Using big data to develop universal and selective suicide prevention strategies

White River Junction Va Medical Center, White River Junction VT

Investigators

Linked publications, trials & patents

Abstract

BACKGROUND: Suicide confers a massive burden to individuals and society, accounting for nearly 20 Veteran deaths every day. After controlling for age, Veterans’ suicide risk is 22% higher than US adult civilians. In response, the US Department of Veterans Affairs (VA) has made suicide prevention its first priority. The VA’s suicide prevention framework prioritizes patients at the highest risk for suicide, including those who previously attempted suicide, were recently released from inpatient mental health treatment, or use opioids. Although this strategy has led to improvements for high-risk patients, the majority of patients that die by suicide are not included in this group. Indeed, less than 3% of recent VA suicide deaths were classified as high-risk by the leading prediction metric. This proposal specifically focuses on this “non-high-risk” majority, i.e., those who died by suicide, but whose risk was not detected by existing prediction mechanisms. This proposal leverages big data to better identify, track, and treat this critical population. It has the potential to have a large impact on Veteran health, broadening the reach of effective suicide prevention services. OBJECTIVES: The long-term goal is for the candidate, Dr. Maxwell Levis, to become an independent clinical researcher focused on developing, testing, and improving suicide prevention resources. His short-term goals are to: 1) acquire skills in population-based approaches to suicide prediction and prevention, 2) improve machine learning and natural language processing ability, and 3) gain experience adapting suicide prevention resources. His research objectives align closely with these goals. Dr. Levis’ proposal’s central hypothesis is that, through leveraging big data, he can better understand non-high-risk suicide decedents, and, in turn, develop targeted suicide prediction and prevention mechanisms. In the award’s last two years, Dr. Levis will submit a VA Merit Award proposal on leveraging psychotherapy to decrease suicide risk in this population. METHODS: The VA’s suicide prevention framework relies on universal strategies to reach all patients (low- risk), selective strategies to reach some patients (moderate-risk), and indicated strategies to reach the relatively few patients with symptoms associated with suicide (high-risk). While strides have been made for indicated strategies, comparable achievements have not been made for universal and selective strategies. Using Corporate Data Warehouse data, Dr. Levis will develop a dataset of recent (2017–2018; n ≈ 4000) non- high-risk suicide decedents (cases) and characterize this sample’s demographics, service and mental health usage, and risk and treatment factors. He will then develop a suicide risk-matched (1:[10]) sample of VA patients that did not die (controls), but shared similar risk, services, demographics, location, and treatment intervals. Dr. Levis will then leverage cases’ and controls’ structured and unstructured electronic health record (EHR) data to develop machine-learning-derived suicide-prediction models. He will also sub-select non-high- risk patients that received psychotherapy in the year before death by suicide (n ≈ 2,750) and risk-matched controls that did not die (1:5), but shared similar risk, psychotherapy usage, [pharmacotherapy], demographics, location, and treatment intervals. Dr. Levis will use time-sensitive machine learning methods to evaluate changes in psychotherapy note text over time, tracking change mechanisms, intervals of heightened changes, and domains associated with reduced suicide risk. These findings will support a subsequent investigator- initiated research proposal using derived information to leverage psychotherapy to reduce suicide risk among Veterans. IMPACT: The goal of this award is to help Dr. Levis become an independent clinical researcher in the suicide prevention field. This award consists of educational, research, and mentorship components. Education components will center on analytic and clinical research skills. Dr. Levis will receive mentorship from suicide prevention and clinical research leaders. These areas will be central for Dr. Levis’ Merit Award proposal.