Skip to main content
PMC home page
Canadian Journal of Psychiatry. Revue Canadienne de Psychiatrie logoLink to Canadian Journal of Psychiatry. Revue Canadienne de Psychiatrie
. 2014 Oct;59(10):531–538. doi: 10.1177/070674371405901006

Acute Risk of Suicide and Suicide Attempts Associated With Recent Diagnosis of Mental Disorders: A Population-Based, Propensity Score–Matched Analysis

Jason R Randall 1,✉, Randy Walld 2, Greg Finlayson 3, Jitender Sareen 4, Patricia J Martens 5, James M Bolton 6
PMCID: PMC4197787  PMID: 25565686

Abstract

Objective:

To determine the degree of risk during the first year after diagnosis with a mental illness.

Methods:

We used propensity scoring to create a matched sample for all identified suicide attempts and suicide deaths in the province of Manitoba from 1996 to 2009. This study identified 2100 suicide deaths and 8641 attempted suicides. Three control subjects were identified for every case and matched on age, sex, income decile, region of residence, and marital status. Five categories of physician-diagnosed mental disorders were tested: schizophrenia, anxiety, depression, dementia, and substance abuse. Logistic regression was used to determine the risk for suicide attempts and suicide deaths overall, and within 3 time periods since initial diagnosis: 1 to 90 days, 91 to 364 days, and 365 or more days.

Results:

All disorders, except dementia, were independently related to death. All disorders were related to suicide attempts. The risk of dying by suicide was particularly high within the first 90 days after initial diagnosis for many disorders, including depression (adjusted odds ratio [AOR] 7.33; 95% CI 4.76 to 11.3), substance use disorders (AOR 4.07; 95% CI 2.43 to 6.82), and schizophrenia (AOR 20.91; 95% CI 2.55 to 172). Depression and anxiety disorders had elevated risk in the first year for suicide attempts.

Conclusions:

These data suggest that several mental disorders independently increase the risk of suicide attempts and death by suicide after controlling for all mental disorders and demographic risk factors. Clinicians should be aware of the heightened risk of suicide and suicidal behaviour within the first 3 months after initial diagnosis.

Keywords: suicide, depression, anxiety, schizophrenia, dementia, substance-related disorders, case–control studies

Suicide is the 10th leading cause of death and it is estimated that a million people die by suicide, worldwide, each year.1,2 Preventing suicide has proven to be problematic so far.2 The causes of suicide and the relation between mental illness and suicidal behaviours are not well understood.3 Understanding these factors is important for determining people most at risk. Previous research has linked mental illness to suicide and attempted suicide.1,4–7 The data suggest that around 90% of suicides occur among people with mental illness.2 There is evidence that the risk of suicide is particularly high shortly after diagnosis with mental illness.5,7–10 During their 18-year follow-up of a sample of patients with deliberate self-harm, De Moore and Robertson11 found that 27% of the suicides occurred within the first year of the study. However, the timing of suicide risk, as it pertains to newly diagnosed mental disorders, is not fully clarified.

Tidemalm et al7 used administrative data to examine the occurrence of suicide 25 years following diagnosis. They found that there was a significantly increased risk for suicide in the first year after diagnosis for most of the disorders. However, they included only patients who had been admitted for a suicide attempt in their study. In that study, the initial suicide attempt was used as the index point, not the onset of mental illness. Nordentoft et al5 also used administrative data to determine the risk over time for suicide by mental disorder. Their study focused on lifetime risk of suicide and provided little statistics about the first few years after diagnosis. They created a healthy population matched only on sex and date of birth, failing to control for other possible confounders. Their study also only assessed the effect of comorbidity for substance use and unipolar affective disorder, while failing to examine the effects of other mental disorders. There does not appear to be a well-controlled analysis of the risk of suicide within the first year of diagnosis, despite the above studies suggesting that the early period of a person’s mental illness is one of substantial vulnerability to suicide. Clarifying the risk in more discrete time intervals within the first year of diagnosis has clinical importance for treatment providers, services delivery, and risk assessment.

Clinical Implications

  • The first year of a new diagnosis of any disorder indicated a high risk for suicidal behaviours.

  • The risk in the first 90 days was elevated, compared with risk after first year, for depression, anxiety, and substance use.

  • All disorders were independently associated with an increased risk of suicidal behaviour.

Limitations

  • The study could not assess severity of disorders.

  • The study was limited to people who received treatment for a mental disorder.

We endeavoured to examine the acute risk of suicide after diagnosis with a mental illness. We used the population-based administrative Data Repository of the province of Manitoba. Propensity score matching was used to address potential biases associated with treatment seeking in previous studies that have used administrative data sources. The primary objectives were to examine the degree of risk for suicide and suicide attempts during the first year after the diagnosis of mental illness. We hypothesized that all of the mental illnesses would increase the risk of suicide to some degree, even after controlling for comorbidity, and that this risk would be highest closer to the point of diagnosis across disorders.

Methods

Sample

Data were obtained from the Data Repository housed at the Manitoba Centre for Health Policy for the years 1995 to 2009, inclusive. The Data Repository is an administrative database that contains anonymized person-level health data for nearly all of the population of Manitoba (n = 1.2 million). Data for some military personnel and members of First Nations may not be available as these groups may receive treatment from providers funded by the federal government of Canada. Data within the Data Repository are coded using ICD-9-CM prior to April 1, 2004, and ICD-10-CA after April 1, 2004.12,13 The registry database contains information on age, sex, region of residence, and other demographic information. The physician claims and hospital admission databases contains information on individual outpatient visits as well as hospital admissions. In Manitoba’s single-payer system, physicians file billing claims with diagnostic codes after treating patients. Hospital abstracts for patients treated are also collected, including the discharge diagnoses made by physicians.

Cohorts

Propensity score matching14 was used to identify 3 control subjects for every identified case of suicide or attempted suicide (subjects). Propensity score matching uses confounding variables to predict the risk of the outcome(s), in this case with logistic regression. Subjects are then matched to control subjects who have similar predicted risk based on the matching variables. In this study, subjects and control subjects were matched on age, sex, income decile, region of residence, and marital status using a greedy matching algorithm.15 Standard differences were calculated before and after matching to ensure that the propensity matching was effective at controlling the sample. The Kolmogorov–Smirnov 2-sample test for difference in distribution of propensity score was performed after matching.

Suicide and Suicide Attempts

The Vital Statistics registry of the province was used to determine cases of suicide. Suicide attempts were determined through analysis of physician claims and hospital admission records. The codes used to determine the occurrence of suicide and suicide attempts are as follows:

Suicide
ICD-9-CM: E850.2, E850.9, E852.9, E862.9, E869.9, E950-E959
ICD-10-CA: X40–X42, X46, X47, X60–X84, Y87.
Suicide attempt
ICD-9-CM: E850–E854, E858, E862, E868 E950–E959; 965, 967, 969, 977.9, 986
ICD-10: T39, T40, T42.3, T42.4, T42.7, T43, T50.9, T58, X40–42, X44, X46, X47, X60–X84, Y10–12, Y16–17.
Open in a new tab

Mental Disorders

The presence of physician-diagnosed depression, anxiety disorders, substance use, schizophrenia, dementia, and other psychosocial disorders was determined for the subjects. These disorders were determined using the definitions of the Manitoba Centre for Health Policy’s concept dictionary.16,17 Owing to the coding structure in these data, it is not possible to differentiate between certain conditions, such as bipolar disorder and unipolar affective disorder. The codes used to define the categories of mental illness are detailed in the Manitoba RHA Indicators Atlas 2009.18 Other psychosocial disorders included all mental disorders not specifically included in the 5 other mental illness categories and was used as a variable to control confounding. The timing of the outcomes following initial physician diagnosis of each mental disorder (within the study period) was determined. This was categorized in the following intervals: initial diagnosis occurring 1 to 90 days, 91 to 364 days, or 365 or more days prior to suicide or suicide attempt. If a mental disorder was diagnosed in more than one of these time periods, it was included only in the earliest period and excluded from the more recent time period(s).

Analysis

Conditional logistic regression was used to obtain the AORs and their 95% confidence intervals. Regression adjusted for the ADGs score.19,20 This score estimates the severity of illness for people based on their history of diagnosis using the ADGs system of classification.19 The presence of all mental disorders was adjusted for simultaneously. Interaction effects were tested for each mental disorder assessed and age, sex, and income level. Analysis was performed for both death by suicide and suicide attempts with AORs derived for both outcomes. AORs were calculated for depression, anxiety disorders, substance use, schizophrenia, and dementia. The AOR was also calculated based on time since diagnosis for each of these disorders, using the time intervals listed above. Analysis was performed using SAS, version 9.2.21

Ethics

This research project received approval by the Research Ethics Board of the University of Manitoba (#H2008:285) and the Health Information Privacy Committee of Manitoba (#2011/2012–24).

Results

Sample

Information on the samples drawn from the Data Repository is provided in Table 1. Standardized differences after matching were less than 2% for all of the matching groups. The results of the Kolmogorov–Smirnov 2-sample tests were nonsignificant for both the suicide control group (KSa: 0.075; P > 0.9) and the attempt control group (KSa: 0.146; P > 0.9). A total of 2100 deaths due to suicide were identified between the years of 1995 and 2009 (12/100 000 per year). We identified 6300 matching control subjects from the Data Repository. The subjects in the cohort were 74.3% male, and 22.7% of all subjects were married, compared with a provincial rate of 37.7%. The Data Repository undercounts people who are married, thus the real percentages are likely higher than the figures obtained in our study.17 We identified 8641 people with an attempted suicide recorded in the Data Repository. Matching at a 3-to-1 ratio provided 25 923 control subjects. In this sample, 36.9% were male and 16.0% were married. Both suicide and suicide attempts were more common in the 3 lowest deciles of income.

Table 1.

Sample characteristics, %

Demographic Province n = 1 175 484 Suicide sample Suicide attempt sample
Subjects n = 2100 Control subjects n = 6300 Subjects n = 8641 Control subjects n = 25 923
Male 49.3 74.3 74.1 36.9 37.2
Married 37.2 22.7 22.4 16.0 16.1
Live in Winnipega 57.9 55.2 55.6 41.4 41.7
Incomeb
  Lowest decile 9.6 21.2 21.2 22.7 23.5
  2 10.0 15.3 15.7 17.4 17.9
  3 10.0 11.3 11.7 13.0 13.2
  4 10.1 8.7 8.8 7.4 7.1
  5 9.9 8.8 8.3 8.5 7.9
  6 10.1 8.2 7.7 7.3 7.0
  7 9.7 6.8 6.8 6.2 6.0
  8 9.5 6.5 6.7 6.2 6.2
  9 9.4 6.3 6.4 5.0 5.0
  Highest decile 9.2 5.7 5.6 5.1 5.1
Open in a new tab
a

Control subjects were selected based on a division of province into 10 regions

b

Percentages do not add to 100% owing to unclassified people in Data Repository

Risk of Suicide by Mental Disorder

All 5 groups of mental disorders were found more often in people who died from suicide (Table 2). After adjusting for covariates, 4 of the 5 disorders examined were found to still be significantly more common among people who died by suicide. The conditions found to be more common among subjects who died by suicide were depression (AOR 3.90; 95% CI 3.37 to 4.52), anxiety disorders (AOR 1.50; 95% CI 1.23 to 1.83), substance use disorders (AOR 3.50; 95% CI 1.91 to 4.18), and schizophrenia (AOR 3.06; 95% CI 2.15 to 4.34).

Table 2.

Risk of suicide by mental disorder

Mental disorder Rates of disorder within each group, %
AOR (95% CI)
Subjects n = 2100 Control subjects n = 6300
Depression 48.5 15.2 3.90 (3.37 to 4.52)
Anxiety 19.9 5.9 1.50 (1.23 to 1.83)
Substance use 27.2 6.7 3.50 (2.92 to 4.18)
Schizophrenia 6.9 1.2 3.06 (2.15 to 4.34)
Dementia 7.7 2.0 1.30 (0.95 to 1.77)
Open in a new tab

Adjusted for Aggregated Diagnosis Group score and all mental disorders simultaneously. Suicide subjects and control subjects were matched 3:1 on age, sex, income decile, region of residence, and marital status.

AOR = adjusted odds ratio

Suicide Risk According to Time Since Diagnosis

Analyzing the rate of death by suicide after initial diagnosis revealed that almost all of the disorders significantly increased the risk of suicide at almost every period after diagnosis (Table 3). There were no control subjects for anxiety patients in the first 90 days following a diagnosis, thus for anxiety only the time periods were defined as 1 to 180 days, 181 to 364 days, and 365 or more days. The only instances that were not significant were the periods of 181 to 365 days from diagnosis for people (subjects and control subjects) with anxiety (AOR 1.33; 95% CI 0.67 to 2.65) and of 365 or more days for people diagnosed with dementia (AOR 1.06; 95% CI 0.74 to 1.51). All of the disorders had higher AORs during the first 90 days after they were diagnosed. Only depression, anxiety, and dementia had significantly increased risk during the first year after diagnosis, compared with their risk after the first year.

Table 3.

Risk of suicide by mental disorder according to time since diagnosis

Mental disorder Time since diagnosis, days
1 to 90 AOR (95% CI) 91 to 364 AOR (95% CI) ≥365 AOR (95% CI)
Depression 7.33 (4.76 to 11.3) 3.17 (2.35 to 4.28) 2.91 (2.49 to 3.39)
Anxietya 7.18 (2.26 to 22.9) 1.33 (0.67 to 2.65) 1.60 (1.31 to 1.95)
Substance abuse 4.07 (2.43 to 6.82) 2.92 (1.99 to 4.28) 3.15 (2.59 to 3.82)
Schizophrenia 20.9 (2.55 to 172) 6.02 (2.41 to 15.1) 2.52 (1.73 to 3.68)
Dementia 2.17 (1.03 to 4.58) 3.14 (1.66 to 5.94) 1.06 (0.74 to 1.51)
Open in a new tab

Adjusted for Aggregated Diagnosis Group score and all mental disorders simultaneously. Suicide subjects and control subjects were matched 3:1 on age, sex, income decile, region of residence, and marital status.

a

There were no control subjects with a first anxiety diagnosis 1 to 91 days prior to case suicide; therefore, for anxiety only, the periods are defined as 1 to 180, 181 to 365 days prior

AOR = adjusted odds ratio

Interactions Between Mental Illness and Demographics

There were several significant interactions found in the regression analysis. Depression had a stronger association for people over the age of 50 (AOR 5.27; 95% CI 3.85 to 7.20, compared with AOR 3.61; 95% CI 3.05 to 4.26 for those less than 50; χ2 = 4.33; df = 1; P = 0.03). Depression was also more strongly linked with suicide for people at higher-income levels. AORs were 5.89 (95% CI 4.36 to 7.95), 4.89 (95% CI 3.65 to 6.53), and 2.84 (95% CI 2.31 to 3.50) for the high-, middle-, and low-income people, respectively. Anxiety disorders increased the risk of suicide only for females (male: AOR 1.15; 95% CI 0.90 to 1.48, female: AOR 2.32; 95% CI 1.65 to 3.27; χ2 = 10.74; df = 1; P = 0.001).

Risk of Attempted Suicide by Mental Disorder

The risk of suicide attempt by mental disorder is located in Table 4. All of the 6 categories examined had significantly higher adjusted rates among those with attempts. After adjusting for ADG score and all mental disorders, all of the disorder clusters were still significantly higher. Depression had the highest AOR (11.26; 95% CI 10.22 to 12.40), while dementia (1.44; 95% CI 1.14 to 1.80) and anxiety (1.48; 95% CI 1.31 to 1.67) had the lowest AORs.

Table 4.

Risk of suicide attempt by mental disorder

Mental disorder Rates of disorder within each group, %
AOR (95% CI)
Subjects n = 8641 Control subjects n = 25 923
Depression 68.9 13.9 11.3 (10.2 to 12.4)
Anxiety 25.0 5.2 1.48 (1.31 to 1.67)
Substance use 41.1 6.6 7.95 (7.15 to 8.83)
Schizophrenia 6.9 0.8 4.26 (3.32 to 5.47)
Dementia 7.8 1.0 1.44 (1.14 to 1.80)
Open in a new tab

Adjusted for Aggregated Diagnosis Group score and all mental disorders simultaneously. Suicide attempt subjects and control subjects were matched 3:1 on age, sex, income decile, region of residence, and marital status.

AOR = adjusted odds ratio

Suicide Risk Attempt According to Time Since Diagnosis

Table 5 depicts the risk of suicide attempt according to time since diagnosis by the 5 clusters of mental illness. The risk of having a suicide attempt was higher for all of the disorder categories at all time periods following their diagnosis, compared with the control subjects. During the first 90 days following a diagnosis, patients with depression and (or) anxiety were at the highest risk. The risk of suicide attempts was higher in the first year for subjects with dementia, substance use, and anxiety, compared with 365 or more days. Depression followed a U-shaped pattern, with increased risk during 1 to 90 days and 365 or more days, compared with 91 to 364 days. There was no significant difference in the odds of having a suicide attempt between the time periods for subjects with schizophrenia. The risk for anxiety disorders decreased the further the patients are from initial diagnosis from an AOR of 7.16 (95% CI 3.45 to 14.9) to 2.81 (95% CI 2.10 to 3.77) during 91 to 364 days, and then an AOR of 1.56 (95% CI 1.39 to 1.75) at 365 or more days.

Table 5.

Risk of suicide attempt by mental disorders according to time since diagnosis

Mental disorder Time since diagnosis, days
1 to 90 AOR (95% CI) 91 to 364 AOR (95% CI) ≥365 AOR (95% CI)
Depression 8.52 (6.83 to 10.6) 2.82 (2.42 to 3.28) 3.49 (3.20 to 3.82)
Anxiety 7.16 (3.45 to 14.9) 2.81 (2.10 to 3.77) 1.56 (1.39 to 1.75)
Substance use 3.81 (2.89 to 5.02) 2.29 (1.88 to 2.78) 2.68 (2.41 to 2.99)
Schizophrenia 5.79 (2.44 to 13.7) 4.28 (2.57 to 7.13) 3.08 (2.39 to 3.96)
Dementia 2.61 (1.36 to 4.98) 2.84 (1.85 to 4.37) 1.29 (1.01 to 1.64)
Open in a new tab

Adjusted for Aggregated Diagnosis Group score and all mental disorders simultaneously. Suicide subjects and control subjects were matched 3:1 on age, sex, income decile, region of residence, and marital status.

AOR = adjusted odds ratio

Interactions Between Mental Illness and Demographics

There were several significant interactions for the factors assessed for the risk of suicide attempts. Depression had a stronger association with suicide attempts for males than females (male: AOR 13.29; 95% CI 11.16 to 15.81, female: AOR 10.47; 95% CI 9.28 to 11.72; χ2 = 4.85; df = 1; P = 0.02). Depression had a higher AOR for subjects over the age of 50 (<50 years: AOR 10.83; 95% CI 9.80 to 11.97, >50 years: 18.94; 95% CI 12.72 to 28.21; χ2 = 7.12; df = 1; P = 0.008). The effect of depression was also higher among subjects at higher-income levels. The AORs were 8.45 (95% CI 7.46 to 9.57), 14.12 (95% CI 11.37 to 17.53), and 19.10 (95% CI 15.17 to 24.05) for the low-, middle-, and high-income subjects, respectively. Substance use had a stronger association for males (males: AOR 9.47; 95% CI 7.94 to 11.29, females: AOR 7.13; 95% CI 6.24 to 8.14; χ2 = 6.34; df = 1; P = 0.01). Schizophrenia had a stronger association with suicide attempts among the middle-class subjects, compared with those with lower incomes (AOR 9.71; 95% CI 4.94 to 19.09, compared with 3.45; 95% CI 2.58 to 4.61; χ2 = 7.60; df = 1; P = 0.01).

Discussion

The large population-based and propensity-matched sample used physician diagnoses, provided an excellent ability to control for risk factors, and had very good statistical power. With these methodological strengths, it yielded novel findings on the acute risk of suicide and suicidal behaviour within the first months following initial diagnosis of mental disorders. Our study illustrates that the risk for suicidal behaviour is high, not only within the first year of diagnosis but also, and especially, within the first 3 months after diagnosis. These results underscore the early period of mental illness as a major risk period for increased morbidity and mortality.

The risk for suicide and suicide attempts was higher for disorders in the first year of diagnosis, the first 90 days in particular for people with depression or anxiety. Substance abuse and schizophrenia were exceptions in the case of suicide. The risk for people with these disorders appears to remain elevated, regardless of how long it has been since diagnosis. These findings agree with previous research that has shown a time-related component to suicide.8–10,22 One explanation is that people are diagnosed as having a mental illness at an acute stage of illness. Therefore, the high risk in the initial period of diagnosis merely reflects that they are currently in an acute stage. The period of 365 or more days may contain more time where the person is less acute or even nonsymptomatic. Another explanation is that this could reveal poor adjustment to the diagnosis. It is possible that people with a recent diagnosis with a mental illness not only are currently experiencing a high level of distress but also may have an expectation that their status of living would remain poor. Their expectations may contribute to a feeling of hopelessness that compounds on their acute illness. Over time, with treatment, expectations may become more optimistic and risk of suicidal behaviour may decrease concurrently.

Using schizophrenia as an example, people with good premorbid functioning (for example, higher intelligence and educational achievement) or good insight into their illness have an increased risk of suicide following diagnosis, compared with those with lower premorbid function.23,24 It has been suggested that this is due to these patients perceiving a greater level of personal deterioration from having schizophrenia.23 It is possible that suicide may be considered to be an option for these patients, particularly early in their diagnosis. The increased AORs associated with schizophrenia in the first year were not significantly different, compared with the AORs for 365 or more days, could be explained by a lack of power. However, it is also possible that our study controlled for the secondary depression hypothesized in previous studies.25,26

Our study detected several interaction effects in the data. There was an interaction effect between sex and anxiety. Only females experience a higher risk of suicide owing to anxiety. This may explain some of the discordant findings27 on the link between anxiety and suicide. Without stratification or adjustment for sex, it is possible that the link between anxiety and suicide among females would be lost. This is due to most suicides occurring among males, males who do not experience increased risk owing to anxiety. This finding has clinical implications for the treatment of anxiety disorder among females. Another interesting interaction is that depression and substance use had higher AORs for people in the middle- and top-income brackets than those in the bottom. Depression, in particular, had a very strong interaction; the AOR for those in the top-income bracket was twice that of those in the low bracket. This finding appears to support previous data suggesting that the effect of psychiatric conditions may be most severe for people with higher socioeconomic status.28 Agerbo28 suggested that a rapid loss of status or social integration for people with higher status may be the cause of this additional risk. It is intuitive that people with more to lose react more negatively to social decline, whether this decline is actual or merely perceived. A similar interpretation would be that there is a risk saturation effect. Specifically, the more risk factors a person has, the less effect adding another risk factor has on their risk. This is supported by the income interaction findings, as well as the sex interaction findings. Females have increased risk from anxiety disorders for suicide, whereas males have increased AORs from some disorders (depression and substance use) for suicide attempts. The increased risk occurs for sex, which is at the lowest baseline risk; females are at lower risk for suicide, males at lower risk for attempts. Another possibility is that there a stress threshold for suicidal behaviour.

Strengths and Limitations

Our study used a population-based sample to detect all cases of suicide attempt and death in Manitoba. Propensity score matching identified 3 control subjects for every subject. These data provide excellent power, and the ability to reduce the confounding effect of demographic factors related to suicide. Importantly, our study controlled for all of the mental illnesses simultaneously to prevent the results from being confounded by comorbidity. The large population-based sample also provided excellent power for analysis. As it captures all confirmed suicides within the province of Manitoba, it should also be representative of the general population. Our study relies on physician-based rather than self-report data. The data are also not subject to recall bias.

The main limitations with these data result from the use of administrative data that were not collected with research purposes in mind. This means that we are limited by the diagnostic coding used by physicians and hospitals, such as coding major depressive disorder and bipolar disorder under the same code. Another limitation is the inability to adjust for factors that are not covered in the administrative data. Factors such as stressful life events, childhood adversity, and similar nonmedical factors could not be measured and adjusted for in the analysis. Other limitations of our study include its inability to adjust for the severity of the disorder, the treatments used by the patients, and whether patients adhered to the treatments they were prescribed by their physicians. Our study is also limited to including only people who receive medical care owing to a suicide attempt or whose death is classified as a suicide. This means that some cases will be missed. Additionally, not all people with mental disorders seek treatment, and these people would not be properly classified in this analysis.29 Any reduction in the risk of suicidal behaviour owing to treatment will reduce the effect size of the conditions. However, this analysis provides data that are reflective of the current risks with the current Manitoba health system. The failure to capture people who are mentally ill but are not diagnosed by physicians will likely lead to a reduction in the apparent effect size of any disorder that increases suicide risk. Using a propensity score to determine control subjects may not entirely control unobserved confounding factors that may be relevant.30 There are also differences between the time-related AORs and the nontime-specific AORs, owing to the inclusion of substance use diagnoses that might have been suicide attempts. Someone who attempted suicide using drugs might have received a substance use diagnosis in hospital that would be included in this diagnosis. As the time-related diagnosis covers the time prior to the outcome event, it would not be affected by this coding and this explains the lower AORs found for substance use in the time-related analysis. It is also possible that patients would be diagnosed with mental illness when they present with their suicide attempt (that is, the doctor assign a diagnosis partially due to the suicide attempt). However, most conditions were significant in the time-based risk analysis, where presentation diagnoses were not used. Therefore, any bias introduced by this coding is not causing the significant test results of the nontime-based analysis.

Acknowledgments

All authors report no competing interests.

Preparation of this article was supported by research grants from the Canadian Institute for Health Information (CIHR) to Mr Randall (agreement #01076–000), the University of Manitoba’s GETS program to Mr Randall, the Thorlakson Foundation Fund Award to Dr Bolton, a Manitoba Health Research Council Award to Dr Bolton, a Manitoba Health Research Council Chair Award to Dr Sareen, a CIHR– Public Health Agency of Canada Applied Public Health Chair Award to Dr Martens, and a CIHR New Investigator Award to Dr Bolton (#113589). The funding sources had no role in the design and conduct of the study; no role in the collection, management, analysis, and interpretation of data; and no role in the preparation, review, and approval of the manuscript.

The authors acknowledge Dan Chateau, PhD, for assistance with statistical analysis.

Abbreviations

ADG

Aggregated Diagnosis Group

AOR

adjusted odds ratio

ICD-9-CM

International Statistical Classification of Diseases and Related Health Problems, Ninth Revision, Clinical Modification

ICD-10-CA

International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Canada

KSa

asymptotic Kolmogorov–Smirnov statistic

References

  • 1.Windfuhr K, Kapur N. Suicide and mental illness: a clinical review of 15 years findings from the UK National Confidential Inquiry into Suicide. Br Med Bull. 2011;100:101–121. doi: 10.1093/bmb/ldr042. [DOI] [PubMed] [Google Scholar]
  • 2.Hawton K, van Heeringen K. Suicide. Lancet. 2009;373(9672):1372–1381. doi: 10.1016/S0140-6736(09)60372-X. [DOI] [PubMed] [Google Scholar]
  • 3.Bolton JM, Robinson J. Population-attributable fractions of Axis I and Axis II mental disorders for suicide attempts: findings from a representative sample of the adult, noninstitutionalized US population. Am J Public Health. 2010;100(12):2473–2480. doi: 10.2105/AJPH.2010.192252. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Cavanagh JTO, Carson AJ, Sharpe M, et al. Psychological autopsy studies of suicide: a systematic review. Psychol Med. 2003;33(3):395–405. doi: 10.1017/S0033291702006943. [DOI] [PubMed] [Google Scholar]
  • 5.Nordentoft M, Mortensen PB, Pedersen CB. Absolute risk of suicide after first hospital contact in mental disorder. Arch Gen Psychiatry. 2011;68(10):1058–1064. doi: 10.1001/archgenpsychiatry.2011.113. [DOI] [PubMed] [Google Scholar]
  • 6.Harris EC, Barraclough B. Suicide as an outcome for mental disorders. A meta-analysis. Br J Psychiatry. 1997;170(3):205–228. doi: 10.1192/bjp.170.3.205. [DOI] [PubMed] [Google Scholar]
  • 7.Tidemalm D, Långström N, Lichtenstein P, et al. Risk of suicide after suicide attempt according to coexisting psychiatric disorder: Swedish cohort study with long term follow-up. BMJ. 2008;337:a2205. doi: 10.1136/bmj.a2205. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Palmer BA, Pankratz VS, Bostwick JM. The lifetime risk of suicide in schizophrenia: a reexamination. Arch Gen Psychiatry. 2005;62(3):247–253. doi: 10.1001/archpsyc.62.3.247. [DOI] [PubMed] [Google Scholar]
  • 9.Rossau CD, Mortensen PB. Risk factors for suicide in patients with schizophrenia: nested case–control study. Br J Psychiatry. 1997;171(4):355–359. doi: 10.1192/bjp.171.4.355. [DOI] [PubMed] [Google Scholar]
  • 10.Malone KM, Haas GL, Sweeney JA, et al. Major depression and the risk of attempted suicide. J Affect Disord. 1995;34(3):173–185. doi: 10.1016/0165-0327(95)00015-f. Also available from: http://www.ncbi.nlm.nih.gov/pubmed/7560545. [DOI] [PubMed] [Google Scholar]
  • 11.De Moore GM, Robertson AR. Suicide in the 18 years after deliberate self-harm a prospective study. Br J Psychiatry. 1996;169(4):489–494. doi: 10.1192/bjp.169.4.489. [DOI] [PubMed] [Google Scholar]
  • 12.Canadian Institute for Health Information (CIHI) The Canadian Enhancement of ICD-10. Ottawa (ON): CIHI; 2001. [Google Scholar]
  • 13.National Center for Health Statistics . International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Hyattsville (MD): National Center for Health Statistics; 2005. [Google Scholar]
  • 14.Cleophas TJ, Zwinderman AH. Statistics applied to clinical studies. Dordrecht (NL): Springer Netherlands; 2012. [DOI] [Google Scholar]
  • 15.Parsons LS. Twenty-Ninth Annual SAS Users Group International Conference [Internet] Cary (NC): SAS Institute; 2004. Performing a 1:N case–control match on propensity score. [cited 2014 Feb 2]. Available from: http://www2.sas.com/proceedings/sugi29/165-29.pdf. [Google Scholar]
  • 16.Manitoba Centre for Health Policy . Concept: mental health disorder / mental health illness classification [Internet] Winnipeg (MB): Manitoba Centre for Health Policy; 2012. [cited 2012 Nov 7]. Available from: http://mchp-appserv.cpe.umanitoba.ca/viewConcept.php?conceptID=1182. [Google Scholar]
  • 17.Manitoba Centre for Health Policy . Concept dictionary and glossary for health services research [Internet] Winnipeg (MB): Manitoba Centre for Health Policy; 2012. [cited 2012 Nov 7]. Available from: http://umanitoba.ca/faculties/medicine/units/mchp/resources/concept_dictionary.html. [Google Scholar]
  • 18.Fransoo R, Martens P, Burland E, et al. Manitoba RHA indicators atlas 2009. Winnipeg (MB): Manitoba Centre for Health Policy; 2009. p. 628. Also available from: http://mchp-appserv.cpe.umanitoba.ca/reference/RHA_Atlas_Report.pdf. [Google Scholar]
  • 19.Austin PC, van Walraven C. The mortality risk score and the ADG score: two points-based scoring systems for the Johns Hopkins aggregated diagnosis groups to predict mortality in a general adult population cohort in Ontario, Canada. Med Care. 2011;49(10):940–947. doi: 10.1097/MLR.0b013e318229360e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Weiner JP, Abrams C, editors. The Johns Hopkins ACG system: technical reference guide. Baltimore (MD): Johns Hopkins University; 2009. [Google Scholar]
  • 21.SAS Institute Inc. SAS version 92. Cary (NC): SAS Institute Inc; 2012. [Google Scholar]
  • 22.Fawcett J, Scheftner WA, Fogg L, et al. Time-related predictors of suicide in major affective disorder. Am J Psychiatry. 1990;147(9):1189–1194. doi: 10.1176/ajp.147.9.1189. Also available from: http://www.ncbi.nlm.nih.gov/pubmed/2104515. [DOI] [PubMed] [Google Scholar]
  • 23.Carlborg A, Winnerbäck K, Jönsson EG, et al. Suicide in schizophrenia. Expert Rev Neurother. 2010;10(7):1153–1164. doi: 10.1586/ern.10.82. [DOI] [PubMed] [Google Scholar]
  • 24.Crumlish N, Whitty P, Kamali M, et al. Early insight predicts depression and attempted suicide after 4 years in first-episode schizophrenia and schizophreniform disorder. Acta Psychiatr Scand. 2005;112(6):449–455. doi: 10.1111/j.1600-0447.2005.00620.x. [DOI] [PubMed] [Google Scholar]
  • 25.Drake RJ, Pickels A, Bentall RP, et al. The evolution of insight, paranoia and depression during early schizophrenia. Psychol Med. 2004;34(2):285–292. doi: 10.1017/S0033291703008821. [DOI] [PubMed] [Google Scholar]
  • 26.Addington D, Addington J, Patten S. Depression in people with first-episode schizophrenia. Br J Psychiatry Suppl. 1998;172(33):90–92. Also available from: http://www.ncbi.nlm.nih.gov/pubmed/9764133. [PubMed] [Google Scholar]
  • 27.Sareen J. Anxiety disorders and risk for suicide: why such controversy? Depress Anxiety. 2011;28(11):941–945. doi: 10.1002/da.20906. [DOI] [PubMed] [Google Scholar]
  • 28.Agerbo E. High income, employment, postgraduate education, and marriage: a suicidal cocktail among psychiatric patients. Arch Gen Psychiatry. 2007;64(12):1377–1384. doi: 10.1001/archpsyc.64.12.1377. [DOI] [PubMed] [Google Scholar]
  • 29.Sareen J, Cox BJ, Afifi TO, et al. Mental health service use in a nationally representative Canadian survey. Can J Psychiatry. 2005;50(12):753. doi: 10.1177/070674370505001204. [DOI] [PubMed] [Google Scholar]
  • 30.Joffe MM, Rosenbaum PR. Invited commentary: propensity scores. Am J Epidemiol. 1999;150(4):327–333. doi: 10.1093/oxfordjournals.aje.a010011. [DOI] [PubMed] [Google Scholar]

Articles from Canadian Journal of Psychiatry. Revue Canadienne de Psychiatrie are provided here courtesy of SAGE Publications

RESOURCES