Association rule mining for risk prediction and stratification: A philips lifeline case study

Abstract

Personal emergency response systems (PERS) such as Philips Lifeline help seniors maintain independence and age in place. PERS can use predictive analytics to help risk stratification and promote response-efficient emergency services. This paper presents a framework for estimating significant associations between Lifeline user characteristics and occurrence of emergency events. Predictive variables including demographics, health conditions, environmental, and user-specific lifeline history were identified and their associations to emergency events were delineated. The predictive variables can help with 1) identifying individuals at high risk and 2) management and prioritization of care and preventive services, which can result in reducing adverse health events and improving user's quality of life.