, 17). Secondly, employees doing person-related work must manage (ie, induce or reduce) their emotions in order to express organization- 1 The National Research Centre for the Working Environment, Copenhagen, Denmark. 2 Institute of Public Health, University of Copenhagen, Denmark. 3 Department of Psychology, University of Copenhagen, Denmark. Correspondence to: Ida EH Madsen, the National Research Centre for the Working Environment, Lerso Park Allé 105, DK-2100 Copenhagen, Denmark. [ [email protected]] Scand J Work Environ Health 2010, vol 36, no 6 435

96 Person-related work and incident use of antidepressants ally required emotions, which according to Hochschild s (18) theory of emotional labor puts them at risk of mental health problems. Emotional labor has been linked to mental health outcomes such as reduced well-being (19), emotional exhaustion (20 23), psychosomatic complaints, and irritation (23). Finally, person-related work has been associated with exposure to work-related threats and violence (24 27), which in turn has been related to depression (26), fatigue (27), and general mental health (28). Inspired by the previous Danish studies relating person-related work to hospitalization for depression (6 8), this paper examines: (i) whether different types of personrelated work, defined by a combination of job groups and self-reported client-contact, are associated with the increased incident use of antidepressants and (ii) whether this association is mediated by one or more of the following work environment exposures: emotional demands, demands for hiding emotions, threats, and violence. As the interaction between employee and client could differ according to the characteristics of the client (9, 29, 30), we distinguished between five types of person-related work: (i) healthcare, (ii) education, (iii) social, (iv) customer service, and (v) other personrelated work. Methods Study design We merged data from the Danish work environment cohort study (DWECS) from 2000 with the register of medicinal product statistics, by the participants unique personal identification number. The register contains all purchases of prescription drugs at pharmacies in Denmark since 1 January 1995 (31). DWECS is a work environment survey administered to representative samples of the working population in Denmark every five years. Data from 2000 were gathered through telephone interviews, between October 2000 and January 2001 (32). The response-rate was 75%. Differences between respondents and non-respondents were minor, and DWECS 2000 is considered representative of the Danish working population (32). A detailed description of DWECS 2000 has been published elsewhere (32). Population There were 6166 participants of DWECS 2000 who were active in the labor market. We excluded selfemployed respondents (N=530) as this group was small and social conditions of self-employed and employees differ (33). We also excluded apprentices (N=233) and those missing data on key variables (N=162). To ensure a prospective study design (ie, that respondents were not suffering from depression at baseline), we additionally excluded 143 respondents with severe self-reported depressive symptoms at baseline, and 140 respondents who had purchased antidepressants during the 12 months prior to the start of follow-up. The final study population was 4958 respondents. The mean age was 40.3 years and 51.5% of the participants were male. Most respondents (65.3%) were whitecollar workers. Operationalizations Person-related work. We created five categories of person-related work: (i) healthcare (eg, doctors, nurses), (ii) education (eg, teachers, pedagogues), (iii) social (eg, police officers, social workers), (iv) customer service (eg, sales personnel, wait staff), and (v) other personrelated work. These were compared to the reference group comprising individuals doing non-person-related work. Respondents were classified as doing personrelated work if they reported contact with clients at least ¾ of the time. Contact with clients was assessed by one item in the interview: Do you, in relation with your work, deal with people who are not employed at the workplace? (eg, customers, clients, passengers, students), with the responses: almost always, ¾ of the time, ½ of the time, ¼ of the time, rarely/ very little, and never. Some contact with clients was common and 64.7% of all respondents reported working with clients at least ¼ of the time (data not shown). In order to define respondents mainly working with clients (ie, doing person-related work), we established a cut-off point for client contact, by examining its distribution in occupations commonly associated with person-related work (ie, doctors, dentists, nurses, teachers, and wait staff). This was skewed towards working with clients ¾ of the time or more (data not shown); this was, therefore, applied as the cut-off point. The type of person-related work was determined by occupation and the type of clients with whom the respondent interacted (see appendix, p444). Thus, healthcare work was defined as working with clients who were ill, educational work as working with clients who were normally functioning students, and social work as working with clients who had social problems. Other person-related work encompassed those occupations where the characteristics of clients were uncertain. The classification of occupation was based on a modified version of the International Standard of Classification of Occupations (ISCO-68) (34). 436 Scand J Work Environ Health 2010, vol 36, no 6

97 Madsen et al Work environment exposures. Emotional demands were measured using a scale of three items: Does your work bring you in emotionally taxing situations?, Is your work emotionally taxing?, and Are you emotionally affected by your work?. The scale was dichotomized into high versus low emotional demands by the median of the DWECS 2000 population. Demands for hiding emotions, as an indication of emotional labour (18), were measured by one item: Does your job require you to hide your emotions?. This was dichotomized into high ( partially, to a high extent, or to a very high extent ) versus low ( to a small extent or to a very small extent ). Exposure to threats and violence was assessed by asking: Have you within the past 12 months been exposed to threats of violence in your workplace? and Have you within the past 12 months been exposed to physical violence in your workplace?. The possible responses were yes or no. Potential confounders. As potential confounders we included: gender, age, cohabitation (ie, living with partner/spouse, yes/no), parental status (ie, having children living at home, yes/no), and socioeconomic position, as these factors have been associated with depression (3, 35, 36). Age and gender were retrieved from the central population register. Cohabitation and parental status were self-reported. Socioeconomic position was obtained from DWECS, by a standard composite measure of five groups (I V), based on employment grade, job title and education (37): group I: executive managers and/or having a university degree, group II: middle managers and/or 3 4 years of further education, group III: other white-collar workers, group IV: skilled bluecollar workers, and group V: semi-skilled or unskilled workers. Depression at baseline. Baseline mental health was assessed in DWECS, using the 5-item mental health inventory of the 36-item short-form (SF-36) health survey (38), which is valid for assessing mood disorders (39). In accordance with previous studies (40 42), respondents scoring 52 points were considered to suffer from severe depressive symptoms, and excluded from the analyses. Incident use of antidepressants. Antidepressants were defined as medications coded N06a by the anatomical therapeutic chemical classifications system (43). Follow-up of purchases of antidepressants was 5 years, starting on 15 March 2001, 1.5 months after the last baseline interviews were conducted. We inserted the lagtime between baseline and the start of follow-up because the outcome is treatment-related, and treatment often is not initiated immediately with the onset of depression (44). Incident use of antidepressants (caseness) was defined as at least one purchase of antidepressants during follow-up. Data analysis Data were analyzed by logistic regressions. Cox regression analysis was considered, but could not be applied due to non-proportional hazards (data not shown). We assessed both the crude association between personrelated work and the use of antidepressants in addition to the risk estimates adjusted for potential confounders (model I). Mediation. Mediation by work environment exposures was assessed in three steps, examining whether (i) person-related work was associated with increased exposure, (ii) the exposures were associated with incident use of antidepressants, and (iii) the risk estimates for person-related work and use of antidepressants were attenuated, when adjusting for the exposures. For this, we applied three statistical models: model IIA (adjusting model I for emotional demands), model IIB (adjusting model I for demands for hiding emotions), and model III (adjusting model I for both emotional demands and demands for hiding emotions). Post hoc analyses. We applied the Hosmer-Lemeshow goodness-of-fit test to evaluate the statistical models. As risks related to person-related work could be modified by gender (6) we furthermore exploratively stratified analyses by this factor. All analyses were carried out in SAS, version (SAS Institute, Cary, NC, USA). Table 1. Type of work and incident use of antidepressants. [Res=respondents; OR=odds ratio; 95% CI=confidence interval] Type of work Cases (N) Res (N) Crude analysis a Model I b OR 95% CI OR 95% CI Healthcare Education Social Customer service Other-personrelated Non-personrelated c 1.00 c Total a Calculated by multivariable logistic regression. b Calculated by multivariable logistic regression and adjusted for gender, age, cohabitation, parental status, and socioeconomic position. Hosmer- Leweshow goodness-of-fit test, P= c Reference Scand J Work Environ Health 2010, vol 36, no 6 437

98 Person-related work and incident use of antidepressants Results A total of 342 respondents (6.9%) started using antidepressants during follow-up (cases). Odds ratios (OR) of caseness for the different types of person-related work, compared to non-person-related work, are shown in table 1 on the previous page. There was an elevated risk for healthcare workers, with an OR of 1.70 [95% confidence interval (95% CI) ]. Also, educational workers had an elevated risk, though not statistically significant, with an OR of 1.37 (95% CI ). Further analyses on the degree of contact with the client (data not shown), showed the greatest use of antidepressants for those healthcare and educational workers who had the most contact with clients. For instance, compared to respondents doing non-person-related work, healthcare workers working with clients all the time had an OR of 1.73 (95% CI ), whereas healthcare workers working with clients ¾ of the time had an OR of 1.45 (95% CI ). Person-related work and work environment exposures Table 2 shows the associations between type of work and work environment exposures, adjusted for confounders. Healthcare, educational, and social workers and those doing other person-related work were more likely than those doing non-person-related work to report all four exposures. Customer service workers only had increased risks of reporting high demands for hiding emotions and exposure to threats. Work environment exposures and antidepressants use Table 3 shows the relations between work environment exposures and the incident use of antidepressants. High emotional demands were related to increased use of antidepressants, with an OR of 1.51 (95% CI ). High demands for hiding emotions were also associated with the use of antidepressants, although not statistically significantly, with an OR of 1.26 (95% CI ). There was no increased use of antidepressants among those exposed to threats or violence. Mediation Risk estimates for the use of antidepressants in relation to the type of work adjusted for emotional demands, demands for hiding emotions, and both factors, respectively, are shown in table 4. As no increased use of antidepressants was found for exposure to threats and violence, these factors were not included as mediators. When adjusted for emotional demands (model IIA), the risk estimate for healthcare work was reduced from 1.70 to 1.47 and became statistically insignificant. Also, the risk estimate for educational work was reduced from 1.37 to Adjustment for demands for hiding emotions (model IIB) resulted in a lesspronounced reduction of risk. When adjusting for the two exposures simultaneously (model III), risk estimates were similar to those obtained when adjusting only for emotional demands. In this final model, emotional demands (OR=1.43) predicted the use of antidepressants, but demands for hiding emotions (OR=1.07) did not. Table 2. Type of work and work environment exposures. [N=number of respondents; OR=odds ratio; 95% CI=95% confidence interval] Type of work High emotional demands High demands for hiding emotions Exposure to threats Exposure to violence N % OR a 95% CI N % OR a 95%CI N % OR a 95%CI N % OR a 95%CI Healthcare Education Social Customer service Other personrelated Non-personrelated b b b b Total Goodness-of-fit, P-value c a Calculated by multivariable logistic regression and adjusted for gender, age, cohabitation, parental status, and socioeconomic position. b Reference c P-value calculated using Hosmer-Leweshow test. 438 Scand J Work Environ Health 2010, vol 36, no 6

99 Madsen et al Post-hoc analyses The Hosmer Lemeshow tests showed acceptable fits for the models, except for the association between personrelated work and demands for hiding emotions (see tables 1 4). This was due to interactions (departure from multiplicativity) between covariates. Stratified analyses yielded patterns similar to the presented results, with the exception that customer service work was not associated with demands for hiding emotions for those of higher socioeconomic position (groups I and II, data not shown). When analyses were stratified by gender, the respective OR for use of antidepressants among men and women were: 2.66 (95% CI ) and 1.53 (95% CI ) for healthcare work; 2.20 (95% CI ) and 1.11 (95% CI ) for educational work; 0.98 (95% CI ) and 1.17 (95% CI ) for customer service work, and 1.59 (95% CI ) and 0.90 (95% CI ) for other person-related work. A gender-stratified risk estimate for social work was not obtained due to lack of statistical power. Table 3. Work environment exposures and use of antidepressants. [N=number of respondents; OR=odds ratio; 95% CI=95% confidence interval] Cases (N) % OR a 95% CI Goodnessof-fit, P-value b Emotional demands High Low c Demands for hiding emotions High Low c Threats Yes No c Violence Yes No c a Calculated by multivariable logistic regression and adjusted for gender, age, cohabitation, parental status, and socioeconomic position. b Value for goodness of fit calculated using Hosmer-Leweshow test. c Reference. Discussion This study showed an increased use of antidepressants among employees doing certain types of person-related work and indicated that this risk is partially mediated by the emotional demands of these types of work. The use of antidepressants was statistically significantly increased among healthcare workers and statistically non-significantly increased among educational workers. We found no increased use of antidepressants among employees doing customer service or other personrelated work. Nor did we find increased use among employees doing social work; the sample size, however, was very small for this group (75 respondents). Human service work Healthcare and educational work (along with social work) have been termed human service work (29). It has been argued that this work is particularly emotion- Table 4. Type of work and use of antidepressants adjusted for work environment exposures. [OR=odds ratio; 95% CI=confidence interval] Type of work Model I Model IIA a Model IIB b Model III c OR d 95% CI OR d 95% CI OR d 95% CI OR d 95% CI Healthcare Educational Social Customer service Other person-related Non-person-related 1.00 e 1.00 e 1.00 e 1.00 e Emotional demands (high versus low) Demands for hiding emotions (high versus low) Goodness-of-fit, P-value f a Adjusted for emotional demands. b Adjusted for demands for hiding emotions. c Adjusted for emotional demands and demands for hiding emotions. d Calculated by multivariable logistic regression and adjusted for gender, age, cohabitation, parental status, and socioeconomic position. e Reference f Value for goodness-of-fit calculated using Hosmer-Leweshow test. Scand J Work Environ Health 2010, vol 36, no 6 439

100 Person-related work and incident use of antidepressants ally demanding (47). Also, such work implies taking responsibility for the fundamental human needs of the client (48), experiencing a lack of reciprocity in relations with clients (49), and being confronted by the clients problems and suffering (10). Our results, in line with previous research (50), support the notion that human service work is more emotionally demanding than other types of work and indicated that these emotional demands are related to a risk of depression. According to Brown & Harris (51, p 233), the inability to hold good thoughts about ourselves, our lives, and those close to us is a central characteristic of depression. It could be reasoned, that witnessing human suffering first hand and having to help and care for others who do not reciprocate accordingly might affect this ability. Further research examining mechanisms between human service work and depression is, therefore, warranted. The results support a distinction between human and customer service work, as the latter was associated with neither emotional demands nor the use of antidepressants. This finding suggests that it is not the interaction with the client per se that put employees who work with people at risk of depression, but rather the quality and content of the relationship. Mediation The results suggest that emotional demands partly mediate the relation between person-related work and depression. However, underlying mechanisms remain unclear as the items used to measure emotional demands in DWECS 2000 were generic. Qualitative research examining the construct of emotional demands in different work-contexts could help further elucidate this issue and explain our results. In contrast, our results do not support an effect of emotional labor on depression. This could be due to a poor operationalization of emotional labor, as this involves not only hiding emotions, but the active management of emotions (18). Alternatively, it could be argued that demands for hiding emotions more purely measures the work environment context than the measure for emotional demands; because the items measuring the latter reflect the experience of the situation as emotionally demanding, this measurement is likely affected by the emotional and cognitive processing of the respondent. However, we found a close relation between type of work and reports of emotional demands, suggesting that this experience is highly related to the work environment of the employee. Because the measure of emotional demands applied was generic, it could be relevant to examine how more specific work environment exposures, such as witnessing the pain and suffering of clients, relate to the experience of emotional demands. Such research could help disclose factors that give rise to emotional demands at work. The correlations between emotional demands and the other work environment exposures included in the present paper were modest; the Pearson correlation coefficients ranged from 0.17 for violence and 0.23 for threats to 0.52 for demands for hiding emotions (data not shown). These correlations suggest that emotional demands and demands for hiding emotions are interrelated constructs, although they have been conceptualized separately. Such interrelation could exist because hiding emotions requires sustained emotional effort (ie, is emotionally demanding). In contrast to a previous Danish study (26) that related threats and violence (measured by job exposure matrix) to depression, we found no increased use of antidepressants for those exposed to threats or violence. We measured threats and violence at the level of the individual and therefore avoided ecological fallacy (52). Our null-finding, however, could be due to a lack of statistical power. Alternatively, threats and violence could be related to anxiety disorders rather than depression, as dangerous events have been related to episodes of anxiety but not depression (53). We plan to examine this issue further in future analyses. We did not examine mediation by classic work environment constructs, such as the demands control support model (54), as the choice of potential mediators was guided by existing literature characterizing personrelated work (eg, 9, 18, 29, 30), rather than a specific theoretical model. However, the increased use of antidepressants associated with person-related work could also be explained by other work environment factors not examined in this study. Following this vein, it should be noted that most Danish employees undertaking healthcare and educational work are employed within the public sector. Due to the high correlation (data not shown) between our exposure and public employment, we did not adjust for employment sector, as it could lead to overadjustment. However, work environment factors disproportionately prevalent within the public sector could bias our results. Methodological considerations We applied the use of antidepressants as a proxy-measure of depression, because it yields three methodological advantages over questionnaire-based symptom rating scales. Firstly, it is objective (ie, not self-reported) and avoids bias due to common method variance (55). Secondly, no active participation from respondents is required after baseline assessment. Attrition during follow-up, which could be differential according to health status (56), is thereby prevented. Thirdly, use of antidepressants is measured continuously and cumulated. Most other measures of depression obtain outcome status at a fixed time, and non-cases could have experienced 440 Scand J Work Environ Health 2010, vol 36, no 6

101 Madsen et al depression during follow-up, were they in remission at the time of follow-up. Misclassification of outcome. However, applying the use of antidepressants as a proxy-measure of depression results in some misclassification of outcome; firstly, antidepressants are used to treat disorders other than depression (eg, anxiety disorders) (57), causing falsepositive classification. If this is non-differential to exposure, it could cause an underestimation of effect (52). Secondly, not all persons suffering from depression are treated with antidepressants (35, 58), affecting the sensitivity of the measure. The use of mental health services has been associated with age, educational level, social support, and physical health (58), and differential sensitivity of antidepressants as a proxy for depression has been shown in relation to gender and socioeconomic position (59). We found the greatest use of antidepressants among healthcare workers. Employees in this sector could be more likely than others to enter treatment with antidepressants when depressed, causing differential misclassification, biasing our estimates. However, because we found the greatest use of antidepressants among healthcare workers with the most contact with clients, we consider it unlikely that treatment-seeking behaviors fully explain our results, as this would imply differential treatmentseeking related to the degree of contact with clients. Misclassification of exposures. Exposure status was only obtained at baseline, and respondents could have changed status during follow-up; if non-differential, this misclassification could bias our results towards the null. Also, reports of work environment exposures could be biased by subclinical depressive symptoms. We excluded respondents suffering from severe depressive symptoms at baseline, and those using antidepressants within 12 months prior to follow-up, but residual depressive symptoms could confound the associations of the use of antidepressants with both client contact and emotional demands. We consider it unlikely that reports of client contact were substantially biased by such symptoms, as we did not see an increased use of antidepressants for all types of personrelated work, but only in relation to certain types of work. Regarding emotional demands, the strong (and expected) associations with human service work suggest that these reflect the external environment. However, selection into human service work, of persons who experience high emotional demands and are particularly at risk of depression, cannot be ruled out. Causality. Although we adjusted for confounding by gender, age, cohabitation, parental status, and socioeconomic position, other risk factors for depression (eg, events in private life, childhood experiences, alcohol, and genetics) could confound results, were they related to type of work. Particularly, individuals who choose to do certain types of person-related work may be especially at risk of depression. It has been proposed that choosing a career within caretaking professions could be motivated by personality traits related to childhood roles of taking care of other family members needs to an excessive extent (parentification) (60). This has been corroborated empirically by increased levels of traumatic childhood events among psychotherapists (61) and psychology students with clinical aspirations (62), in addition to increased levels of parentification among counseling psychology trainees (63). Other studies have shown high degrees of parentification (60, 64) and childhood separation from parents (64) among social workers; unfortunately, these studies did not apply appropriate reference populations. However, one should take into account that the results of our study could reflect pre-existing characteristics of individuals who do certain types of person-related work, rather than a causal effect of these types of work. Concluding remarks We found an increased incident use of antidepressants among employees doing certain types of person-related work (ie, healthcare and educational work) in a representative sample of the Danish working population. These effects were partly mediated by the emotional demands of the work. The methodological issues discussed above should be taken into consideration when interpreting the findings. Acknowledgements The Danish Data Protection Agency approved this study (45). [Approval from the Danish National Committee on Biomedical Research Ethics is not required for Danish questionnaire- and register-based studies (46).] This work was supported by the Danish Working Environment Authority, the Ministry of Labor, and the Danish Working Environment Research Fund (grant numbers and ). The authors would like to thank Dr Harald Hannerz for his assistance, facilitating the analyses of this paper. References 1. Simon GE. Social and economic burden of mood disorders. Biol Psychiatry. 2003;54: Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJL. Global and regional burden of disease and risk factors, 2001: systematic analysis of population health data. Lancet. 2006;367: Scand J Work Environ Health 2010, vol 36, no 6 441

102 Person-related work and incident use of antidepressants 3. Harris T. Recent developments in understanding the psychosocial aspects of depression. Br Med Bull. 2001;57: Bonde JPE. Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med. 2008;65: Netterstrøm B, Conrad N, Bech P, Fink P, Olsen O, Rugulies R, et al. The relation between work-related psychosocial factors and the development of depression. Epidemiol Rev. 2008;30: Wieclaw J, Agerbo E, Mortensen PB, Bonde JP. Risk of affective and stress related disorders among employees in human service professions. Occup Environ Med. 2006;63: Hannerz H, Tüchsen F, Pedersen BH, Dyreborg J, Rugulies R, Albertsen K. Work-relatedness of mood disorders in Denmark. Scand J Work Environ Health. 2009;35(4): Wieclaw J, Agerbo E, Mortensen PB, Bonde JP. Occupational risk of affective and stress-related disorders in the Danish workforce. Scand J Work Environ Health. 2005;31(5): Zapf D. Emotion work and psychological well-being: a review of the literature and some conceptual considerations. Hum Resour Manage Rev. 2002;12: van Vegchel N, de Jonge J, Söderfeldt M, Dormann C, Schaufeli W. Quantitative versus emotional demands among Swedish human service employees: moderating effects of job control and social support. Int J Stress Manag. 2004;11: Wieclaw J, Agerbo E, Mortensen PB, Burr H, Tüchsen F, Bonde JP. Psychosocial working conditions and the risk of depression and anxiety disorders in the Danish workforce. BMC Public Health. 2008;8: Andrea H, Bültmann U, Beurskens AJHM, Swaen GMH, van Schayck CP, Kant IJ. Anxiety and depression in the working population using the HAD Scale. Soc Psychiatry Psychiatr Epidemiol. 2004;39: Bültmann U, Kant IJ, Schröer CAP, Kasl SV. The relationship between psychosocial work characteristics and fatigue and psychological distress. Int Arch Occup Environ Health. 2002;75: Vanroelen C, Levecque K, Moors G, Gadeyne S, Louckx F. The structuring of occupational stressors in a Post-Fordist work environment: moving beyond traditional accounts of demand, control and support. Soc Sci Med. 2009;68: Burr H, Albertsen K, Rugulies R, Hannerz H. Do dimensions from the Copenhagen Psychosocial Questionnaire predict vitality and mental health over and above the job strain and effort-reward imbalance models? Scand J Public Health. 2010;38: Borritz M, Rugulies R, Bjorner JB, Villadsen E, Mikkelsen OA, Kristensen TS. Burnout among employees in human service work: design and baseline findings of the PUMA study. Scand J Public Health. 2006;34: Sundin L, Hochwälder J, Bildt C, Lisspers J. The relationship between different work-related sources of social support and burnout among registered and assistant nurses in Sweden: a questionnaire survey. Int J Nurs Stud. 2007;44: Hochschild A. The managed heart : commercialization of human feeling. London: University of California Press; Giardini A, Frese M. Reducing the negative effects of emotion work in service occupations: Emotional competence as a psychological resource. J Occup Health Psychol. 2006;11: Brotheridge CM, Grandey AA. Emotional labor and burnout: comparing two perspectives of people work. J Vocat Behav. 2002;60: Grandey AA. When the show must go on : surface acting and deep acting as determinants of emotional exhaustion and peer-rated service delivery. Acad Manage J. 2003;46: Johnson HAM, Spector PE. Service with a smile: do emotional intelligence, gender, and autonomy moderate the emotional labor process? J Occup Health Psychol. 2007;12: Zapf D, Vogt C, Seifert C, Mertini H, Isic A. Emotion work as a source of stress: the concept and development of an instrument. Eur J Work Organ Psychol. 1999;8: Menckel E, Viitasara E. Threats and violence in Swedish care and welfare magnitude of the problem and impact on municipal personnel. Scand J Caring Sci. 2002;16: Rippon TJ. Aggression and violence in health care professions. J Adv Nurs. 2000;31: Wieclaw J, Agerbo E, Mortensen PB, Burr H, Tüchsen F, Bonde JP. Work related violence and threats and the risk of depression and stress disorders. J Epidemiol Community Health. 2006;60: Høgh A, Borg V, Mikkelsen KL. Work-related violence as a predictor of fatigue: a 5-year follow-up of the Danish Work Environment Cohort Study. Work Stress. 2003;17: Høgh A, Mikkelsen EGE. Is sense of coherence a mediator or moderator of relationships between violence at work and stress reactions? Scand J Psychol. 2005;46: Hasenfeld Y, English RA. Human service organizations: a conceptual overview. In: Hasenfeld Y, English RA, editors. Human service organizations: a book of readings. Ann Arbor (MI): University of Michigan Press; p Morris JA, Feldman DC. The dimensions, antecedents, and consequences of emotional labor. Acad Manage Rev. 1996;21: Lægemiddelstyrelsen [Danish Medicines Agency]. About the Register of Medicinal Product Statistics [Internet]. Copenhagen: Lægemiddelstyrelsen [Danish Medicines Agency]; [cited 1 December 2009]. Available from: Burr H, Bjorner JB, Kristensen TS, Tüchsen F, Bach E. Trends in the Danish work environment in and their associations with labor-force changes. Scand J Work Environ Health. 2003;29(4): Rose D, Harrison E. The European Socio-Economic Classification: a new social class schema for comparative European research. Eur Soc. 2007;9: International Labour Organization (ILO). ISCO-68 major, minor and unit groups [Internet]. ILO; [cited 1 December 2009]. Available from: bureau/stat/isco/isco68/major.htm 442 Scand J Work Environ Health 2010, vol 36, no 6

103 Madsen et al 35. Olsen LR, Mortensen EL, Bech P. Prevalence of major depression and stress indicators in the Danish general population. Acta Psychiatr Scand. 2004;109: Simon RW. Revisiting the relationships among gender, marital status, and mental health. AJS. 2002;107: Borg V, Kristensen TS. Social class and self-rated health: can the gradient be explained by differences in life style or work environment? Soc Sci Med. 2000;51: Bjorner JB, Thunedborg K, Kristensen TS, Modvig J, Bech P. The Danish SF-36 Health Survey: translation and preliminary validity studies. J Clin Epidemiol. 1998;51: Rumpf HJ, Meyer C, Hapke U, John U. Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard. Psychiatry Res. 2001;105: Strand BH, Dalgard OS, Tambs K, Rognerud M. Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36). Nord J Psychiatry. 2003;57: Holmes WC. A short, psychiatric, case-finding measure for HIV seropositive outpatients: performance characteristics of the 5-item Mental Health Subscale of the SF-20 in a male, seropositive sample. Med Care. 1998;36: Rugulies R, Bültmann U, Aust B, Burr H. Psychosocial work environment and incidence of severe depressive symptoms: prospective findings from a 5-year follow-up of the Danish Work Environment Cohort Study. Am J Epidemiol. 2006;163: World Health Organization (WHO). The Anatomical Therapeutic Chemical Classification System with Defined Daily Doses (ATC/DDD) [Internet]. Geneva: WHO [cited 7 September 2009]. Available from: classifications/atcddd/en/ 44. Angst J, Gamma A, Gastpar M, Lépine JP, Mendlewicz J, Tylee A. Gender differences in depression. Eur Arch Psychiatry Clin Neurosci. 2002;252: Datatilsynet [The Danish Data Protection Agency]. Introduction to the Danish Data Protection Agency [Internet]. Copenhagen: Datatilsynet; [cited 1 December 2009]. Available from: Den Centrale Videnskabsetiske Komité [The Danish National Committee on Biomedical Research]. Frequently asked questions on application [Internet]. Copenhagen: Den Centrale Videnskabsetiske Komité; [cited 1 December 2009]. Available from: Söderfeldt B, Söderfeldt M, Jones K, O Campo P, Muntaner C, Ohlson CG, et al. Does organization matter?: a multilevel analysis of the demand-control model applied to human services. Soc Sci Med. 1997;44: Dybbroe B. Omsorg i skæringspunktet mellem arbejde og liv [Care in the intersection between work and life]. Tidsskr Arbejdsliv. 2006;8: Schaufeli WB. The balance of give and take: toward a social exchange model of burnout. Int Rev Soc Psychol. 2006;19: Kristensen TS, Hannerz H, Høgh A, Borg V. The Copenhagen Psychosocial Questionnaire a tool for the assessment and improvement of the psychosocial work environment. Scand J Work Environ Health. 2005;31(): Brown GW, Harris TO. Social origins of depression. A study of psychiatric disorder in women. London: Tavistock; Rothman KJ, Greenland S, Lash TL. Modern epidemiology. Philadelphia (PA): Wolters Kluwer; Kendler KS, Hettema JM, Butera F, Gardner CO, Prescott CA. Life event dimensions of loss, humiliation, entrapment, and danger in the prediction of onsets of major depression and generalized anxiety. Arch Gen Psychiatry. 2003;60: Karasek R, Theorell T. Healthy work: stress, productivity, and the reconstruction of working life. New York (NY): Basic Books; Podsakoff PM, MacKenzie SB, Jeong-Yeon L, Podsakoff NP. Common method biases in behavioral research: a critical review of the literature and recommended remedies. J Appl Psychol. 2003;88: Goldberg M, Chastang JF, Zins M, Niedhammer I, Leclerc A. Health problems were the strongest predictors of attrition during follow-up of the GAZEL cohort. J Clin Epidemiol. 2006;59: Trifirò G, Barbui C, Spina E, Moretti S, Tari M, Alacqua M, et al. Antidepressant drugs: prevalence, incidence and indication of use in general practice of Southern Italy during the years Pharmacoepidemiol Drug Saf. 2007;16: Gadalla TM. Comparison of users and non-users of mental health services among depressed women: a national study. Women Health. 2008;47: Thielen K, Nygaard E, Andersen I, Rugulies R, Heinesen E, Bech P, et al. Misclassification and the use of registerbased indicators for depression. Acta Psychiatr Scand. 2009;119: Lackie B. The families of origin of social-workers. Clin Soc Work J. 1983;11: Fussell FW, Bonney WC. A comparative study of childhood experiences of psychotherapists and physicists: implications for clinical practice. Psychother Theory Res Pract Train. 1990;27: Nikcevic AV, Kramolisova-Advani J, Spada MM. Early childhood experiences and current emotional distress: what do they tell us about aspiring psychologists? J Psychol. 2007;141: DiCaccavo A. Investigating individuals motivations to become counselling psychologists: the influence of early caretaking roles within the family. Psychol Psychother. 2002;75: Vincent J. Why ever do we do it?: unconscious motivation in choosing social work as a career. J Soc Work Pract. 1996;10:63 9. Received for publication: 14 December 2009 Scand J Work Environ Health 2010, vol 36, no 6 443

104 Person-related work and incident use of antidepressants Appendix. Categorization of type of person-related work Type of work Healthcare Education Social work Customer service Other Job group according to International Standard of Classification of Occupations (ISCO-68) Caring personnel (hospital); caring personnel (retirement homes); caring personnel (homecare); caring personnel (low skilled); doctors and dentists; nurses; physio-/ergotherapists; firefighters; and paramedics. Primary school teachers; secondary school teachers; teachers on educations of medium length; pedagogues (day care); pedagogues (low skilled); and daycare workers. Social workers; pedagogues (24-hour institution); and police and prison personnel. Librarians and museum employees; doctors secretaries; bank assistants; sales assistants; spouses helping (sales); waiting personnel; hairdressers; bus drivers; taxi drivers; mechanics; plumbers; electricians; carpenters; construction workers (skilled); construction workers (unskilled); and truck drivers. Academics (natural sciences); academics (human and social sciences); engineers and architects; IT employees; researchers (university); technicians and constructers; laboratory technicians; media employees; managing clerks; office assistants (private sector); office assistants (public sector); accountants; mail deliverers; warehouse clerks; kitchen staff; cleaning staff; janitors; agricultural workers; farmers; spouses helping (agriculture); forest workers; tool-room workers; metalworkers (low skilled); electronics workers (low skilled); industrial workers (wood); slaughterhouse workers; food workers; wrappers (bottlery workers); warehouse and harbor workers; seamstresses; factory workers (concrete); trainees and students (service, office etc.); trainees and students (industrial, crafts); managers (private sector); managers (public sector); store managers; foremen; and others. 444 Scand J Work Environ Health 2010, vol 36, no 6

105 APPENDIX 3. ARTICLE 3 Madsen IEH, Aust B, Burr H, Carneiro IG, Diderichsen F, Rugulies R. Paid care work and depression: a longitudinal study of antidepressant treatment in female eldercare workers before and after entering their profession. Depression and Anxiety 2012; Online First.

106 Research Article DEPRESSION AND ANXIETY 00:1 9 (2012) PAID CARE WORK AND DEPRESSION: A LONGITUDINAL STUDY OF ANTIDEPRESSANT TREATMENT IN FEMALE ELDERCARE WORKERS BEFORE AND AFTER ENTERING THEIR PROFESSION Ida E. H. Madsen, 1 Birgit Aust, 1 Hermann Burr, 2 Isabella Gomes Carneiro, 1 Finn Diderichsen, 3 and Reiner Rugulies 1,3,4 Background: Previous studies have reported that employees in paid care work (e.g., child, health, and elderly care) have increased rates of hospitalization with depression and treatment with antidepressants. It is unclear, however, whether these findings reflect a causal effect of the work on employee mental health or a selection into these professions. Methods: We examined prevalences of antidepressant purchases during in a cohort of female eldercare workers who entered their profession in 2004 (n = 1,946). These yearly prevalences were compared to those of a representative sample of the female Danish working population (n = 4,201). Trends in antidepressants prevalences were examined using generalized estimation equations. Further, to account for bias by treatment seeking, we compared self-reported depressive symptoms in 2005 measured by the mental health scale from the SF-36. Results: Female eldercare workers had consistently higher prevalence of antidepressant treatment than the general female working population. The eldercare workers were also more likely to suffer from depressive symptoms in 2005 (standardized prevalence ratio = 1.28, 95% CI = ). Prevalences of antidepressant treatment increased during follow up for both cohorts, with similar estimated odds ratios of about 1.15 per year. The trend in the antidepressant prevalences for the eldercare workers was unchanged by entering eldercare work. Conclusions: These findings indicate that female eldercare workers are at increased risk of depression. Further, as the trend in the antidepressant prevalences among the eldercare workers was similar before and after entering their profession, the results suggest that this increased risk is due to selection into the profession. Depression and Anxiety 00:1 9, C 2012 Wiley Periodicals, Inc. Key words: occupational health; mood disorder; cohort study; psychotropic drugs 1 The National Research Centre for the Working Environment, Copenhagen, Denmark 2 Federal Institute of Occupational Safety and Health, Berlin, Germany 3 Department of Public Health, University of Copenhagen, Denmark 4 Department of Psychology, University of Copenhagen, Denmark The authors disclose the following financial relationships within the past 3 years: Contract grant sponsor: Danish Working Environment Research Fund; Contract grant number: Conflict of interests: The funding source had no further role in study design, data collection, analyses, interpretation of data, or the decision to submit the paper for publication Correspondence to: Ida E. H. Madsen, National Research Centre for the Working Environment, Lersø Parkallé 105, DK-2100 Copenhagen, Denmark. [email protected] Received for publication 21 November 2011; Revised 13 January 2012; Accepted 04 February 2012 DOI /da Published online in Wiley Online Library (wileyonlinelibrary.com). C 2012 Wiley Periodicals, Inc.

107 2 Madsen et al. INTRODUCTION Depressive disorders are highly prevalent and are one of the leading causes of disease burden worldwide. [1] Recently, a large US study estimated the 12-month prevalence of major depressive disorder to 3.2%, of which 26.5% had a chronic course (episode duration 2 years or more). [2] The etiology of depression involves an interplay of biological, psychological, and social factors, [3,4] and several studies have shown longitudinal associations between psychosocial work environment factors and depression. [5,6] Correspondingly, the prevalence of depressive disorders varies considerably across occupations [7 9] and studies show that employees in paid care work (e.g. child, health, and elderly care) have increased prevalence of antidepressant treatment [10] and hospitalization with depressive and anxiety disorders. [7,11] This increased treatment prevalence may indicate an increased risk of depression. If such increased risk is due to causal effects of care work associated work environment factors, the effect might be related to the concept of emotional labor. [12] This theory proposes that employment in paid care work can be detrimental to the mental health of the worker, due to an alienation of the worker from their feelings. [12, 13] Also, care workers more often face emotionally demanding situations in their work, [10] for instance when dealing with their clients problems and suffering. [14, 15] Emotionally demanding work has been associated with psychological distress, [16, 17] depressive symptoms, [18] antidepressant treatment, [10] and hospitalization with depression. [19] Alternatively, an increased risk of depression among care workers may not be caused by the work, but explained by selection into these professions. Such selection could occur because individuals from disadvantaged backgrounds may be more likely to enter care work professions. Choosing such employment may be motivated by high levels of childhood parentification (i.e. excessively taking care of other family members needs), [20] a condition which may increase risk of depression. [21] This hypothesis is corroborated by studies among social workers, psychotherapists, and psychology students. [22 24] Further, as care work professions are generally low status, [25] individuals with less available resources due to either socioeconomically disadvantaged backgrounds or existing mental health problems may be selected into these professions. Previous studies [7, 10, 11] have examined the risk of depression in care workers already employed in the field. Hence, the temporal order between care work and depression cannot be established and it is unclear whether the association is explained by selection or may be due to a causal effect of the work. As an indicator of depression, the present paper examines the prevalence of antidepressant treatment in female eldercare workers during 9 years before and 4 years after entering their profession, using a Danish cohort of individuals graduating as social and health care helpers or assistants in These prevalences are compared to those of a representative sample of the female Danish working population to account for general population trends. Because most depressive episodes in the general population are never treated, [26] we also compare the prevalence of self-reported depressive symptoms in the populations. This comparison assesses whether any difference in antidepressant treatment reflects differences in morbidity, or may be due to treatment-related factors such as help-seeking behaviors or physicians prescribing patterns. MATERIALS AND METHODS STUDY DESIGN This paper uses data from three sources: (1) the Danish Health Care Worker Cohort Class of 2004 (DHCWC-2004), (2) the Danish Work Environment Cohort Study (DWECS), and (3) the Danish National Prescription Registry (DNPR). The care worker cohort (CWC) was established from the DHCWC-2004; a three-wave study of individuals graduating as social and health care helpers or assistants in Denmark in The Danish education as social and health care helper takes 14 months and qualifies for working with care of the elderly. [27] The education as social and health care assistant requires a social and health care helper education and takes 20 months. Although social and health care assistants may also work in hospital care, they are primarily trained for the eldercare sector. [27] All 28 Danish schools providing these types of training were invited to participate in the study, and 27 schools accepted. [28] The baseline questionnaire was distributed to 6,365 graduating students in 2004, shortly before their final exams, and had a response rate of 89.5%. [28] The DHCWC-2004 also encompassed followup questionnaires in 2005 and 2006, and the CWC consists of women reporting employment in care work in 2005 (response rate = 65.1% [28] ). As reference population we use DWECS; an ongoing cohort study of the Danish work environment with follow up every 5 years since At each wave, the population is supplemented with age and migration panels to remain representative of the Danish workforce. We use data from 2005 which were collected through questionnaire or telephonic interview (10% random sample) with a response rate of 62.5%. From the representative sample, we included women reporting gainful employment in Details of DWECS are published elsewhere. [29] Data on antidepressant treatment were obtained from the DNPR; a national Danish registry containing all purchases of prescription medications at Danish pharmacies since [30] We included data from the years , and linked survey and registry data using the unique personal identifier assigned to all Danish residents. The study was approved by the Danish Data Protection Agency. [31] POPULATION From the DHCWC-2004, we selected all individuals who reported employment in care work in 2005 (n = 2,699). For the reference population, we selected all individuals gainfully employed (employed, selfemployed, or studying) DWECS participants in 2005 (n = 8,883). We excluded all men (n = 117 for the CWC, n = 4,375 for DWECS) as there were few men in the CWC and gender may modify associations between work environment and antidepressant treatment. [32] We further excluded 307 individuals from DWECS and 636 from the CWC with missing data. The CWC exclusions were primarily due to missing data on depressive symptoms (n = 377). The final study populations consisted of 1,946 participants in the CWC and 4,201 in DWECS. Depression and Anxiety

108 Antidepressant Rates in Eldercare Workers 3 MEASURES Antidepressant treatment. We defined individuals as treated with antidepressants if they during a given year purchased any medication coded N06A according to the Anatomical Therapeutic Chemical classification system developed by the World Health Organization. [33] Depressive symptoms. We assessed depressive symptoms in 2005 by the five-item mental health inventory (MHI-5) of the 36-item short-form (SF-36) health survey. [34] Although the scale was originally developed to measure general mental health, a validation study from 2001 showed that the scale is a valid screening tool for mood disorder. [35] At a cut-off point of 55, the scale had a sensitivity of 0.72 and a specificity of 0.86 for detecting mood disorder according to the fourth edition of the Diagnostic Statistical Manual of Mental Disorders. The five items in MHI-5 were scored from (higher scores indicating better mental health), a mean calculated, and all individuals with a score of 55 or less were categorized as suffering from depressive symptoms. Covariates. We included data on migration and death from the civil registration system [36] for calculating antidepressant prevalences. For descriptive purposes, we also included data on: age, education, cohabitation, and dependent children. Age at inclusion (year 2005) was obtained from the civil registration system. [36] Highest education attained was assessed self reportedly with two items recording highest completed school education and vocational training and categorized as: basic education only (primary school up to 11 years), high school, shorter vocational training (less than 3 years), and longer vocational training (3 years or more). For the care workers, we used education before care worker training. Educational background was used as an indicator of socioeconomic position which is associated with antidepressant treatment. [32] Cohabitation (living with a partner or spouse, yes/no) and dependent children (children living in the household, yes/no) were also self-reported. For the care workers, we further included self-reported data on years of previous work experience and whether previous care work experience was a motivation for undertaking their education (yes/no). STATISTICAL ANALYSIS Prevalence of antidepressant treatment. We calculated yearly prevalences of antidepressant treatment for the CWC and DWECS participants, accounting for migration and death (number of antidepressants treated individuals divided by number of individuals at risk). To account for age differences between the cohorts, we standardized data from DWECS by the age distribution of the CWC (direct standardization) and constructed 95% confidence intervals for the yearly prevalence following Morris and Gardner. [37] The standardization used the age groups 18 29, 30 39, 40 49, 50+ and was conducted in R (version 2.10, Standardized prevalence ratios (SPRs). The cohort prevalences were compared using age-standardized SPRs, comparing the observed prevalence of the outcome in the CWC, to the prevalence expected if the outcome was as likely within age strata as it was observed in DWECS. We calculated the ratios as the observed number of cases divided by the expected number of cases and constructed 95% confidence intervals following Breslow and Day. [38] An SPR greater than 1 indicate higher prevalence among the care workers than in DWECS. All SPRs were calculated using R (version 2.10, Time trends. We assessed time trends in antidepressant treatment as the effect of time on odds of antidepressant treatment using generalized estimation equations (GEE) for binary outcomes with a logit link function, thus accounting for the clustering of data within each individual. The outcome was antidepressant treatment, and we used two time-dependent variables: (1) A variable assessing the overall effect of time and counting years since start of follow up in 1995 (year 0). (2) A dichotomous variable coded 0 in years before 2004 and 1 from 2004 onwards assessed the effect of entering care work (period after 2003). The analysis assessing the effect of period after 2003 included both time variables to control for the overall time trend. All GEE analyses were conducted separately for the two cohorts in SAS 9.2 (SAS Institute, Inc., Cary, North Carolina) and the DWECS analysis was age standardized by weighting. All statistical tests were two sided with a significance level of P <.05. Sensitivity analyses. We conducted five sensitivity analyses. First, we restricted DWECS to participants with shorter vocational training, to compare the CWC to women with similar socioeconomic position. Second, we excluded CWC-participants graduating as health care assistants, as they could have previous exposure to care work, which might bias the time-trend analysis. Third, we excluded care workers reporting that previously working in the field was a motivation to undertake the education, as their previous care work experience might also bias the time-trend analysis. Fourth, we restricted DWECS to data obtained through questionnaire due to potential reporting differences between questionnaires and telephonic interviews which might bias the comparison of depressive symptoms. Fifth, we assessed whether patterns were consistent across age in an age-stratified analysis (age-groups 18 29, 30 39, 40 49, and 50+). RESULTS Table 1 shows the characteristics of the participants. The CWC had a mean age of 36 years. Prior to eldercare training, most CWC-participants (43%) had completed basic education only, whereas 38% had completed shorter, and 6% longer vocational training. Among CWC participants with shorter vocational training, most were studying for social and health care assistants, so the prior vocational training may in many cases have been the social and health care helper education (data not shown in tables). The mean age of DWECS was 42, and when age standardized, 7% of DWECS participants had completed basic education only, 41% had completed shorter and 46% longer vocational training. Figure 1 shows the prevalence of antidepressants throughout the 14-year period. The prevalence for the care workers was consistently greater than that of DWECS, and the prevalence increased over time for both cohorts. There was no indication, though, of an effect of the year 2004, when the care workers entered their profession. Figure 2 presents the SPRs for antidepressant treatment in the CWC. The ratios are consistently greater than 1, indicating that CWC participants were more likely to be treated with antidepressants than DWECS participants. For depressive symptoms, the SPR for the CWC-participants was 1.28 (95% CI = , data not shown in tables), indicating that these individuals were also more likely to suffer from depressive symptoms at the time of survey. The regression analyses (data not shown in tables) showed a statistically significant effect of time on antidepressant prevalence for both cohorts. The effect of time since start of follow up was estimated to odds ratios (OR) of 1.15 per year increase (95% CI: ) for the CWC and 1.16 (95% CI: ) for DWECS. Depression and Anxiety

109 4 Madsen et al. TABLE 1. Characteristics of participants CWC DWECS, age standardised DWECS, observed N (%) Mean (SD) N (%) Mean (SD) N (%) Mean (SD) Total 1,946 4,201 4,201 Age in (10.2) 41 (5.0) 42 (10.6) (28.6) 1,200 (28.6) 507 (12.1) (29.9) 1,254 (29.9) 1,061 (25.3) (28.5) 1,198 (28.5) 1,293 (30.8) (13.1) 548 (13.1) 1,340 (31.9) Educational level a Basic education 828 (42.6) 311 (7.4) 367 (8.7) High school 267 (13.7) 250 (6.0) 133 (3.17) Shorter vocational training (<3 years) 730 (37.5) 1,717 (40.9) 1,812 (43.13) Longer vocational training (3+ years) 121 (6.2) 1,922 (45.8) 1,889 (44.97) Previous work experience, length 0 5 years 477 (24.5) years 300 (15.4) or more years 1,018 (52.3) - - Missing 151 (7.8) - - Living with partner or spouse Yes 1,525 (78.4) 3,205 (76.3) 3,341 (79.5) No 421 (21.6) 995 (23.7) 860 (20.5) Children in the household Yes 1,119 (57.5) 2,241 (53.4) 2,170 (51.6) No 827 (42.5) 1,959 (46.6) 2,031 (48.4) Mental health, score b 79.1 (15.4) 79.7 (6.9) 80.0 (14.0) Depressive symptoms Yes 163 (8.4) 275 (6.6) 254 (6.1) No 1,783 (91.62) 3,925 (93.4) 3,947 (94.0) Note: CWC, Care Worker Cohort; DWECS, Danish Work Environment Cohort Study. a Educational level for the care workers was assessed in 2004, before graduation. b Higher score indicates better mental health. There was no additional effect of period after 2003 among the care workers (OR = 0.99, 95% CI: ), whereas the DWECS prevalence tended towards decelerated growth (OR = 0.88, 95% CI: ). Sensitivity analyses The patterns of antidepressant use were effectively unchanged when: (1) restricting DWECS to individuals with shorter vocational training, (2) excluding CWC participants graduating as eldercare assistants, (3) excluding CWC participants reporting previous experience in the field as a motivation to undertake the eldercare education (data not shown in tables). The SPR for depressive symptoms was slightly decreased when restricting DWECS to questionnaire data (SPR = 1.17, 95% CI: ). The age-stratified analysis revealed that the difference in antidepressant prevalence increased with age (see Fig. 3). DISCUSSION This study finds that the prevalence of antidepressant treatment among Danish female eldercare workers was higher than in the general female working population even before the care workers started doing care work, and that there was no increased growth in the prevalence after the care workers entered their profession. Also, the care workers were more likely to suffer from depressive symptoms in These findings are consistent with a selection of vulnerable individuals into eldercare work, rather than a causal effect of the work on the mental health of the care workers. Antidepressant treatment prevalences increased for both cohorts throughout the period. This trend corresponds to that of the general population, where national figures show increasing treatment prevalence during this period. [39] The antidepressant prevalence in both cohorts were, however, lower than those of the general population. National figures show that 8.1% of Danish women aged were treated with antidepressants in [40] The comparative figures were 4.8% for DWECS and 5.8% for the CWC. These lower prevalences likely reflect healthy worker selection, as the analyzed cohorts consisted of individuals gainfully employed in 2005, and employed individuals have better mental health than the general population. [8] Further, the cohorts included only individuals responding to a survey, which may induce further health selection. [41] For the CWC, additional healthy worker selection may have been introduced by including only individuals employed Depression and Anxiety

110 Antidepressant Rates in Eldercare Workers 5 Figure 1. Antidepressant treatment in the Care Worker Cohort (CWC) and Danish Work Environment Cohort Study (DWECS), , antidepressant-treated individuals per 1,000 individuals, individuals treated and nontreated. in care work a year after graduation, as mental distress predicted drop out from care work in the DHCWC [27] We examined the prevalence of antidepressant treatment as an indicator of depressive disorder. Antidepressants are, however, used to treat a number of conditions [39, 42] including depression, anxiety, and pain. Also, a range of factors influence the probability of being treated when suffering from any disorder, and antidepressant treatment when depressed is more likely in individuals who are for instance female and older. [32, 43, 44] Additionally, attitudes toward antidepressants are associated with the probability of treatment, [45] and if individuals who become eldercare workers have more positive attitudes toward pharmaceutical treatment, this may result in treatment differences although there are no underlying differences in morbidity. To assess whether the differences in antidepressant prevalence reflect differences in depression, we compared self-reported data on depressive symptoms. The care workers were also more likely than DWECS participants to suffer from depressive symptoms. This finding corroborates that the differences in antidepressant prevalence reflect differences in depression risk, although the measure of self-reported depressive symptoms does not determine diagnosed depressive disorder. Selection of individuals at increased risk of depression into eldercare work may be related to psychological factors influencing the motivation to enter care work. Parentification is one condition that has been related to motivations for employment in care work, [22 24] though not eldercare specifically. Parentification occurs when children excessively take on roles of caring for other family members to an extent beyond what is developmentally appropriate. [46] This pattern is thought to be caused by unfulfilled parental needs that are directed toward the child, [47] and to possibly affect adult psychological functioning including relational capabilities [47] and risk of depression and anxiety. [21] A second factor possibly selecting individuals into eldercare work is socioeconomic position. As the care work professions are generally low status, [25] individuals from disadvantaged backgrounds and with fewer available resources may enter these professions from necessity rather than choice. This mechanism may be two-fold: First, individuals from socially disadvantaged Depression and Anxiety

111 6 Madsen et al. Figure 2. Age-standardized prevalence ratios for antidepressant treatment in the Care Worker Cohort, backgrounds may have poorer academic achievements and more mental health problems in childhood, and in turn enter low-status occupations in adulthood. [48] The antidepressant prevalence among the care workers in the present analysis were, however, still increased when restricting the reference population to individuals with similar educational level, as an indicator of socioeconomic position. This finding suggests that educational achievement may not be a key explanatory factor for the present results. It should be noted, though, that material living conditions, an important dimension of socioeconomic position, [49] are not directly reflected when using educational level to indicate socioeconomic position. [49] As data on household income were unavailable for the present analyses, we could not assess the importance of material living conditions for the reported relations. Second, individuals with existing mental health conditions may enter low-status professions as a consequence of their disorder. Although evidence for this effect is generally sparse, [48] the age distribution of the care workers and their relatively high levels of past job experience suggest that some may be selected out of other occupations. Our sensitivity analyses revealed that results were similar after excluding individuals graduating as eldercare assistants and individuals reporting previous experience in the field as a motivation to enter the education. These findings may indicate that the results were not substantially biased by previous experience in the eldercare sector. The age-stratified analysis, however, showed that the increased prevalence of antidepressant treatment was most pronounced among care workers older than 30 years at graduation. These findings may indicate, that the observed differences are related to previous worklife experiences in other occupations. Strengths and limitations We used data on antidepressant treatment from a national registry of sales of prescription medications. This data substantially strengthens the study as it allowed a lengthy follow up period (14 years in total), and avoided recall and social desirability bias, which may affect reports of psychotropic treatment. [44] Further, the study applied a representative sample of the Danish workforce Depression and Anxiety

112 Antidepressant Rates in Eldercare Workers 7 Figure 3. Age-stratified antidepressant prevalences in the Care Worker Cohort (CWC) and the Danish Work Environment Cohort Study (DWECS), as a reference population, thus accounting for general trends in antidepressant treatment. Some limitations of the study must, however, be noted. First, using a representative sample of the workforce as reference population may have led to an underestimation of the SPRs, as the examined exposure (employment in paid care work) also occurs within this reference population. Second, we selected participants for the CWC who were employed in care work in As there is substantial drop out from the eldercare sector, [27] some CWC participants may not have stayed within the eldercare sector throughout the remaining followup period. Such drop out may be seen as misclassification of exposure, possibly causing an underestimation of an effect of care work on antidepressant use. However, as drop out from elder care work is related to mental distress, [27] we retained these individuals in the CWC cohort to avoid further healthy worker selection. Consequently, the estimated effect may be interpreted as the average population effect of entering care work. Third, data on antidepressant use were only available until Hence, the participants were only followed 4 years after the elder care workers entered their profession, and this study can only establish the short-term effects of entering care work. Future studies applying a longer follow up are needed to establish whether eldercare work may affect mental health over the long term. Fourth, the self-reported data for the CWC were collected throughout the year 2005, whereas data for DWECS were mainly collected during the winter (November January). As depressive symptoms are Depression and Anxiety

113 8 Madsen et al. more prevalent during winter, [50] the differential timing of data collection may have biased the SPRs towards the null. Fifth, we used direct and indirect age standardization to account for the different age distributions in the two populations. This method may, however, not have been fully effective, as indicated by the age-standardized mean age of DWECS (41 years) which was higher than the mean age of the CWC (36 years). Any residual confounding by age is, however, likely to be a minor concern for the analyses on antidepressant treatment, as this outcome increases with age [32] and bias would be toward an underestimation of the prevalence ratios. Sixth, the analyses include only antidepressants dispensed through pharmacies, as the applied registry of sales of prescription medications does not register antidepressants dispensed through hospitals at the individual level. However, hospitalization for depressive disorder is rare and the misclassification from omitting this data is likely to be a minor concern. Finally, this study examined antidepressant treatment in a group of female eldercare workers. Other types of paid care work include for instance child and health care, and the results of the present study may not be generalizable to these types of care work. Also, it should be underlined that the presented results may not be generalizable to male employees, as gender may modify the examined relations. CONCLUSION This paper found that the prevalence of antidepressant treatment was consistently higher among a cohort of female eldercare workers compared to the general female working population, and that there was no effect of entering care work on the treatment prevalence of the care workers. Also, the care workers had an increased risk of suffering from depressive symptoms in These findings suggest that eldercare workers are a high-risk group for depression, and that this increased risk is due to selection into these professions rather than a causal effect of this work on employee mental health. REFERENCES 1. Lopez AD, Mathers CD, Ezzati M, et al. Global and regional burden of disease and risk factors, 2001: systematic analysis of population health data. Lancet 2006;367: Rubio JM, Markowitz JC, Alegría A, et al. Epidemiology of chronic and nonchronic major depressive disorder: results from the national epidemiologic survey on alcohol and related conditions. Depress Anxiety 2011;28: Harris T. Recent developments in understanding the psychosocial aspects of depression. Brit Med Bull 2001;57: Caspi A, Hariri AR, Holmes A, et al. Genetic sensitivity to the environment: the case of the serotonin transporter gene and its implications for studying complex diseases and traits. Am J Psychiatry 2010;167: Netterstrøm B, Conrad N, Bech P, et al. The relation between work-related psychosocial factors and the development of depression. Epidemiol Rev 2008;30: Bonde JPE. Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med 2008;65: Hannerz H, Tüchsen F, Pedersen BH, et al. Work-relatedness of mood disorders in Denmark. Scand J Work Environ Health 2009;35: Rugulies R, Madsen IE, Nielsen MB, et al. Occupational position and its relation to mental distress in a random sample of Danish residents. Int Arch Occup Environ Health 2010;83: Harlow BL, Cohen LS, Otto MW, et al. Demographic, family, and occupational characteristics associated with major depression: the Harvard study of moods and cycles. Acta Psychiatr Scand 2002;105: Madsen IE, Diderichsen F, Burr H, Rugulies R. Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study. Scand J Work Environ Health 2010;36: Wieclaw J, Agerbo E, Mortensen PB, Bonde JP. Risk of affective and stress related disorders among employees in human service professions. Occup Environ Med 2006;63: Hochschild A. The Managed Heart : Commercialization of Human Feeling. 1st. London: University of California Press; England P. Emerging theories of care work. Annu Rev Sociol 2005;31: Söderfeldt B, Söderfeldt M, Muntaner C, et al. Psychosocial work environment in human service organizations: a conceptual analysis and development of the demand-control model. Soc Sci Med 1996;42: Zapf D. Emotion work and psychological well-being: a review of the literature and some conceptual considerations. Hum Resou Manage Rev 2002;12: Burr H, Albertsen K, Rugulies R, Hannerz H. Do dimensions from the Copenhagen Psychosocial Questionnaire predict vitality and mental health over and above the job strain and effort-reward imbalance models? Scand J Public Health 2010;38: Bültmann U, Kant IJ, Schröer CA, Kasl SV. The relationship between psychosocial work characteristics and fatigue and psychological distress. Int Arch Occup Environ Health 2002;75: Andrea H, Bültmann U, Beurskens AJHM, et al. Anxiety and depression in the working population using the HAD Scale. Soc Psychiatry Psychiatr Epidemiol 2004;39: Wieclaw J, Agerbo E, Mortensen PB, et al. Psychosocial working conditions and the risk of depression and anxiety disorders in the Danish workforce. BMC Public Health 2008;8: Lackie B. The families of origin of social-workers. Clin Soc Work J 1983;11: Jacobvitz DB, Bush NF. Reconstructions of family relationships: parent child alliances, personal distress, and self-esteem. Dev Psychol 1996;32: Fussell FW, Bonney WC. A comparative study of childhood experiences of psychotherapists and physicists: Implications for clinical practice. Psychother: Theory, Research, Practice, Training 1990;27: Nikcevic AV, Kramolisova-Advani J, Spada MM. Early childhood experiences and current emotional distress: what do they tell us about aspiring psychologists? J Psychol 2007;141: Vincent J. Why ever do we do it? Unconscious motivation in choosing social work as a career. J Soc Wor Pract 1996;10: England P, Budig M, Folbre N. Wages of virtue: the relative pay of care work. Soc Probl 2002;49: Wittchen HU, Jacobi F. Size and burden of mental disorders in Europe a critical review and appraisal of 27 studies. Eur Neuropsychopharmacol 2005;15: Depression and Anxiety

114 Antidepressant Rates in Eldercare Workers Giver H, Faber A, Hannerz H, et al. Psychological well-being as a predictor of dropout among recently qualified Danish eldercare workers. Scand J Public Health 2010;38: Carneiro IG, Faber A, Flyvholm M, et al. SOSU er årgang 2004 Det første år [Social and Health Care Workers Year 2004 The First year]. Copenhagen: National Research Centre for the Working Environment; Feveile H, Olsen O, Burr H, Bach E. Danish work environment cohort study 2005: from idea to sampling design. Statistics in Transition 2007;8: Kildemoes HW, Sørensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health 2011;39: Datatilsynet [The Danish Dataprotection Agency]. Introduction to the Danish Data Protection Agency. Datatilsynet [The Danish Dataprotection Agency] Thielen K, Nygaard E, Andersen I, et al. Misclassification and the use of register-based indicators for depression. Acta Psychiatr Scand 2009;119: World Health Organization. The Anatomical Therapeutic Chemical Classification System with Defined Daily Doses (ATC/DDD). World Health Organization; Bjorner JB, Thunedborg K, Kristensen TS, et al. The Danish SF- 36 Health Survey: translation and preliminary validity studies. J Clin Epidemiol 1998;51: Rumpf HJ, Meyer C, Hapke U, John U. Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard. Psychiatry Res 2001;105: Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011;39: Morris JA, Gardner MJ. Calculating confidence-intervals for relative risks (odds ratios) and standardized ratios and rates. BMJ 1988;296: Breslow N, Day N. The Design and Analysis of Cohort Studies. Breslow, NE and Day, NE Lyon, World Health Organization. Statistical Methods in Cancer Research. 39. Lægemiddelstyrelsen [Danish Medicines Agency]. Forbruget af antidepressiva i Danmark i den primære sundhedssektor i perioden [The use of antidepressants in Denmark in the primary healthcare sector during ] Lægemiddelstyrelsen Lægemiddelstyrelsen [Danish Medicines Agency]. Lægemiddelsalget i Danmark [Sales of medicines in Denmark]. Lægemiddelstyrelsen [Danish Medicines Agency] Goldberg M, Chastang JF, Zins M, et al. Health problems were the strongest predictors of attrition during follow-up of the GAZEL cohort. J Clin Epidemiol 2006;59: Trifirò G, Barbui C, Spina E, et al. Antidepressant drugs: prevalence, incidence and indication of use in general practice of Southern Italy during the years Pharmacoepidemiol Drug Saf 2007;16: Boenisch S, Kocalevent RD, Matschinger H, et al. Who receives depression-specific treatment? A secondary data-based analysis of outpatient care received by over 780,000 statutory health-insured individuals diagnosed with depression. Soc Psychiatry Psychiatr Epidemiol 2012;47: Hämäläinen J, Isometsä E, Sihvo S, et al. Treatment of major depressive disorder in the Finnish general population. Depress Anxiety 2009;26: Aromaa E, Tolvanen A, Tuulari J, Wahlbeck K, et al. Personal stigma and use of mental health services among people with depression in a general population in Finland. BMC Psychiatry 2011;11: Hooper LM, Doehler K, Wallace SA, Hannah NJ. The parentification inventory: development, validation, and cross-validation. Am J Family Therapy 2011;39: Mayseless O, Bartholomew K, Henderson A, Trinke S. I was more her Mom than she was mine: Role reversal in a community sample. Family Relations 2004;53: Eaton WW, Muntaner C, Bovasso G, Smith C. Socioeconomic status and depressive syndrome: The role of inter- and intragenerational mobility, government assistance, and work environment. J Health Soc Behav 2001;42: Galobardes B, Shaw M, Lawlor DA, et al. Indicators of socioeconomic position (part 1). J Epidemiol Community Health 2006;60: Magnusson A. An overview of epidemiological studies on seasonal affective disorder. Acta Psychiatr Scand 2000;101: Depression and Anxiety

115 APPENDIX 4. ARTICLE 4 Hanson LLM, Madsen IEH, Westerlund H, Theorell T, Burr H, Rugulies R. Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish gainfully employed. Journal of Affective Disorders 2012; Online first. Madsen 2012 Social Work Environment and Mental Health

116 Journal of Affective Disorders ] (]]]]) ]]] ]]] Contents lists available at SciVerse ScienceDirect Journal of Affective Disorders journal homepage: Research Report Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish gainfully employed Linda L. Magnusson Hanson a,n, Ida E.H. Madsen b, Hugo Westerlund a,c,töres Theorell a, Hermann Burr d, Reiner Rugulies b,e a Research division of epidemiology, Stress Research Institute, Stockholm University, Stockholm, Sweden b Danish National Research Centre for the Working Environment, Copenhagen, Denmark c Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden d Federal Institute for Occupational Safety and Health, Berlin, Germany e Department of Public Health and Department of Psychology, University of Copenhagen, Denmark article info Article history: Received 20 March 2012 Received in revised form 13 June 2012 Accepted 15 August 2012 Keywords: Antidepressants Conflicting demands Depressive disorder Emotional demands Psychosocial work characteristics Work fast abstract Background: Although depression is common, prevalence estimates of antidepressant use among the workforce and undisputed evidence relating psychosocial work characteristics to depression is scarce. This study cross-sectionally assesses the prevalence of antidepressant use among employed in Sweden and Denmark and prospectively examines associations between work characteristics and antidepressant use. Methods: Data on work demands, influence and learning possibilities was collected from two representative samples of employed aged years from Sweden (n¼4351) and Denmark (n¼8064) and linked to purchases of antidepressants through national prescription drug registries. Standardized 12-month prevalences were calculated. Cox regressions on work characteristics and incident use were performed separately and estimates pooled. Results: Employed Swedish residents had higher standardized prevalence than Danish, 6.0% compared to 5.0%. Working fast and conflicting demands were associated with incident use when estimates were pooled, but adjustment for baseline health attenuated these estimates. Emotionally disturbing situations were related to any incident use, and more strongly to use 4179 defined daily dosages/ year, even after adjustment for various covariates. Limitations: Statistics based on national prescription drug registries are influenced by, e.g., treatment seeking behaviours and other reasons for prescription than depression. Selective drop-out may also affect prevalence estimates. Conclusions: The study indicates that use of antidepressants among the workforce is relatively high and that employed Swedish residents had higher prevalence of antidepressant use than Danish. Relationships between work characteristics and antidepressant use were, however, similar with emotional demands showing the strongest association, indicating that particular groups of employees may be at increased risk. & 2012 Elsevier B.V. All rights reserved. 1. Introduction Depressive disorders affect around 6 7% of the adult population in Western countries yearly (Wittchen and Jacobi, 2005; Kessler et al., 2008), and are associated with considerable morbidity and disability (Ustun et al., 2004; Paykel et al., 2005). Workforce participation, despite suffering from depression, is common (Sanderson and Andrews, 2006; Kessler et al., 2008) leading to substantial work performance loss (Birnbaum et al., 2010). Some n Corresponding author at: Research division of epidemiology, Stress Research Institute, Stockholm University, Stockholm SE-10691, Sweden Tel.: þ ; fax: þ address: [email protected] (L.L. Magnusson Hanson). evidence indicates that work stressors such as psychological demands and poor social support at work could increase the risk of depression (Netterstrom et al., 2008; Bonde, 2008). However, clear evidence regarding psychosocial working conditions is lacking and there are few high-quality studies on other measures than the demand-control-support model (Netterstrom et al., 2008). Problems like self-reports of both explanatory and dependent variables complicate causal inference, which may be ameliorated by objective or independent assessments (Bonde, 2008). Antidepressant use from official registers may be a good independent indicator, indicating a depression diagnosis requiring treatment, although antidepressants are underused among people with depressive disorders and prescribed for other diagnoses as well (Demyttenaere et al., 2008b) /$ - see front matter & 2012 Elsevier B.V. All rights reserved. Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

117 2 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] Some recent studies have associated register data on antidepressant use with factors such as socioeconomic position (Andersen et al., 2009), unemployment/job insecurity (Rugulies et al., 2010), workload (overcrowding in hospital wards) (Virtanen et al., 2008b), team climate (Sinokki et al., 2009), and social capital measured at the individual-level (Kouvonen et al., 2008). However, social capital and work climate measured at department level did not predict antidepressant use (Kouvonen et al., 2008; Bonde et al., 2009). In another study though, work climate as measured by constant bullying did predict antidepressant purchases as did mental strenuousness of work (Laaksonen et al., 2012). One study also linked working with clients and patients to antidepressant use which seemed to be partially explained by emotional demands (Madsen et al., 2010). A few studies have examined relations between dimensions of the job demand-control-support model and antidepressant use. Findings have been largely inconsistent (Virtanen et al., 2007; Bonde et al., 2009; D errico et al., 2010; Thielen et al., 2011), with the possible exception of social support (Sinokki et al., 2009, Bonde et al., 2009; Thielen et al., 2011). As most of these studies assessed use of antidepressants prospectively by official register data they avoided the problem of self-reports of both explanatory and dependent variables. However, given the isolated or inconsistent findings on a range of factors and that few of these are based on samples representative of the general working population (Virtanen et al., 2007; Andersen et al., 2009; Sinokki et al., 2009; Rugulies et al., 2010; Madsen et al., 2010; Thielen et al., 2011), more studies are needed to strengthen the evidence on psychosocial work characteristics and its relation to antidepressant treatment. Some studies also assessed prevalence of antidepressant use among employed, but have been limited to public sector employees (Kivimaki et al., 2007; Virtanen et al., 2008a; Bonde et al., 2009). The Nordic countries have national prescription databases on dispensed drugs containing reliable data and offer a unique potential for record linkage and cross-national studies (Furu et al., 2010). Nordic registry statistics showed that sales of antidepressants varied between the countries, and was highest in Iceland, Sweden and Denmark during (Nomesco, 2009). However, few studies so far have used register data from more than one country which could shed more light on the public health burden of antidepressant use, potential cross-country differences in prevalence and associations with psychosocial work characteristics, as well as strengthen the generalizability of the findings. The present study uses national prescription data on antidepressants to assess 12-month prevalence of antidepressant use among employed Swedish and Danish residents, and examine whether exposure to psychosocial work characteristics was prospectively associated with incident use of antidepressants. We hypothesized that, even though prevalences might differ between the two countries, the patterns of associations would be similar. 2. Methods 2.1. Study samples The study is based on representative samples of gainfully employed individuals years of age from Sweden and Denmark participating in the Swedish Longitudinal Occupational Survey of Health (SLOSH) and the Danish Work Environment Cohort Study (DWECS). SLOSH is a follow-up of participants in the Swedish Work Environment Survey (SWES) , in turn based on the Swedish Labor Force Survey (Magnusson Hanson et al., 2008). A stratified random sample of Swedish residents was first contacted, and of those a subsample of employed aged years were invited to SWES (64% response rate). Out of 9154 eligible participants from SWES 2003, 5985 (65%) responded to the first SLOSH follow-up survey during March May Questionnaires were distributed; one for those who worked 30% or more during the past 3 months and one for those working less or not at all. In total, 5141 responded to the former and were regarded gainfully employed, of which 4351 were aged years. DWECS is a cohort study of the working population in Denmark, with follow-up every 5 years. Participants were randomly drawn from Danish residents aged in 1990, and at each wave the population is supplemented with age and migration panels to ensure representativity of the Danish working population (Feveile et al., 2007). This paper uses data from the 2005 wave, collected between October 2005 and May 2006 using either postal or web-based questionnaire (90%) or telephonic interviews (10%, randomly distributed). Of the 19,855 eligible participants, 12,413 (62.5%) responded, of which 8478 had full information on the personal identification number and were gainfully employed. Of these, 8064 participants were aged years. For the analyses of associations with incident antidepressant use, 263 (SLOSH) and 349 (DWECS) individuals with previous purchases, and another 248 (SLOSH) and 356 (DWECS) with considerable baseline depressive symptoms plus 179 (SLOSH) and 941 (DWECS) with missing data on work characteristics and covariates were excluded. This resulted in subsamples of 3661 (SLOSH) plus 6418 (DWECS) respondents. The research was conducted in accordance with the Helsinki Declaration. All individuals gave informed consent to participate in the respective studies by responding to the postal-, web-based questionnaires or telephonic interviews. SLOSH has been approved by the regional ethics board. DWECS has been approved by the Danish Data Protection Agency. Approval from the Danish National Committee on Biomedical Research Ethics is not required for Danish questionnaire- and register-based studies Work characteristics The Swedish and Danish surveys collected similar data on a number of items reflecting job demands, influence and learning possibilities (Table 3). These included working very fast and conflicting demands, part of the demand dimension of the demand-control model (Theorell et al., 1988) and measures of work pace and role conflicts in the Copenhagen Psychosocial Questionnaire (COPSOQ) (Kristensen et al., 2005; Pejtersen et al., 2010), as well as emotionally disturbing situations, part of the emotional demands scale in COPSOQ (Kristensen et al., 2005; Pejtersen et al., 2010). Moreover, learning new things and influence over what to do at work were similarly assessed, part of the control dimension of the demand-control model (Theorell et al., 1988) and measures of possibilities for development and influence in COPSOQ (Kristensen et al., 2005; Pejtersen et al., 2010). The response categories for the surveys were largely similar, although there were minor differences (Table 3). The items were dichotomized at the median creating cohort specific estimates of high/low scores (Table 3) Use of antidepressants Data on antidepressants were retrieved from the Swedish National Prescribed Drug Register from July 2005 (when personal identification numbers started to be registered) until April 2009 and the Danish register of medicinal product statistics from 2005 until December All redeemed prescriptions coded N06a according to the Anatomical therapeutic chemical (ATC) system Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

118 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] 3 with exact dates of purchase were extracted, and defined daily dosages (DDDs) for each purchase were calculated. 12-month prevalence was defined as at least one purchase of antidepressants between 6 months before and 6 months after responding to the surveys. At least one purchase of antidepressants within 960 day from, and no previous purchase within 263 day prior to, responding to the surveys was considered any incident use. The periods used to define incident use were chosen because they were the maximum available for both studies. In addition, we regarded respondents as incident users of antidepressants 4179 DDDs, if they had no previous purchase but had purchased more than 179 DDDs within any given 12-month period during the follow up, to approximate use for a 6 months duration (Thielen et al., 2009) Covariates Covariates included baseline sex, age (either in five-year intervals or four categories: 20 29, 30-39, 40 49, 50 59), survey data on cohabitation (yes/no), full/part time work, and education. Educational level was assessed through five categories in the Swedish questionnaire roughly corresponding to: 9, 10 11, 12, 13 14, 15þ years of education, and the Danish data was categorized analogously. Yearly gross income from work was obtained from Statistics Sweden, and gross total income from Statistics Denmark; divided into cohort-specific quartiles. Both studies assessed self-rated health by one question with somewhat different wordings (Table 2) and number of days of sick leave during the past 12 months. Finally, depressive symptoms were assessed by 6 questions in SLOSH from the (Hopkins) Symptom Checklist 90 (SCL-90) (Magnusson Hanson et al., 2009) and the 5 item Mental Health Inventory (MHI-5) in DWECS (Bjorner et al., 1998). About 5% of the study samples were classified as having considerable depressive symptoms based on previous research (Magnusson Hanson et al., Manuscript) (Strand et al., 2003; Rugulies et al., 2006). No gender stratified analyses was performed as tests of interaction with sex indicated that associations were not significantly dissimilar for men and women Data analysis Prevalences, both crude and directly standardized for sex and age (5 year intervals) with 95% confidence intervals (CIs) were first calculated separately, using an aggregate of the two populations as standard. Direct standardization was applied to calculate the prevalence that would have been observed if the samples had had the same sex and age structure. Prevalences were then compared by a ratio between the standardized prevalences with 95% CIs (Armitage et al., 2002). Next, associations between work characteristics and incident use of antidepressants were examined using Cox proportional hazard regression models to compute hazard ratios (HRs) with 95% CIs. The hazard ratio may be interpreted as an incidence rate ratio an estimate of the incidence rate of entering antidepressant treatment in a group experiencing a high level of a certain work characteristic compared to the group experiencing a low level of the particular work characteristic. Two sets of analyses were conducted, first examining associations with any incident use and second with incident use 4179 DDDs. Follow up time was calculated from survey response to the first date of purchase, date of death or end of follow up, whichever came first. Dates of deaths were retrieved from the Swedish Register of Causes of Deaths and the Danish Civil Registration System. HRs were first calculated separately with and without adjustments for sex, age, cohabitation, full/part time work, education, income, and additionally sickness absence, self-rated health and depressive symptoms. SAS 9 was used to perform the analyses. A summary HR was then computed by a meta-analytic approach combining the cohort specific estimates using a random-effects model in Stata 11. Heterogeneity was estimated using the I 2 statistic, which is independent of the number of studies and represents the proportion of variability in point estimates due to between-study heterogeneity rather than sampling error (Higgins and Thompson, 2002). Sensitivity analyses were done by alternative dichotomizations, when considered reasonable, and modelling the work characteristics as linear predictors. 3. Results Table 1 presents 12-month prevalences of antidepressant use. In total, 273 Swedish employed (6.3%) and 398 Danish employed (4.9%) were classified as prevalent users, with a higher prevalence among women than men. When directly standardized for age and sex, the country estimates were altered to 6.0% and 5.0%, respectively with a standardized prevalence ratio (Sweden/Denmark) of 1.21 (CI¼ ). Descriptive statistics for the final study samples (SLOSH n¼3661 and DWECS n¼6418) are presented in Table 2. The proportion of men and women, single versus married/cohabiting, and full versus part time workers were similar, as were the levels of self-rated health. However, the age and educational distributions differed somewhat, with a higher proportion of older participants and people with fewer years of education in the Swedish sample. Furthermore, short term sickness absence was more common in the Danish sample. In total, 148 (4.0%) Swedish and 273 (4.3%) Danish employed were incident users of antidepressants. Incident use was more common among women, older and those with lower income, poorer self-rated health and more sickness absence (Table 2). Fewer years of education were associated with more incident use among Danish employed, whereas the pattern was opposite in the Swedish sample. Table 1 Prevalence of antidepressant use in the Swedish (n¼4351) and Danish (n¼8064) sample of employed. Prevalent users Crude prevalence Standardized prevalence (95% CI) a Standardized prevalence ratio (95% CI) Total Swedish sample ( ) Danish sample ( ) Swedish vs. Danish sample 1.21 ( ) Men Swedish sample ( ) Danish sample ( ) Swedish vs.danish sample 1.14 ( ) Women Swedish sample ( ) Danish sample ( ) Swedish vs.danish sample 1.24 ( ) a Standardized by age (5 year categories) and sex. Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

119 4 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] Table 2 Descriptive information and any incident use of antidepressants according to covariates in the Swedish (n¼3661) and Danish (n¼6418) sample of employed. Swedish sample Danish sample n (%) n new users (%) n (%) n new users (%) Sex Men Women Age groups Cohabitation Single Married/cohabiting Education 9 years years years years years Yearly gross income Quartile Quartile Quartile Quartile Full time/part time work Full time Part time Self-rated health a Very good Good Neither good nor bad b /Somewhat good c Somewhat poor or very poor b /Poor or very poor c Sickness absence day day day or more a How would you rate your general state of health? (Swedish Longitudinal Occupational Survey of Health) and Overall, how would you rate your health? (Danish Work Environment Cohort Study). b Swedish Longitudinal Occupational Survey of Health. c Danish Work Environment Cohort Study. Table 3 presents distributions of the work characteristics and incident use of antidepressants. A total of 80 respondents (2.2%) were incident users of 4179 DDDs in SLOSH and 122 (1.9%) in DWECS. Incident use was more common among employed with high scores on working fast, conflicting- and emotional demands. Table 4 presents the results from the country-specific Cox regression analyses and the pooled estimates of associations between work characteristics and incident use of antidepressants. Working fast and conflicting demands tended to be associated with antidepressant use. For working fast the pooled estimates, and for conflicting demands the country-specific and pooled HRs, were statistically significant before and after adjustment for age, cohabitation, full/part time work, education, and income, but became insignificant when baseline health was added to the models. Moreover, all models indicated a significant relation between emotionally disturbing situations and antidepressant use, with a fully adjusted pooled HR of 1.45 (CI¼ ). When adjusting for baseline health, the HRs were generally attenuated markedly bydepressive symptoms. Addition of sickness absence and self-rated health only slightly further attenuated the HRs. No association with learning possibilities and influence was, however, observed. The I 2 statistics indicated heterogeneity for adjusted estimates regarding learning possibilities (Model 2) and influence (Model 1 2), which never reached more than 20%. Only the unadjusted estimates for emotional demands showed considerable between study heterogeneity (78%) which was no longer apparent when adjusting for sex. The results were similar when using incident use of 4179 DDDs as outcome (Table 5), and indicated an even stronger association with emotional demands. Sensitivity analyses did not essentially alter results (data not shown). When workplace factors were modelled as linear predictors, the results were also similar to the main results, although most items showed no clear linear trend. The items referring to working fast, conflicting demands and emotional demands appeared to have an approximate J or U shaped association with antidepressants. Having these types of demands never/almost never appeared worse than having them seldom, except for emotional demands in the Danish sample. It should be noted, though, that the groups responding never/ almost never were relatively small. 4. Discussion This study indicated a higher 12-month prevalence of antidepressant use among Swedish than Danish employed. However, relationships between the work characteristics and incident use were similar in the two countries. High emotional demands, in terms of emotionally disturbing situations, were most clearly linked to antidepressant use, but conflicting demands and to some extent working fast also tended to be associated Prevalence of antidepressant use To our knowledge, this is the first study to assess and compare use of antidepressants among employed from non-selected and similar populations in two neighbouring countries. The Nordic registers contain reliable data on all dispensed antidepressants, from all pharmacies except, e.g., nursing homes (Sweden) and inpatient care (Furu et al., 2010). However, a number of factors influence dispensed drug statistics such as underrecognition of depressive disorders, treatment seeking behaviours, differences in treatment needs, and poor compliance (Henriksson et al., 2006; Kivimaki et al., 2007; Demyttenaere et al., 2008a; Thielen et al., 2009). Moreover, about one third of antidepressant prescriptions are for other reasons like anxiety, pain, and sleeping problems Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

120 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] 5 Table 3 Work characteristics, number and percentages with high/low scores as well as corresponding number of incident users of antidepressants (any incident use or incident users of 4179 DDDs) in the Swedish (n¼3661) and Danish (n¼6418) sample of employed. Swedish sample Response options Scores n (%) n new users % n new users of 4179 DDDs (%) Working fast Never/almost never, seldom, or Low sometimes Do you have to work very fast? a Often High Conflicting demands Never/almost never or seldom Low Does your work often involve conflicting demands? a Sometimes or often High Emotional demands Never/almost never or seldom Low Does your work put you in emotionally disturbing situations? b Sometimes or often High Learning possibilities Never/almost never, seldom, or Low sometimes Do you have the possibility to learn new things through your job? a Often High Influence Never/almost never or seldom Low Do you have a choice in deciding what you do at work? a Sometimes or often High Danish sample Working fast Never/hardly ever, seldom or sometimes Low Do you have to work very fast? b Often or always High Conflicting demands Completely incorrect Low How correct or incorrect are the following statements about your Sometimes correct or completely correct High role in your work? Conflicting demands are placed on me in my work Emotional demands Never/hardly ever or seldom Low Does your work put you in emotionally disturbing situations? b Sometimes, often or always High Learning possibilities To a very small or small extent or Low somewhat Do you have the possibility of learning new things through your To a large or very large extent High work? b Influence Never/hardly ever, seldom or sometimes Low Do you have any influence on what you do at work? b Often or always High DDDs¼defined daily dosages. a Theorell, T., Perski, A., Akerstedt. T., et al Changes in job strain in relation to changes in physiological state. A longitudinal study. Scandinavian Journal of Work, Environment and Health. Jun;14(3): b Kristensen, T.S., Hannerz, H., Hogh, A., Borg V., The Copenhagen Psychosocial Questionnaire a tool for the assessment and improvement of the psychosocial work environment. Scandinavian Journal of Work, Environment and Health. Dec;31(6):438 49, and Pejtersen, J.H., Kristensen, T.S., Borg, V., Bjorner, J.B., The Second version of the Copenhagen Psychosocial Questionnaire. Scandinavian Journal of Public Health. 38(Suppl 3):8-24. Table 4 Results of the Cox regressions and random-effect models on the association between work characteristics (high versus low scores) and any incident use of antidepressants in the Swedish (n¼3661) and Danish (n¼6418) sample of employed. Swedish sample Danish sample Pooled result Model a HR 95(%) CI HR 95(%) CI HR 95(%) CI I 2 (95% CI) b Working fast Conflicting demands Emotional demands (3 95) Learning possibilities (NA) Influence (NA) (NA) a Models were fitted separately for the 5 work characteristics: model 0¼crude, model 1¼adjusted for sex, age, cohabitation, full time/part time work, education, income, model 2¼model 1þdepressive symptoms, sickness absence, self-rated health. b Confidence intervals for I 2 are based on the statistical significance of Q (Cochrans Q test), NA¼Not applicable. (Henriksson et al., 2003, 2006; Gardarsdottir et al., 2007), which could not be taken into account. The samples had similar distributions except for age, although educational level was not directly comparable since the educational systems differ considerably. However, methodological inconsistencies cannot be completely ruled out and a higher proportion of people with depression may have dropped out affecting prevalence estimates. Country variations at the time may also explain the difference in prevalence. Sales of antidepressants among the general population in were higher in Sweden than Denmark. Since Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

121 6 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] Table 5 Results of the Cox regressions and random-effect models on the association between work characteristics (high versus low scores) and incident use of 4179 defined daily dosages (DDDs) of antidepressants in the Swedish (n¼3661) and Danish (n¼6418) sample of employed. Swedish sample Danish sample Pooled results Model a HR 95 (%) CI HR 95(%) CI HR 95(%) CI I 2 (95% CI) b Working fast Conflicting demands (NA) Emotional demands (NA) Learning possibilities (NA) (NA) Influence (NA) (NA) (NA) a Models were fitted separately for the 5 work characteristics: model 0¼crude, model 1¼adjusted for sex, age, cohabitation, full time/part time work, education, income, model 2¼model 1þdepressive symptoms, sickness absence, self-rated health. b Confidence intervals for I 2 are based on the statistical significance of Q (Cochrans Q test), NA¼Not applicable. then any country variation might, however, have diminished. Sales seem to have levelled off during recent years and Danish levels approached Swedish levels in 2007 and 2008 (Nomesco, 2009). It is possible that different prescription cultures may have developed. Alternatively, there was a true difference in the prevalence/incidence of depression between the two countries. However, a higher prevalence of antidepressant use does not necessarily indicate that depressive disorders were more common among Swedish employees. Comparative 12-month prevalence estimates of depressive disorders among the working population are scarce. A study among 16 European countries summarized in 2005 did not indicate large cultural/cross-country variations in depressive disorders, though the estimated 12 months prevalence for major depression varied between 3.1 and 10.1% in the different studies. The study from Denmark on a sample of the Danish population reported a prevalence of 3.3% year 2000 (Olsen et al., 2004) whereas the Swedish study reported 4.1% among men and 6.6% among women based on questionnaire responses from a Stockholm county sample (Hällström et al., 2003). These estimates are, however, likely to be uncertain due to lack of standardization in terms of methods, design and instruments (Wittchen and Jacobi, 2005). Findings from other studies on antidepressant use report 12- month prevalences for public sector employees from Finland (5.3%) (Virtanen et al., 2008a) and Denmark (6.5%) (Bonde et al., 2009), within similar range as in the present study though they varied somewhat between the different regions Association with work characteristics We found similar associations between specific work characteristics and use of antidepressants among Swedish and Danish employed. To our knowledge only one other study has investigated associations between work characteristics and depression from cross-national samples. The authors found that effortreward imbalance but not job control was consistently associated with depression in three Eastern European countries (Pikhart et al., 2004). The samples were, however, not nationally representative and the design cross-sectional. We used a prospective design, rendering reverse causality unlikely. Included subjects were free from considerable depressive symptoms at baseline and recent use of antidepressants, making preexisting disease unlikely to influence the perception of the work environment. Adjustment for baseline health and depressive symptoms should additionally reduce the influence of sub-threshold symptoms, but could result in over-adjustment. Estimates robust to this adjustment may point to associations with treatment needs other than depression. Another advantage is independent outcome measures avoiding common method variance. It should, however, be acknowledged that the measure based on DDDs, which was used to approximate 6 months of treatment, rest on the assumption that the average maintenance dose of the drug for its main indication is used. This conveys a greater risk of misclassification, which is probably nondifferential leading to a dilution of risk estimates. The selected work characteristics depended on similarity of items in the two studies. Since only a few items from the original JCQ/DCQ scales were available, we did not investigate the full scales of demand, job control or the job strain hypothesis. Constructing composite scales of the limited items available and especially a job strain measure was expected to result in misclassification (Fransson et al., 2012). Hence, results are not directly comparable to much of the previous literature. The main results were concordant with some previous findings (Virtanen et al., 2007, D errico et al., 2010; Thielen et al., 2011), but discordant with others (Bonde et al., 2009; Thielen et al., 2011). Although non standard measures may be viewed as a limitation, previous work has documented multidimensionality, particularly for the psychological demands scale. Some argue for leaving conflicting demands and working fast out of the scale (Kristensen et al., 2004; Choi et al., 2008) and suggest working fast to be a measure of work pace in contrast to work hours/ overtime (Kristensen et al., 2004) or physical demands (Choi et al., 2008). The available items are thus not necessarily fully representative of the complete demand, and control constructs. Furthermore, single items may in fact represent more precise measures and have different associations to health dimensions (Choi et al., 2008; Eller et al., 2009). Additional demand dimensions such as emotional demands have been found predictive of mental health adding predictive power above the standard models (Burr et al., 2010). Emotional demands have also been related to antidepressant use in a previous study based on DWECS 2000 (Madsen et al., 2010) though contrasting findings have been reported (Thielen et al., 2011). Work pace has previously been found unrelated to antidepressant use (Thielen et al., 2011), but was weakly cross-sectionally associated with psychiatric disorder Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

122 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] 7 (Stansfeld et al., 1995). The latter study also found a strong association with conflicting demands, but no relation with objectively assessed work characteristics. Finally, externally assessed influence on what to do was not associated to depression diagnoses in a previous study (Waldenstrom et al., 2008), which is in line with our findings. Some potential explanations for lack of a clear association with working fast, conflicting demands, influence and learning possibilities include non-linear associations and dichotomization at the median which could represent too crude measures. Another potential limitation is that work characteristics might have changed prior to/after the surveys. A measure involving a mix of short term/long term or present/past working conditions could mask existing associations. The observed associations were independent of age, sex, cohabitation, full time/part time work, socioeconomic indicators and prior health problems. However, possibly other unmeasured covariates could have confounded the observed relationships. Some candidates are personality and private life/lifestyle factors which we did not adjust for as they were measured dissimilarly, and/or considered possible mediators. Exact wording of the questions, response options and response patterns in the two countries may also influence the results. Limited impact of non-random between-study heterogeneity on the results was nevertheless indicated, giving relatively consistent evidence regarding specific work demands and incident use of antidepressants. The fact that this study is based on samples of working populations of two Nordic countries which are principally nonselected with respect to age, sex, geography, or labour market sector strengthens the generalizability of our results Implications of the findings The study highlight that use of antidepressants among the workforce is relatively high. The findings in the present study concerning a difference in prevalence of antidepressant use between Swedish and Danish employed warrant more in-depth studies in which geographical differences (between and within countries) in diagnostic procedures and indications for antidepressive prescription are studied. This could have major implications for pharmacological practice. The results also corroborate previous findings on an association between demands and particularly emotional demands, and mental health problems. This implicates that particular groups of gainfully employed, for example in human service work may be at higher risk for developing mental health problems. Reducing this type of psychosocial working conditions, or identifying measures which could moderate the effect of these stressors, might help to prevent treatment with antidepressants. 5. Conclusion Employed Swedish residents had a higher prevalence of antidepressant use than Danish residents during The relationships between work demands, influence and learning possibilities and incident antidepressant use were similar. The results suggested an association particularly between emotional demands and incident use of antidepressants. Role of funding source The funders had no role in the study design, data collection, and analysis, decision to publish or preparation of the manuscript. Conflict of interest The authors declare that they have no conflict of interest. Acknowledgements This work was supported by the Swedish Council for Working Life and Social Research (FAS) [grant number , ], who also supports research based at Stress Research Institute through the Stockholm Stress Centre [grant number ], and by the Danish Working Environment Research Fund [grant number ]. References Andersen, I., Thielen, K., Nygaard, E., Diderichsen, F., Social inequality in the prevalence of depressive disorders. Journal of Epidemiol Community Health 63, Armitage, P., Berry, G., Matthews, J.N.S., Statistical Methods in Medical Research. Blackwell Science Ltd.. Birnbaum, H.G., Kessler, R.C., Kelley, D., Ben-Hamadi, R., Joish, V.N., Greenberg, P.E., Employer burden of mild, moderate, and severe major depressive disorder: mental health services utilization and costs, and work performance. Depress Anxiety 27, Bjorner, J.B., Thunedborg, K., Kristensen, T.S., Modvig, J., Bech, P., The Danish SF-36 Health Survey: translation and preliminary validity studies. Journal of Clinical Epidemiology 51, Bonde, J.P., Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occupational and Environmental Medicine 65, Bonde, J.P., Munch-Hansen, T., Wieclaw, J., Westergaard-Nielsen, N., Agerbo, E., Psychosocial work environment and antidepressant medication: a prospective cohort study. BMC Public Health 9, 262. Burr, H., Albertsen, K., Rugulies, R., Hannerz, H., Do dimensions from the Copenhagen Psychosocial Questionnaire predict vitality and mental health over and above the job strain and effort-reward imbalance models? Scandinavian Journal of Public Health 38, Choi, B., Kawakami, N., Chang, S., Koh, S., Bjorner, J., Punnett, L., Karasek, R., A cross-national study on the multidimensional characteristics of the five-item psychological demands scale of the job content questionnaire. International Journal of Behavioral Medicine 15, D errico, A., Cardano, M., Landriscina, T., Marinacci, C., Pasian, S., Petrelli, A., Costa, G., Workplace stress and prescription of antidepressant medications: a prospective study on a sample of Italian workers. International Archives of Occupational and Environmental Health. Demyttenaere, K., Adelin, A., Patrick, M., Walthere, D., De, Katrien, Michele, S., B., 2008a. Six-month compliance with antidepressant medication in the treatment of major depressive disorder. International Clinical Psychopharmacology 23, Demyttenaere, K., Bonnewyn, A., Bruffaerts, R., De Girolamo, G., Gasquet, I., Kovess, V., Haro, J.M., Alonso, J., 2008b. Clinical factors influencing the prescription of antidepressants and benzodiazepines: results from the European study of the epidemiology of mental disorders (ESEMeD). Journal of Affective Disorders 110, Eller, N.H., Netterstrom, B., Gyntelberg, F., Kristensen, T.S., Nielsen, F., Steptoe, A., Theorell, T., Work-related psychosocial factors and the development of ischemic heart disease: a systematic review. Cardiology in Review 17, Feveile, H., Olsen, O., Burr, H., Bach, E., Danish work environment cohort study 2005: from idea to sampling design. Statistics in Transition 8, Fransson, E.I., Nyberg, S.T., Heikkila, K., Alfredsson, L., De, Bacquer, Batty, D., Bonenfant, G.D., Casini, S., Clays, A., Goldberg, E., Kittel, M., Koskenvuo, F., Knutsson, M., Leineweber, A., Magnusson Hanson, C., Nordin, L.L., Singh- Manoux, M., Suominen, A., Vahtera, S., Westerholm, J., Westerlund, P., Zins, H., Theorell, M., Kivimaki, M., T., Comparison of alternative versions of the job demand-control scales in 17 European cohort studies: the IPD-Work consortium. BMC Public Health 12, 62. Furu, K., Wettermark, B., Andersen, M., Martikainen, J.E., Almarsdottir, A.B., Sorensen, H.T., The Nordic countries as a cohort for pharmacoepidemiological research. Basic & Clinical Pharmacology & Toxicology 106, Gardarsdottir, H., Heerdink, E.R., Van Dijk, L., Egberts, A.C., Indications for antidepressant drug prescribing in general practice in the Netherlands. Journal of Affective Disorders 98, Henriksson, S., Asplund, R., Boethius, G., Hallstrom, T., Isacsson, G., Infrequent use of antidepressants in depressed individuals (an interview and prescription database study in a defined Swedish population ). European Psychiatry 21, Henriksson, S., Boethius, G., Hakansson, J., Isacsson, G., Indications for and outcome of antidepressant medication in a general population: a prescription database and medical record study, in Jamtland county, Sweden, Acta Psychiatrica Scandinavica 108, Higgins, J.P., Thompson, S.G., Quantifying heterogeneity in a meta-analysis. Statistics in Medicine 21, Hällström, T., Damström Thakker, K., Forsell, Y., Lundberg, I., Tinghög, P., The PART Study. A population based study on mental health in the Stockholm county: study design. Phase I, Kessler, R.C., Merikangas, K.R., Wang, P.S., The prevalence and correlates of workplace depression in the national comorbidity survey replication. Journal of Occupational and Environmental Medicine 50, Kivimaki, M., Gunnell, D., Lawlor, D.A., Davey Smith, G., Pentti, J., Virtanen, M., Elovainio, J., Klaukka, T., Vahtera, J., Social inequalities in antidepressant Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

123 8 L.L. Magnusson Hanson et al. / Journal of Affective Disorders ] (]]]]) ]]] ]]] treatment and mortality: a longitudinal register study. Psychological Medicine 37, Kouvonen, A., Oksanen, T., Vahtera, J., Stafford, M., Wilkinson, R., Schneider, J., Vaananen, A., Virtanen, M., Cox, S.J., Pentti, J., Elovainio, M., Kivimaki, M., Low workplace social capital as a predictor of depression: the finnish public sector study. American Journal of Epidemiology 167, Kristensen, T.S., Bjorner, J., Christensen, K.B., Borg, V., The distinction between work pace and working hours in the measurement of quantitative demands at work. Work & Stress 18, Kristensen, T.S., Hannerz, H., Hogh, A., Borg, V., The Copenhagen Psychosocial Questionnaire a tool for the assessment and improvement of the psychosocial work environment. Scandinavian Journal of Work, Environment & Health 31, Laaksonen, M., Lallukka, T., Lahelma, E., Partonen, T., Working conditions and psychotropic medication: a prospective cohort study. Social Psychiatry and Psychiatric Epidemiology 47, Madsen, I.E.H., Diderichsen, F., Burr, H., Rugulies, R., Person-related work and incident use of antidepressants: relations and mediating factors from the Danish work environment cohort study. Social Psychiatry and Psychiatric Epidemiology 36, Magnusson Hanson, L.L., Theorell, T., Bech, P., Rugulies, R., Burr, H., Hyde, M., Oxenstierna, G., Westerlund, H., Psychosocial working conditions and depressive symptoms among Swedish employees. International archives of occupational and environmental health 82, Magnusson Hanson, L.L., Theorell, T., Oxenstierna, G., Hyde, M., Westerlund, H., Demand, control and social climate as predictors of emotional exhaustion symptoms in working Swedish men and women. Scandinavian Journal of Public Health 36, Netterstrom, B., Conrad, N., Bech, P., Fink, P., Olsen, O., Rugulies, R., Stansfeld, S., The relation between work-related psychosocial factors and the development of depression. Epidemiologic Reviews 30, Nomesco Medicines Consumption in the Nordic Countries Nordic Medico-Statistical Committee (NOMESCO). Olsen, L.R., Mortensen, E.L., Bech, P., Prevalence of major depression and stress indicators in the Danish general population. Acta Psychiatrica Scandinavica 109, Paykel, E.S., Brugha, T., Fryers, T., Size and burden of depressive disorders in Europe. European Neuropsychopharmacology 15, Pejtersen, J.H., Kristensen, T.S., Borg, V., Bjorner, J.B., The second version of the Copenhagen Psychosocial Questionnaire. Scandinavian Journal of Public Health 38, Pikhart, H., Bobak, M., Pajak, A., Malyutina, S., Kubinova, R., Topor, R., Sebakova, H., Nikitin, Y., Marmot, M., Psychosocial factors at work and depression in three countries of Central and Eastern Europe. Social Science & Medicine 58, Rugulies, R., Bültmann, U., Aust, B., Burr, H., Psychosocial work environment and incidence of severe depressive symptoms: prospective findings from a 5-year follow-up of the Danish work environment cohort study. American Journal of Epidemiology 163, Rugulies, R., Thielen, K., Nygaard, E., Diderichsen, F., Job insecurity and the use of antidepressant medication among Danish employees with and without a history of prolonged unemployment: a 3.5-year follow-up study. Journal of Epidemiology & Community Health 64, Sanderson, K., Andrews, G., Common mental disorders in the workforce: recent findings from descriptive and social epidemiology. Canadian Journal of Psychiatry 51, Sinokki, M., Hinkka, K., Ahola, K., Koskinen, S., Klaukka, T., Kivimaki, M., Puukka, P., Lonnqvist, J., Virtanen, M., The association between team climate at work and mental health in the Finnish Health 2000 Study. Occupational and Environmental Medicine 66, Stansfeld, S.A., North, F.M., White, I., Marmot, M.G., Work characteristics and psychiatric disorder in civil servants in London. Journal of Epidemiology & Community Health 49, Strand, B.H., Dalgard, O.S., Tambs, K., Rognerud, M., Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36). Nordic Journal of Psychiatry 57, Theorell, T., Perski, A., Akerstedt, T., Sigala, F., Ahlberg-Hulten, G., Svensson, J., Eneroth, P., Changes in job strain in relation to changes in physiological state. A longitudinal study. Scandinavian Journal of Work Environ Health 14, Thielen,K.,Nygaard,E.,Andersen,I.,Rugulies,R.,Heinesen,E.,Bech,P.,Bultmann,U., Diderichsen, F., Misclassification and the use of register-based indicators for depression. Acta Psychiatrica Scandinavica 119, Thielen, K., Nygaard, E., Rugulies, R., Diderichsen, F., Job stress and the use of antidepressant medicine: a 3.5-year follow-up study among Danish employees. Occupational and Environmental Medicine 68, Ustun, T.B., Ayuso-Mateos, J.L., Chatterji, S., Mathers, C., Murray, C.J., Global burden of depressive disorders in the year British Journal of Psychiatry 184, Virtanen, M., Honkonen, T., Kivimaki, M., Ahola, K., Vahtera, J., Aromaa, A., Lonnqvist, J., Work stress, mental health and antidepressant medication findings from the Health 2000 Study. Journal of Affective Disorders 98, Virtanen, M., Kivimaki, M., Ferrie, J.E., Elovainio, M., Honkonen, T., Pentti, J., Klaukka, T., Vahtera, J., 2008a. Temporary employment and antidepressant medication: a register linkage study. Journal of Psychiatric Research 42, Virtanen, M., Pentti, J., Vahtera, J., Ferrie, J.E., Stansfeld, S.A., Helenius, H., Elovainio, M., Honkonen, T., Terho, K., Oksanen, T., Kivimaki, M., 2008b. Overcrowding in hospital wards as a predictor of antidepressant treatment among hospital staff. American Journal of Psychiatry 165, Waldenstrom, K., Ahlberg, G., Bergman, P., Forsell, Y., Stoetzer, U., Waldenstrom, M., Lundberg, I., Externally assessed psychosocial work characteristics and diagnoses of anxiety and depression. Occupational and Environmental Medicine 65, Wittchen, H.U., Jacobi, F., Size and burden of mental disorders in Europe a critical review and appraisal of 27 studies. European Neuropsychopharmacology 15, Please cite this article as: Magnusson Hanson, L.L., et al., Antidepressant use and associations with psychosocial work characteristics. A comparative study of Swedish and Danish... Journal of Affective Disorders (2012),

124 APPENDIX 5. ARTICLE 5 Madsen IEH, Hanson LLM, Rugulies R, Theorell T, Burr H, Diderichsen F, Westerlund H. Does good leadership buffer effects of high emotional demands at work on risk of antidepressant treatment? A prospective study from two Nordic countries. (Submitted).