Combining Patient Survey Data With Diagnosis Codes Improved Medicare Advantage Risk-Adjustment Accuracy

Under the current Medicare Advantage (MA) risk-adjustment system, plans are incentivized to report diagnosis codes on enrollees' medical claims reflecting additional and more severe health conditions to increase enrollees' risk scores and corresponding plan payments. To improve the integrity of risk adjustment, researchers have proposed four alternative methods to construct risk scores: calculate Hierarchical Condition Categories (HCC) scores excluding diagnosis codes from health risk assessments and chart reviews, calculate HCC scores excluding diagnosis codes most subject to score inflation, use pharmaceutical claims alone, and use self-reported survey responses alone or in combination with diagnosis codes. Using 2016-19 medical and pharmaceutical claims linked to Consumer Assessment of Healthcare Providers and Systems survey responses from 151,432 MA enrollees, we compared the predictive accuracy of each alternative strategy with the standard HCC approach. Relative to the standard HCC model, models combining HCC scores with survey responses were more predictive of health care use, explaining 5.8-6.0 percent of individual variation in total price-standardized MA utilization, compared with 5.1 percent. These findings suggest that survey responses can be used in tandem with diagnosis codes to strengthen MA risk adjustment.