Journal of Public Health Medicine ol., No., pp. - Printed in Great Britain Causes of death associated with psychiatric illness P. Prior, C. Hassall and K. W. Cross Abstract Background A prospective cohort analysis of mortality, among entrants to a population-based psychiatric case register, was undertaken to identify specific causes of death responsible for the increased risk of mortality previously reported in this large group of unselected patients. Methods The analysis was based on a study population of cases, aged - years, from Worcester and Kidderminster Health Districts, entering the case register between and and generating a total of patient-years (PYR) of observation. The underlying cause of death was coded to the relevant revision of the International Classification of Diseases (ICD). Numbers of deaths observed in the study population were compared with the number of deaths expected on the basis of mortality rates for England and Wales. Comparisons were made for eight main causes of death, aggregated at Chapter level of the ICD, and categories of psychiatric diagnoses. Two indices of mortality were used for evaluation: relative risk () "observed deaths/expected deaths; and excess mortality rate () = (observed - expected deaths)/pyr. Results s were significantly raised for accidents, including suicides, as anticipated, and for various main causes of death. The increased risk of accidental death was found across the majority of the psychiatric diagnostic groups although the s were low at less than / PYR. Deaths from respiratory disorders gave rise to the highest s after accidental deaths, and were responsible for substantial excess mortality among in-patients and patients with psychotic illnesses (especially dementia). The largest numbers of deaths of both sexes were due to diseases of the circulatory system, with a per cent excess of observed over expected values in the whole series. The excess was due mainly to deaths of in-patients and of patients with psychotic diagnoses. No excess of deaths owing to neoplasms was found for either in-patients or out-patient groups. Conclusions The findings that psychiatric illness is associated with an increased risk of death from 'natural' causes and that the level of risk was related to the severity and to the diagnostic category of the illness have implications for patterns of care and use of resources. Keywords: psychiatric, case register, mortality, causes Introduction The psychiatric case register which formed the basis of this study was described in a previous report. Briefly, the case register was a complete record of every patient using the psychiatric services in the catchment area (Worcester and Kidderminster Health Districts). Data recorded comprised not only the usual personal characteristics (sex, date of birth, area of residence, treatment, etc.) but also details of each contact with the ten psychiatric services available. Hence the data base gave a complete record of all psychiatric patients in a defined geographical area with a known population, and their use of the services. A preliminary study of mortality of all patients using the psychiatric services provided by the two health districts over a ten-year period revealed that deaths among these patients exceeded those expected for the general population by per cent, and that the relative risk was maximal during the first year after registration and was significantly raised in both sexes and in patients of all age groups. This large series of unselected psychiatric patients from defined geographical areas with known populations also provides the opportunity to study mortality by service use and diagnostic group. Thus it was shown that the excess mortality of 'in-patients' (i.e. those patients who experienced at least one episode of psychiatric in-patient care) both in relative and absolute terms exceeded those of 'out-patients' (i.e. Department of Public Health and Epidemiology, The Medical School, University of Birmingham, Edgbaiton, Birmingham B. P. PRIOR, Research Fellow C. HASSALL, Senior Research Fellow K. W. CROSS, Senior Lecturer Address correspondence to Dr P. Prior, Centre for Cancer Epidemiology, Christie Hospital NHS Trust, Kinnaird Road, Withington, Manchester M QL. Oxford University Press
JOURNAL OF PUBLIC HEALTH MEDICINE the remainder). Differences in the excess mortality of patients with different psychiatric diagnoses, and the related different patterns with respect to sex, age and time after registration were illustrated by considering two diagnostic groups of patients: those with dementia and those with schizophrenia and paranoid psychosis. It was clear that an analysis with respect to cause of death was essential to pursue some of thefindingsfrom this preliminary study. For example, it seemed likely that accidental deaths and suicide would prove to be a major factor in the excess of deaths of patients, but the extent of the contributions of this cause and indeed of other causes needs to be established. This paper is therefore concerned with analysis by cause of death of this series of psychiatric patients. Separate analyses have again been made of 'in-patients' and 'outpatients', and results are also given for patients in various diagnostic groups including the two mentioned above. Material and method The Worcester psychiatric case register which formed the basis of this study was described in the previous report. A total of patients aged between and years, registered between January and December, have been followed-up until death or to December, when the analysis was terminated. The names of those who had died were submitted to the Office of Population Censuses and Surveys (OPCS) and death certificates were obtained for per cent. Causes were coded to rubrics of the International Classification of Diseases (ICD) or according to year of death, but those referable to deaths in and were coded back to the rules as operative in, for compatibility with the last quinquennial mortality rate used in the analysis. The numbers of deaths under each Chapter of the ICD that might be expected to occur in a given series were computed from sex- and age-specific 'person-years' (PYR) at risk, and sex-, age- and -specific mortality rates for England and Wales. Quinquennial rates for estimating expected numbers were derived from annual publications of deaths and populations for - (OPCS, Mortality Statistics, Series DH). It should be noted that throughout this paper the term 'accidental' deaths will include deaths attributable to all rubrics within Chapter of the ICD, i.e. accidents, poisonings and violence. Two summary indices were used to assess the level of mortality in the series: a relative risk () defined as a standardized mortality ratio, obtained from observed/expected numbers; and an excess mortality rate () defined as (observed minus expected/ PYR). The former is a measure of the risk of death in the patient series relative to the risk in the general population of the same sex and age constitution; the latter is taken as a measure of mortality attributable to the psychiatric illness or factors associated with it. The following groups of patients have been investigated: () In-patients ( cases); Out-patients: (i) one-day service ( cases); (ii) + days service ( cases). () Psychotics - ICD -: (i) dementia - ICD ; (ii) affective psychosis - ICD ; (iii) schizophrenia and paranoid psychosis - ICD,. () Alcoholism and drug dependence - ICD, -. () Non-psychotics: (i) depression - ICD., +. (th Revision); (ii) anxiety states - ICD.; (iii) personality disorders - ICD -. The distinction between out-patients who received a service on one day only and the remainder calls for comment. The former comprise a group of patients who experienced only a single contact with one of the psychiatric services: in some cases, the contact was made to assess whether or not the patient was mentally ill (with negative results); in others, a domiciliary visit was made to a physically ill (usually elderly) patient to assess whether he or she had an organic psychosis. This group of patients was therefore very heterogeneous and has been included only for completeness of the analysis. Such patients are not included in the analysis by psychiatric diagnosis. Significance testing for individual s assumed that the observed numbers followed a Poisson distribution. When the expected number of deaths was less than the exact probability of obtaining the observed number or more by chance was computed directly from a Poisson distribution with mean equal to the expected number. For larger expected values the distributions were assumed to approximate the Normal (i.e. with variance equals mean). For both Poisson and normally distributed estimates, two-tailed significance levels (/><, p<\ andp<) have been used. The more stringent two-tailed tests of significance were used to make some allowance for spuriously 'significant' results which may arise when carrying out multiple testing. The per cent confidence intervals ( per cent CI) for s were obtained from Byar's approximation when deaths numbered less than. For more than
PSYCHIATRIC ILLNESS AND RISK OF DEATH deaths the Normal approximation was again invoked but, in this instance, taking the observed number as the mean. Results Total series Considering all male patients (Table ) there was an per cent increase in mortality compared with the general population. Relative risks (s) were significantly raised for patients with causes of death subsumed under s of mental disorders ( = -), nervous diseases (-), diseases of the circulatory system (), respiratory diseases (-), diseases of digestive system (-), and accidental deaths including suicides (-). It is noteworthy that the for malignant neoplasms was less than unity. For all female patients there was a per cent increase in mortality (Table ). Similar s as for males were obtained for mental disorders (-), diseases of the circulatory system () and respiratory diseases (). The s for females for nervous diseases (-), diseases of the digestive system (-) and accidental deaths (-) were substantially lower than for men, and of these three causes only the for accidental deaths was highly significant. However, the s for endocrine, nutritional and metabolic diseases, and diseases of the genito-urinary tract were significant at the per cent level, and that for diseases of the nervous system at the per cent level. The observed and expected deaths from malignant neoplasms were almost identical. The overall excess death rates were and per PYR for males and females, respectively. For each sex the for diseases of the respiratory system was maximal, followed by that for diseases of the circulatory system. 'Accidental' deaths contributed and per PYR to the male and female overall excess rates, respectively. In-patients Table gives corresponding results for patients who had at some time received in-patient care. The overall relative risks were - ( per cent CI --) and ( per cent CI -- ) for males and females, respectively, and the patterns of risk by were essentially the same as for the total series but with increased s and s. The overall male was per PYR with respiratory diseases (), circulatory diseases () and accidental deaths () being the major contributors. Deaths from diseases subsumed under these three Chapters also accounted for the major part of the excess female deaths (overall = ). Out-patients Patients who had not received in-patient care at any time have been divided into those who only used a psychiatric service on one day and the remainder. Results for both groups are given in Table. For male one-day service patients a relative risk of - was obtained, most of the excess being due to TABLE Relative risk and excess mortality rates by sex and ; total series Males (n - ) Females (n - ) % CI % CI, Neoplasms I, Endocrine, nutrition and metabolic, Mental disorders I, Nervous system, Circulatory system I, Respiratory system, Digestive system, Genito-urinary system E, Accidental death - - -'" " # -"* -" - "' -" -"' --- --- --- --- --- --- --- --- --- --- --- - -" -*" -* -"* - -" -"* - - --- --- --- --- -- --- --- -- --- --- --, Observed deaths;, relative risk;, excess mortality rate per PYR;, cause of death not known. ><; "p<; " # p<-.
JOURNAL OF PUBLIC HEALTH MEDICINE TABLE Relative risk and excess mortality rates by sex and ; inpatients Males (,, = ) Females (n = ) I I I E % Cl -* -*" -" - - -" - - -*" - (---) <! -" - * - - - - (---) <-, */><; "p<; p<. accidental deaths ( =, --), respiratory diseases ( = -, ---) and diseases of the circulatory system (=l-, --); together, these contributed to the overall of per PYR. One-day service female patients had a relative risk of - and an of per PYR. Deaths from circulatory diseases and from diseases of the respiratory system gave rise to s of - (--) and - (---), respectively, and largely accounted for the excess of mortality. The maximum occurred for mental disorders (-, ---) but, although highly significant, onlyfivedeaths were observed. Out-patients who received services on more than one day experienced lower relative risks and smaller excess mortality rates than the above group. The overall male was - and significant s were obtained only for respiratory diseases (-, ---), accidental deaths (-, ---), mental disorders (, -- -), and nervous diseases (-, ---). The TABLE Relative risks and excess mortality rates by sex and ; out-patients One-day service + day service Males (rt-) Females (n-) Males (n = ) Females (n = ) I I I E % Cl - - -" - - " - (---) " " - -" - - (---) - - " -' - " - " (---) - - - * -- - - (---) *p<-; "p<; '"p<-.
overall female (-) differed little from the male value, but the s for respiratory disease (-, ---) and accidental deaths (-, ---) were lower. The only other with a significant was diseases of the genito-urinary system (-, ---). Psychoses (patients with more than one day of service use) For all male psychotic patients the overall was - and the excess annual mortality rate was per PYR (Table ). Excess deaths attributable to diseases of the respiratory and circulatory systems largely accounted for the latter rate. In addition to these two Chapters, significant s were obtained for accidental deaths (-, ---) and mental disorders (-, --). Diseases of the digestive and nervous systems were also significant, accounting for deaths but contributing only per PYR to the overall. The same picture emerged for all female psychotic patients, with values of s and s being generally somewhat smaller than for males. Exceptions to this parallel were the significant female for diseases of the genito-urinary tract (-, --) and the non-significance of the for diseases of the digestive system (-). Selected psychotic diagnoses Dementia For male patients the overall was - and significant s were obtained for mental disorders, diseases of the circulatory system and of the respiratory PSYCHIATRIC ILLNESS AND RISK OF DEATH system (Table ). Although the numbers were small, deaths from endocrine, nutritional and metabolic disorders, diseases of the digestive system and of the nervous system also gave rise to significant s. The pattern of risk for female dementia patients was very similar, although the s were somewhat smaller, and those for endocrine, nutritional and metabolic disorders, and for diseases of the digestive system were not significant. The overall male and female excess mortality rates were very high: and per PYR, respectively, largely owing to excess deaths from mental disorders, diseases of the circulatory system and of the respiratory system. Notably, there were comparatively few accidental deaths of patients of either sex with this condition. Affective psychoses The overall of - for males was significantly high (Table ). Nearly half the observed deaths were due to diseases of the circulatory system (=l-, - -). There were only ten accidental deaths, although the (-, ---) was highly significant with the () contributing one-third of the overall value. The male deaths from malignant neoplasms only marginally exceeded the expected number (-). For females, the overall ( -) was also significant; however, it was less than that for males, and the () was about half the male rate. The for accidental deaths (-, ---) was similar to the male value. The for respiratory diseases (-, -) was significant at only the per cent level, whereas the small excess for the circulatory system was not found to be significant. TABLE Relative risk and excess mortality rates by sex and ; all psychoses Males (/-) Females (n-) I I I E % Cl - - - -' -*" -" - - -* - (---) - -" - - -- - (---) *p<; "/><-; "*p<-.
JOURNAL OF PUBLIC HEALTH MEDICINE TABLE Relative risk and excess mortality rates by sex and ; dementia Males (n-) Females (n-) I I I E % Cl " ' -*" -" - -* - - - - (---) - - -' - "* - - -- - - (---) p<; "p<; "*p<. Schizophrenia or paranoid psychosis Male patients had a relative risk of - (nonsignificant) and the eight accidental deaths provided the only significant ( = -, - ) and the only () worthy of mention. On the other hand, female patients had a significant overall (). Deaths from circulatory diseases accounted for half the total deaths, although the (-, --) was significant at only the per cent level. The for respiratory diseases (-, ---) was more significant and the deaths subsumed under these two Chapters together accounted for most of the () for female patients with this psychiatric diagnosis. Miscellaneous psychoses The mortality from all causes was significantly high in males ( = -, --) and the was substantial ( per PYR). E>eaths from diseases of the respiratory ( =, ---) and circulatory (= -, --) systems largely accounted for the overall excess deaths. Although very small numbers were involved, the for nervous disorders ( = -, ---) was also significant. For female patients the overall and were smaller ( = -, ---, = ). Only diseases of the respiratory system ( = -, --, = ) gave rise to a substantial number of deaths. ery small TABLE Relative risks and excess mortality rates by sex and Affective pyschoses Schizophrenia or paranoid psychosis Males (n = ) Females (n = ) Males (/? = ) Females (n = ) O O I E % Cl - - "* - ( --) - -- - - " (---) - - - - - (-- ) - -' -" - - (---) p<-; "p<; p<-.
PSYCHIATRIC ILLNESS AND RISK OF DEATH numbers of deaths attributable to s mental disorders, nervous diseases and diseases of the genito-urinary system gave rise to moderately raised s but the confidence intervals were wide. Alcoholism and drug dependence The overall was significantly raised for males ( = -). The increase was due mainly to diseases of the respiratory system ( = -,---), digestive system ( =, ---) and accidental deaths ( = -, ---). Both the overall () and () for females were higher than for males, with high s for diseases of the circulatory system (-, ---) and of the digestive system ( = -), but the latter referred to only seven deaths. A small excess of respiratory system deaths was also observed. Non-psychotic disorders (patients with more than one day of service use) The overall relative risk for males was significantly raised [=, =, ---, /?<-, = ] and for female patients (O =, =l-, ---, /><, = ). The most significant s were obtained for accidental deaths to males (O =, = -, --, p< ), and to females ( =, = -, ---, p< -); the of per PYR was the same for both sexes. Diseases of the respiratory system was the only other of significance for females (O =, =l-, -, p<, ). Although the for males was similar (O =, = -, --, ), it did not achieve statistical significance. Three specific non-psychotic diagnoses have been considered: anxiety states, depression and personality disorders. Of these, significantly raised s for all causes were obtained only for depression for males (O=, = -,---,/><-, = )and for females (O =, =l-, ---, /><, = ). Accidental deaths were also high in this group for males (O=, = -, --, = ) and for females ( =, = -, -- -, = ). Only accidental deaths were of significance among male patients with a diagnosis of anxiety state ( =, = -, ---, p<, = ) and males with personality disorders (O=, =, ---,/?<-, = ). Discussion Most studies of mortality of psychiatric patients have been confined to selected groups, for example, those with a given psychiatric diagnosis or long-stay inpatients. This study is of a total unselected population of psychiatric patients who, for the most part, made theirfirstcontact with the psychiatric services during an -year period, and for whom complete data of their psychiatric care was available. The previously reported excess of observed deaths over expectation was evident for various causes of death after aggregation at Chapter level of the ICD. It was anticipated that some of the excess mortality of both in-patients and out-patients would be due to accidental death including suicide. This has been confirmed and shown to be true for several psychiatric diagnostic groups. Relative risks were significantly raised for of the sex and diagnostic groups; the exceptions were: dementia (male); schizophrenia and paranoid psychosis (female); the miscellaneous group of psychoses (males and females); alcoholism and drug dependence (females) and personality disorder (females). For these six sex- diagnostic groups the numbers of deaths were very small (< ). Both male and female patients with depression had high values (> ) of relative risk, as did those suffering from other non-psychotic disorders, namely, alcohol and drug dependence and personality disorders (males). Despite the high relative risks, the excess mortality rates for all diagnostic groups here considered were very small, and although there was some variation between diagnostic categories, differences between males and females (the former having higher values in most instances) were the predominant feature of thefindingsfor this measure of mortality. Causes of death subsumed under s I, - and Chapter (males only) also made substantial contributions to the increase of mortality (measured in excess or relative terms or both) of patients in the psychiatric categories here considered. In the total series, the maximum relative risk was that for mental disorders; however, few () deaths were recorded as such and the excess mortality was less than per PYR. The majority () of these deaths were of patients who had received in-patient care; were suffering from dementia. Future studies may disclose even higher rates, particularly for dementia, because of the change of rules for coding causes of death on certificates post, when dementia, appearing in Part of the death certificate (contributory cause), may in some instances be coded as the underlying cause of death. Deaths from respiratory diseases accounted for per cent of all deaths and gave rise to the maximum excess mortality rates of all Chapters in the whole series, and to the highest relative risks after accidental deaths and mental disorders. This was a major group of causes of death especially for in-patients and for
JOURNAL OF PUBLIC HEALTH MEDICINE TABLE Relative risks and excess mortality rates by sex and Miscellaneous ipsychoses Alcoholism and drug dependence Males (/)- ) Females (n ) Males (/?- ) Females (n- ) I IM E % Cl - -" " -" -*" - - - - -*" (---) - * * -" - - -* - - (-- ) - - - - - -'" (---) - - -" -* - - (--) *p<-; "p<; *"p<-o. patients with psychotic illnesses,resultingin substantial excess mortality; this was particularly true for dementia. Patients who suffered from alcoholism and drug dependence also had significantly raised relativerisksof death from respiratory disease. The - diseases of the circulatory system - provided the largest number of deaths of both sexes. Although very highly significant because of the numbers involved, the relative risks for the whole series were rather small (- for both sexes). Both in-patients and out-patients (with a one-day service), had high risks for diseases included in this Chapter, as did all psychotic patients combined; those patients with dementia and the other individual psychotic diagnoses here considered were notable in this respect. Outpatients with more than one day of service did not suffer an excess of circulatory system deaths. The results for deaths from neoplasms call for special mention in that no significant excess over expectation was found for any of the groups here considered. Similar findings have been obtained by others for selected psychiatric populations; for example, by Sims and Prior in respect of severely neurotic patients in the United Kingdom, by Casadebaig and Quemada for psychiatric in-patients in France, and by Mortenssen for schizophrenic in-patients in Denmark. The increased relative risks of psychiatric in-patient deaths from natural causes (i.e. other than accidental deaths) found in our study has also been reported by other workers from national studies in France, Netherlands and Norway; the last two papers were in reference to long-stay in-patient populations which have been excluded from our analysis. We might therefore infer that premature mortality is a general feature of the more severe forms of mental illness. Few studies have been made of psychiatric outpatients and these have usually been based upon small numbers. ' In our study of much larger numbers, there was a pronounced excess of deaths from natural causes over expectations for patients with a 'one-day' service. These were patients for whom thefirstcontact with the psychiatric services (when they were registered) proved to be their only contact during the study period. This often took the form of a psychiatric assessment of a patient with concurrent physical illness. It is not surprising, therefore, that raised mortality should be observed in this group. The reluctance of psychiatrists to admit patients with primary physical problems to psychiatric care, and the concern of consultants in general hospitals to discharge patients as soon as possible after treatment for physical conditions, led to large numbers of assessments. Many of these patients aged years or over died within months of their assessments from circulatory and respiratory conditions. Three-quarters of the deaths of those () outpatients who received more than one day's service were from neoplasms, and circulatory and respiratory diseases. However, only patients who died from respiratory diseases significantly exceeded those expected from the experience of the general population. In fact, only half the number of males died from neoplasms compared with expectation, most of the deficit being due to the relatively small number of
PSYCHIATRIC ILLNESS AND RISK OF DEATH deaths from respiratory cancer; this finding may be due to inconclusive investigation of the patients' physical conditions and probably contributed towards the excess of deaths from respiratory diseases. This study has demonstrated that the excess mortality from natural causes occurs across a broad spectrum of patients with psychiatric disorders with a gradient which parallels the severity of the disorder: from the highest for organic psychotic conditions to the lowest for neurotic conditions which did not require inpatient care. Furthermore, most of this excess mortality was from common conditions - cardiovascular and respiratory diseases - and hence is of considerable importance for public health. Improvement in the health and well-being of the mentally ill is one of the key areas named in Health of the nation. l We have shown that there is a considerable burden of premature mortality in people with mental health problems and, therefore, room for improvement. Strategies for the general population in combating the common causes of death should also be applied to the mentally ill. They may, however, prove more difficult to implement in individuals who, by reason of their condition, are perhaps less accessible or receptive to health education initiatives. The mentally ill may also have increased exposure to known causes of premature death, such as tobacco, alcohol and other drug abuse. Although the patients in our series received good community care, the increased mortality from natural causes, and the implied co-morbidity, suggests that health professionals should be alerted to their need for general medical care. For all these circumstances the reduction in premature mortality among psychiatric patients presents a formidable challenge to the public health services, and our data provide a basis for assessing the efficacy of any public health policies. Acknowledgement Financial support for this study was provided by the Medical Research Council. References Hassall C, Prior P, Cross K. A preliminary study of excess mortality using a psychiatric case register. J Epidemiol Commun tilth ; : -. Rothman KJ, Boice JD, Jr. Epidemiologic analysis with a programmable calculator. Publication -. Washington, DC: National Institutes of Health,. Sims A, Prior P. The pattern of mortality in severe neuroses. Br J Psychiat ; : -. Casadebaig F, Quemada N. Mortality in psychiatric inpatients. Ada Psychiat Scand ; : -. Mortenssen PB. The incidence of cancer in schizophrenic patients. J Epidemiol Commun Hlth ; : -. Brook OM. Mortality in the long-stay population of Dutch mental hospitals. Ada Psychiat Scand ; : -. Sangstad LF, Odegard O. Mortality in psychiatric hospitals in Norway. Ada Psychiat Scand ; : -. Martin RL, Cloninger CR, Guze SB, Clayton PJ. Mortality in a follow-up of psychiatric out-patients. Causespecific mortality. Arch Gen Psychiat ; : -. Rorsman B. Mortality amongst psychiatric patients. Ada Psychiat Scand ; : -. Health of the nation. A strategy for health in England. London: HMSO,. Accepted on February