Modeling the impact of care transition programs on patient outcomes and 30 day hospital readmissions

The increasing adoption of care transition programs - interventions designed to reduce hospital readmissions - has introduced a new challenge of evaluating such programs, i.e., assessing their impact on patient outcomes and care quality. This is difficult given the limited availability of program outcome data and analytical feedback exchange between providers. Moreover, the temporal nature of the effects of scheduled interventions on patient health raises the question of selecting and applying methodological tools appropriate for scientific research in this area. Our aim is to provide such methodological guidance and assist analysts, healthcare providers, and policy makers with extracting meaningful insights regarding the impact of care transition programs based on available data. We explore two well-known modeling approaches, Cox models and Markov chains, and using an illustrative example, demonstrate how they can be translated into informative analytic models with sufficient flexibility to analyze programs with diverse structures. We show that Cox Proportional Hazard models are particularly useful for identifying variables with the greatest impact on readmissions and quantifying the benefits of patient participation in a readmission reducing program. Extended Cox models provide an understanding of the effects of influential variables on readmissions as they change throughout the recovery period, allowing assessment of the relative benefits of care transition programs on different patient populations at specific times following a hospital discharge. Discrete Time Markov Chain models are particularly useful for assessing the impact of care transition programs in terms of expected time to readmission, facilitating the comparison of alternative program designs on patient outcomes.