Association for Public Policy Analysis & Management

The proliferation of predictive modeling in health care has led to an environment in which the opportunities for (and the pressures on) public payers’ adopting such methods are growing rapidly. There are vast academic literatures regarding the use of predictive modeling; however, there are few if any practical tools available to health policy leaders to assist them in making their own judgments regarding the suitability and effectiveness of predictive methods. The goal of this policy brief is to provide the requisite translational bridge. The specific objectives are: 1. To provide guidance regarding which types of health policy questions are well-suited for predictive modeling methods and which are not; 2. To create a visual aid that facilitates an intuitive understanding of traditional statistical measures of predictive model performance; and 3. To develop a checklist that guides policymakers in appraising the quality of a particular predictive analytics application.  In summary, this brief translates academic predictive modeling methods into practical, usable tools supporting health policy leaders.