Effectiveness of Violence Prevention Interventions: Umbrella Review of Research in the General Population.

To address the societal harms of violence, many violence prevention interventions have been developed, tested, and implemented in the general population. These have been reported in systematic reviews and meta-analyses, which have typically focused on one type of intervention or outcome. We aimed to provide a comprehensive overview of the current evidence regarding the effectiveness of different psychosocial interventions in reducing all forms of violence toward others. We have conducted an umbrella review of previous meta-analyses using standard approaches and converted findings on effectiveness into odds ratios. We tested for the underlying quality of the meta-analytic evidence by examining heterogeneity, excess statistical significance, prediction intervals, and small study effects. We identified 16 meta-analyses, including nine investigating psychosocial interventions, and five legislative and policy changes. Most meta-analyses reported positive effects of tested interventions. The strongest effects were found for sports-based initiatives, and the weakest for general population programs aimed at early childhood, youth development, and reducing sexual assault perpetration by men. Legislative changes had varying effectiveness. We conclude that simple, scalable, and cost-efficient programs, such as sport-based initiatives, have the clearest empirical support as population-based approaches to violence prevention.

Individual-level risk factors for suicide mortality in the general population: an umbrella review.

BACKGROUND: Deaths by suicide remain a major public health challenge worldwide. Identifying and targeting risk factors for suicide mortality is a potential approach to prevention. We aimed to summarise current knowledge on the range and magnitude of individual-level risk factors for suicide mortality in the general population and evaluate the quality of the evidence. METHODS: In this umbrella review, five bibliographic databases were systematically searched for articles published from database inception to Aug 31, 2022. We included meta-analyses of observational studies on individual-level risk factors for suicide mortality in the general population. Biological, genetic, perinatal, and ecological risk factors were beyond the scope of this study. Effect sizes were synthesised and compared across domains. To test robustness and consistency of the findings, evidence for small-study effects and excess significance bias (ie, the ratio between the overall meta-analysis effect size and that of its largest included study) was examined, and prediction intervals were calculated. Risk of bias was assessed by the Risk of Bias in Systematic Reviews instrument. The protocol was pre-registered with PROSPERO (CRD42021230119). FINDINGS: We identified 33 meta-analyses on 38 risk factors for suicide mortality in the general population. 422 (93%) of the 454 primary studies included in the meta-analyses were from high-income countries. A previous suicide attempt and suicidal ideation emerged as strong risk factors (with effect sizes ranging from 6 to 16). Psychiatric disorders were associated with a greatly elevated risk of suicide mortality, with risk ratios in the range of 4-13. Suicide risk for physical illnesses (such as cancer and epilepsy) and sociodemographic factors (including unemployment and low education) were typically increased two-fold. Contact with the criminal justice system, state care in childhood, access to firearms, and parental death by suicide also increased the risk of suicide mortality. Among risk factors for which sex-stratified analyses were available, associations were generally similar for males and females. However, the quality of the evidence was limited by excess significance and high heterogeneity, and prediction intervals suggested poor replicability for almost two-thirds of identified risk factors. INTERPRETATION: A wide range of risk factors were identified across various domains, which underscores suicide mortality as a multifactorial phenomenon. Prevention strategies that span individual and population approaches should account for the identified factors and their relative strengths. Despite the large number of risk factors investigated, few associations were supported by robust evidence. Evidence of causal inference will need to be tested in high-quality study designs. FUNDING: Wellcome Trust.

Development and Validation of a Prediction Tool for Reoffending Risk in Domestic Violence.

Importance: Current risk assessment tools for domestic violence against family members were developed with small and selected samples, have low accuracy with few external validations, and do not report key performance measures. Objective: To develop new tools to assess risk of reoffending among individuals who have perpetrated domestic violence. Design, Setting, and Participants: This prognostic study investigated a national cohort of all individuals arrested for domestic violence between 1998 and 2013 in Sweden using information from multiple national registers, including National Crime Register, National Patient Register, Longitudinal Integrated Database for Health Insurance and Labour Market Studies Register, and Multi-Generation Register. Data were analyzed from August 2022 to June 2023. Exposure: Arrest for domestic violence. Main Outcomes and Measures: Prediction models were developed for 3 reoffending outcomes after arrest for domestic violence: conviction of a new violent crime (including domestic violence), conviction of any new crime, and rearrest for domestic violence at 1 year, 3 years, and 5 years. The prediction models were created using sociodemographic factors, criminological factors, and mental health status-related factors, linking data from multiple population-based longitudinal registers. Cox proportional hazard multivariable regression was used to develop prediction models and validate them in external samples. Key performance measures, including discrimination at prespecified cutoffs and calibration statistics, were investigated. Results: The cohort included 27 456 individuals (mean [SD] age, 39.4 [11.6] years; 24 804 men [90.3%]) arrested for domestic violence, of whom 4222 (15.4%) reoffended and were convicted for a new violent crime during a mean (SD) follow-up of 26.5 (27.0) months, 9010 (32.8%) reoffended and were convicted for a new crime (mean [SD] follow-up, 22.4 [25.1] months), and 2080 (7.6%) were rearrested for domestic violence (mean [SD] follow-up, 25.7 [30.6] months). Prediction models were developed with sociodemographic, criminological, and mental health factors and showed good measures of discrimination and calibration for violent reoffending and any reoffending. The area under the receiver operating characteristic curve (AUC) for risk of violent reoffending was 0.75 (95% CI, 0.74-0.76) at 1 year, 0.76 (95% CI, 0.75-0.77) at 3 years, and 0.76 (95% CI, 0.75-0.77) 5 years. The AUC for risk of any reoffending was 0.76 (95% CI, 0.75-0.77) at 1 year and at 3 years and 0.76 (95% CI, 0.75-0.76) at 5 years. The model for domestic violence reoffending showed modest discrimination (C index, 0.63; 95% CI, 0.61-0.65) and good calibration. The validation models showed discrimination and calibration performance similar to those of derivation models for all 3 reoffending outcomes. The prediction models have been translated into 3 simple online risk calculators that are freely available to use. Conclusions and Relevance: This prognostic study developed scalable, evidence-based prediction tools that could support decision-making in criminal justice systems, particularly at the arrest stage when identifying those at higher risk of reoffending and screening out individuals at low risk of reoffending. Furthermore, these tools can enhance treatment allocation by enabling criminal justice services to focus on modifiable risk factors identified in the tools for individuals at high risk of reoffending.

The predictive performance of criminal risk assessment tools used at sentencing: Systematic review of validation studies.

Although risk assessment tools have been widely used to inform sentencing decisions, there is uncertainty about the extent and quality of evidence of their predictive performance when validated in new samples. Following PRISMA guidelines, we conducted a systematic review of validation studies of 11 commonly used risk assessment tools for sentencing. We identified 36 studies with 597,665 participants, among which were 27 independent validation studies with 177,711 individuals. Overall, the predictive performance of the included risk assessment tools was mixed, and ranged from poor to moderate. Tool performance was typically overestimated in studies with smaller sample sizes or studies in which tool developers were co-authors. Most studies only reported area under the curve (AUC), which ranged from 0.57 to 0.75 in independent studies with more than 500 participants. The majority did not report key performance measures, such as calibration and rates of false positives and negatives. In addition, most validation studies had a high risk of bias, partly due to inappropriate analytical approach used. We conclude that the research priority is for future investigations to address the key methodological shortcomings identified in this review, and policy makers should enable this research. More sufficiently powered independent validation studies are necessary.

Prediction of reoffending risk in men convicted of sexual offences: development and validation of novel and scalable risk assessment tools (OxRIS).

Background: Current risk assessment tools have a limited evidence base with few validations, poor reporting of outcomes, and rarely include modifiable factors. Methods: We examined a national cohort of men convicted of sexual crimes in Sweden. We developed prediction models for three outcomes: violent (including sexual), any, and sexual reoffending. We used Cox proportional hazard regression to develop multivariable prediction models and validated these in an external sample. We reported discrimination and calibration statistics at prespecified cut-offs. Findings: We identified 16,231 men convicted of sexual offences, of whom 14.8% violently reoffended during a mean follow up of 38 months, 31.4% for any crime (34 months), and 3.6% for sexual crimes (42 months). Models for violent and any reoffending showed good discrimination and calibration. At 1, 3, and 5 years, the area under the curve (AUC) was 0.75-0.76 for violent reoffending and 0.74-0.75 for any reoffending. The prediction model for sexual reoffending showed modest discrimination (AUC = 0.67) and good calibration. We have generated three simple and web-based risk calculators, which are freely available. Interpretation: Scalable evidence-based risk assessment tools for sexual offenders in the criminal justice system and forensic mental health could assist decision-making and treatment allocation by identifying those at higher risk, and screening out low risk persons.

Risk factors for suicide in adults: systematic review and meta-analysis of psychological autopsy studies.

QUESTION: Effective prevention of suicide requires a comprehensive understanding of risk factors. STUDY SELECTION AND ANALYSIS: Five databases were systematically searched to identify psychological autopsy studies (published up to February 2022) that reported on risk factors for suicide mortality among adults in the general population. Effect sizes were pooled as odds ratios (ORs) using random-effects models for each risk factor examined in at least three independent samples. FINDINGS: A total of 37 case-control studies from 23 countries were included, providing data on 40 risk factors in 5633 cases and 7101 controls. The magnitude of effect sizes varied substantially both between and within risk factor domains. Clinical factors had the strongest associations with suicide, including any mental disorder (OR=13.1, 95% CI 9.9 to 17.4) and a history of self-harm (OR=10.1, 95% CI 6.6 to 15.6). By comparison, effect sizes were smaller for other domains relating to sociodemographic status, family history, and adverse life events (OR range 2-5). CONCLUSIONS: A wide range of predisposing and precipitating factors are associated with suicide among adults in the general population, but with clear differences in their relative strength. PROSPERO REGISTRATION NUMBER: CRD42021232878.

Evaluating the Risk of Suicide and Violence in Severe Mental Illness: A Feasibility Study of Two Risk Assessment Tools (OxMIS and OxMIV) in General Psychiatric Settings.

Background: Two OxRisk risk assessment tools, the Oxford Mental Illness and Suicide (OxMIS) and the Oxford Mental Illness and Violence (OxMIV), were developed and validated using national linked registries in Sweden, to assess suicide and violence risk in individuals with severe mental illness (schizophrenia-spectrum disorders and bipolar disorders). In this study, we aim to examine the feasibility and acceptability of the tools in three different clinical services. Method: We employed a two-step mixed-methods approach, by combining quantitative analyses of risk scores of 147 individual patients, and thematic analyses of qualitative data. First, 38 clinicians were asked to use OxMIS and OxMIV when conducting their routine risk assessments in patients with severe mental illness. The risk scores for each patient (which provide a probability of the outcome over 12 months) were then compared to the unstructured clinical risk assessment made by the treating clinician. Second, we carried out semi-structured interviews with the clinicians on the acceptability and utility of the tools. Thematic analysis was conducted on the qualitative data to identify common themes, in terms of the utility, accuracy, and acceptability of the tools. The investigations were undertaken in three general adult psychiatric clinics located in the cities of Barcelona and Sevilla (Spain), and Changsha (China). Results: Median risk probabilities over 12 months for OxMIS were 1.0% in the Spanish patient sample and 1.9% in the Chinese sample. For OxMIV, they were 0.7% (Spanish) and 0.8% (Chinese). In the thematic analysis, clinicians described the tools as easy to use, and thought that the risk score improved risk management. Potential additions to predictors were suggested, including family history and the patient's support network. Concordance rates of risk estimates between the tools and clinicians was high for violence (94.4%; 68/72) and moderate for suicide (50.0%; 36/72). Conclusion: Both OxMIS and OxMIV are feasible and practical in different general adult psychiatric settings. Clinicians interviewed found that both tools provide a useful structured approach to estimate the risk of suicide and violence. Risk scores from OxMIS and OxMIV can also be used to assist clinical decision-making for future management.

Predicting Violent Reoffending in Individuals Released From Prison in a Lower-Middle-Income Country: A Validation of OxRec in Tajikistan.

Background: Although around 70% of the world's prison population live in low- and middle-income countries (LMICs), risk assessment tools for criminal recidivism have been developed and validated in high-income countries (HICs). Validating such tools in LMIC settings is important for the risk management of people released from prison, development of evidence-based intervention programmes, and effective allocation of limited resources. Methods: We aimed to externally validate a scalable risk assessment tool, the Oxford Risk of Recidivism (OxRec) tool, which was developed in Sweden, using data from a cohort of people released from prisons in Tajikistan. Data were collected from interviews (for predictors) and criminal records (for some predictors and main outcomes). Individuals were first interviewed in prison and then followed up over a 1-year period for post-release violent reoffending outcomes. We assessed the predictive performance of OxRec by testing discrimination (area under the receiver operating characteristic curve; AUC) and calibration (calibration statistics and plots). In addition, we calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for different predetermined risk thresholds. Results: The cohort included 970 individuals released from prison. During the 12-month follow-up, 144 (15%) were reincarcerated for violent crimes. The original model performed well. The discriminative ability of OxRec Tajikistan was good (AUC = 0.70; 95% CI 0.66-0.75). The calibration plot suggested an underestimation of observed risk probabilities. However, after recalibration, model performance was improved (Brier score = 0.12; calibration in the large was 1.09). At a selected risk threshold of 15%, the tool had a sensitivity of 60%, specificity of 65%, PPV 23% and NPV 90%. In addition, OxRec was feasible to use, despite challenges to risk prediction in LMICs. Conclusion: In an external validation in a LMIC, the OxRec tool demonstrated good performance in multiple measures. OxRec could be used in Tajikistan to help prioritize interventions for people who are at high-risk of violent reoffending after incarceration and screen out others who are at lower risk of violent reoffending. The use of validated risk assessment tools in LMICs could improve risk stratification and inform the development of future interventions tailored at modifiable risk factors for recidivism, such as substance use and mental health problems.

Associations between prisons and recidivism: A nationwide longitudinal study.

OBJECTIVES: To examine differences in recidivism rates between different prisons using two designs-between-individual and within-individual-to account for confounding factors. METHODS: We examined recidivism rates among 37,891 individuals released from 44 Swedish prisons in three security levels, and who were followed from 2006 to 2013. We used longitudinal data from nationwide registers, including all convictions from district courts. First, we applied a between-individual design (Cox proportional hazards regression), comparing reconviction rates between individuals released from prisons within the same security level, while adjusting for a range of individual-level covariates. Second, we applied a within-individual design (stratified Cox proportional hazards regression), comparing rates of reconviction within the same individuals, i.e., we compared rates after release from one prison to the rates in the same individual after release from another prison, thus adjusting for all time-invariant confounders within each individual (e.g. genetics and early environment). We also adjusted for a range of time-varying individual-level covariates. RESULTS: Results showed differences in the hazard of recidivism between different prisons in between-individual analyses, with hazards ranging from 1.22 (1.05-1.43) to 4.99 (2.44-10.21). Results from within-individual analyses, which further adjusted for all time-invariant confounders, showed minimal differences between prisons, with hazards ranging from 0.95 (0.87-1.05) to 1.05 (0.95-1.16). Only small differences were found when violent and non-violent crimes were analyzed separately. CONCLUSIONS: The study highlights the importance of research designs that more fully adjust for individual-level confounding factors to avoid over-interpretation of the variability in comparisons across prisons.

Validation and recalibration of OxMIV in predicting violent behaviour in patients with schizophrenia spectrum disorders.

Oxford Mental Illness and Violence (OxMIV) addresses the need in mental health services for a scalable, transparent and valid tool to predict violent behaviour in patients with severe mental illness. However, external validations are lacking. Therefore, we have used a Dutch sample of general psychiatric patients with schizophrenia spectrum disorders (N = 637) to evaluate the performance of OxMIV in predicting interpersonal violence over 3 years. The predictors and outcome were measured with standardized instruments and multiple sources of information. Patients were mostly male (n = 493, 77%) and, on average, 27 (SD = 7) years old. The outcome rate was 9% (n = 59). Discrimination, as measured by the area under the curve, was moderate at 0.67 (95% confidence interval 0.61-0.73). Calibration-in-the-large was adequate, with a ratio between predicted and observed events of 1.2 and a Brier score of 0.09. At the individual level, risks were systematically underestimated in the original model, which was remedied by recalibrating the intercept and slope of the model. Probability scores generated by the recalibrated model can be used as an adjunct to clinical decision-making in Dutch mental health services.

Effectiveness of psychological interventions in prison to reduce recidivism: a systematic review and meta-analysis of randomised controlled trials.

BACKGROUND: Repeat offending, also known as criminal recidivism, in people released from prison has remained high over many decades. To address this, psychological treatments have been increasingly used in criminal justice settings; however, there is little evidence about their effectiveness. We aimed to evaluate the effectiveness of interventions in prison to reduce recidivism after release. METHODS: For this systematic review and meta-analysis, we searched the Cochrane Central Register of Controlled Trials, Embase, Global Health, MEDLINE, PsycINFO, and Google Scholar for articles published from database inception to Feb 17, 2021, without any language restrictions. We searched for randomised controlled trials (RCTs) that evaluated the effect of psychological interventions, delivered to adolescents and adults during incarceration, on recidivism outcomes after release. We excluded studies of solely pharmacological interventions and of participants in secure psychiatric hospitals or special residential units, or attending therapies mainly delivered outside of the prison setting. We extracted summary estimates from eligible RCTs. Data were extracted and appraised according to a prespecified protocol, with effect sizes converted to odds ratios. We used a standardised form to extract the effects of interventions on recidivism and estimated risk of bias for each RCT. Planned sensitivity analyses were done by removing studies with fewer than 50 participants. Our primary outcome was recidivism. Data from individual RCTs were combined in a random-effects meta-analysis as pooled odds ratios (ORs) and we explored sources of heterogeneity by comparing effect sizes by study size, control group, and intervention type. The protocol was pre-registered with PROSPERO, CRD42020167228. FINDINGS: Of 6345 articles retrieved, 29 RCTs (9443 participants, 1104 [11·7%] females, 8111 [85·9%] males, and 228 [2·4%] unknown) met the inclusion criteria for the primary outcome. Mean ages were 31·4 years (SD 4·9, range 24·5-41·5) for adult participants and 17·5 years (SD 1·9; range 14·6-20·2) for adolescent participants. Race or ethnicity data were not sufficiently reported to be aggregated. If including all 29 RCTs, psychological interventions were associated with reduced reoffending outcomes (OR 0·72, 95% CI 0·56-0·92). However, after excluding smaller studies (<50 participants in the intervention group), there was no significant reduction in recidivism (OR 0.87, 95% CI 0·68-1·11). Based on two studies, therapeutic communities were associated with decreased rates of recidivism (OR 0·64, 95% CI 0·46-0·91). These risk estimates did not significantly differ by type of control group and other study characteristics. INTERPRETATION: Widely implemented psychological interventions for people in prison to reduce offending after release need improvement. Publication bias and small-study effects appear to have overestimated the reported modest effects of such interventions, which were no longer present when only larger studies were included in analyses. Findings suggest that therapeutic communities and interventions that ensure continuity of care in community settings should be prioritised for future research. Developing new treatments should focus on addressing modifiable risk factors for reoffending. FUNDING: Wellcome Trust, Fonds de recherche du Québec - Santé.

Factors associated with suicide risk among Chinese adults: A prospective cohort study of 0.5 million individuals.

BACKGROUND: Suicide is a leading cause of death in China and accounts for about one-sixth of all suicides worldwide. The objective of this study was to examine the recent distribution of suicide and risk factors for death by suicide. Identifying underlying risk factors could benefit development of evidence-based prevention and intervention programs. METHODS AND FINDINGS: We conducted a prospective study, the China Kadoorie Biobank, of 512,715 individuals (41% men, mean age 52 years) from 10 (5 urban, 5 rural) areas which are diverse across China in geographic locations, social economic developmental stages, and prevalence of disease patterns. After the baseline measurements of risk factors during 2004 to 2008, participants were followed up for suicide outcomes including suicide and possible suicide deaths. Risk factors, such as sociodemographic factors and physical and mental health status, were assessed by semistructured interviews and self-report questionnaires. Suicide and possible suicide deaths were identified through linkage to the local death registries using ICD-10 codes. We conducted Cox regression to calculate hazard ratios (HRs) for suicide and for possible suicide in sensitivity analyses. During an average follow-up period of 9.9 years, 520 (101 per 100,000) people died from suicide (51.3% male), and 79.8% of them lived in rural areas. Sociodemographic factors associated with increased suicide risk were male gender (adjusted hazard ratios [aHR] = 1.6 [95% CI 1.4 to 2.0], p < 0.001), older age (1.3 [1.2 to 1.5] by each 10-yr increase, p < 0.001), rural residence (2.6 [2.1 to 3.3], p < 0.001), and single status (1.7 [1.4 to 2.2], p < 0.001). Increased hazards were found for family-related stressful life events (aHR = 1.8 [1.2 to 1.9], p < 0.001) and for major physical illnesses (1.5 [1.3 to 1.9], p < 0.001). There were strong associations of suicide with a history of lifetime mental disorders (aHR = 9.6 [5.9 to 15.6], p < 0.001) and lifetime schizophrenia-spectrum disorders (11.0 [7.1 to 17.0], p < 0.001). Links between suicide risk and depressive disorders (aHR = 2.6 [1.4 to 4.8], p = 0.002) and generalized anxiety disorders (2.6 [1.0 to 7.1], p = 0.056) in the last 12 months, and sleep disorders (1.4 [1.2 to 1.7], p < 0.001) in the past month were also found. All HRs were adjusted for sociodemographic factors including gender, age, residence, single status, education, and income. The associations with possible suicide deaths were mostly similar to those with suicide deaths, although there was no clear link between possible suicide deaths and psychiatric factors such as depression and generalized anxiety disorders. A limitation of the study is that there is likely underreporting of mental disorders due to the use of self-report information for some diagnostic categories. CONCLUSIONS: In this study, we observed that a range of sociodemographic, lifestyle, stressful life events, physical, and mental health factors were associated with suicide in China. High-risk groups identified were elderly men in rural settings and individuals with mental disorders. These findings could form the basis of targeted approaches to reduce suicide mortality in China.

Assessment of violence risk in 440 psychiatric patients in China: examining the feasibility and acceptability of a novel and scalable approach (FoVOx).

BACKGROUND: Violence risk assessment is a routine part of clinical services in mental health, and in particular secure psychiatric hospitals. The use of prediction models and risk tools can assist clinical decision-making on risk management, including decisions about further assessments, referral, hospitalization and treatment. In recent years, scalable evidence-based tools, such as Forensic Psychiatry and Violent Oxford (FoVOx), have been developed and validated for patients with mental illness. However, their acceptability and utility in clinical settings is not known. Therefore, we conducted a clinical impact study in multiple institutions that provided specialist mental health service. METHODS: We followed a two-step mixed-methods design. In phase one, we examined baseline risk factors on 330 psychiatric patients from seven forensic psychiatric institutes in China. In phase two, we conducted semi-structured interviews with 11 clinicians regarding violence risk assessment from ten mental health centres. We compared the FoVOx score on each admission (n = 110) to unstructured clinical risk assessment and used a thematic analysis to assess clinician views on the accuracy and utility of this tool. RESULTS: The median estimated probability of violent reoffending (FoVOx score) within 1 year was 7% (range 1-40%). There was fair agreement (72/99, 73% agreement) on the risk categories between FoVOx and clinicians' assessment on risk categories, and moderate agreement (10/12, 83% agreement) when examining low and high risk categories. In a majority of cases (56/101, 55%), clinicians thought the FoVOx score was an accurate representation of the violent risk of an individual patient. Clinicians suggested some additional clinical, social and criminal risk factors should be considered during any comprehensive assessment. In addition, FoVOx was considered to be helpful in assisting clinical decision-making and individual risk assessment. Ten out of 11 clinicians reported that FoVOx was easy to use, eight out of 11 was practical, and all clinicians would consider using it in the future. CONCLUSIONS: Clinicians found that violence risk assessment could be improved by using a simple, scalable tool, and that FoVOx was feasible and practical to use.

Risk factors for suicide in prisons: a systematic review and meta-analysis.

BACKGROUND: Rates of suicide among people in prison are elevated compared with people of similar age and sex who are living in the community. Improving assessments and interventions to reduce suicide risk requires updated evidence on risk factors. We aimed to examine risk factors associated with suicide in prisoners. METHODS: We did an updated systematic review and meta-analysis of risk factors for suicide among people in prison. We searched five biblographic databases for articles published between Jan 1, 2006, and Aug 13, 2020, and one database for articles published between Jan 1, 1973, and Aug 13, 2020. Eligible studies reported risk factors in individuals who died by suicide while in prison and in controls from the general prison population. Two reviewers independently extracted data for each study using a standardised form. We calculated random-effects pooled odds ratios (ORs) for the association of suicide with demographical, clinical, criminological, and institutional risk factors, and investigated heterogeneity using subgroup and meta-regression analyses. This systematic review is registered with PROSPERO, CRD42020137979. FINDINGS: We identified 8041 records through our searches, and used 77 eligible studies from 27 countries, including 35 351 suicides, in the main analysis. The strongest clinical factors associated with suicide were suicidal ideation during the current period in prison (OR 15·2, 95% CI 8·5-27·0), a history of attempted suicide (OR 8·2, 4·4-15·3), and current psychiatric diagnosis (OR 6·4, 3·6-11·1). Institutional factors associated with suicide included occupation of a single cell (OR 6·8, 2·3-19·8) and having no social visits (OR 1·9, 1·5-2·4). Criminological factors included remand status (OR 3·6, 3·1-4·1), serving a life sentence (OR 2·4, 1·3-4·6), and being convicted of a violent offence, in particular homicide (OR 3·1, 2·2-4·2). INTERPRETATION: Several modifiable risk factors, such as psychiatric diagnosis, suicidal ideation during the current period in prison, and single-cell occupancy, are associated with suicide among people in prison. Preventive interventions should target these risk factors and include improved access to evidence-based mental health care. Understanding other factors associated with suicide might improve risk stratification and resource allocation in prison services. FUNDING: Wellcome Trust, National Institute for Health Research Applied Research Collaboration Oxford and Thames Valley.

Ms. Beaudry et al. Reply.

We thank Atilola et al.1 for their thoughtful letter in response to our systematic review and meta-analysis of the prevalence of mental health problems among adolescents in juvenile detention and correctional facilities. We agree with Dr. Atilola that search strategies for systematic reviews should specifically look for low- and middle-income countries. In our systematic review, we used the bibliographic index Global Health to address this.

An Updated Systematic Review and Meta-regression Analysis: Mental Disorders Among Adolescents in Juvenile Detention and Correctional Facilities.

OBJECTIVE: To synthesize evidence on the prevalence of mental disorders in adolescents in juvenile detention and correctional facilities and examine sources of heterogeneity between studies. METHOD: Electronic databases and relevant reference lists were searched to identify surveys published from January 1966 to October 2019 that reported on the prevalence of mental disorders in unselected populations of detained adolescents. Data on the prevalence of a range of mental disorders (psychotic illnesses, major depression, attention-deficit/hyperactivity disorder [ADHD], conduct disorder, and posttraumatic stress disorder [PTSD]) along with predetermined study characteristics were extracted from the eligible studies. Analyses were reported separately for male and female adolescents, and findings were synthesized using random-effects models. Potential sources of heterogeneity were examined by meta-regression and subgroup analyses. RESULTS: Forty-seven studies from 19 countries comprising 28,033 male and 4,754 female adolescents were identified. The mean age of adolescents assessed was 16 years (range, 10-19 years). In male adolescents, 2.7% (95% CI 2.0%-3.4%) had a diagnosis of psychotic illness; 10.1% (95% CI 8.1%-12.2%) major depression; 17.3% (95% CI 13.9%-20.7%) ADHD; 61.7% (95% CI 55.4%-67.9%) conduct disorder; and 8.6% (95% CI 6.4%-10.7%) PTSD. In female adolescents, 2.9% (95% CI 2.4%-3.5%) had a psychotic illness; 25.8% (95% CI 20.3%-31.3%) major depression; 17.5% (95% CI 12.1%-22.9%) ADHD; 59.0% (95% CI 44.9%-73.1%) conduct disorder; and 18.2% (95% CI 13.1%-23.2%) PTSD. Meta-regression found higher prevalences of ADHD and conduct disorder in investigations published after 2006. Female adolescents had higher prevalences of major depression and PTSD than male adolescents. CONCLUSION: Consideration should be given to reviewing whether health care services in juvenile detention can address these levels of psychiatric morbidity.

Drug Use Disorders and Violence: Associations With Individual Drug Categories.

We conducted a systematic review that examined the link between individual drug categories and violent outcomes. We searched for primary case-control and cohort investigations that reported risk of violence against others among individuals diagnosed with drug use disorders using validated clinical criteria, following Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. We identified 18 studies published during 1990-2019, reporting data from 591,411 individuals with drug use disorders. We reported odds ratios of the violence risk in different categories of drug use disorders compared with those without. We found odds ratios ranging from 0.8 to 25.0 for most individual drug categories, with generally higher odds ratios among individuals with polydrug use disorders. In addition, we explored sources of between-study heterogeneity by subgroup and meta-regression analyses. Cohort investigations reported a lower risk of violence than case-control reports (odds ratio =  2.7 (95% confidence interval (CI): 2.1, 3.5) vs. 6.6 (95% CI: 5.1, 8.6)), and associations were stronger when the outcome was any violence rather than intimate partner violence (odds ratio = 5.7 (95% CI: 3.8, 8.6) vs. 1.7 (95% CI: 1.4, 2.1)), which was consistent with results from the meta-regression. Overall, these findings highlight the potential impact of preventing and treating drug use disorders on reducing violence risk and associated morbidities.

Risk factors for self-harm in prison: a systematic review and meta-analysis.

BACKGROUND: Self-harm is a leading cause of morbidity in prisoners. Although a wide range of risk factors for self-harm in prisoners has been identified, the strength and consistency of effect sizes is uncertain. We aimed to synthesise evidence and assess the risk factors associated with self-harm inside prison. METHODS: In this systematic review and meta-analysis, we searched four electronic databases (PubMed, Embase, Web of Science, and PsycINFO) for observational studies on risk factors for self-harm in prisoners published from database inception to Oct 31, 2019, supplemented through correspondence with authors of studies. We included primary studies involving adults sampled from general prison populations who self-harmed in prison and a comparison group without self-harm in prison. We excluded studies with qualitative or ecological designs, those that reported on lifetime measures of self-harm or on selected samples of prisoners, and those with a comparison group that was not appropriate or not based on general prison populations. Data were extracted from the articles and requested from study authors. Our primary outcome was the risk of self-harm for risk factors in prisoners. We pooled effect sizes as odds ratios (OR) using random effects models for each risk factor examined in at least three distinct samples. We assessed study quality on the basis of the Newcastle-Ottawa Scale and examined between-study heterogeneity. The study protocol was registered with PROSPERO, CRD42018087915. FINDINGS: We identified 35 independent studies from 20 countries comprising a total of 663 735 prisoners, of whom 24 978 (3·8%) had self-harmed in prison. Across the 40 risk factors examined, the strongest associations with self-harm in prison were found for suicide-related antecedents, including current or recent suicidal ideation (OR 13·8, 95% CI 8·6-22·1; I2=49%), lifetime history of suicidal ideation (8·9, 6·1-13·0; I2=56%), and previous self-harm (6·6, 5·3-8·3; I2=55%). Any current psychiatric diagnosis was also strongly associated with self-harm (8·1, 7·0-9·4; I2=0%), particularly major depression (9·3, 2·9-29·5; I2=91%) and borderline personality disorder (9·2, 3·7-22·5; I2=81%). Prison-specific environmental risk factors for self-harm included solitary confinement (5·6, 2·7-11·6; I2=98%), disciplinary infractions (3·5, 1·2-9·7; I2=99%), and experiencing sexual or physical victimisation while in prison (3·2, 2·1-4·8; I2=44%). Sociodemographic (OR range 1·5-2·5) and criminological (1·8-2·3) factors were only modestly associated with self-harm in prison. We did not find clear evidence of publication bias. INTERPRETATION: The wide range of risk factors across clinical and custody-related domains underscores the need for a comprehensive, prison-wide approach towards preventing self-harm in prison. This approach should incorporate both population and targeted strategies, with multiagency collaboration between the services for mental health, social care, and criminal justice having a key role. FUNDING: Wellcome Trust.

Identifying Predictors of Suicide in Severe Mental Illness: A Feasibility Study of a Clinical Prediction Rule (Oxford Mental Illness and Suicide Tool or OxMIS).

BACKGROUND: Oxford Mental Illness and Suicide tool (OxMIS) is a brief, scalable, freely available, structured risk assessment tool to assess suicide risk in patients with severe mental illness (schizophrenia-spectrum disorders or bipolar disorder). OxMIS requires further external validation, but a lack of large-scale cohorts with relevant variables makes this challenging. Electronic health records provide possible data sources for external validation of risk prediction tools. However, they contain large amounts of information within free-text that is not readily extractable. In this study, we examined the feasibility of identifying suicide predictors needed to validate OxMIS in routinely collected electronic health records. METHODS: In study 1, we manually reviewed electronic health records of 57 patients with severe mental illness to calculate OxMIS risk scores. In study 2, we examined the feasibility of using natural language processing to scale up this process. We used anonymized free-text documents from the Clinical Record Interactive Search database to train a named entity recognition model, a machine learning technique which recognizes concepts in free-text. The model identified eight concepts relevant for suicide risk assessment: medication (antidepressant/antipsychotic treatment), violence, education, self-harm, benefits receipt, drug/alcohol use disorder, suicide, and psychiatric admission. We assessed model performance in terms of precision (similar to positive predictive value), recall (similar to sensitivity) and F1 statistic (an overall performance measure). RESULTS: In study 1, we estimated suicide risk for all patients using the OxMIS calculator, giving a range of 12 month risk estimates from 0.1-3.4%. For 13 out of 17 predictors, there was no missing information in electronic health records. For the remaining 4 predictors missingness ranged from 7-26%; to account for these missing variables, it was possible for OxMIS to estimate suicide risk using a range of scores. In study 2, the named entity recognition model had an overall precision of 0.77, recall of 0.90 and F1 score of 0.83. The concept with the best precision and recall was medication (precision 0.84, recall 0.96) and the weakest were suicide (precision 0.37), and drug/alcohol use disorder (recall 0.61). CONCLUSIONS: It is feasible to estimate suicide risk with the OxMIS tool using predictors identified in routine clinical records. Predictors could be extracted using natural language processing. However, electronic health records differ from other data sources, particularly for family history variables, which creates methodological challenges.