Excess Mortality in Bipolar and Unipolar Disorder in Sweden

Background  Selected groups of patients with bipolar and unipolar disorder have an increased mortality rate from suicide and natural causes of death. However, there has been no population-based study of mortality of patients followed up from the onset of the illness.

Methods  All patients with a hospital diagnosis of bipolar (n = 15 386) or unipolar (n = 39 182) disorder in Sweden from 1973 to 1995 were identified from the inpatient register and linked with the national cause-of-death register to determine the date and cause of death. Overall and cause-specific standardized mortality ratios (SMRs) and numbers of excess deaths were calculated by 5-year age classes and 5-year calendar periods.

Results  The SMRs for suicide were 15.0 for males and 22.4 for females with bipolar disorder, and 20.9 and 27.0, respectively, for unipolar disorder. For all natural causes of death, SMRs were 1.9 for males and 2.1 for females with bipolar disorder, and 1.5 and 1.6, respectively, for unipolar disorder. For bipolar disorder, most excess deaths were from natural causes, whereas for unipolar disorder, most excess deaths were from unnatural causes. The SMR for suicide was especially high for younger patients during the first years after the first diagnosis. Increasing SMR for suicide during the period of study was found for female patients with unipolar disorder.

Conclusions  This population-based study of patients treated in the hospital documented increased SMRs for suicide in patients with bipolar and unipolar disorder. The SMR for all natural causes of death was also increased, causing about half the excess deaths.

INCREASED mortality is one of the major adverse effects in individuals with mood disorders. In bipolar disorder, many studies found a mortality rate approximately 2 times^1-8 and a suicide mortality rate approximately 10 times that of the general population, with suicide being the leading single cause of excess mortality. In major depression, several studies found an increased total mortality,^8-16 with a suicide mortality about 20 times that of the general population.

To obtain accurate risk estimates of the increased mortality among patients with bipolar and unipolar disorder, large cohorts are needed to identify small differences in mortality ratios for natural causes of death, which nevertheless may represent large numbers of excess deaths. Patients need to be followed up from their first diagnosis, since suicide mortality is especially likely to be higher in the first phases of the illness and also among younger patients. Cohorts should be population based to enable inferences to be made from the findings to large patient groups.

The aim of the present study was to assess mortality, compared with the general population, in 2 large cohorts comprising all patients in Sweden with an inpatient diagnosis of bipolar or unipolar disorder. Patients were identified from the national patient register during 1973 through 1995, and followed up from the first psychiatric hospital diagnosis. Suicide and other unnatural causes of death were analyzed as well as cardiovascular death and other specific natural causes. Standardized mortality ratios (SMRs) and the numbers of excess deaths were calculated, and the effects of sex, age at first diagnosis, and duration of follow-up were assessed.

The Swedish psychiatric inpatient register starts in 1971 and covers all inpatient treatments since 1973. For each hospitalization, the unique national registration number, date of admission and discharge, and diagnosis are registered. No private inpatient facilities exist in Sweden, so the psychiatric inpatient register is therefore population based. The diagnoses were recorded according to the International Classification of Diseases, Eighth Revision (ICD-8), from 1971 to 1986, and according to the International Classification of Diseases, Ninth Revision (ICD-9), from 1987 to 1995. All diagnoses in the register are made by a consultant in psychiatry at the time of the patient's discharge from the hospital. Several diagnoses can be recorded at each admission, but only the first (main) diagnosis was used in this study. To exclude readmitted cases, admissions in the inpatient register were observed until the end of 1972, and all individuals with a relevant diagnosis before 1973 were excluded. An upper age limit at the first diagnosis was set at 70 years. All patients with a first diagnosis of bipolar disorder or unipolar disorder from January 1, 1973, to December 31, 1995, were identified in the psychiatric inpatient register. Bipolar disorder was assessed from ICD-8 as manic-depressive psychosis manic type (296.10), manic-depressive psychosis circular type (296.30), and reactive excitative psychosis (298.10), and from ICD-9 as unipolar affective psychosis manic form (296 A), bipolar affective psychosis manic phase (296 C), bipolar affective psychosis mixed form (296 E), and reactive excitative psychosis (298 B). Unipolar disorder was defined from ICD-8 as melancholic involution psychosis (296.00), manic-depressive psychosis depressive type (296.20), manic-depressive psychosis alia definita (296.88), manic-depressive psychosis not otherwise specified (296.99), and reactive depressive psychosis (298.00), and from ICD-9 as unipolar affective psychosis melancholic form (296 B), bipolar affective psychosis melancholic phase (296 D), unspecified affective psychosis (296 X), and reactive depressive psychosis (298 A).

All patients were linked to the national cause-of-death register to determine the date and cause of death, using the national registration number. The cause-of-death register is based on the death certificates. Deaths are classified according to the ICD codes, and on each certificate there is one underlying cause of death and a possibility to add contributing causes. For patients who die in the hospital, the death certificate is made by the consultant in charge, and for deaths out of the hospital, the physician in charge of the patient certifies the cause of death. When the cause of death is unclear, an autopsy is performed. In undetermined deaths, there is always a forensic autopsy. If an autopsy is performed, the death certificate is considered preliminary until the information from the autopsy is also included. During 1973-1995, the autopsy frequency for natural causes of death was 86% in persons aged 15 to 49 years, 68% in persons aged 50 to 69 years, and 45% in persons 70 years and older. For unnatural causes of death, the autopsy frequency was 90% in persons aged 15 to 49 years, 88% in persons aged 50 to 69 years, and 53% in persons aged 70 years and older. The cause-of-death register covers more than 99% of all deaths occurring in Sweden (Statistics Sweden [federal agency for national statistics], 1998).

We calculated person-years of follow-up by sex, 5-year age class, and 5-year calendar period, from the date of admission for a first hospitalization with an eventual bipolar or unipolar disorder diagnosis, occurring from January 1, 1973, until December 31, 1995, or a possible death before this date. Thus, inpatient periods from the first and later admissions were included in the person-years count, and deaths could occur both in and out of the hospital. A bipolar diagnosis was considered more severe than a unipolar diagnosis. Patients with an initial unipolar diagnosis who were discharged later with a bipolar diagnosis were, from the date of admission for a hospitalization with a bipolar diagnosis, excluded from the unipolar group and included among the bipolar patients. Mortality rates for the Swedish population from 1973 through 1995 retrieved from the cause-of-death register (Statistics Sweden, 1998) were used to estimate the expected number of deaths by 5-year age class and 5-year calendar periods. The SMRs were calculated as the observed number of deaths divided by the expected number of deaths, with 95% confidence intervals.¹⁷ We calculated the SMRs for each ICD class causes of death and for natural (ICD classes I-XVI) and unnatural (ICD class XVII; suicide, accidents, homicide, undetermined) causes, in males and females separately. The number of excess deaths for natural and unnatural causes was calculated by subtracting the expected number of deaths from the observed number of deaths.

The SMRs by age at admission and time of follow-up were calculated using Poisson regression models,¹⁸ controlling for calendar time of the first admission. Age at admission and calendar time were divided into 5-year intervals, except for the first calendar interval, which was 3 years. The SMR for suicide was calculated according to age at admission in the following age groups: younger than 30 years, 30 to 44 years, 45 to 64 years, and 65 years and older; time of follow-up was divided into less than 1 year, 1 to 3 years, more than 3 years to 5 years, and more than 5 years. The relative risks for different methods of suicide among patients with bipolar and unipolar disorder compared with methods of suicide in the general population in 1973 to 1995 were calculated. Time trends in suicide mortality for the first 3 years of follow-up were calculated according to 5-year calendar intervals from 1976 to 1995, for bipolar and unipolar disorder groups.

A total of 6578 males and 8808 females in Sweden with a bipolar disorder diagnosis during 1973 to 1995 were included in the study (Table 1). For unipolar disorder, a total of 15 829 males and 23 353 females were included. The number of deaths for patients with a bipolar disorder was 1716 for males and 1747 for females, while there were 4119 male and 4902 female deaths among patients with unipolar disorder. There was an overlap between the 2 diagnostic groups; 3109 individuals (20% of the patients with bipolar disorder) were initially discharged with a unipolar diagnosis. The mean follow-up was about 10 years, irrespective of sex or psychiatric diagnosis (Table 1).

The most frequent cause of death was cardiovascular disease, followed by suicide and cancer in both bipolar and unipolar disorder groups (Table 2). For bipolar disorder, the SMRs for all deaths were 2.5 in males and 2.7 in females. All natural causes of death (ICD I-XVI) were increased, except cancer and diseases of the nervous system for males. For unipolar disorder, SMRs for all deaths were 2.0 for both sexes. All natural causes of death except cancer were significantly increased. In patients with bipolar disorder, SMRs for death from unnatural causes were highest for suicide (15.0 in males and 22.4 in females) and undetermined violent death (10.3 in males and 14.2 in females). The SMRs for suicide were higher in the unipolar group compared with the bipolar group, while they tended to be lower for deaths due to undetermined violence. There were 2130 excess deaths in patients with bipolar disorder and 4585 in those with unipolar disorder. In the bipolar disorder group, there were more excess deaths from natural than from unnatural causes, both for males (ratio: 561/470 = 1.19) and for females (ratio: 668/430 = 1.55), while in the unipolar disorder group, the number of excess deaths was higher from natural causes for females only (ratio: 1305/1183 = 1.10), but not for males (ratio: 865/1232 = 0.70) (data not shown).

For both bipolar and unipolar disorder, the SMR for all deaths was highest in patients with their first admission at younger ages (Figure 1). The SMR decreased with increasing age but was still significantly increased for patients with a first diagnosis at 65 to 69 years of age. An increased SMR was also most pronounced during the first years of follow-up after the first diagnosis, with a significant increase still observed after 15 years of follow-up (Figure 2). There were no sex differences in risk ratios by age at first admission or time of follow-up.

The SMRs for suicide were elevated for both bipolar and unipolar disorder in all age classes and follow-up intervals (Table 3). In the bipolar disorder group, SMRs for the younger-than-30-year age class in the first year of follow-up were 81.6 for males and 71.7 for females, and SMRs were still 4.7 for males and 13.4 for females in the 65-years-and-older age class with more than 5 years of follow-up. In all age groups, SMR decreased with increasing time of follow-up. An interaction between age and duration of follow-up was observed; the younger the age and shorter the follow-up, the higher the SMR. However, formal tests for multiplicative interaction effects did not render significant results.

Compared with the general population, jumping as a method of suicide was more common in patients with bipolar disorder, whereas shooting was less common for male patients and poisoning less common for female patients (Table 4). Hanging and drowning were more common as methods of suicide for male patients with unipolar disorder than in the general population, while shooting was less common. Poisoning was less common in both male and female patients with unipolar disorder.

An increasing time trend in suicide mortality during the first 3 years of follow-up was found for females with unipolar disorder (P<.001), but not for males or for those with bipolar disorder (Table 5).

The main finding of this study was an increased suicide mortality rate in patients with bipolar and unipolar disorder, which was most pronounced at younger ages and in the first years after the initial diagnosis. In patients with unipolar disorder, suicide mortality for female patients even increased during the study period. The number of excess deaths from natural causes was very high, indicating the somatic health of the patients as an important area for improved treatment besides suicide prevention.

Validation of the clinical diagnosis is a problem with register-based studies. The diagnostic system changed in Sweden from ICD-8 to ICD-9 in 1987, with accompanying changes in the diagnostic criteria for bipolar and unipolar disorders. It is likely that the patients with bipolar disorder in this study had a history of manic states, since those diagnostic subclassifications were used to define the bipolar group. The psychotic depressive subclassifications used to define the unipolar group may also include some patients with bipolar disorder only diagnosed in a depressive state. Thus, the bipolar group is more strictly defined than the unipolar group.⁸ Of the bipolar patients in our study, 20% had a previous diagnosis of unipolar disorder. This may lead to an underestimation of suicide mortality for bipolar disorder, since some patients with bipolar disorder who were initially classified as having unipolar disorder will die before they get the correct diagnosis. No validation has been made in Sweden of the clinical affective psychosis diagnoses, but a validation based on medical records of clinical schizophrenia diagnosis in Stockholm County, comprising approximately 20% of the Swedish population, estimated that 80% to 85% of patients with clinical schizophrenia diagnosis met DSM-III criteria for schizophrenia.¹⁹ Another problem with this study is that only patients with a hospital admission are included, which leads to a selection of severely ill patients, especially in unipolar disorder, where patients with the less severe forms are expected to have been in outpatient treatment only. Thus, the mortality ratios found in this study for unipolar disorder may be exaggerated by a selection of more severely ill patients, who may have higher mortality rates than patients with unipolar disorder in general.

In the patients with bipolar disorder, suicide mortality was lower and mortality from natural causes was higher than in the patients with unipolar disorder, which is in line with earlier findings.^5,8 When suicide mortality in our bipolar group was compared with the aggregate mortality ratios of a meta-analysis of mortality studies in psychiatric disorders,²⁰ our rates were several times higher (male-female: 15.05/22.37 vs 5.71/5.88). Only one previous study of mortality in mood disorder has used national population-based patient data, followed up from the first diagnosis,⁸ and the reported suicide mortality from that study was in accordance with our findings. Since mortality in suicide is particularly increased in the first years after the first admission and will tend to decline with time of follow-up, first-episode cohort studies provide better estimates of the excess suicide mortality than studies based on cases identified later during the course of the illness. Suicide mortality in our patients with unipolar disorder, however, was not different from that reported in the meta-analysis (male-female: 20.93/26.98 vs 15.73/27.81). The increased mortality rate in undetermined death in both bipolar and unipolar disorder groups is to a large extent likely to be due to suicides.^21,22 In our study, an increase in SMR in suicide during the period of study was apparent among female patients with unipolar disorder. There are other previous findings of increasing time trends in suicide mortality in mood disorders⁸ and also in schizophrenia.^23,24 In the Swedish population, suicide as a cause of death has decreased during the past decades, hypothesized to be an effect of increased use of antidepressant drugs.²⁵ The finding of increasing SMR for suicide indicates that the outcome among patients with unipolar disorder has not improved, although a selection of more severely ill patients with higher suicide risks as inpatients over time cannot be excluded. The proportion of violent methods of suicide such as jumping (bipolar disorder) and hanging and drowning (unipolar disorder) were higher, while the use of poisoning was lower in patients with unipolar disorder. Shooting as a method of suicide was lower for male patients with bipolar and unipolar disorder, which may reflect a reduced access to guns, possibly indicating that the efforts of society and the psychiatric services to reduce access to guns have been successful. A similar result was found for patients with schizophrenia.²⁶

There are several studies suggesting that patients with bipolar disorder selected for and compliant with long-term treatment at specialized lithium clinics have lower-than-expected mortality from suicide.^27-32 Unfortunately, there is no information about treatment in the patient registry. However, before such programs are initiated for all patients, one should keep in mind that patients in special lithium programs constitute a selected group, who are probably more compliant compared with other patients with such disorders. Thus, more studies are needed in population-based samples to assess the impact of such measures on the risk of suicide.

Considering natural causes of death, the increased mortality for cardiovascular and respiratory disease found in both bipolar and unipolar disorder patients is in accordance with the findings of other studies.^7,20,33 One possible explanation is higher smoking rates among this population, which also may be the explanation for the increased mortality due to cerebrovascular disease. However, there are also studies that did not find an increased mortality rate from cardiovascular disease,¹² notably studies of patients receiving long-term lithium treatment. Whether this difference is an effect of the specific treatment or the selection of patients is not yet clear. In unipolar disorder, there are findings of an increased risk for coronary heart disease in both men and women.^34-36 Depression must be considered as a risk factor for coronary heart disease, although the mechanism is not known. It may be a direct effect on heart rate or increasing platelet aggregation, or an indirect effect by poor self-care or social isolation.

The number of excess deaths, rather than increased relative mortality, should be the target for preventive programs, since prevention should focus on the number of saved lives. In bipolar disorder, the total number of excess deaths was slightly higher for natural than for unnatural causes of death, while in unipolar disorder the total number of excess deaths was larger for unnatural causes. Suicide was the specific cause of death that caused most excess deaths in both bipolar and unipolar disorder groups, but due to the higher prevalence of unipolar disorder there were approximately 3 times as many excess deaths from suicide in patients with unipolar disorder. Studies of suicide risks have found that the strongest risk factor is mental illness necessitating hospital admission.^37-39 Suicide prevention programs should target patients in the first years after the first diagnosis, but there is also a need for adequate somatic care and general health measures to be improved, especially in bipolar disorder.

In conclusion, this study found a markedly increased mortality rate for patients with bipolar and unipolar disorder. Future studies should focus on the effects of specific interventions such as lithium therapy and other specific treatments. Our results also underscore the need to continuously monitor mortality in those patient groups to improve treatment.

Accepted for publication April 19, 2001.

This study was supported by grant 1998 7289 from the Stockholm County Council.

The study was conducted within the Swedish Schizophrenia Sibling Pair study. We are also indebted to Lars Terenius, MD, PhD, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, for useful opinions and critical revisions of the manuscript.

Corresponding author and reprints: Urban Ösby, MD, PhD, Karolinska sjukhuset S4, S-171 76 Stockholm, Sweden (e-mail: Urban.Osby@nvso.sll.se).