Over 44,000 people died by suicide in the U.S. in 2016 and national rates continue to increase. The majority of people who died by suicide had contact with the health care system in the year prior to their death. Major hurdles to implementing suicide prevention in healthcare settings include the lack of scalable and sustainable methods for training routine healthcare providers in suicide prevention. Innovations in machine learning and artificial intelligence may overcome these hurdles as it is now possible for technology to assess the quality of provider skill in intervention delivery and provide opportunities for skill acquisition and practice. The candidate's long-term goal is to harness technological advances in artificial intelligence, natural language processing, and machine learning to improve the scalability and sustainability of training among general medical providers in suicide prevention. The proposed research and training activities will take place at the University of Washington at Harborview Medical Center in Seattle, WA, a county safety-net hospital and level I trauma center serving patients across Washington, Wyoming, Alaska, Montana and Idaho.
The research aims to adapt and deploy existing scalable technology to train frontline trauma center providers (e.g., nurses) to collaboratively engage patients in a suicide safety planning intervention (SPI) and conduct a pilot feasibility trial of the resultant training.
Aim 1 includes focus groups with trauma nurses to identify individual, setting, and organizational-level implementation barriers and facilitators based on the Theoretical Domains Framework and inform strategies for engaging nurses in training and delivery of the SPI with patients.
Aim 1 also includes the user-centered design method of contextual inquiry, including task analysis, with nurses to inform workflow- integration.
Aim 2 includes user-centered design methods to identify technology refinements and adaptations based on nurse preferences to increase usability. The technologies are a 1) conversational agent, with simulated patient role-play and real-time feedback, and 2) AI-based feedback of counseling performance from SPI audio recordings.
Aim 3 is to conduct a pilot randomized trial of a technology-enhanced provider training as compared to a web-based didactic only condition. The longitudinal trial will include 20 nurses (10 per condition), each with 3 patients, and support submission of an NIMH R01 full-scale trial. The K23 training goals include building knowledge and skills in 1) technology-focused team science, 2) the application and integration of implementation science, user-centered design, and adult learning theory for technology adaptation and integration for nurse training, 3) acute care suicide prevention clinical trials research, including the responsible conduct of research with patients at-risk for suicide, and statistical methods for low base-rate outcomes and nested longitudinal clinical trials data. This K23 application addresses the NIMH Strategic Plan by developing strategies incorporating information technology and pragmatic feedback systems for suicide prevention efforts in real-world practice, reaching the full breadth of patients presenting to the health care system after injury.
Each year in the United States approximately 30 million individuals require emergency department visits; 1.5- 2.5 million are so severely injured that they require inpatient hospitalization and over 650,000 patients present for self-inflicted injury. Acute care medical settings, such as emergency departments and inpatient trauma units, present critical opportunities for suicide prevention, which is stymied by a lack of scalable and sustainable methods for training providers in high quality suicide prevention interventions. The proposed K23 harnesses advances in artificial intelligence, natural language processing, and machine learning to improve the scalability and sustainability of suicide prevention training among acute medical care providers to better reach the population of patients at-risk for suicide with potentially life-saving interventions.
