Co-Leads: Anthony Rothschild, Feifan Liu; Collaborator: Ben Gerber

The Automated, Data-Driven, Adaptable, and Transferable Learning for Suicide Risk Prediction Project (ADAPT), aims to systematically assessing and improving a suicide risk algorithm’s generalizability and adaptability from an original development setting to a new healthcare system.

Machine learning algorithms predicting suicide and suicide attempts have demonstrated promising results. However, lack of large-scale external validations, transfer guidance, and automated learning-based adaptation impedes adoption in clinical practice. ADAPT aims to address the translation gap from research to clinical practice by assessing and improving the generalizability and adaptability of the NIMH-funded Mental Health Research Network (MHRN) suicide risk prediction algorithm when transferred to a different healthcare system, by building an innovative adaptation pipeline, ADAPT, for an scalable and automated end-to-end solution to optimally adapt existing suicide risk prediction models to a new healthcare context and in different clinical settings. We will explore strategies to fill potential performance gaps via model adaptation and advanced deep learning methods. The Specific Aims are to (1) assess the generalizability and adaptability of MHRN risk algorithms for suicide risk prediction (2) develop a unified pipeline of Automated, Data-driven, AdaPtable, and Transferable learning (ADAPT) for suicide risk prediction, (3a) to explore an innovative deep learning approach for suicide risk prediction, and (3b) to engage diverse stakeholders to assess ADAPT’s usability, acceptability, and feasibility, as well as potential barriers and facilitators to implementation.

Additional information about ADAPT is available on NIH RePORTER.