An integrated framework for suicide risk prediction

Suicide is a major concern in society. Despite of great attention paid by the community with very substantive medico-legal implications, there has been no satisfying method that can reliably predict the future attempted or completed suicide. We present an integrated machine learning framework to tackle this challenge. Our proposed framework consists of a novel feature extraction scheme, an embedded feature selection process, a set of risk classifiers and finally, a risk calibration procedure. For temporal feature extraction, we cast the patient’s clinical history into a temporal image to which a bank of one-side filters are applied. The responses are then partly transformed into mid-level features and then selected in 1-norm framework under the extreme value theory. A set of probabilistic ordinal risk classifiers are then applied to compute the risk probabilities and further re-rank the features. Finally, the predicted risks are calibrated. Together with our Australian partner, we perform comprehensive study on data collected for the mental health cohort, and the experiments validate that our proposed framework outperforms risk assessment instruments by medical practitioners.

HealthMap: a visual platform for patient suicide risk review

Misjudging suicide risk can be fatal. Risk assessment is complicated by multiplicity of risk factors, none of which individually can reliably predict risk. This paper addresses the need for better clinical support, visualising risk factors scattered in raw electronic medical records. HealthMap is a visual tool that helps clinicians effectively examine patient histories during a suicide risk assessment. We characterise the information visualisation problems accompanying suicide risk assessments. A design driven by visualisation principles was implemented. The prototype was evaluated by clinicians and accepted into daily clinical work-flow.

Does gender explain the relationship between occupation and suicide? Findings from a meta-analytic study

This report investigated whether suicide risk by occupational groups differed for males and females. We examined this using a sub-set of articles examined in a recent meta-analysis and stratified by gender. For certain occupational groups, males and females had a similar risk of suicide (the military, community service occupations, managers, and clerical workers). There was some indication of gender differences for other occupations (technicians, plant and machine operators and ship’s deck crew, craft and related trades workers, and professionals), although these did not reach statistical significance. These findings highlight the complexity of the relationship between occupation and suicide and suggest the possible role of a range of individual, work-related and social-environmental risk factors that may differ for males and females.

Suicide in veterinarians and veterinary nurses in Australia: 2001-2012

BACKGROUND: Whether veterinarians have an elevated suicide rate compared with the general population is controversial. METHODS: Reported cases of suicide among veterinarians and veterinary nurses in Australia over the period 2001 to 2012 were investigated in a retrospective case-series study. RESULTS: The standardised mortality ratio of veterinarians (n = 18) was 1.92 (95% CI 1.14-3.03) and that of veterinary nurses (n = 7) to the general population was 1.24 (95% CI 0.80-1.85). Overdosing on drugs (pentobarbitone) was the main method of suicide in these occupations. CONCLUSION: The reasons for veterinary suicides are likely to be multifactorial, including work- and life-related stressors, and individual characteristics. This research highlights the need for targeted suicide prevention and intervention for veterinarians.

Occupational class differences in suicide: evidence of changes over time and during the global financial crisis in Australia

BACKGROUND: Previous research showed an increase in Australian suicide rates during the Global Financial Crisis (GFC). There has been no research investigating whether suicide rates by occupational class changed during the GFC. The aim of this study was to investigate whether the GFC-associated increase in suicide rates in employed Australians may have masked changes by occupational class.

METHODS: Negative binomial regression models were used to investigate Rate Ratios (RRs) in suicide by occupational class. Years of the GFC (2007, 2008, 2009) were compared to the baseline years 2001-2006.

RESULTS: There were widening disparities between a number of the lower class occupations and the highest class occupations during the years 2007, 2008, and 2009 for males, but less evidence of differences for females.

CONCLUSIONS: Occupational disparities in suicide rates widened over the GFC period. There is a need for programs to be responsive to economic downturns, and to prioritise the occupational groups most affected.

The relationship between suicide and violence in schizophrenia: analysis of the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) dataset

Background

Suicide and violence often co-occur in the general population as well as in mentally ill individuals. Few studies, however, have assessed whether these suicidal behaviors are predictive of violence risk in mental illness.

Aims

The aim of this study is to investigate whether suicidal behaviors, including suicidal ideation, threats, and attempts, are significantly associated with increased violence risk in individuals with schizophrenia.

Method

Data for these analyses were obtained from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) trial, a randomized controlled trial of antipsychotic medication in 1460 adults with schizophrenia. Univariate Cox regression analyses were used to calculate hazard ratios (HRs) for suicidal ideation, threats, and attempts. Multivariate analyses were conducted to adjust for common confounding factors, including: age, alcohol or drug misuse, major depression, antisocial personality disorder, depression, hostility, positive symptom, and poor impulse control scores. Tests of discrimination, calibration, and reclassification assessed the incremental predictive validity of suicidal behaviors for the prediction of violence risk.

Results

Suicidal threats and attempts were significantly associated with violence in both males and females with schizophrenia with little change following adjustment for common confounders. Only suicidal threats, however, were associated with a significant increase in incremental validity beyond age, diagnosis with a comorbid substance use disorder, and recent violent behavior.

Conclusions

Suicidal threats are independently associated with violence risk in both males and females with schizophrenia, and may improve violence risk prediction.

Letters, green cards, telephone calls and postcards: systematic and meta-analytic review of brief contact interventions for reducing self-harm, suicide attempts and suicide

BACKGROUND: There is growing interest in brief contact interventions for self-harm and suicide attempt. AIMS: To synthesise the evidence regarding the effectiveness of brief contact interventions for reducing self-harm, suicide attempt and suicide. METHOD: A systematic review and random-effects meta-analyses were conducted of randomised controlled trials using brief contact interventions (telephone contacts; emergency or crisis cards; and postcard or letter contacts). Several sensitivity analyses were conducted to examine study quality and subgroup effects. RESULTS: We found 14 eligible studies overall, of which 12 were amenable to meta-analyses. For any subsequent episode of self-harm or suicide attempt, there was a non-significant reduction in the overall pooled odds ratio (OR) of 0.87 (95% CI 0.74-1.04, P = 0119) for intervention compared with control. The number of repetitions per person was significantly reduced in intervention v. control (incidence rate ratio IRR = 066, 95% CI 0.54-0.80, P<0001). There was no significant reduction in the odds of suicide in intervention compared with control (OR = 0.58, 95% CI 0.24-1.38). CONCLUSIONS: A non-significant positive effect on repeated self-harm, suicide attempt and suicide and a significant effect on the number of episodes of repeated self-harm or suicide attempts per person (based on only three studies) means that brief contact interventions cannot yet be recommended for widespread clinical implementation. We recommend further assessment of possible benefits in well-designed trials in clinical populations.

The impact of a suicide prevention strategy on reducing the economic cost of suicide in the New South Wales construction industry

BACKGROUND: Little research has been conducted into the cost and prevention of self-harm in the workplace. AIMS: To quantify the economic cost of self-harm and suicide among New South Wales (NSW) construction industry (CI) workers and to examine the potential economic impact of implementing Mates in Construction (MIC). METHOD: Direct and indirect costs were estimated. Effectiveness was measured using the relative risk ratio (RRR). In Queensland (QLD), relative suicide risks were estimated for 5-year periods before and after the commencement of MIC. For NSW, the difference between the expected (i.e., using NSW pre-MIC [2008-2012] suicide risk) and counterfactual suicide cases (i.e., applying QLD RRR) provided an estimate of potential suicide cases averted in the post-MIC period (2013-2017). Results were adjusted using the average uptake (i.e., 9.4%) of MIC activities in QLD. Economic savings from averted cases were compared with the cost of implementing MIC. RESULTS: The cost of self-harm and suicide in the NSW CI was AU $527 million in 2010. MIC could potentially avert 0.4 suicides, 1.01 full incapacity cases, and 4.92 short absences, generating annual savings of AU $3.66 million. For every AU $1 invested, the economic return is approximately AU $4.6. CONCLUSION: MIC represents a positive economic investment in workplace safety.

An investigation of neighborhood-level social, economic and physical factors for railway suicide in Victoria, Australia

BACKGROUND: This study investigates the associations between railway suicide and neighborhood social, economic, and physical determinants using postcode-level data. It also examines whether the associations are influenced by having high concentration of high-risk individuals in a neighborhood area. METHODS: Railway suicide cases from Victoria, Australia for the period of 2001-2012, their age, sex, year of death, usual residential address and suicide location were obtained from the National Coronial Information System. Univariate negative binomial regression models were used to estimate the association between railway suicide and neighborhood-level social, economic and physical factors. Variables which were significant in these univariate models were then assessed in a multivariate model, controlling for age and sex of the deceased and other known confounders. RESULTS: Findings from the multivariate analysis indicate that an elevated rate of railway suicide was strongly associated with neighborhood exposure of higher number of railway stations (IRR=1.30 95% CI=1.16-1.46). Other significant neighborhood risk factors included patronage volume (IRR=1.06, 95% CI=1.02-1.11) and train frequency (IRR=1.02, 95% CI=1.01-1.04). An increased number of video surveillance systems at railway stations and carparks was significantly associated with a modest reduction in railway suicide risk (IRR=0.93, 95% CI=0.88-0.98). These associations were independent of concentration of high-risk individuals. LIMITATIONS: Railway suicide may be under-reported in Australia. CONCLUSIONS: Interventions to prevent railway suicide should target vulnerable individuals residing in areas characterized by high station density, patronage volume and train frequency.

Specific mood symptoms confer risk for subsequent suicidal ideation in bipolar disorder with and without suicide attempt history: multi-wave data from STEP-BD

BACKGROUND: Little is known about specific mood symptoms that may confer risk for suicidal ideation (SI) among patients with bipolar disorder (BD). We evaluated prospectively whether particular symptoms of depression and mania precede the onset or worsening of SI, among adults with or without a history of a suicide attempt. METHODS: We examined prospective data from a large (N = 2,741) cohort of patients participating in the Systematic Treatment Enhancement Program for BD (STEP-BD). We evaluated history of suicide attempts at baseline, and symptoms of depression and mania at baseline and follow-up visits. Hierarchical linear modeling tested whether specific mood symptoms predicted subsequent levels of SI, and whether the strength of the associations differed based on suicide attempt history, after accounting for the influence of other mood symptoms and current SI. RESULTS: Beyond overall current depression and mania symptom severity, baseline SI, and illness characteristics, several mood symptoms, including guilt, reduced self-esteem, psychomotor retardation and agitation, increases in appetite, and distractibility predicted more severe levels of subsequent SI. Problems with concentration, distraction, sleep loss and decreased need for sleep predicted subsequent SI more strongly among individuals with a suicide attempt history. CONCLUSIONS: Several specific mood symptoms may confer risk for the onset or worsening of SI among treatment-seeking patients with BD. Individuals with a previous suicide attempt may be at greater risk in part due to greater reactivity to certain mood symptoms in the form of SI. However, overall, effect sizes were small, suggesting the need to identify additional proximal predictors of SI.

Perception of mattering and suicide ideation in the Australian working population: evidence from a cross-sectional survey

Thoughts about suicide are a risk factor for suicide deaths and attempts and are associated with a range of mental health outcomes. While there is considerable knowledge about risk factors for suicide ideation, there is little known about protective factors. The current study sought to understand the role of perceived mattering to others as a protective factor for suicide in a working sample of Australians using a cross-sectional research design. Logistic regression analysis indicated that people with a higher perception that they mattered had lower odds of suicide ideation than those with lower reported mattering, after controlling for psychological distress, demographic and relationship variables. These results indicate the importance of further research and intervention studies on mattering as a lever for reducing suicidality. Understanding more about protective factors for suicide ideation is important as this may prevent future adverse mental health and behavioural outcomes.

Psychosocial working conditions and suicide ideation: evidence from a cross-sectional survey of working Australians

OBJECTIVES: This study examined the relationship between psychosocial working factors such as job control, job demands, job insecurity, supervisor support, and workplace bullying as risk factors for suicide ideation. METHODS: We used a logistic analytic approach to assess risk factors for thoughts of suicide in a cross-sectional sample of working Australians. Potential predictors included psychosocial job stressors (described above); we also controlled for age, gender, occupational skill level, and psychological distress. RESULTS: We found that workplace bullying or harassment was associated with 1.54 greater odds of suicide ideation (95% confidence interval 1.64 to 2.05) in the model including psychological distress. Results also suggest that higher job control and security were associated with lower odds of suicide ideation. CONCLUSIONS: These results suggest the need for organizational level intervention to address psychosocial job stressors, including bullying.

Suicide among male road and rail drivers in Australia: A retrospective mortality study

Objectives: This paper aims to describe the epidemiology of suicide among males employed in driving occupations (road and rail) compared to other male occupations in Australia. Methods: Suicide cases among road and rail drivers were extracted from a national dataset of occupationally coded suicide cases for the period 2001 to 2010. Suicide rates per 100 000 were calculated and standardised using the Australian standard population (2001). Incidence rate ratios (IRR) with 95% confidence intervals were calculated using Mantell Haenszel rates and compared to all employed suicide cases. Results: The majority of suicides in this occupational category occurred in truck drivers, followed by road and rail drivers. 98% of these suicides were among males; hence only males were included in further analyses. The age-standardised rate of male suicide among Road and Rail drivers over the period 2001 to 2010 was 22.6 per 100 000 (95% CI 19.2 to 25.9). The IRR of suicide in this occupational group compared to other male occupations was 1.42 (95% CI 1.26 to1.60). Conclusions: Suicide among Road and Rail drivers is higher than in the other male occupations. Suicide prevention initiatives addressing these risk factors, while also providing access to treatment for those at risk, are clearly needed.

Clusters of suicides and suicide attempts: detection, proximity and correlates.

BACKGROUND: A suicide cluster is defined as a higher number of observed cases occurring in space and/or time than would typically be expected. Previous research has largely focused on identifying clusters of suicides, while there has been comparatively limited research on clusters of suicide attempts. We sought to identify clusters of both types of behaviour, and having done that, identify the factors that distinguish suicide attempts inside a cluster from those that were outside a cluster. METHODS: We used data from Western Australia from 2000 to 2011. We defined suicide attempts as admissions to hospital for deliberate self-harm and suicides as deaths due to deliberate self-harm. Using an analytic strategy that accounted for the repetition of attempted suicide within a cluster, we performed spatial-temporal analysis using Poisson discrete scan statistics to detect clusters of suicide attempts and clusters of suicides. Logistic regression was then used to compare clustered attempts with non-clustered attempts to identify risk factors for an attempt being in a cluster. RESULTS: We detected 350 (1%) suicide attempts occurring within seven spatial-temporal clusters and 12 (0.6%) suicides occurring within two spatial-temporal clusters. Both of the suicide clusters were located within a larger but later suicide attempt cluster. In multivariate analysis, suicide attempts by individuals who lived in areas of low socioeconomic status had higher odds of being in a cluster than those living in areas of high socioeconomic status [odds ratio (OR) = 29.1, 95% confidence interval (CI) = 6.3-135.5]. A one percentage-point increase in the proportion of people who had changed address in the last year was associated with a 60% increase in the odds of the attempt being within a cluster (OR = 1.60, 95% CI = 1.29-1.98) and a one percentage-point increase in the proportion of Indigenous people in the area was associated with a 7% increase in the suicide being within a cluster (OR = 1.07, 95% CI = 1.00-1.13). Age, sex, marital status, employment status, method of harm, remoteness, percentage of people in rented accommodation and percentage of unmarried people were not associated with the odds of being in a suicide attempt cluster. CONCLUSIONS: Early identification of and responding to suicide clusters may reduce the likelihood of subsequent clusters forming. The mechanisms, however, that underlie clusters forming is poorly understood.

Predicting risk of suicide attempt using history of physical illnesses from electronic medical records

BACKGROUND: Although physical illnesses, routinely documented in electronic medical records (EMR), have been found to be a contributing factor to suicides, no automated systems use this information to predict suicide risk.

OBJECTIVE: The aim of this study is to quantify the impact of physical illnesses on suicide risk, and develop a predictive model that captures this relationship using EMR data.

METHODS: We used history of physical illnesses (except chapter V: Mental and behavioral disorders) from EMR data over different time-periods to build a lookup table that contains the probability of suicide risk for each chapter of the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) codes. The lookup table was then used to predict the probability of suicide risk for any new assessment. Based on the different lengths of history of physical illnesses, we developed six different models to predict suicide risk. We tested the performance of developed models to predict 90-day risk using historical data over differing time-periods ranging from 3 to 48 months. A total of 16,858 assessments from 7399 mental health patients with at least one risk assessment was used for the validation of the developed model. The performance was measured using area under the receiver operating characteristic curve (AUC).

RESULTS: The best predictive results were derived (AUC=0.71) using combined data across all time-periods, which significantly outperformed the clinical baseline derived from routine risk assessment (AUC=0.56). The proposed approach thus shows potential to be incorporated in the broader risk assessment processes used by clinicians.

CONCLUSIONS: This study provides a novel approach to exploit the history of physical illnesses extracted from EMR (ICD-10 codes without chapter V-mental and behavioral disorders) to predict suicide risk, and this model outperforms existing clinical assessments of suicide risk.