Journal of Affective Disorders

Transcription

1 Journal of Affetive Disorders 146 (2013) Contents lists available at SiVerse SieneDiret Journal of Affetive Disorders journal homepage: Researh report Risk fators at medial shool for later severe depression: A 15-year longitudinal, nationwide study (NORDOC) $ Kjersti Støen Grotmol n, Tore Gude, Torbjørn Moum, Per Vaglum, Reidar Tyssen Department of Behavioural Sienes in Mediine, Institute of Basi Medial Sienes, Faulty of Mediine, University of Oslo, PO Box 1111 Blindern, 0317 Oslo, Norway artile info Artile history: Reeived 11 April 2012 Reeived in revised form 27 August 2012 Aepted 27 August 2012 Available online 25 September 2012 Keywords: Depressive symptoms Physiian health Longitudinal study abstrat Bakground: Dotors have an inreased risk of suiide, and depressive symptoms are prominent among young dotors. We lak prospetive studies that identify risk fators to be targeted in medial shools. Methods: From 1993 to 2008, graduating medial students (n¼631) from all four Norwegian universities partiipated in the Longitudinal Study of Norwegian Medial Students and Dotors (NORDOC). After their graduating term (T1), they were followed up 1 (T2), 4 (T3), 10 (T4), and 15 (T5) years later with postal surveys. Severe depressive symptoms were measured by the General Health Questionnaire-28 (T1, T2, T3, T4, and T5) and analyzed by generalized estimating equations. Results: At T1 and T5, 13.7% and 7.2%, respetively, of the dotors reported severe depressive symptoms; a signifiant redution over time (p¼0.001) in both genders (response rates 56 83%). Independent risk fators for future depressive symptoms were: young age (odds ratio [OR] 1.1, 95% onfidene interval [CI] , p¼0.003); high neurotiism (OR 3.4, 95% CI , p¼0.003); high reality weakness (OR 2.3, 95% CI , p¼0.008); and severe depressive symptoms at T1 (OR 3.6, 95% CI , po0.001). Limitations: Seletion bias and onurrent life and work stress may have influened the results. Conlusions: In addition to low age, high neurotiism yielded a threefold inreased risk over the 15-year follow-up, whereas high reality weakness, whih is linked to personality pathology, doubled the risk. These fators are linially relevant for identifiation of students at risk. & 2012 Elsevier B.V. All rights reserved. 1. Introdution Dotors are at inreased risk of suiide ompared with other aademis and health professionals (Shernhammer and Colditz, 2004; Hem et al., 2005), and depressive symptoms are prominent among medial students (Dyrbye et al., 2006) and young dotors (Sen et al., 2010). Therefore, studies that establish the risk fators for depression in dotors are required to identify individuals already at risk in medial shool so that they an be offered support and treatment (Center et al., 2003). Individual harateristis suh as personality traits, in partiular neurotiism, may be important (Tyssen and Vaglum, 2002). Furthermore, early individual fators are more generalizable aross ountries in this oupational group than are work-related stress and other onurrent fators, whih may vary more between different settings and ountries. Our researh group has previously identified a personality harateristi related to personality pathology ( reality weakness ) as a risk fator for suiidal ideation (Tyssen et al., $ The Longitudinal Study of Norwegian Medial Students and Dotors. n Corresponding author. Tel.: þ address: k.s.grotmol@medisin.uio.no (K. Støen Grotmol). 2004a). Other possible risk fators for depression are female sex (Sen et al., 2010) and, aording to developmental psyhology (Lerner, 2002) and the depression literature (Ingram and Prie, 2010), a diffiult early environment, suh as poor parental bonding (Grotmol et al., 2010). Researh has speifially pointed to the importane of negative parental experienes for gifted hildren, who may later beome physiians (Miller, 1979; Johnson, 1991). Medial-shool-speifi fators, suh as pereived medial shool stress (Vitaliano et al., 1984) and pereived linial ompetene and skills (Chandavarkar et al., 2007), have also been found to be assoiated positively and negatively, respetively, with depressive symptoms in medial shool. Beause medial shool is highly performane-oriented, we an assume that a high degree of self-pereived ompetene and skills would be partiularly important in avoiding depression. Aording to the injury hypothesis, alohol problems an inrease the risk of depression (Fergusson et al., 2009). All suh fators should be inluded as preditors in a longitudinal study that inludes onsiderable time periods in the areers of established dotors. Only in this way an we study how these fators ontribute to the development of depressive symptoms over time. Based on previous researh, we would expet all the above-mentioned fators to be important in the development of depressive symptoms. However, the long-term and adjusted effets /$ - see front matter & 2012 Elsevier B.V. All rights reserved.

2 K. Støen Grotmol et al. / Journal of Affetive Disorders 146 (2013) of these fators have not yet been inluded in preditor model studies of this oupation or in any other population studies. In the present study, we aimed to identify risk fators that an be measured in medial shool. Therefore, ontextual fators suh as work stress and negative life events were not inluded in the present study, although they are known to influene the risk of depression. This is the first nationwide, long-term prospetive study of dotors to identify possible early risk fators for severe depressive symptoms beyond the very first postgraduate years. We used data from a 15-year longitudinal study of Norwegian graduating medial students (NORDOC; n¼631), assessed five times for depressive symptoms, to obtain the odds ratios for individual and/or medial shool-related fators, measured in the final term of medial shool, that predit future severe depressive symptoms. 2. Methods 2.1. Sample and study design The Young Dotor Cohort of the Longitudinal Study of Norwegian Medial Students and Dotors (NORDOC) onsisted of all students who graduated in 1993 and 1994 from all four medial faulties in Norway (n¼631) (Tyssen et al., 2000). The respondents were surveyed by postal questionnaire at five time points: T1, the final year of medial shool (n¼522/631, 83% of the eligible sample); T2, internship, 1 year after graduation (n¼371/631, 56% of the eligible sample); T3, the fourth postgraduate year (n¼422/631, 67% of the eligible sample); T4, the tenth postgraduate year (n¼390/631, 62% of the eligible sample); and T5, the 15th postgraduate year (n¼330/ 631, 52% of the eligible sample). Two hundred and nineteen students (35% of the eligible sample) responded at all five time points. The mean ages at T1, T2, T3, T4, and T5 were 28 (s.d.¼2.8), 29 (2.8), 31 (2.8), 37 (2.7) and 42 (2.7) years, respetively, with no signifiant sex differenes. The mean observation times for T2, T3, T4 and T5 were 1.2 (s.d.¼0.2), 3.5 (0.3), 9.5 (0.6) and 14.6 (0.6) years after graduation (T1), respetively. An attrition analysis showed no signifiant assoiations between severe depressive symptoms, sex or age and the number of times of survey partiipation. Therefore, our sample of responders was not biased with respet to previous severe depressive symptoms, sex or age Measures Dependent variables Severe depressive symptoms were measured at all five observation points using the seven items that onstitute the severe depression subsale of the General Health Questionnaire-28 (Goldberg and Williams, 1988). This questionnaire has been validated for sreening linially signifiant mental health problems in general population samples (Goldberg and Williams, 1988; Goldberg et al., 1998). Four of the seven items relate to suiidal ideation. The respondent is asked whether the symptoms have been present during the past 2 weeks, and the response ategories range from (1) less than usual to (4) muh more than usual. The values (1) and (2) were sored 0, and (3) and (4) were sored 1. Respondents with a sum sore equal to or greater than 1 were onsidered severely depressed, whereas those soring 0 were onsidered not severely depressed (Vaglum and Falkum, 1999). The internal validity of our severe depressive symptom ut-off value of 1 was tested by alulating the overlapping responses with other linially relevant ategorial variables, viz. mental health treatment required and help-seeking (Tyssen et al., 2000), suiidal planning (Tyssen et al., 2004a) and impat on work apaity. The latter variable was assessed by one item: How often has your mental health made it diffiult to master your work in the past 4 weeks? This item was ategorized from (1) not at all to (5) all of the time but was dihotomized to test for its assoiation with severe depressive symptoms (1 was sored as 0; 2 5 were sored as 1) Preditor variables Age and sex. Age was measured as a ontinuous variable in years (with reversed soring), and in analyzing sex, females were oded 0 and males 1. Personality traits were measured in half the sample in the graduating semester (T1) and in the remaining half of the sample in the year after graduation (T2), as desribed elsewhere (Tyssen et al., 2000). Personality traits were measured using the 36-item version of Torgersen s Basi Charater Inventory (Torgersen, 1980; Torgersen and Alnæs, 1989), whih has been desribed previously (Tyssen et al., 2000). The vulnerability dimension losely resembles the lassi neurotiism sale and will be referred to here as neurotiism. It inludes items suh as I m very touhy about ritiism and It often seems that others do things muh better than me. The intensity sale is a measure of extraversion/introversion, and the ontrol dimension desribes the degree of obsessiveness or onsientiousness. Reality weakness inludes pereptions and ideations on the borderline between reality and fantasy. It measures hroni illusions, paranoid traits and problems with identity-inseurity and relationships, traits that are assoiated with severe personality disorders. Examples of these items are: It is diffiult for me to trust people beause they so often turn against me or leave me in the lurh, Sometimes I feel like I live in a fog and Every now and then, I get strange thoughts in my head that I annot get rid of. Pereived parental bonding was measured in half the sample at T1 and in the remaining half at T2, for the same reasons that applied as to the measurement of personality traits (Parker et al., 1979). Pereived parental bonding was measured with the Parental Bonding Instrument, whih asks respondents to rate the attitudes and behaviours of their parents on two dimensions: are and overprotetion. Our researh group has demonstrated that maternal are has the strongest effet on depressive symptoms among dotors (Grotmol et al., 2010). Therefore, only this dimension was inluded in the analysis. Pereived medial shool stress was measured at T1 with an instrument that has been desribed previously (Vitaliano et al., 1984; Tyssen et al., 2000). This instrument onsists of stressor items suh as: Medial training ontrols my life and leaves too little time for other ativities and Medial shool is more of a threat than a hallenge. Pereived diagnosti skills were measured at T1 with an instrument onstruted by our researh group and desribed in detail elsewhere (Tyssen et al., 2000). This variable, asertained with 16 items, desribes how ertain/unertain the respondent feels about the way to approah a ase history and a physial examination based on some speified medial diagnosis. Pereived reording skills were measured at T1 by six items based on previous patient interviews and medial reords (Tyssen et al., 2000). The items measured the respondent s onfidene in his/her work during the preeding patient interviews and when medial reords were written up during the previous term. Hazardous drinking was measured at T1 with the alohol use disorder identifiation test (AUDIT) (Saunders et al., 1993). One of the ten items (item 2: How many drinks ontaining alohol do you have on a typial day when you are drinking? ) was omitted from the questionnaire based on the rationale that this item may be misunderstood in Norwegian settings. A typial day when you are drinking may be interpreted as a typial day with a party or

3 108 K. Støen Grotmol et al. / Journal of Affetive Disorders 146 (2013) any other day when you are drinking alohol (Gulbrandsen and Aasland, 2002). The present version has previously been used to study hazardous drinking among Norwegian dotors, polie and ambulane offiers (Sterud et al., 2007) Statistial analysis A preparatory analysis revealed that the rates of severe depressive symptoms were relatively unstable. Of the respondents who reported severe depressive symptoms at any observational point, 63.5% (35/56) reported them only one, 23.2% (13/56) reported them twie, 8.9% (5/56) three times, 3.6% (2/56) four times and 1.8% (1/56) reported severe depressive symptoms at all five observational points. Therefore, to apture a reliable measure of the respondents tendeny to experiene severe depressive symptoms throughout the extensive follow-up period, we hose to assess the effets of the T1 variables on severe depressive symptoms at T2 T5 using generalized estimating equations (GEE). GEE is an extension of the generalized linear model that aounts for orrelated repeated ategorial measures within subjets (Diggle et al., 1994). To alulate the variane terms for repeated measures, the GEE approah requires that a working orrelation struture be speified. An unstrutured working orrelation was speified for these analyses. The GEE approah uses all available data for eah subjet and is therefore less affeted by missing data than are other statistial strategies. To alulate easily omprehensible and meaningful risk estimates (odds ratios), an important aim of the study, we dihotomized the ontinuous preditor variables (exept age), despite the potential loss of information entailed by the appliation of this strategy. The preditor variables were divided into groups of high (at median or above) and low (below median) sores. Beause we expeted severe depressive symptoms at T1 to be a strong preditor of future severe depressive symptoms, two preditor models were applied. The first adjusted model inluded all signifiant univariate preditors, exept severe depressive symptoms at T1. Severe depressive symptoms at T1 was inluded in the seond model whih means that the independent variables in this model predit hanges in depressive symptoms from baseline. The interations between the different variables and time were tested to identify possible hanges in the effets of the preditors over time. To identify any potential modifying effets of sex, interations with sex were also tested. Finally, to test the possible modifying effets within pairs of variables, we performed twoway interation analyses between all preditor variables. Differenes in the prevalene of severe depressive symptoms between the five observational points were assessed by GEE. Statistial signifiane was set at the 0.05 level Ethis The study was approved by the Ethial Committee for Medial Researh and the National Data Inspetorate of Norway. Table 1 desribes the independent variables. 3. Results Severe depressive symptoms at T1 were reported by 13.7% of all subjets. There were signifiant redutions (p¼0.001) in severe depressive symptoms from T1 to 7.7% at T2, and to 7.5% at T5. There were no signifiant sex-based differenes in the prevalene of severe depressive symptoms at any of the five observation points. Table 1 Desription of the independent variables with ut-off thresholds. N Range Mean (s.d.) Median (used for ut-off) Age (years) (2.83) Sex % females Neurotiism (2.31) 3 Reality weakness (1.42) 1 Intensity (2.46) 6 Control (2.03) 3 Pereived medial shool stress (6.85) 20 Pereived diagnosti skills (10.24) 86 Pereived reording skills (4.75) 29 Maternal are (6.52) 30 Hazardous drinking (2.60) 3 Our internal validation of the outome measure showed that over the various observation points, 46 87% of the subjets with severe depressive symptoms reported a need for mental health treatment ompared with 8 20% of subjets without symptoms; 13 39% of students with severe depressive symptoms reported suiidal planning ompared with 0 6% of students without symptoms; and 10 13% of students with severe depressive symptoms reported using antidepressant mediations ompared with 0 3% of students without symptoms. At T5, 82% of subjets with severe depressive symptoms reported that their mental health ondition had a negative effet on their work ompared with 19% of subjets without symptoms (all w 2 : po0.001) Univariate preditors of severe depressive symptoms Table 2 shows that the following variables were signifiant univariate preditors of severe depressive symptoms: young age (OR 1.10 [ ], p¼0.01), high neurotiism (OR 6.88 [ ], po0.001), high reality weakness (OR 4.56 [ ], po0.001), low maternal are (OR 2.64 [ ], po0.001), high pereived medial shool stress (OR 2.4 [ ], p¼0.001), low pereived diagnosti skills (OR 2.15 [ ], p¼0.002), low reording skills (OR 1.92 [ ], p¼0.01) and severe depressive symptoms at T1 (OR 6.69 [ ], po0.001). Hazardous drinking was almost signifiant in the univariate analysis (OR 1.73 [ ], po0.07) Adjusted preditors of severe depressive symptoms The following variables were signifiant adjusted preditors of severe depressive symptoms in the first multiple regression model: young age (OR 1.14 [ ], p¼0.01), high neurotiism (OR 4.56 [ ], p¼0.001), high reality weakness (OR 2.55 [ ], p¼0.008) and low maternal are (OR 1.97 [ ], p¼0.01). In the seond model, in whih we inluded the severe depressive symptom status at T1, the following variables remained signifiant: young age (OR 1.12 [ ], p¼0.003), high neurotiism (OR 3.40 [ ], p¼0.003), high reality weakness (OR 2.28 [ ], p¼0.008) and severe depressive symptoms at T1 (OR 3.62 [ ], po0.001). No interations were identified with either time or sex, so that the other preditor effets were not dependent on sex or the observational point during the follow-up period.

4 K. Støen Grotmol et al. / Journal of Affetive Disorders 146 (2013) Table 2 Risk of future severe depressive symptoms (T2 T5). Univariate Adjusted Model 1 Model 2 OR (95% CI) p value OR (95% CI) p value OR (95% CI) p value Young age 1.10 ( ) ( ) ( ) Male sex 1.05 ( ) ( ) ( ) 0.5 Neurotiism 6.88 ( ) ( ) ( ) Reality weakness 4.56 ( ) ( ) ( ) Intensity 0.92 ( ) 0.74 a a a a Control 1.16 ( ) 0.56 a a a a Pereived medial shool stress 2.4 ( ) ( ) 0.89 Low pereived diagnosti skills 2.15 ( ) ( ) 0.06 Low reording skills 1.92 ( ) ( ) 0.95 Low maternal are 2.64 ( ) ( ) ( ) 0.13 Hazardous drinking 1.73 ( ) 0.07 a a a a Severe depressive symptoms at T ( ) b b 3.62 ( ) OR¼Odds ratio. CI¼Confidene interval. a Variable not signifiant in the univariate analysis, therefore not inluded in the adjusted analysis. b Variable not inluded in the first model. Variable not signifiant in the first model, so was not inluded in the seond model. 4. Disussion 4.1. Main findings The prevalene of severe depressive symptoms dereased from the graduating semester in medial shool to the first postgraduate year and then stayed at about the same level until the 15th postgraduate year. There were no differenes in the prevalene of severe depressive symptoms aording to sex. However, subjets with severe depressive symptoms or high sores for neurotiism or reality weakness in medial shool ran a two- to fourfold inreased risk of future severe depressive symptoms. Further, younger medial students were at greater risk of futureseveredepressivesymptoms. This is the first nationwide, long-term prospetive study of dotors to estimate the early risk fators for severe depressive symptoms beyond the very first postgraduate years. We observed a fourfold inreased risk of future severe depressive symptoms among those who experiened these symptoms in medial shool, whih undersores the importane of early intervention among medial students to prevent future depressive episodes. It is also in keeping with a previous study that demonstrated that undergraduate depression is a long-term preditor of depression in male dotors (Firth-Cozens, 1998). However, to the best of our knowledge, the present study is the first to show this long-term effet among female dotors. Depression is not only devastating for the affeted dotor, but an also interfere negatively with his/her patient are (Fahrenkopf et al., 2008). However, partly beause of the high stigmatization of depressive illness, medial students and dotors are less likely than the general population to seek and reeive appropriate treatment (Bramness et al., 1991; Töyry et al., 2000). Therefore, it must be ensured that they reeive early and proper treatment. Beause rates of suiidal ideation are elevated in this group (Dyrbye et al., 2008), some dotors may also benefit from antidepressant mediation. From our linial experiene, we have reason to believe that bipolar disorder type II is relatively prevalent in this oupational group (Gabbard and Myers, 2008). The relatively low stability of the depression measure observed in our sample indiates flutuations in mood levels over the years, as seen in bipolar disorder, but this onjeture requires further study. Young age predited severe depressive symptoms. The risk of depressive symptoms inreased by 10% for eah year younger the respondents were when they graduated from medial shool. This is an important finding, and should diret our attention to the youngest students, who have less experiene in both work and relationships and who are exposed to stressful medial eduation and a subsequent stressful areer. Until reently, one ould apply for admission to Norwegian medial shools after earning extra points through work experiene or by improving high shool grades. Today, students very often enter medial shool diretly after high shool (aged 19 years), so they are younger than the present study sample and may be even more vulnerable to severe depressive symptoms. These onsiderations may also apply to other ountries that enroll quite young students. Theories of psyhologial development suggest that identity formation is still ongoing in the early twenties (Lerner, 2002), and our findings suggest that some young adults enounter the hallenges of medial shool somewhat prematurely. For instane, demanding and time-onsuming medial shool studies may negatively affet their personal life by impeding the development of soial skills and soial support (Kjeldstadli et al., 2006). These findings suggest that the admission age to medial shool should be reonsidered. Would it be wise to delay entrane to the rather overwhelming medial urriulum for a ouple of years? We found that reality weakness was an independent risk fator for subsequent severe depressive symptoms. To some extent, this is onsistent with the finding that personality psyhopathology inreases the risk of suiidal ideation (Brezo et al., 2006). As reality weakness also predits a lak of help-seeking, independent of the level of emotional distress (Tyssen et al., 2004b), the medial faulty and olleagues of dotors in trouble should be aware that this trait might be a marker of severe future problems and even serious suiidal ideation (Tyssen et al., 2004a). Whether personality traits should be used as seletion riteria might also be an issue worth onsidering. Previously, a NORDOC study onluded that interpersonal or therapeuti support is probably more important than seletion based on normal personality traits (Tyssen et al., 2007). Nevertheless, appliants with high sores for reality weakness, indiating pathology, may benefit from areer advie that prompts them towards an option other than the medial profession. However, this issue requires further investigation. Contrary to studies of other populations, whih have laimed that alohol disorders inrease the risk of depression (Fergusson

5 110 K. Støen Grotmol et al. / Journal of Affetive Disorders 146 (2013) et al., 2009), we found no signifiant preditive effet of hazardous drinking in its own right. One explanation of this may be our deision to dihotomize the hazardous drinking variable at the median value, hene we may not have aurately measured hazardous alohol onsumption. The finding that high neurotiism predits depression onurs with previous (Tyssen and Vaglum, 2002) and reent (Sen et al., 2010) studies of interns. The effet of neurotiism was only slightly affeted by severe depressive symptoms in the graduating semester, illustrating that neurotiism and severe depressive symptoms are distint onepts. Therefore, self-ritiism and neurotiism should be targeted in psyhotherapy (Jorm, 2000). For preventive purposes, the risks assoiated with these traits should also be taught to students as part of their urriulum. Parental bonding was the only variable that lost its independent effet when depressive symptoms at T1 was inluded in the final analysis. This may be beause the effet was absorbed by depressive symptoms at T1, so that the effet of suboptimal bonding exerts its effet on future depressive symptoms through depressive symptoms at baseline. There are disrepanies between previous findings among medial students and young dotors with regard to genderbased differenes: some studies have reported no differenes, whereas the majority have found more depression among women (Tyssen and Vaglum., 2002; Grotmol et al., 2010). However, a robust finding in the general population that we failed to repliate was the gender-based differene in disfavour of the women (Kringlen et al., 2001). The lak of gender-based differenes in the present study may be attributable to: (1) the seletion of female students with good mental health; (2) the seletion of male students with poor mental health; or (3) both. To some extent, the findings of the present study onur with those of longitudinal general population studies of depression. Neurotiism and (earlier) depressive symptoms are also known to be potent risk fators for future depression in the general population (Kendler et al., 2006; Mattisson et al., 2009). Furthermore, studies of general population samples have reported that depressive symptoms tend to derease from early adulthood to mid-life (Jorm, 2000), and the deline in the prevalene of severe depressive symptoms observed in the present study is onsistent with this Strengths and limitations The strengths of this study inlude the nationwide sample, the longitudinal design and the appliation of a widely used measure to apture severe depressive symptoms, whih has also been onfirmed to be of linial importane. Beause the rates of severe depressive symptoms were unstable over time, we attempted to establish a reliable measure of severe depressive symptoms over time by assessing the population on five oasions. The possible effets of missing data were also onsiderably redued by using a GEE approah. We are aware that ontextual fators, suh as work stress and negative life events, an influene depressive symptoms, and further longitudinal researh should larify suh o-ourrent fators. There are obvious limitations assoiated with the use of self-report measures, but these were ounterated, to some extent, by ontrolling for neurotiism in our preditor model, beause this trait resembles negative affetivity (Depue and Monroe, 1986). However, medial students may also be relutant to express their vulnerability (Chew-Graham et al., 2003), and the effets doumented in our results ould be underestimated. An attrition analysis found no greater prevalene of severe depressive symptoms in the graduating semester among later responders, whih tends to validate our findings. Although some researhers have found that people with mental health problems partiipate in longitudinal studies to a lesser extent than healthy subjets (Knudsen et al., 2010), attrition has a stronger influene on prevalene estimates than on estimates of the assoiations between variables. Regarding generalizability, physiians working onditions vary greatly aross the world. However, as mentioned in the Introdution, beause the present study fouses on individual fators, our findings are presumed to be generalizable beyond Norway. Finally, it must also be noted that we aimed to identify depressive symptoms, whih do not neessarily orrespond to a diagnosis of a depressive disorder. Severe depressive symptoms in medial shool are a risk fator for future severe depressive symptoms, as are young age and partiular personality traits. Neurotiism and reality weakness an be treated with psyhotherapy. Finally, medial students who suffer from severe depressive symptoms should be identified and offered the appropriate treatment, perhaps with extra attention paid to the youngest students. Role of funding soure Funded by the Norwegian Researh Counil (NRC) Projet no: /H10 and running expenses support from the Norwegian Medial Assoiation (NMA). Conflit of interest No onflits of interest are reported. The study was funded by the Norwegian Researh Counil and the Norwegian Medial Assoiation. The funding soures played no role in any aspet of the study. No pharmaeutial ompany or other ageny paid to have this artile written. The orresponding author, Grotmol, had full aess to all data in the study and had the final responsibility for the deision to submit the manusript for publiation. Aknowledgements The authors thank Magne Thoresen (PhD), Department of Biostatistis, Institute of Basi Medial Sienes, Faulty of Mediine, University of Oslo, for statistial advie. Referenes Bramness, J.G., Fixdal, T.C., Vaglum, P., Effet of medial shool stress on the mental health of medial students in early and late linial urriulum. Ata Psyhiatria Sandinavia 84, Brezo, J., Paris, J., Tremblay, R., Vitaro, F., Zoolillo, M., Hébert, M., Tureki, G., Personality traits as orrelates of suiide attempts and suiidal ideation in young adults. Psyhologial Mediine 36, Center, C., Davis, M., Detre, T., Ford, D.E., Hansbrough, W., Hendin, H., Laszlo, J., Litts, D.A., Mann, J., Mansky, P.A., Mihels, R., Miles, S.H., Proujansky, R., Reynolds III, C.F., Silverman, M.M., Confronting depression and suiide in physiians: a onsensus statement. Journal of the Amerian Medial Assoiation 289, Chandavarkar, U., Azzam, A., Mathews, C.A., Anxiety symptoms and pereived performane in medial students. Depression and Anxiety 24, Chew-Graham, C.A., Rogers, A., Yassin, N., I wouldn t want it on my CV or their reords: medial students experienes of help-seeking for mental health problems. Medial Eduation 37, Depue, R.A., Monroe, S.M., Coneptualization and measurement of human disorder in life stress researh: the problem of hroni disturbane. Psyhologial Bulletin 99, Diggle, P.J., Liang, K.Y., Zeger, Z.L., Analysis of Longitudinal Data. Clarendon Press, Oxford, England. Dyrbye, L.N., Thomas, M.R., Shanafelt, T.D., Systemati review of depression, anxiety, and other indiators of psyhologial distress among U.S. and Canadian medial students. Aademi Mediine 81, Dyrbye, L.N., Thomas, M.R., Massie, F.S., Power, D.V., Eaker, A., Harper, W., Durning, S., Moutier, C., Szydlo, D.W., Novotny, P.J., Sloan, J.A., Shanafelt, T.D., Burnout and suiidal ideation among U.S. medial students. Annals of Internal Mediine 149, Fahrenkopf, A.M., Setish, T.C., Barger, L.K., Sharek, P.J., Lewin, D., Chiang, V.W., Edwards, S., Wiedermann, B.L., Landrigan, C.P., Rates of mediation errors among depressed and burnt out residents: prospetive ohort study. British Medial Journal 336, Fergusson, D.M., Boden, J.M., Horwood, L.J., Tests of ausal links between alohol abuse or dependene and major depression. Arhives of General Psyhiatry 66, Firth-Cozens, J., Individual and organizational preditors of depression in general pratitioners. British Journal of General Pratie 48,

6 K. Støen Grotmol et al. / Journal of Affetive Disorders 146 (2013) Gabbard, G.O., Myers, M.F., The Physiian as a Patient: A Clinial Handbook for Mental Health Professionals. Amerian Psyhiatri Publishing In., Washington, DC. Goldberg, D., Williams, P., A User s Guide to the General Health Questionnaire. NFER-Nelson, Windsor. Goldberg, D., Privett, M., Ustun, B., Simon, G., Linden, M., The effets of detetion and treatment on the outome of major depression in primary are: a naturalisti study in 15 ities. British Journal of General Pratie 48, Grotmol, K.S., Ekeberg, Ø., Finset, A., Gude, T., Moum, T., Vaglum, P., Tyssen, R., Parental bonding and self-esteem as preditors of severe depressive symptoms: a 10-year follow-up study of Norwegian physiians. The Journal of Nervous and Mental Disease 198, Gulbrandsen, P., Aasland, O.G., Changes in drinking habits among Norwegian physiians Tidsskrift for Den Norske Legeforening 122, Hem, E., Haldorsen, T., Aasland, O.G., Tyssen, R., Vaglum, P., Ekeberg, O., Suiide among physiians. Amerian Journal of Psyhiatry 162, Ingram, R.E., Prie, J.M., Vulnerability to Psyhopathology: Risk Aross the Lifespan, seond ed. The Guilford Press, New York, London. Johnson, W.D., Predisposition to emotional distress and psyhiatri illness amongst dotors: the role of unonsious and experiential fators. British Journal of Medial Psyhology 64, Jorm, A.F., Does old age redue the risk of anxiety and depression? A review of epidemiologial studies aross the adult life span. Psyhologial Mediine 30, Kendler, K.S., Gatz, M., Gardner, C.O., Pedersen, N.L., Personality and major depression: a Swedish longitudinal, population-based twin study. Arhives of General Psyhiatry 63, Kjeldstadli, K., Tyssen, R., Finset, A., Hem, E., Gude, T., Gronvold, N.T., Ekeberg, O., Vaglum, P., Life satisfation and resiliene in medial shool a six-year longitudinal, nationwide and omparative study. BMC Medial Eduation 6, 48. Knudsen, A.K., Hotopf, M., Skogen, J.C., Overland, S., Mykletun, A., The health status of nonpartiipants in a population-based health study: the Hordaland Health Study. Amerian Journal of Epidemiology 172, Kringlen, E., Torgersen, S., Cramer, V., A Norwegian psyhiatri epidemiologial study. Amerian Journal of Psyhiatry 158, Lerner, R.M., The differential approah. In: Webber, B. (Ed.), Conepts and Theories of Human Development. Lawrene Erlbaum Assoiates In., Mahwah, pp Mattisson, C., Bogren, M., Horstmann, V., Tambs, K., Munk-Jorgensen, P., Nettelbladt, P., Risk fators for depressive disorders in the Lundby ohorta 50 year prospetive linial follow-up. Journal of Affetive Disorders 113, Miller, A., Prisoners of Childhood: the Drama of the Gifted Child and the Searh for the True Self. Basi Books, New York. Parker, G., Tupling, H., Brown, L.B., A parental bonding instrument. British Journal of Medial Psyhology 52, Saunders, J.B., Aasland, O.G., Babor, T.F., de la Fuente, J.R., Grant, M., Development of the alohol use disorders identifiation test (AUDIT): WHO ollaborative projet on early detetion of persons with harmful alohol onsumption II. Addition 88, Shernhammer, E.S., Colditz, G.A., Suiide rates among physiians: a quantitative and gender assessment (meta-analysis). Amerian Journal of Psyhiatry 161, Sen, S., Kranzler, H.R., Krystal, J.H., Speller, H., Chan, G., Gelernter, J., Guille, C., A prospetive ohort study investigating fators assoiated with depression during medial internship. Arhives of General Psyhiatry 67, Sterud, T., Hem, E., Ekeberg, O., Lau, B., Oupational stress and alohol use: a study of two nationwide samples of operational polie and ambulane, personnel in Norway. Journal of Studies on Alohol and Drugs 68, Torgersen, S., Hereditary environmental differentiation of general neuroti, obsessive, and impulsive hysterial personality traits. Ata Genetiae Mediae et Gemellologiae (Roma) 29, Torgersen, S., Alnæs, R., Loalizing DSM-III personality disorders in a threedimensional strutural spae. Journal of Personality Disorders 3, Töyry, S., Räsänen, K., Kujala, S., Aärimaa, M., Juntunen, J., Kalimo, R., Luhtala, R., Mäkelä, P., Myllymäki, K., Seuri, M., Husman, K., Self-reported health, illness, and self-are among Finnish physiians: a national survey. Arhives of Family Mediine 9, Tyssen, R., Vaglum, P., Gronvold, N.T., Ekeberg, O., The impat of job stress and working onditions on mental health problems among junior house offiers. A nationwide Norwegian prospetive ohort study. Medial Eduation 34, Tyssen, R., Vaglum, P., Mental health problems among young dotors: an updated review of prospetive studies. Harvard Review of Psyhiatry 10, Tyssen, R., Hem, E., Vaglum, P., Gronvold, N.T., Ekeberg, O., 2004a. The proess of suiidal planning among medial dotors: preditors in a longitudinal Norwegian sample. Journal of Affetive Disorders 80, Tyssen, R., Rovik, J.O., Vaglum, P., Gronvold, N.T., Ekeberg, O., 2004b. Help-seeking for mental health problems among young physiians: is it the most ill that seeks help? A longitudinal and nationwide study. Soial Psyhiatry and Psyhiatri Epidemiology 39, Tyssen, R., Dolatowski, F.C., Rovik, J.O., Thorkildsen, R.F., Ekeberg, O., Hem, E., Gude, T., Grønvold, N.T., Vaglum, P., Personality traits and types predit medial shool stress: a six-year longitudinal and nationwide study. Medial Eduation 41, Vaglum, P., Falkum, E., Self-ritiism, dependeny and depressive symptoms in a nationwide sample of Norwegian physiians. Journal of Affetive Disorders 52, Vitaliano, P.P., Russo, J., Carr, J.E., Heerwagen, J.H., Medial shool pressures and their relationship to anxiety. The Journal of Nervous and Mental Disease 172,