Development of an early-warning system for high-risk patients for suicide attempt using deep learning and electronic health records

Abstract

Suicide is the tenth leading cause of death in the United States (US). An early-warning system (EWS) for suicide attempt could prove valuable for identifying those at risk of suicide attempts, and analyzing the contribution of repeated attempts to the risk of eventual death by suicide. In this study we sought to develop an EWS for high-risk suicide attempt patients through the development of a population-based risk stratification surveillance system. Advanced machine-learning algorithms and deep neural networks were utilized to build models with the data from electronic health records (EHRs). A final risk score was calculated for each individual and calibrated to indicate the probability of a suicide attempt in the following 1-year time period. Risk scores were subjected to individual-level analysis in order to aid in the interpretation of the results for health-care providers managing the at-risk cohorts. The 1-year suicide attempt risk model attained an area under the curve (AUC ROC) of 0.792 and 0.769 in the retrospective and prospective cohorts, respectively. The suicide attempt rate in the "very high risk" category was 60 times greater than the population baseline when tested in the prospective cohorts. Mental health disorders including depression, bipolar disorders and anxiety, along with substance abuse, impulse control disorders, clinical utilization indicators, and socioeconomic determinants were recognized as significant features associated with incident suicide attempt.

Recurrence of SARS-CoV-2 PCR positivity in COVID-19 patients: a single center experience and potential implications

“…Univariable analysis was performed on z-score-normalized features, and logistic regression was used to calculate the odds ratios and P values for feature filtering. For multivariate model building, a gradient boosting tree algorithm XGBoost was used for constructing a multivariable prediction model [10][11][12][13][14][15][16] . The baseline learner is the classification and regression tree and the number of trees is selected via cross-validation to avoid over-fitting.…”

Deep neural networks detect suicide risk from textual facebook posts

“…However, the applicability of machine-learning models may still be limited because their predictions are often based on medical sources, such as neuroimaging data, counselling transcripts, and clinical reports. A recent study, for example, managed to develop a highly accurate suicide prediction model (0.769 ≤ AUC ≤ 0.792), based on the health records of patients who visited one of the Berkshire Health System hospitals 8 . Although valuable, these sources do not capture first-hand the patients’ natural behavior, nor do they include data from non-treated or non-diagnosed individuals.…”

Deep Learning for Medicine and Remote Sensing: A Brief Review

“…The study of Zheng and his colleagues is one of the newest examples of deep learning applications in medicine (Zheng et al, 2020). They have developed an early-warning-system for high-risk suicide attempt patients.…”