A longitudinal machine learning approach providing clinicians timely detection to prevent military suicide

项目总结

PROJECT SUMMARY The suicide rates among U.S. military service members and Veterans (MV) remain alarmingly high. The suicide rate for active military service members has increased from 20.4 suicides in 2014 to 28.7 suicides in 2020 per 100,000. Veterans’ suicide rates have remained high, approximately 2 times higher than the general population (14.5 per 100,000). Unfortunately, the current suicide approaches from the Department of Defense and the Department of Veterans Affairs are insufficient. Further, recent literature shows inconsistent findings of suicide causes and suicide attempts across measures and time points, and lack of effectiveness of suicide screening and interventions. This is problematic for proactively and effectively preventing and stopping suicide among the MV populations. Additionally, less research has focused on suicide ideation than suicide completion/deaths, which means we are ultimately missing the first chance to stifle suicide and address risk factors. We will use secondary datasets and innovative machine learning (ML) to develop early screening and intervention modeling to address military suicide issues. The study will apply data-driven ML to improve MV healthcare quality by accelerating the implementation of patient-centered outcomes research, using several personalized-contextual variables of 10 clinically applicable dimensions, to predict suicide risk levels. Further, we will develop person-centered, context-sensitive ML modeling for suicide ideation (SI) and suicide attempt (SA) data-visualization profiles, which will assist in clinical screening, evaluation, and intervention. Our specific aims are to (1) establish ML algorithms detecting SI/SA at different military statuses to inform clinicians and (2) develop an SI/SA cross-sectional and longitudinal risk data-visualization profile for clinicians. Our overarching goals are to demonstrate (1) a new SI/SA screening paradigm and (2) a new SI/SA prevention, evidence-based intervention, and policy-making model for the MV populations. We harness big data and innovative ML applications to provide a 360-degree view of MV patients, which will improve healthcare quality and MV patient outcomes, specifically decreasing SI/SA. Our project will exemplify the Healthcare Effectiveness and Outcomes Research mission to make healthcare safer, higher quality, more accessible, equitable, and affordable. Most importantly, we will ensure that clinical professionals and relevant stakeholders who serve the MV populations can understand and apply the study’s findings.