Poster Paper: Incorporating Prescription Drug Utilization Information into the Marketplace Risk Adjustment Model Increases Payment Accuracy

Saturday, November 10, 2018
Exhibit Hall C - Exhibit Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Jianhui Xu1, Erin E Trish1 and Geoffrey Joyce2, (1)University of Southern California, (2)Schaeffer Center for Health Policy and Economics

The Affordable Care Act (ACA) imposed guaranteed issue and adjusted community rating in the individual market, prohibiting insurers from denying coverage or adjusting premiums based on health status or gender. In order to mitigate insurers’ incentives to select only healthier enrollees as a result of these premium rating policies, the ACA also created a risk adjustment policy, where funds are transferred from plans with low-risk enrollees to those with high-risk enrollees. While initial results suggest that this risk adjustment program has helped to neutralize these incentives overall, plans have been predictably over- or under-compensated for certain patient subgroups.

Thus, to improve payment accuracy, the Marketplace risk adjustment model started incorporating information on beneficiaries’ prescription drug utilization in the 2018 benefit year (prior to 2018, risk score calculation relied on demographic information and diagnoses from medical claims only). The model now incorporates information on the utilization of 12 drug classes (RXCs) in order to account for the under-recording of diagnoses on medical claims, especially for some conditions that do not need frequent doctor visits, and to reduce plans’ disincentive to enroll individuals taking expensive prescription drugs. Each class includes drugs that are clinically and empirically shown to be associated with its corresponding Hierarchical Condition Category (HCC) and that are not primarily prophylactic.

We evaluate claims data from a large private insurer to examine the spending of patients identified by HCCs and RXCs and the impact of including prescription drug utilization on plans’ selection incentives. We simulate risk scores both pre- and post-inclusion of RXCs for adults ages 21 to 64, focusing on three chronic conditions: HIV/AIDS, multiple sclerosis (MS), and diabetes. We find that the share of beneficiaries identified only by RXC (the RXC-only) ranges from 3 percent (MS) to 11 percent (HIV/AIDS). Compared to those identified only by HCC, RXC-only beneficiaries have lower risk scores in the absence drug utilization information, while incurring similar or higher plan spending. This under-compensation pre-inclusion could lead to considerable selection incentives, even with risk adjustment in place. After the inclusion of prescription information, risk scores for RXC-only beneficiaries increase substantially, and their predictive ratios (the ratio of mean group risk score to mean sample score over the ratio of mean group plan spending to mean sample plan spending) rise from well below 1 to above 1. Additionally, incorporating prescription drug information increases the predictive power of the model, explaining more variation in plan spending among patients (R-squared increases by 0.07, 0.14, and 0.01 for HIV/AIDS, MS, and diabetes, respectively).

These findings suggest that the inclusion of prescription drug utilization in the Marketplace risk adjustment model increases the risk scores of RXC-only patients, neutralizing under-payments for these patients and enhancing the predictive power of the model, thereby increasing risk adjustment payment accuracy. However, concerns remain about whether the new algorithm may incentivize plans to shift patients from multi-indication drugs that are not included in the model to drugs that are included, potentially resulting in other unintended effects on patient access, care delivery, and health outcomes.