Digital Monitoring of Agitation for Short-Term Suicide Risk Prediction

点击翻译按钮获取中文摘要

Suicide is a prevalent and burdensome public health problem that warrants immediate attention. As the tenth leading cause of death in the United States, suicide claims the lives of more than 44,000 Americans each year. There is an urgent need to identify objective and clinically informative markers of imminent risk for suicidal behavior. Agitation, defined in DSM-5 as excessive motor activity associated with a feeling of inner tension, is listed as a warning sign for suicide by leading organizations and in widely used risk assessment protocols. Yet, prior research on the association between agitation and suicide has key methodological limitations (including related to the operationalization of agitation), which has resulted in minimal empirical evidence to support agitation as a proximal risk factor for suicide. Addressing this gap in knowledge has the potential for significant impact, including informing both the clinical assessment of suicide risk and the development of just-in-time interventions for detecting and responding to acute suicide risk. This project will overcome the limitations of prior suicide risk factor research by assessing multiple behavioral (motor activity and vocal features [e.g., volume, speaking rate, pitch]) and subjective components of agitation and suicidal thoughts and behaviors in a sample at elevated risk for suicide over a short, high-risk period. We will test the hypotheses that (1) objectively measured real-time indicators of agitation correlate with both momentary subjective ratings and validated, gold standard measures of agitation, and (2) both subjective and objective indicators of agitation improve prediction of short-term increases in suicide ideation, plan, and attempt above and beyond other distal and proximal risk factors. We propose to collect high-resolution self-report (e.g., ecological momentary assessment) and passive (e.g., accelerometer) data on agitation using smartphones and wearable sensors from psychiatric inpatients admitted for suicide ideation or attempt during inpatient treatment and the four weeks after discharge. Multi- level modeling and machine learning approaches will be implemented to examine (1) associations between objective and subjective real-time indicators of agitation and validated measures of agitation, and (2) the degree to which real-time indicators of agitation predict momentary fluctuations in suicidal ideation and suicide plan and attempt above and beyond other distal and proximal risk factors. The scientific aims of this study map onto the candidate’s training in three primary areas: (1) digital monitoring of high-risk patients, (2) advanced longitudinal multivariate data analysis, and (3) identification of behavioral and vocal biomarkers. The candidate’s training plan includes mentorship from Dr. Matthew Nock (primary mentor), Dr. Jordan Smoller (co- mentor), Dr. Maurizio Fava (co-mentor), and Drs. Rosalind Picard, Evan Kleiman, and Thomas Quatieri (consultants), as well as quantitative coursework at the Harvard School of Public Health and Massachusetts Institute of Technology. This mentored five-year award will propel the candidate to an independent patient- oriented research career focused on using scalable methods to advance suicide prediction and prevention.