Predicting patient risk of readmission with frailty models in the Department of Veteran Affairs

Reducing potentially preventable readmissions has been identified as an important issue for decreasing Medicare costs and improving quality of care provided by hospitals. Based on previous research by medical professionals, preventable readmissions are caused by such factors as flawed patient discharging process, inadequate follow-ups after discharging, and noncompliance of patients on discharging and follow up instructions. It is also found that the risk of preventable readmission also may relate to some patient's characteristics, such as age, health condition, diagnosis, and even treatment specialty. In this study, using both general demographic information and individual past history of readmission records, we develop a risk prediction model based on hierarchical nonlinear mixed effect framework to extract significant prognostic factors associated with patient risk of 30-day readmission. The effectiveness of our proposed approach is validated based on a real dataset from four VA facilities in the State of Michigan. Simultaneously explaining both patient and population based variations of readmission process, such an accurate model can be used to recognize patients with high likelihood of discharging non-compliances, and then targeted post-care actions can be designed to reduce further rehospitalization.