Intelligent selection of frequent emergency department patients for case management: A machine learning framework based on claims data

ABSTRACT

Frequent emergency department (ED) users impose a significant burden on the healthcare system. Case management (CM) can target potential frequent users to reduce their ED utilization. As CM is costly, it is essential to enroll individuals who will achieve improved health outcomes. We present a novel machine learning framework for effectively selecting enrollees for CM. Unlike traditional methods that only target current frequent users, our approach selects members for enrollment based on their likelihood of frequent use and their potential benefit from the program. We develop predictive models for two types of future frequent users—“jumpers” whose current ED usage is low but will increase significantly in the future, and “repeaters” whose ED usage remains consistently high. We propose a strategy to select optimal combinations of these two types of users, and compare the cost effectiveness. We demonstrate that the traditional enrollment strategy works well only for targeting potential repeaters, yet it will not result in positive savings unless the CM program is very effective in reducing ED usage. Including jumpers helps to improve cost effectiveness, because of the strength of the machine learning models, and jumpers are more likely to achieve successful outcomes from participation in a CM program.

KEYWORDS:

• Emergency department
• case management
• machine learning
• cost-effectiveness
• selection strategy