A Machine Learning Framework to Prevent 'Gaming' in Plan Payment Risk Adjustment
*Sherri Rose, Harvard Medical School
Keywords: machine learning, risk adjustment
Risk adjustment models for plan payment are typically estimated using classical linear regression models. These models are designed to predict plan spending, often as a function of age, gender, and diagnostic conditions. The trajectory of risk adjustment methodology in the federal government has been largely frozen since the 1970s, failing to incorporate methodological advances that could yield improved formulas. The use of novel machine learning techniques may improve estimators for risk adjustment, including reducing the ability of insurers to “game” the system with aggressive diagnostic upcoding. This upcoding has been recently estimated to cost more than $11 billion in excess payments in Medicare Advantage annually. We present a nonparametric machine learning framework for risk adjustment in the Truven MarketScan database and assess whether use of these procedures improves risk adjustment in general, as well as in upcoding settings. This framework accommodates variable screening, as well as effect measures when causal inference is of primary research interest, instead of prediction.