Washington Post Praises VA Suicide Prediction Technology Compared to Silicon Valley
December 21, 2018
According to an article in the Washington Post, more than one million Americans attempted suicide last year – 47,000 succeeded. Predictive modeling can support the prevention of suicide-related behavior, as it can be used to identify patients at risk for suicide before they engage in self-harming behaviors. Medical providers and tech companies, including VA and Facebook, are increasingly applying artificial intelligence (AI) to the problem of suicide prediction. The Post article notes that AI research is moving along two tracks: academic/medical (i.e., VA) and one skewing toward the commercial (i.e., Facebook). While private-sector efforts are “completely unregulated, potentially putting at risk people’s privacy, safety and autonomy,” VA is striving to make sure that this line of research complies with health laws and ethical standards, as well as the need to demonstrate efficacy at each step.1
In partnership with VA’s Office of Mental Health and Suicide Prevention (OMHSP), HSR&D investigators are conducting a large national study that will evaluate the implementation of the Recovery Engagement and Coordination for Health – Veterans Enhanced Treatment (REACH VET). REACH VET is a predictive modeling system that uses a combination of demographic, prior suicide attempt, mental and physical health diagnoses, and VA healthcare use information from Veterans’ medical records to identify those at highest risk for various adverse events, including suicide. REACH VET utilizes a dashboard that provides facility-level REACH VET coordinators with the names of Veterans who have been identified as being at highest risk once a month.2
REACH VET study coordinators at each of the 140 VA healthcare systems are responsible for monitoring the REACH VET dashboard that identifies those at high risk and tracks next steps for coordinators and providers. Following identification of patients at risk, coordinators notify each patient's provider of their high-risk status and orient the provider to the dashboard. Providers are required to re-evaluate the patient's care, determine if care enhancements are needed, and contact the patient. The evaluation team is evaluating numerous implementation outcomes, including:
- Reach: Proportion of patients identified at each facility who receive the REACH VET intervention;
- Adoption: Proportion of mental health and primary care providers in each facility that participate;
- Implementation fidelity: Whether facilities implemented all components of the intervention as directed by the memos and the REACH VET program website; and
- Cost of implementation: Amount of effort and time needed to offer virtual external facilitation.
OMHSP recently completed a preliminary evaluation of the impact of REACH VET on patient outcomes by examining six-month outcomes for Veterans identified by REACH VET. In comparison to control groups, Veterans engaged by REACH VET had more health and mental healthcare appointments, decreases in percent of missed appointments, fewer inpatient mental health admissions, and lower all-cause mortality.3
- Marks M. Suicide prediction technology is revolutionary. It badly needs oversight. Washington Post. December 20, 2018.
- Peterson K, Anderson J, Bourne, D. Evidence Brief: Suicide Prevention in Veterans. VA ESP Project #09-199; 2018.
- Evaluation of a new program that uses predictive modeling in the fight against Veteran suicide. FORUM. Spring 2018.