The aetiology of suicide is complex and may clinically be classified as continuum that ranges from suicidal ideation and suicide plans towards (non-violent or violent) attempted and completed suicide (for review see Reference Maris). Overall, suicide has a strong association with mental disorders Reference Harris and Barraclough, particularly with schizophrenia and affective disorders [22,26]. However, despite considerable efforts in research and progresses in drug treatment over the last decades, suicide rates have not dropped significantly in patients with schizophrenia or affective disorders [15,21]. Patients suffering from such diagnoses seem to be at increased risk for suicide in the presence of an early-disturbed psychosocial adjustment Reference Modestin, Zarro and Waldvogel and distinct demographic characteristics (e.g. male gender, [6–Reference Marusic, Tavcar, Dernovsek and Steblaj14]), while one of the most potent risk factor for inpatients is given by a history of previous suicide attempts [25,20]. Protective factors include being young, non-white, female and pregnancy or motherhood [12,1,9], despite the risk of postpartum depression and psychosis. Inclusion of demographic factors such as age proved to increase the predictive ability of suicide risk assessment scales (e.g. ‘Suicide Assessment Scale’; SUAS) in suicide attempters Reference Nimeus, Alsen and Traskman-Bendz. However, demographical and clinical risk factors may differ importantly between genders as evidenced for example by a study from Qin et al. Reference Qin, Agerbo, Westergard-Nielsen, Eriksson and Mortensen, who performed joint and separate analyses with demographical (or socioeconomic) and clinical (or health) characteristics in a population based sample including patients who committed suicide compared to controls. For both genders, a history of hospitalised mental illness was the most marked risk factor. Although the authors concluded that the observed gender differences could not be explained by differential exposure to demographical and clinical risk factors alone, analyses revealed that suicide risk was reduced in females when being a parent of a child less than 2 years old. In males, the relationship between socioeconomic factors (e.g. being single, unemployment etc.) and suicide was more pronounced than in females.
In light of these findings, we analysed the association between suicide attempts and specific demographical and clinical characteristics in male and female individuals exclusively affected with schizophrenic (including schizoaffective) or affective disorders. Since all of our included study subjects belonged to a high-risk group for suicide, we were interested which risk factor may particularly be associated with suicide attempts and if specific factors may have a stronger impact on one of both genders.
2. Subjects and methods
2.1. Study population
Patients with a diagnosis of schizophrenia or affective disorder admitted to the psychiatric departments at the University of Bonn or the Mental State Hospital (Rheinische Kliniken Bonn) were routinely asked for participation in our study program as were patients' relatives. Written informed consent was obtained from all participants prior to examination. Thus, we were able to perform face-to-face interviews on 500 individuals personally.
2.2. Clinical assessments and data collection
Lifetime consensus diagnoses according to DSM-IV criteria  were given to each individual and carried out by trained psychiatrists on the basis of a multi-dimensional phenotype characterization inventory including a personal structured interview (SADS-L) Reference Endicott and Spitzer, family history method (FISC) Reference Mannuzza, Fyer, Endicott and Klein, OPCRIT documentation Reference McGuffin, Farmer and Harvey and a systematic review of medical records. Patient relatives were interviewed using the same methods. Out of this sample patients were grouped into those who had (i.e. case) and those who had not attempted suicide (i.e controls). Lifetime occurrence of suicide attempt was defined as a potential life threatening act intended to result in death.
We analysed demographical variables (age, gender, children status and educational level) and clinical variables with respect to diagnosis, course and severity of the disorder (DSM-IV diagnosis, comorbidity, lowest level of the ‘Global Assessment Scale’ (GAS) Reference Endicott, Spitzer, Fleiss and Cohen, age at onset, premorbid social adjustment, and lack of insight) for association with suicide attempts.
The level of education was divided into two classes according to the common school system in Germany, those with up to 10 years of scholastic education, and those with more. Comorbidity refers to all other psychiatric diagnoses other than when the main diagnosis was schizophrenia (DSM-IV: 295.XX), affective disorder (major depression or bipolar disorder, DSM-IV: 296.XX). In those patients suffering from both schizophrenia and affective disorders, schizophrenia was considered to be the main diagnosis and affective disorder to be the comorbid disorder. The lowest lifetime score on the GAS was chosen to reflect the highest impairment that was clinically attributable to the main disorder. Age at onset of psychiatric disease was defined as the age when the criteria of the main psychiatric disorder were fulfilled for the first time. The variables ‘poor premorbid social adjustment’ and ‘lack of insight’ were based on the criteria applied in the OPCRIT checklist.
2.3. Statistical analysis
Each variable was tested for association with suicide attempt(s) using univariate analyses (χ 2-tests for categorical and t-tests for continuous variables) and by calculating odds ratios. Subsequently, a logistic regression analysis was performed which included all variables, i.e., no successive inclusion or exclusion procedure took place. An inclusive analysis was chosen in order to evaluate the combined effects of all variables. Finally, this analysis was separately performed for both genders. For all analyses the significance level was set at 2-sided P < 0.05. Data were analysed using the ‘Statistical Package for the Social Sciences’ (SPSS).
The sample comprised 500 individuals (277 females and 223 males), 170 with schizophrenia, 315 with affective disorder (109 with bipolar disorder, 206 with depressive disorders) while 15 had depressive or psychotic symptoms in the context of other psychiatric disorders without reaching thresholds for full-blown episodes of major depression or schizophrenia. Out of this sample, 147 individuals had a lifetime history of attempted suicide (i.e. cases) and 353 individuals had not (i.e. controls). Results of our analyses are shown in Table 1. Comorbidity, lower age at interview and poor social premorbid adjustment were significantly associated with suicide attempts in univariate analyses (P < 0.05). However, we did not perform correction for multiple testing. The strongest associations were found for the lowest lifetime GAS scores and for age at onset of the main psychiatric diagnosis that were significantly higher in controls than in suicide attempters (P < 0.001). In a next step, variables were adjusted for each other performing logistic regression analysis. Odds ratios for the variables ‘comorbidity’, ‘lowest level of GAS’, ‘age at onset of the main psychiatric diagnosis’, and ‘poor premorbid social adjustment’ remained significant, while odds ratio for ‘age at interview’ became insignificant.
a Cases (or controls): individuals with (or without) a lifetime history of attempted suicide.
b Results of logistic regression analysis including all variables.
c Results of logistic regression analysis including all variables for males.
d Results of logistic regression analysis including all variables for females.
e Reference category.
f Mean (standard deviation).
g t (df, P), * P < 0.05 (Wald's χ 2), ** P < 0.01 (Wald's χ 2).
h OR refers to the increased risk when the continuous variable increases by one unit.
Finally, logistic regression analysis was performed separately for both genders. Males and females were found to be concordant in that lower GAS scores were associated with higher rates of suicide attempts although the odds ratios suggest small effects. Males with schizophrenic disorders were less likely to attempt suicide than males with affective or other disorders. In females, lower age at onset was significantly related to suicide attempts. Additionally, females were more likely to attempt suicide if they were childless.
4. Discussion and conclusion
Preventing suicide remains a difficult task in clinical practice. The main purpose of this study was to analyse demographical and clinical characteristics and their relation to suicide attempts in a large German sample of individuals affected with schizophrenia and/or affective disorders. Since risk (or protective) factors are known to differ between males and females, we performed additional gender specific analyses. The selection of health and socioeconomic indicators and statistical methods were mainly based on a previous study that included gender specific analyses (Qin et al. Reference Qin, Agerbo, Westergard-Nielsen, Eriksson and Mortensen), although we did include some other characteristics in addition (e.g., lack of insight). The major difference, however, is given by the fact that Qin et al. Reference Qin, Agerbo, Westergard-Nielsen, Eriksson and Mortensen performed their analyses in suicide completers using Danish longitudinal register databases while our study comprised suicide attempters and was mainly based on an inpatient sample.
In the total sample, individuals with a history of suicide attempts were more often affected by more than one psychiatric disorder (i.e. comorbidity), low scores on the Global Assessment Scale (GAS), low age at onset, and poor premorbid social adjustment. Overall, these characteristics are indicative for more severe and chronic forms of psychiatric disorders and these results are consistent with previous findings [12,17,6,20,10]. Analyses stratified by gender revealed that males affected with schizophrenia were less likely to attempt suicide when compared to males with diagnoses other than schizophrenia. However, we cannot exclude that this finding might be due to an ascertainment bias in our sample. Schizophrenic males have been reported to be at particular risk for completed suicide Reference De Hert, McKenzie and Peuskens and this particular group may be under represented in our sample of suicide attempters. In females, lower age at onset was a risk factor not seen in males, however, the strongest factor associated with suicide attempt(s) in females was childlessness. Despite methodological differences between both studies, our finding is consistent with that of Qin et al. Reference Qin, Agerbo, Westergard-Nielsen, Eriksson and Mortensen who reported that having a child less than 2 years is a protective factor against suicide in females.
One important limitation of our study is that the assessment of variables was conducted for lifetime occurrence and that a direct or causal relationship between suicide attempt and some specific variables (e.g., having children, lack of insight) may not have been present in all cases. Moreover, as GAS scores include suicidality, patients with a history of attempted suicide will be given lower GAS scores and thus our significant association between lower GAS scores and suicide attempts may not allow concluding that lower GAS scores are risk factors for suicide attempts.
In conclusion, we here report that in our sample suicide attempts were associated with poor premorbid social adjustment, low age at onset, low GAS scores and childlessness in females. Odds ratios, albeit reaching statistical significance, were predominantly small in our study. In other words, the effect sizes (or the predictive value) of clinical demographical characteristics per se are likely to be of little relevance in clinical practice. This is in accordance with previous findings Reference Busch, Fawcett and Jacobs, however, demographical and clinical risk factors should not be ignored but need to be applied in addition and in context with standard clinical procedures Reference Morgan and Stanton. Only then may the consideration of characteristics associated with suicide be beneficial and help reducing the excess mortality in the mentally ill. However, future studies are needed that are designed to focus on demographical and clinical characteristics in suicide attempters and completers.
DJM is the recipient of a postdoctoral fellowship award supported by the Canadian Institutes of Health Research (CIHR). Thanks to Matthew Lanktree for assistance in preparation of this manuscript.