Predicting point-of-care outcomes for text message crisis interventions in teens

Project Summary/Abstract Over the last decade, suicidal ideation and suicide death have increased significantly across all age groups, but have shown the most dramatic rise in teens and young adults. There is clear evidence that greater access to mental health care is associated with reduced suicide risk, but increasing mental health care using conventional

models faces multiple barriers; ensuring access to mental health care in rural areas is especially challenging. Accordingly, innovative approaches to provide mental health care, and especially crisis intervention, to teens are essential. SafeUT is a text-messaging app developed in conjunction with state government agencies that links Utah

teens who have a mental health crisis to a counselor who can provide support, assess suicide risk, triage to emergency services, and refer for additional treatment. Our project will utilize a large, rigorously anonymized repository of text-message data (>130,000 encounters and >2.3 million messages) contained within SafeUT to

predict important outcomes of using the app, such as being referred to emergency services, staying engaged with the counselor, and receiving a thorough risk assessment. We will examine SafeUT data with cutting-edge machine-learning techniques, including natural language processing, to develop ways of

predicting SafeUT users’ outcomes. These predictive systems will, ultimately, be used to produce real-time feedback systems that monitor and help improve the quality of SafeUT services. The current proposal squarely addresses Notice of Special Interest NOT-MH-22-110 (Priority Research Opportunities in Crisis Response Services), which emphasizes the importance of projects

addressing crisis care in children and which requests applications focused on the development of “assessment strategies and decision-making aides (e.g., predictive algorithms) that incorporate demographic, clinical…and contextual data…to guide tailored strategies for resolution of distress, referral, and engagement in appropriate

follow-up services.” These are precisely the goals of the project described here. Evaluating SafeUT’s impact is essential to improving the service over time through counselor training, policy changes, and in future, the development of real-time decision-support systems that enhance counselors’ interventions.