Community OHCA Risk Prediction Using Machine Learning: A Multi-Domain Approach to Prevention and Preparedness

About This Grant

PROJECT SUMMARY Out-of-hospital cardiac arrest (OHCA) is a leading cause of death in the United States (US). Advances in resuscitation science have improved survival rates in some communities, but wide variability in incidence and survival outcomes persist. Multiple geographic and temporal factors contribute to this variability such as comorbidity burden, place effects, access to healthcare, poverty, community resources, and variation in clinical care policies. Given the substantial public health burden of OHCA and marked geographic variability in incidence and survival, developing a targeted framework to identify and measure OHCA incident and outcome risks is essential. We will employ a participatory and mixed methods approach that combines machine learning (ML) and artificial intelligence with qualitative research methods to develop and evaluate an OHCA risk score and a virtual laboratory (VL) environment as decision support tools to inform community-level interventions to improve OHCA outcomes. We will first engage community representatives and officials, involved in the OHCA system of care (e.g. community service organizations, emergency medical service providers, hospital quality assurance officers, public health officials, and cardiac arrest survivors) to participate in focus groups and key informant interviews to identify optimal and efficient data elements to define a scalable and usable OHCA risk score (Aim 1). Based on this information, we will then employ ML methods to develop the OHCA risk score and VL environment (Aim 2), which will then be discussed and evaluated by community representatives (Aim 3). These elements of participatory ML will provide important context for data interpretation while building trust in the OHCA risk score and VL environment as pre-implementation tools to diagnose local delivery capabilities and develop implementation strategies to overcome any barriers identified. The OHCA risk score and VL environments resulting from this project can inform public health messaging, aid local public health departments and hospitals to identify areas where surveillances needs to be heightened, and inform government agencies where to direct funding and resource allocation as it pertains both to the chain of survival as well as prevention and early identification of patients at risk for OHCA. This work is a necessary first step to direct strategic investments in emergency response infrastructure and community-level interventions to improve preparedness and optimize OHCA survival outcomes.