SEARCH

SEARCH BY CITATION

Keywords:

  • epidemiology;
  • nested case–control design;
  • physical illness;
  • population study;
  • suicide risk

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

Objective

To examine suicide risk in relation to physical illness across a broad range of illnesses, including hospitalization history, specific organ or system illness and comorbidity.

Design

A nested case–control study.

Setting

Data were retrieved from five Danish national registers.

Subjects

On the basis of the entire population of Denmark, this study included 27 262 suicide cases, and 468 007 live controls matched for sex and date of birth.

Main outcome measures

Risk of suicide was assessed using conditional logistic regression.

Results

In the study population, 63.5% of suicide cases and 44.5% of comparison controls had a history of hospitalization for physical illness. A physical illness significantly increased the risk of subsequent suicide (incidence rate ratios 2.13, 95% CI 2.07–2.18) with a substantially greater effect in women than in men (< 0.01). The elevated risk increased progressively with frequency and recency of hospitalization and was significant for diseases occurring in all organs or systems of the body. Comorbidity involving several organs or systems increased the risk substantially. The associated estimates were to some extent reduced but remained highly significant after adjustment for psychiatric history and socio-economic status. Taking into account both prevalence and adjusted effect size, physical illness accounted for 24.4%, 21.0% and 32.3% of population attributable risk for suicide in total, male and female populations, respectively.

Conclusions

Physical illness constitutes a significant risk factor for suicide independent of psychiatric and socio-economic factors. Clinicians treating physically ill patients should be aware of the risk, especially amongst those with multiple or recent hospitalizations, or multiple comorbidities.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

It is well known that people with physical illness are at increased risk of suicide, but the strength of the effect and the differences across a broad range of diagnoses are poorly understood. In general, studies have been conducted to examine whether suicide is associated with a single specific physical illness or generally ill condition. For instance, previous epidemiological studies have indicated an increased suicide risk associated with cancer [1], diabetes [2], epilepsy [3], multiple sclerosis [4, 5], stroke [6], myocardial infarction [7] and allergy [8]. Rarely has any large population been systematically examined for suicide risk across a range of physical conditions simultaneously, to explore the effect of level of comorbidity and to assess the association whilst taking into account personal socio-economic status and psychiatric history. A major reason for this could be the difficulty of obtaining the necessary data. In this large population study, using data from Danish national registers, we have investigated the risk of suicide in relation to physical illness, examining the effects of frequency and recency of hospitalization, diagnosis according to the categories of the International Classification of Diseases (ICD) and number of comorbid conditions.

Material and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

Setting

We conducted this study in Denmark (population 5.4 million), where hospital treatment is provided free of charge to all residents. Individual-level data for the entire population were retrieved from several national longitudinal registries and merged using the unique personal identifier (CPR number) given to all Danes by the Civil Registration System at birth and to new residents upon immigration [9].

The Cause-of-Death Register [10] contains information on cause and date of all deaths that have occurred in Denmark since 1969. The Danish National Hospital Register [11] contains individual-level data on diagnoses and dates of contact with all nonpsychiatric hospitals in Denmark since 1978. The Danish Psychiatric Central Register [12], covering all psychiatric facilities in Denmark, maintains a cumulative record of all hospital admissions and discharges since 1969. Data on outpatient contact (visits to emergency departments, hospital outpatient clinics and ambulance calls) have been included in both general and psychiatric hospital registers from 1995. Diagnoses of illness and causes of death were coded according to the ICD, 8th revision (ICD-8), until the end of 1993 and the 10th revision (ICD-10) thereafter. The ICD 9th revision has never been used in Denmark.

The IDA Database (Danish acronym for ‘the Integrated Database for Labour Market Research’) [13] records annual information on labour market conditions (employment status, type of occupation, job function, employer) and sociodemographic status for all Danish residents. Individual socio-economic data for a given calendar year are complete only for individuals living in Denmark on 31 December of that year.

Study design and participants

We utilized a nested case–control design to select study participants from the national population of Denmark. The Cause-of-Death Register [10] allowed us to identify all suicides (= 27 380; ICD-8 codes E950–959 and ICD-10 X60–84) that occurred during the period 1981–2006. In line with previous studies of suicide in Denmark [8, 14], we included as study cases only individuals for whom the cause of death was a definite suicide. We excluded individuals (= 118) who were not residing in Denmark on 31 December in the preceding year because of the incompleteness of their socio-economic data in the IDA database. Using incidence density sampling [15], we randomly selected up to 20 live controls per suicide case, matched by sex and date of birth, from a 25% representative sample of the national population in the Civil Registration System [9]. The rationale for sampling 20 controls per case was to enable the examination of uncommon exposures with reasonable statistical precision. In total, 27 262 suicide cases, representing 99.6% of all national suicides during the study period, and 468 007 live population controls were included in the study.

Study variables and covariates

We assessed personal history of physical illness through patient records in the Danish National Hospital Register [11]. For each study subject, we retrieved the records of inpatient treatment for physical illness prior to the date of suicide or the matching date for controls and classified the diagnosis of illness by areas of organ or system of the body according to categories of the ICD-10 (chapter headings). We excluded the following conditions from our definition of a physical illness: mental and behavioural disorders (Chapter V), pregnancy and childbirth (Chapter XV), certain conditions of the prenatal period (Chapter XVI) and external causes (chapters XIX and XX, covering all accidental or intentional poisonings and injuries). Details regarding diagnostic codes in ICD-10 and conversion from ICD-8 to ICD-10 codes are provided in the Appendix. We also constructed a few variables of interest including the total number of hospitalizations, time elapsed from the most recent hospitalization and number of comorbid physical illnesses.

At the same time, we obtained personal data on psychiatric history, including both inpatient and outpatient care, from the Danish Psychiatric Central Register [12]. We also extracted individual data on socio-economic status, that is, marital status, annual gross income, place of residence and citizenship, from the IDA database [13]. These data were included for adjustment purposes in the analyses [14].

Statistical analysis

We constructed contingency tables for the main study variables and computed the associated risk of suicide using conditional logistic regression with the PhReg procedure available in SAS version 9 [16] with each case forming a separate stratum. Because the controls were sampled from individuals at risk of suicide (i.e. incidence density sampling), the estimated odds ratios (ORs) from the analyses were unbiased estimates of incidence rate ratios (IRRs) [17]. We estimated IRRs from crude analyses and from adjusted models controlling for effects of psychiatric history and socio-economic factors. For estimates associated with a specific organ or system illness, we used a generic reference group, that is, individuals without hospitalization for any of the physical illnesses included in the study, and adjusted the estimates for the effect of any illness other than the specified one.

The Wald test was used to determine the significance of estimates relative to the reference group and to examine the overall effect difference across the range of organ or system illnesses. Interactions between sex and overall physical illness and between sex and specific organ or system illness were additionally examined through the likelihood ratio test. Test for trend associated with the number of comorbid illnesses was performed by treating number of physical illnesses as a continuous variable (0–6 corresponding to seven categories; see Table 2). Population attributable risk (PAR) associated with physical illness was calculated based on the adjusted IRRs estimated from the conditional logistic regression analyses and the prevalence in suicide cases of the physical illnesses considered in the study [18]. All P values were two sided.

Ethical approval

We obtained approval to conduct the study from the Danish Data Protection Agency.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

The study included a total of 27 262 suicide cases and 468 007 live population controls, comprising 18 099 male and 9163 female suicide cases and 307 729 male and 160 278 female control subjects. In the study population, 63.5% of suicide cases and 44.5% of comparison controls had a history of hospitalization because of physical illness (Table 1). For both suicide cases and controls, physical illness was generally more common in women (70.8% of cases; 50.0% of controls) than men (59.7% of cases; 41.6% of controls). A psychiatric history was more common in individuals with physical illness compared with those without: 47.2% of physically ill suicide cases, 40.8% of physically healthy suicide cases, 7.7% of physically ill controls and 3.6% of healthy controls.

Table 1. Distribution and incidence rate ratio (IRR) of suicide associated with frequency and recency of hospitalization for physical illness
Physical IllnessSuicide casesMatched controlsSuicide risk, crudeaSuicide risk, adjusted IbSuicide risk, adjusted IIc
n % n %IRR95% CIIRR95% CIIRR95% CI
  1. a

    Crude model IRRs were adjusted for sex and age (through matching).

  2. b

    Model I IRRs were further adjusted for psychiatric history.

  3. c

    Model II IRRs were additionally adjusted for marital status, income, citizenship and place of residence.

History of hospitalization
No996236.54259 80455.511 111 
Yes17 30063.46208 20344.492.132.07–2.181.651.60–1.691.621.58–1.67
Frequency of hospitalization
0996236.54259 80455.511 1 1 
1–2861931.62134 64028.771.711.66–1.771.441.40–1.491.441.39–1.48
3–5471817.3148 97010.462.572.47–2.671.821.75–1.901.791.72–1.87
6–1025819.4717 5803.763.873.68–4.062.362.23–2.492.272.15–2.40
11–2010753.9457351.235.054.70–5.432.622.42–2.842.472.28–2.68
>203071.1312780.276.565.76–7.462.742.37–3.182.532.18–2.94
Time since last hospitalization
No history of hospitalization996236.54259 80455.511 1 1 
Inpatient or the day of discharge10363.8019200.4113.4112.37–14.5310.579.63–11.6010.149.24–11.13
1–7 days5111.8711420.2410.659.54–11.889.017.95–10.208.687.66–9.83
8–30 days8953.2834290.736.275.80–6.784.704.30–5.144.594.20–5.02
1–3 months13544.9777201.654.314.05–4.593.072.86–3.303.022.81–3.24
3–6 months13234.8598112.103.343.14–3.562.332.18–2.502.292.14–2.46
6–12 months17736.5016 3293.492.712.57–2.861.921.81–2.041.871.77–1.99
1–3 years391114.3547 81710.222.071.99–2.151.531.47–1.601.501.44–1.57
3–5 years20367.4733 5747.171.551.48–1.631.211.15–1.281.191.13–1.26
>5 years446116.3686 46118.471.371.32–1.421.141.09–1.191.131.09–1.18

Frequency and recency of hospitalization for serious physical illness

Multiple hospitalizations for physical illness as well as a recent hospitalization were more common amongst suicide cases, relative to the control group (Table 1). Of suicide cases with physical illness, 50.2% had been hospitalized three or more times, and 39.8% had been hospitalized within the year prior to suicide; the corresponding values for the comparison controls were 35.3% and 19.4%, respectively.

Using conditional logistic regression to analyse the data, we found that a history of physical illness significantly increased the risk of suicide (crude IRR 2.13, 95% CI 2.07–2.18) and that the effect was greater amongst women than men (crude IRR 1.97, 95% CI 1.91–2.03 for men; crude IRR 2.50, 95% CI 2.38–2.62 for women; test of sex differences: χ2 = 64.64, < 0.001). We also noted a progressively increased suicide risk associated with both the frequency of previous hospitalizations and the recency of the last hospitalization – the greater the number of hospitalizations or the more recent the last hospitalization, the higher the risk of suicide (Table 1). The observed pattern persisted in the analyses stratified by sex although the associated IRRs were generally higher in women than that in men (data not shown). When the data were adjusted for history of psychiatric illness, the associated IRRs were reduced to some extent, especially in the presence of multiple hospitalizations for physical illness. Still, the increase in associated risks remained highly elevated. Further adjustment for socio-economic status had minimal effect on the risk estimates (adjusted IRR 1.62, 95% CI 1.58–1.67 for total cases; adjusted IRR 1.54, 95% CI 1.49–1.60 for men; adjusted IRR 1.84, 95% CI 1.74–1.94 for women; test of sex differences: χ= 51.04, < 0.001).

Diagnosis by ICD main categories and number of physical illnesses

When diagnoses of physical illness were grouped according to ICD-10 categories (Chapter headings; Table 2), a significantly higher prevalence of physical illness was observed in suicide cases than in controls across all organs or systems of the body. Digestive, circulatory and genitourinary system diseases were amongst the most common illnesses in both suicide cases and comparison controls. It is interesting that a substantial proportion of suicide cases (19.1%) had been hospitalized at least once because of symptoms, signs and abnormal clinical and laboratory findings not included in other categories.

Table 2. Distribution and incident rate ratio (IRR) of suicide associated with diagnostic groups according to ICD-10 category
ICD -10 chapterDiagnostic category: chapter headingSuicide casesMatched controlsSuicide riska,b, crudecSuicide riska,b, adjustedd
n % n %IRR95% CIIRR95% CI
  1. a

    The group of individuals with no record of hospitalization for any physical illness was used as the generic reference group for the estimate associated with each specific type of illness.

  2. b

    Test of differences in overall effect size across diagnostic groups: χ2 = 640.21, df = 13, < 0.001 for the crude estimates; χ= 130.09, df = 13, < 0.001 for the adjusted estimates.

  3. c

    Crude IRRs were adjusted for sex and age (through matching) and the presence of a physical illness other than the specified one.

  4. d

    IRRs were further adjusted for psychiatric history, marital status, income, citizenship and place of residence.

IInfectious and parasitic diseases12764.6812 0602.582.812.64–2.991.761.64–1.88
IINeoplasms277810.1930 0626.422.362.25–2.471.921.83–2.02
IIIBlood and immune mechanism disorders3191.1726530.572.822.50–3.191.921.68–2.20
IVEndocrine, nutritional and metabolic diseases11094.0711 6932.502.432.28–2.601.611.50–1.73
VINervous system diseases16325.9911 2872.413.833.62–4.052.021.89–2.15
VII, VIIISensory organ diseases11364.1712 0352.572.152.01–2.301.591.48–1.71
IXCirculatory system diseases454516.6745 3239.682.362.27–2.461.721.65–1.80
XRespiratory system diseases307311.2728 5906.112.762.64–2.881.901.81–1.99
XIDigestive system diseases573521.0456 05911.982.572.48–2.661.731.66–1.80
XIISkin diseases11804.3311 0412.362.882.70–3.071.741.62–1.86
XIIIMusculoskeletal system diseases341612.5335 6217.612.482.38–2.591.821.73–1.90
XIVGenitourinary system diseases440016.1449 30110.532.282.20–2.381.711.64–1.79
XVIICongenital malformations4111.5156061.202.071.87–2.291.561.39–1.74
XVIIIUnspecified symptoms, signs, and abnormal clinical and laboratory findings519719.0638 6038.253.513.39–3.652.041.96–2.13

With the group of individuals with no hospitalization for any physical illness included in the study as a reference, there was a significantly increased risk of suicide associated with specific diseases in all organs or systems of the body. The IRRs associated with the different specific organ or system illnesses varied significantly in magnitude (< 0.001 in both crude and adjusted analyses), from 2.07 to 3.83 in the crude analysis and from 1.56 to 2.04 in the analysis adjusted for psychiatric history and socio-economic status (Table 2). The highest IRRs were associated with nervous system diseases and the group of unspecified symptoms, signs and abnormal clinical and laboratory findings; the remaining groups of diseases had a comparable effect on suicide risk.

Although analyses stratified by sex showed somewhat similar patterns (Fig. 1), the effects of some organ or system illnesses differed significantly between men and women. Combining the results from adjusted IRRs and the test for sex interaction showed that infectious and parasitic diseases (Chapter I), nervous system diseases (Chapter VI), respiratory system diseases (Chapter X), digestive system diseases (Chapter XI), skin diseases (Chapter XII), musculoskeletal system diseases (Chapter XIII), congenital malformations (Chapter XVII) and unspecified symptoms, signs and abnormal clinical and laboratory findings (Chapter XVII) increased suicide risk significantly more in women than in men, whereas neoplasms (Chapter II) had a significantly greater effect on suicide in men. The increased suicide risk for the other diagnostic categories was comparable for both sexes.

image

Figure 1. Incidence rate ratio (IRR) of suicide associated with diagnostic groups by ICD-10 chapter headings: stratified analyses for men and women separately. The group of individuals with no record of hospitalization for any physical illness was used as the reference group for the estimate associated with each specific type of illness. Crude IRRs were adjusted for sex and age (through matching) and the presence of a physical illness other than the specified one; adjusted IRRs were further adjusted for psychiatric history, marital status, income, citizenship and place of residence. Test of sex differences: < 0.01 for infectious and parasitic diseases (Chapter I), neoplasms (Chapter II), nervous system diseases (Chapter VI), respiratory system diseases (Chapter X), digestive system diseases (Chapter XI), skin and subcutaneous diseases (Chapter XII), musculoskeletal system diseases (Chapter XIII), congenital malformations (Chapter XVII) and unspecified symptoms, signs and abnormal clinical and laboratory findings (Chapter XVII).

Download figure to PowerPoint

Comorbid illnesses involving multiple organs or systems were more common in suicide cases than in controls (Table 3). Overall, 35.5% of suicide cases, but only 18.4% of controls, had been hospitalized for diseases occurring in more than one organ or system. The associated suicide risk increased progressively with the number of comorbid organ or system diseases, and the test of trend was highly significant (analogous IRR 1.44, 95% CI 1.42–1.45; < 0.0001). The highly elevated risks associated with multiple comorbidities were largely reduced when the data were adjusted for psychiatric and socio-economic status. Still, the progressive increase in suicide risk associated with multiple comorbidities remained prominent for both sexes together (analogous IRR for trend test 1.23, 95% CI 1.21–1.24; < 0.0001) (Table 3) as well as for men and women separately (data not shown).

Table 3. Distribution and incident rate ratio (IRR) associated with comorbidity of physical illnesses according to number of ICD-10 categories
Number of physical illnessesSuicide casesMatched controlsSuicide risk, crudeaSuicide risk, adjustedb
n % n %IRR95% CIIRR95% CI
  1. a

    Crude IRRs were adjusted for sex and age (through matching);

  2. b

    IRRs were further adjusted for psychiatric history, marital status, income, citizenship and place of residence.

  3. c

    Test for trend was performed by including number of physical illnesses as a continuous variable, that is, from 0 to 6 corresponding to the seven categories.

0 996236.54259 80455.511 1 
1761827.94121 990 26.071.671.62–1.731.42 1.37–1.47
24703 17.2552 20311.152.432.34–2.521.76 1.69–1.83
32579 9.4620 7404.433.353.20–3.522.07 1.97–2.19
41300 4.7782311.764.243.98–4.532.21 2.06–2.38
5638 2.3432580.705.304.84–5.802.35 2.12–2.61
>5462 1.7017810.387.076.34–7.892.71 2.40–3.07
Test of trendc    1.441.42–1.451.23 1.21–1.24

Population attributable risk

To assess the proportion of suicides in the population that is attributable to exposure to the physical illnesses considered in this study, we computed the PAR – a measurement that takes into account both the prevalence of physical illness in the population and the magnitude of its effect on suicide risk (Table 4). We estimated a PAR of 24.4% for the total population, 21.0% for men and 32.3% for women, including a contribution because of a recent hospitalization for physical illness within 30 days of 7.7%, 7.2% and 8.6%, respectively. Illness categories with a high PAR included the group of unspecified symptoms, signs and abnormal clinical and laboratory findings, as well as digestive, circulatory and genitourinary system diseases.

Table 4. Population attributable risk (PAR) of suicide associated with physical illness
ICD -10 chapterDiagnostic category: chapter headingsTotalMale casesFemale cases
  1. PAR was calculated based on the adjusted IRR using model II (see legend to Table 1) and the distribution of exposure in the cases. PARs associated with specific diagnostic groups were estimated independently of each other and are not additive. All values are given as percentages.

IInfectious and parasitic diseases2.01.82.6
IINeoplasms4.93.97.2
IIIBlood and immune mechanism disorders0.60.50.7
IVEndocrine, nutritional and metabolic diseases1.51.32.3
VINervous system diseases3.02.44.4
VII, VIIISensory organ diseases1.61.32.2
IXCirculatory system diseases7.06.87.6
XRespiratory system diseases5.34.86.4
XIDigestive system diseases8.97.911.0
XIISkin diseases1.81.72.1
XIIIMusculoskeletal system diseases5.64.77.7
XIVGenitourinary system diseases6.74.411.9
XVIICongenital malformations0.50.50.7
XVIIIUnspecified symptoms, signs and abnormal clinical and laboratory findings9.77.514.5
Any physical illness included in the study24.421.032.3 

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

The findings of this large study in the entire population of Denmark show that hospitalization for a physical illness significantly increases the risk of subsequent suicide, with a substantially greater effect in women than men. The elevated risk increased progressively with the frequency and recency of hospitalization and was significant for all diagnostic groups of disease in any organ or system defined by ICD-10 categories. Comorbidity involving several organs or systems greatly increased suicide risk. At the same time, the observed effect of physical illness was somewhat reduced after adjustment for history of psychiatric illness, which was especially apparent for the estimated risks associated with multiple hospitalizations and multiple comorbidities. Nevertheless, the elevated risks remained highly significant after adjusting for psychiatric history and socio-economic status. Moreover, hospitalization for physical illness accounted for 24.4% of the PAR for suicide in the general population, with a higher attributable fraction for female (32.3%) than for male (21.0%) suicide.

These robust findings support a strong link between physical illness and risk of subsequent suicide in the general population. There have been a number of reports of an increased suicide risk associated with several severe diseases, including amongst others cancer and diabetes [1-8]. The findings of this study extend the existing evidence by demonstrating that a disease of any organ or physical system significantly increases the risk of subsequent suicide, regardless of sex and after controlling for individual psychiatric history and socio-economic status.

A serious physical illness is often accompanied by pain, disability, limited social activity and worry about prognosis and financial loss and can lead to many problems with respect to physical, psychological and social well-being. Although we were not able to assess the severity of specific diagnoses, the observed progressive increase in suicide risk associated with frequency of hospitalization and multiple comorbidities indicates that the severity of physical illness plays an important role in the risk of subsequent suicide. The observed effects of these factors were apparently reduced after adjustment for psychiatric history [19]. Still, being physically severely ill, as evidenced by either multiple hospitalizations or the involvement of several organs or systems, confers a strong risk of subsequent suicide. Patients may choose to end their life to obtain release from a long-term debilitating condition or to avoid experiencing the progression of an illness.

Our finding that variation in effect size across illnesses in various organs or systems is relatively small differs somewhat from the findings of previous studies of suicide in relation to single-disease entities [1-8, 20, 21]. One methodological explanation could be that diagnoses in this study were broadly grouped on the basis of ICD-10 categories, covering a diversity of illnesses that occur in both severe and less severe forms or have a strong or weak relation to suicide. For instance, the group of nervous system diseases includes both diagnostic entities such as multiple sclerosis with a high risk of suicide [4, 5] and other diagnoses such as Parkinson's disease with a low suicide risk [22]. Another possible explanation is that the variation in risk reported from the previous studies of single-disease groups may be subject to differential selection and information biases. A key strength of this register-based population study was the ability to estimate relative risk across a broad range of different physical illness types within the same study cohort compared with a generic reference category (i.e. individuals without any of the physical illnesses included in the study). This excluded the possibility of differential types and levels of bias, making comparison of risk between diagnostic groups more straightforward.

We observed that a psychiatric history was significantly more common in suicide cases with physical illness compared with those without (47.2% vs. 40.8%) and also that the elevated suicide risk associated with physical illness weakened to some extent after adjustment for history of contact with psychiatric services. These findings are consistent with previous reports of high comorbidity of psychiatric illness, especially depression, in people with physical diseases [19, 23], and suggest that psychiatric illness is an important confounding or mediating factor in the link between physical illness and suicide risk.

To our knowledge, this study is the first to demonstrate a strong impact of the ICD-10 category of unspecified symptoms, signs and abnormal clinical and laboratory findings, in terms of both increased IRR and contributed PAR of suicide. This is noteworthy and consistent with reports indicating that the presence of unexplained somatic symptoms is associated with psychiatric disorders such as major depression and anxiety [24, 25] and that complaints of illness or feeling unwell may provide clues to patients’ desire for emotional support [26]. Although our estimates were controlled for the effects of psychiatric history and socio-economic status as well as of specified diagnoses from other ICD chapters, it is still possible that complaints of unspecified symptoms and signs may be linked to undiagnosed psychological symptoms, stress or somatization.

Our observation of a generally stronger effect of physical illness on suicide risk in women than in men suggests that women probably react more strongly to poor health than their male counterparts. This notion is also supported by previous evidence that psychiatric history has a significantly stronger impact on suicide in women than in men [14]. At the same time, some physical illnesses and their influences on suicide may be more sex specific. For instance, we found that illnesses in most organs or systems have a stronger effect on increasing suicide risk in women than in men, but neoplasms increased the risk significantly more in men. Further investigation of detailed diagnoses stratified by sex is needed for better understanding of this issue.

Relative to the intensive research and preventive efforts focusing on suicide in patients with psychiatric illness, subsequent suicide related to physical illness has received little attention. This study is the first to report that around a quarter of all suicides in the general population is attributable to exposure to physical illnesses that require hospital admission for treatment, including a 7.6% contribution because of a recent hospitalization for physical illness within 30 days. These PAR estimates could be subject to revision if more factors that may underlay the link between physical illness and subsequent suicide are taken into consideration. It does, however, indicate that suicide in patients with physical illness is clearly an important public health problem and warrants more investment of social resources and prevention efforts.

Limitations and strengths

A limitation of this study is that we considered only serious physical illnesses resulting in hospitalization. This may provide a slightly skewed picture as the influence of milder physical illnesses was not ascertained and because some severe medical conditions are treated in outpatient care. Similarly, we used psychiatric service contact as a proxy for psychiatric status. This underestimates the full extent of psychiatric disorders and reflects more serious disorders. Moreover, we were only able to obtain individual data on serious physical illness starting in 1978 when the Danish National Hospital Register was established. Thus, our data cover a relatively long period, but do not represent the lifetime of older subjects. Additionally, we detected a strong modification effect of psychiatric history on the link between physical illness and suicide risk, but did not address specifically the temporal relationship between physical illness and comorbid psychiatric disorders in this study. Furthermore, we grouped the diagnosis of illness based on areas of organ or system using the ICD-10 categories (chapter headings), which captures a mixed effect of illnesses occurring in one organ or body system but is not able to separate individual effects of specific clinical diagnoses on risk of suicide.

A major strength of this study is the coverage of the entire population of Denmark and the availability of precise data collected systematically without any research purpose. This ensures the full representation of a national population and minimizes possible biases that are often induced by recall of information and selected access to healthcare. The results should also be highly applicable to the general public, especially in countries or regions with comparable social settings.

To our knowledge, this study is the first to document suicide risk in relation to physical illness, encompassing hospitalization history, diagnosis and comorbidity and to assess the effect in the context of personal psychiatric and socio-economic status. It is also the first to evaluate the proportion of suicides in the population that is attributable to exposure to the physical illnesses included in the study. These findings will be informative for making strategies to reduce suicide in the general population.

Clinical implications

The results from this population study indicate the need to integrate suicide prevention within hospital treatment as well as within general medical practice [27]. Patients hospitalized for physical illness not only form a well-defined population group at high risk of suicide, but also maintain frequent contact with their general practitioner after hospital discharge. This provides opportunities for both risk assessment and prevention. According to our estimates, almost a quarter of all suicides in the general population could be avoided if prevention efforts were sufficiently implemented in services and settings dealing with people who are hospitalized because of physical illness. Although it is impossible to prevent all suicides in physically ill people, a considerable number of lives could be saved. Clinicians treating people with physical illness should be aware of the risk of suicide, especially amongst patients with multiple or recent hospitalizations or with comorbid physical illnesses.

At the same time, our findings underscore the necessity of further investigation. Whilst physical illness is common, only a small proportion of physically ill patients commit suicide. Therefore, we need to explore more thoroughly the link between various specific diagnoses of physical illness and subsequent suicide and to better understand the characteristics, medical conditions and outlook amongst individuals who are physically ill but do not commit suicide.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

The study was funded by the Sygekassernes Helsefonden in Denmark (2009B063 to PQ). The study sponsor had no involvement in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review and approval of the manuscript; and the decision to submit the article for publication. The corresponding author had full access to the study data (including statistical reports and tables), takes responsibility for the integrity of the data and the accuracy of the analysis and had final responsibility for the decision to submit for publication.

Authors’ contributions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

PQ designed the study, compiled the data, conducted the analysis and wrote the report. RW, NK and HTS participated in all discussions about the study design and interpretation of data and made critical revisions to the manuscript for important intellectual content. All authors have seen and approved the final version of the manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix
  • 1
    Allebeck P, Bolund C, Ringback G. Increased suicide rate in cancer patients. A cohort study based on the Swedish Cancer-Environment Register. J Clin Epidemiol 1989; 42: 6116.
  • 2
    Tseng CH. Mortality and causes of death in a national sample of diabetic patients in Taiwan. Diabetes Care 2004; 27: 16059.
  • 3
    Christensen J, Vestergaard M, Mortensen PB, Sidenius P, Agerbo E. Epilepsy and risk of suicide: a population-based case-control study. Lancet Neurol 2007; 6: 6938.
  • 4
    Bronnum-Hansen H, Stenager E, Nylev SE, Koch-Henriksen N. Suicide among Danes with multiple sclerosis. J Neurol Neurosurg Psychiatry 2005; 76: 14579.
  • 5
    Fredrikson S, Cheng Q, Jiang GX, Wasserman D. Elevated suicide risk among patients with multiple sclerosis in Sweden. Neuroepidemiology 2003; 22: 14652.
  • 6
    Teasdale TW, Engberg AW. Suicide after a stroke: a population study. J Epidemiol Community Health 2001; 55: 8636.
  • 7
    Larsen KK, Agerbo E, Christensen B, Sondergaard J, Vestergaard M. Myocardial infarction and risk of suicide: a population-based case-control study. Circulation 2010; 122: 238893.
  • 8
    Qin P, Mortensen PB, Waltoft BL, Postolache TT. Allergy is associated with suicide completion with a possible mediating role of mood disorder - a population-based study. Allergy 2011; 66: 65864.
  • 9
    Pedersen CB, Gotzsche H, Moller JO, Mortensen PB. The Danish Civil Registration System. A cohort of eight million persons. Dan Med Bull 2006; 53: 4419.
  • 10
    Juel K, Helweg-Larsen K. The Danish registers of causes of death. Dan Med Bull 1999; 46: 3547.
  • 11
    Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull 1999; 46: 2638.
  • 12
    Munk-Jorgensen P, Mortensen PB. The Danish Psychiatric Central Register. Dan Med Bull 1997; 44: 824.
  • 13
    Danmarks Statistik. IDA — En Integreret Database for Arbejdsmarkedsforskning. København: Danmarks Statistiks trykkeri, 1991.
  • 14
    Qin P, Agerbo E, Mortensen PB. Suicide risk in relation to socioeconomic, demographic, psychiatric, and familial factors: a national register-based study of all suicides in Denmark, 1981–1997. Am J Psychiatry 2003; 160: 76572.
  • 15
    Clayton D, Hills M. Statistical Models in Epidemiology. Oxford, New York, Tokyo: Oxford University Press, 1993.
  • 16
    SAS Institute Inc. The PHREG Procedure. SAS/STAT User's Guide, Version 8. Cary, NC: SAS Institute Inc, 1999.
  • 17
    King G, Zeng L. Estimating risk and rate levels, ratios and differences in case-control studies. Stat Med 2002; 21: 140927.
  • 18
    Bruzzi P, Green SB, Byar DP, Brinton LA, Schairer C. Estimating the population attributable risk for multiple risk factors using case-control data. Am J Epidemiol 1985; 122: 90414.
  • 19
    Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B. Depression, chronic diseases, and decrements in health: results from the World Health Surveys. Lancet 2007; 370: 8518.
  • 20
    Harris EC, Barraclough BM. Suicide as an outcome for medical disorders. Medicine (Baltimore) 1994; 73: 28196.
  • 21
    Stenager EN, Stenager E. Physical illness and suicidal behaviour. In: Hawton K, van Heeringen K, eds. The International Handbook of Suicide and Attempted Suicide. Chichester: Wiley, 2000; 40520.
  • 22
    Mainio A, Karvonen K, Hakko H, Sarkioja T, Rasanen P. Parkinson's disease and suicide: a profile of suicide victims with Parkinson's disease in a population-based study during the years 1988–2002 in Northern Finland. Int J Geriatr Psychiatry 2009; 24: 91620.
  • 23
    Koenig HG, George LK, Peterson BL, Pieper CF. Depression in medically ill hospitalized older adults: prevalence, characteristics, and course of symptoms according to six diagnostic schemes. Am J Psychiatry 1997; 154: 137683.
  • 24
    van BK, Lucassen P, van RH et al. Do unexplained symptoms predict anxiety or depression? Ten-year data from a practice-based research network Br J Gen Pract 2011; 61: 31625.
  • 25
    Olde Hartman TC, Borghuis MS, Lucassen PL, van de Laar FA, Speckens AE, van WC. Medically unexplained symptoms, somatisation disorder and hypochondriasis: course and prognosis. A systematic review. J Psychosom Res 2009; 66: 36377.
  • 26
    Salmon P, Ring A, Humphris GM, Davies JC, Dowrick CF. Primary care consultations about medically unexplained symptoms: how do patients indicate what they want? J Gen Intern Med 2009; 24: 4506.
  • 27
    Druss B, Pincus H. Suicidal ideation and suicide attempts in general medical illnesses. Arch Intern Med 2000; 160: 15226.

Appendix

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. Authors’ contributions
  10. References
  11. Appendix

ICD-10 and ICD-8 codes

ICD-10 chapterDiagnostic category: chapter headingsICD-10 codesICD-8 codes
  1. a

    Diagnostic groups are not normally viewed as physical diseases and therefore were excluded from our definition of a physical illness.

IInfectious and parasitic diseasesA00–B99001–136
IINeoplasmsC00–D48140–239
IIIBlood and immune mechanism disordersD50–D89280–289
IVEndocrine, nutritional and metabolic diseasesE00–E90240–279
VaMental disordersF00–F99290–315
VINervous system diseasesG00–G99320–359
VII, VIIISensory organ diseasesH00–H95360–389
IXCirculatory system diseasesI00–I99390–458
XRespiratory system diseasesJ00–J99460–519
XIDigestive system diseasesK00–K93520–577
XIISkin and subcutaneous tissue diseasesL00–L99680–709
XIIIMusculoskeletal system diseasesM00–M99710–738
XIVGenitourinary system diseasesN00–N99580–629
XV, XVIaPregnancy, childbirth and prenatal conditionsO00–P99630–678, 760–779
XVIICongenital malformationsQ00–Q96740–759
XVIIIUnspecified symptoms, signs and abnormal clinical and laboratory findingsR00–R99780–796
XIX, XXaExternal causesS00–T98, V01–Y98800–999
SuicideX60–X84E950–E959