Dataset Information

Machine Learning to Differentiate Risk of Suicide Attempt and Self-harm After General Medical Hospitalization of Women With Mental Illness.

ABSTRACT:

Background

Suicide prevention is a public health priority, but risk factors for suicide after medical hospitalization remain understudied. This problem is critical for women, for whom suicide rates in the United States are disproportionately increasing.

Objective

To differentiate the risk of suicide attempt and self-harm following general medical hospitalization among women with depression, bipolar disorder, and chronic psychosis.

Methods

We developed a machine learning algorithm that identified risk factors of suicide attempt and self-harm after general hospitalization using electronic health record data from 1628 women in the University of California Los Angeles Integrated Clinical and Research Data Repository. To assess replicability, we applied the algorithm to a larger sample of 140,848 women in the New York City Clinical Data Research Network.

Results

The classification tree algorithm identified risk groups in University of California Los Angeles Integrated Clinical and Research Data Repository (area under the curve 0.73, sensitivity 73.4, specificity 84.1, accuracy 0.84), and predictor combinations characterizing key risk groups were replicated in New York City Clinical Data Research Network (area under the curve 0.71, sensitivity 83.3, specificity 82.2, and accuracy 0.84). Predictors included medical comorbidity, history of pregnancy-related mental illness, age, and history of suicide-related behavior. Women with antecedent medical illness and history of pregnancy-related mental illness were at high risk (6.9%-17.2% readmitted for suicide-related behavior), as were women below 55 years old without antecedent medical illness (4.0%-7.5% readmitted).

Conclusions

Prevention of suicide attempt and self-harm among women following acute medical illness may be improved by screening for sex-specific predictors including perinatal mental health history.

SUBMITTER: Edgcomb JB

PROVIDER: S-EPMC7810157 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:CONTEXT Previous studies have suggested that children who experience parental suicide at earlier ages are at higher risk of future hospitalization for suicide attempt. However, how the trajectories of risk differ by offspring age at the time of parental suicide is currently unknown. OBJECTIVE To study time at risk to suicide attempt hospitalization among offspring of suicide decedents as compared with offspring of unintentional injury decedents by their developmental period at the time of parental death. DESIGN Population-based retrospective cohort study. SETTING Sweden. PARTICIPANTS Twenty-six thousand ninety-six offspring who experienced parental suicide and 32 395 offspring of unintentional injury decedents prior to age 25 years between the years 1973 and 2003. MAIN OUTCOME MEASURE Parametric survival analysis was used to model the time to hospitalization for suicide attempt among offspring who lost a parent during early childhood (0-5 years old), later childhood (6-12 years old), adolescence (13-17 years old), and young adulthood (18-24 years old). RESULTS The risk in offspring who lost a parent to suicide or an unintentional injury during childhood surpassed the other age groups' risk approximately 5 years after the origin and, for the youngest group, continued to rise over decades. Offspring who lost a parent during adolescence or young adulthood were at greatest risk within 1 to 2 years after parental death, and risk declined over time. Offspring who lost a parent to suicide in childhood and young adulthood had earlier onset of hospitalization for suicide attempt compared with offspring who lost a parent to an unintentional injury. CONCLUSIONS The hospitalization risk for suicide attempt in offspring who lost a parent during their childhood is different from those who lost a parent later in development. The results suggest critical windows for careful monitoring and intervention for suicide attempt risk, especially 1 to 2 years after parental death for the older age groups and over decades for childhood survivors of parental death.

Project description:ObjectiveThe majority of suicide decedents have had contact with health services close to their death. Some of these contacts include admissions to hospitals for physical and mental health conditions, injury and intentional self-harm. This study aims to establish and quantify the risks of suicide following hospital admission for a range of mental and physical illnesses.MethodsA retrospective analysis was carried out on existing morbidity and mortality data in Victoria. Data was extracted from the Victorian Admitted Episodes Dataset and the Victorian Suicide Register. Unplanned hospital admissions among adult patients (> = 15 years of age), discharged between 01 January 2011 and 31 December 2016 (2,430,154 admissions), were selected. Standardised Mortality Ratios were calculated for conditions with at least five linked suicides within one year of discharge from hospital.ResultsForty-three conditions defined at the three-digit level of the International Statistical Classification of Diseases and Related Health Problems 10th Revision, were associated with at least five subsequent suicides (within one year of hospital discharge); 14 physical illnesses, 5 symptoms, signs and abnormal clinical and laboratory findings, 12 mental health conditions, and 12 types of injury and poisonings. The highest Standardised Mortality Ratios were for poisonings (range; 27.8 to 140.0) and intentional self-harm (78.8), followed by mental health conditions (range; 15.5 to 72.9), symptoms, signs and abnormal clinical and laboratory findings (range; 1.4 to 43.2) and physical illnesses (range; 0.7 to 4.9).ConclusionsHospital admissions related to mental health conditions and injury and poisonings including self-harm were associated with a greater risk of suicide than physical conditions. Mental health conditions such as depressive episodes, personality disorders and psychotic episodes, injuries caused by intentional-self-harm and poisonings by certain types of drugs, carbon monoxide and hormones such as insulin can be prioritised for targeting suicide prevention initiatives for persons discharged from hospitals.

Project description:BackgroundEmpirical evidence supporting the distinction between suicide attempt (SA) and non-suicidal self-harm (NSSH) is lacking. Although NSSH is a risk factor for SA, we do not currently know whether these behaviours lie on a continuum of severity, or whether they are discrete outcomes with different aetiologies. We conducted this exploratory genetic epidemiology study to investigate this issue further.MethodsWe explored the extent of genetic overlap between NSSH and SA in a large, richly-phenotyped cohort (the Avon Longitudinal Study of Parents and Children; N = 4959), utilising individual-level genetic and phenotypic data to conduct analyses of genome-wide complex traits and polygenic risk scores (PRS).ResultsThe single nucleotide polymorphism heritability of NSSH was estimated to be 13% (SE 0.07) and that of SA to be 0% (SE 0.07). Of the traits investigated, NSSH was most strongly correlated with higher IQ (rG = 0.31, SE = 0.22), there was little evidence of high genetic correlation between NSSH and SA (rG = - 0.1, SE = 0.54), likely due to the low heritability estimate for SA. The PRS for depression differentiated between those with NSSH and SA in multinomial regression. The optimal PRS prediction model for SA (Nagelkerke R2 0.022, p < 0.001) included ADHD, depression, income, anorexia and neuroticism and explained more variance than the optimal prediction model for NSSH (Nagelkerke R2 0.010, p < 0.001) which included ADHD, alcohol consumption, autism spectrum conditions, depression, IQ, neuroticism and suicide attempt.ConclusionsOur findings suggest that SA does not have a large genetic component, and that although NSSH and SA are not discrete outcomes there appears to be little genetic overlap between the two. The relatively small sample size and resulting low heritability estimate for SA was a limitation of the study. Combined with low heritability estimates, this implies that family or population structures in SA GWASs may contribute to signals detected.

Project description:Importance:Some reports have raised concerns regarding a potential psychiatric worsening associated with first-time use of aripiprazole in patients already treated with other antipsychotic medications. Whether aripiprazole use, particularly in the long term, increases the risk for serious psychiatric events is unclear. Objective:To assess whether switching to or adding aripiprazole is associated with serious psychiatric treatment failure compared with switching to or adding another antipsychotic drug in patients previously exposed to antipsychotic medications. Design, Setting, and Participants:This population-based cohort study was conducted from January 1, 2005, to March 31, 2015. Data were obtained from the United Kingdom Clinical Practice Research Datalink, one of the world's largest computerized databases linked to the Hospital Episodes Statistics repository and the Office for National Statistics (ONS) mortality database. Within a base cohort of new users of antipsychotic drugs, patients switching or adding aripiprazole (n = 1643) were propensity matched 1:1 to patients switching to or adding another antipsychotic medication (n = 1643). All patients were followed up until psychiatric treatment failure, for 365 days (1 year) after cohort entry, until death from any cause other than suicide, until end of registration or linkage with the databases, or end of the study period (March 31, 2016). Main Outcomes and Measures:Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% CIs of serious events of psychiatric treatment failure (psychiatric hospitalizations, self-harm, or suicide) associated with switching to or adding aripiprazole compared with other antipsychotic drugs. In addition to propensity score matching, all models were adjusted for age, number of psychiatric hospitalizations or self-harm events in the 6 months before cohort entry, number of different antipsychotic drugs before cohort entry, and quintiles of the Index of Multiple Deprivation. Results:The study cohort included 1643 patients (949 [57.8%] were women with a mean [SD] age of 42.1 [16.8] years) starting aripiprazole use; they were matched 1:1 to 1643 patients (871 [53.0%] were women with a mean [SD] age of 42.4 [17.1] years) starting use of another antipsychotic drug. During 2692 person-years of follow-up, 391 incident serious psychiatric treatment failures occurred, with a crude incidence rate of 14.52 (95% CI, 13.16-16.04) per 100 person-years. First-time use of aripiprazole was not associated with an increased rate of psychiatric treatment failure (HR, 0.87; 95% CI, 0.71-1.06), psychiatric hospitalizations (HR, 0.85; 95% CI, 0.69-1.06), or self-harm or suicide (HR, 0.96; 95% CI, 0.68-1.36) compared with starting use of another antipsychotic drug. Results were consistent across several sensitivity analyses. Conclusions and Relevance:Initiating aripiprazole use, compared with another antipsychotic medication, after a previous antipsychotic exposure did not appear to be associated with psychiatric hospitalization, self-harm, or suicide; these findings warrant replication in large observational studies.

Project description:ObjectivesFirst, to examine general practitioner (GP) knowledge about the care (needs) of their patients; second, to examine the quality of GP follow-up care; third, to examine the transmission of patient care information from hospitals/emergency services (ES) to GPs.Setting105 general practices from the representative Belgian Network of Sentinel General Practices (SGP) in Flanders, the largest region of Belgium, during 2013-2016.Participants245 suicide attempts by regular patients.Outcomes measuresTen care-related measures, including three indicators of quality of follow-up care, were based on data reported by the SGP on structured forms at baseline and at two follow-up points in time.ResultsAs for GP knowledge, 10.5% of SGP failed to report whether suicidal risk was noticed in patients seen in the month preceding the attempt; 9.0% whether there were previous attempts; 22.5% whether the patient was receiving mental health treatment at follow-up and 22.0% whether suicidal behaviour was repeated at follow-up. Relatively more patients≥65 years had no suicide risk evaluation (OR 3.54; 95% CI 1.11 to 11.26). As for quality of follow-up care, there was a GP-patient contact following 90.5% of the attempts, follow-up appointments were planned following 43.4% of the attempts and there was a GP contact with patient proxies following 62.8% of the attempts. Patient age ≥65 years (OR 4.09; 95% CI 1.79 to 9.33), a recent GP-patient contact preceding the attempt (OR 1.97; 95% CI 1.13 to 3.43), depression of patient (OR 1.96; 95% CI 1.14 to 3.37) and a suburban SGP area (OR 2.34; 95% CI 1.13 to 4.82) were determinants of an increased quality of care sum. GPs received patient care information from a hospital (ES) for 67.8% of eligible attempts, with SGP practice location being a determinant.ConclusionsGPs are highly involved in the care of suicide attempters but there is room for improvement, also in informational continuity from hospital (ES) to GPs.

Project description:BackgroundOxford 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.MethodsIn 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).ResultsIn 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).ConclusionsIt 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.