Actuarial Notes

HHS-Operated Risk Adjustment Technical Paper on Possible Model Changes

By Center for Consumer Information and Insurance Oversight, Centers for Medicare & Medicaid Services · Centers for Medicare & Medicaid Services · 2021 · GH301-107-25

Study guide: Exam GH-301

A technical paper of October 26, 2021 from the Center for Consumer Information and Insurance Oversight (CCIIO) of the Centers for Medicare & Medicaid Services on possible changes to the HHS-operated ACA risk adjustment models. It gives further detail and analyses on the model updates proposed in the 2022 Payment Notice, and information on HHS's ongoing evaluation of the cost-sharing reduction induced demand factors in the state payment transfer formula.

On the syllabus

  • Exam GH-301 — Risk Adjustment, outcome 3a; Ch. 1, 5 & Appendix A.

Overview

  • The ACA established a permanent risk adjustment program to minimize the negative effects of adverse selection, transferring funds from issuers that enroll lower-than-average risk populations to issuers that enroll higher-than-average risk populations.
  • Chapter 1 explains how the models currently work and the focus on improving predictive accuracy for certain subpopulations; Chapters 2 to 4 explore improvements for the lowest-risk enrollees, partial-year enrollees and the highest-risk enrollees; Chapter 5 gives the combined results.

1 Overview and Current Status of Risk Adjustment Models

  • 1.1 Overview of Risk Adjustment
    • 1.1.1 History
    • 1.1.2 Purpose and Goals of Risk Adjustment
  • 1.2 The Current HHS Risk Adjustment Models
    • 1.2.1 Principles of Risk Adjustment
    • 1.2.2 Calculation of the Risk Score
    • 1.2.3 Calculations Under the State Payment Transfer Formula
  • 1.3 Previous Risk Adjustment Model and Transfer Formula Updates
  • 1.4 Predictive Accuracy of the Current HHS Risk Adjustment Models
    • 1.4.1 Purpose of this Paper

2 Improving the Models' Predictive Accuracy for Lowest-Risk Enrollees – Two-Stage Weighted Approach

  • 2.1 Current Models' Prediction of Costs for Lowest-Risk Enrollees
  • 2.2 Improving Prediction for Lowest-Risk Enrollees – Two-Stage Weighted Model Specification
  • 2.3 Impact of the Two-Stage Weighted Approach
  • 2.4 Conclusion

3 Improving the Models' Predictive Accuracy for Partial Year Enrollees – Enrollment Duration Factors

  • 3.1 Background on the Enrollment Duration Factors
  • 3.2 Potential Changes to the Enrollment Duration Factors
  • 3.3 Testing Other Options for the Enrollment Duration Factors
    • 3.3.1 Partial-Year Enrollment Type
    • 3.3.2 Market
    • 3.3.3 Specific HCCs
  • 3.4 Obtaining Diagnosis Codes for Partial-Year Enrollees
  • 3.5 Conclusion

4 Improving the Predictive Accuracy for the Very Highest-Risk Enrollees – Interacted HCC Counts

  • 4.1 Current Models' Prediction of Costs for the Very Highest-Risk Enrollees
  • 4.2 Improving Prediction for the Very Highest Risk Enrollees – Interacted HCC Counts Approach
  • 4.3 How the Interacted HCC Counts Approach Works
  • 4.4 Interacted HCC Counts Approach Change Impact
  • 4.5 Conclusion

5 Results and Estimated Impact of Model Specification Changes Under Consideration

  • 5.1 Combined Impact of Model Specification Changes
    • Describes model performance metrics of the combined proposed model specification changes using the 2018 enrollee-level EDGE dataset.
  • 5.2 EDGE Server Transfer Simulation
    • Outlines the methodology of a simulation applying the proposed 2022 model specification changes to 2020 benefit year data from issuers' EDGE servers to estimate what 2020 transfers would have been.
  • 5.3 Conclusion

Appendix A Cost-Sharing Reduction Induced Demand Factors

  • A.1 Background on the CSR Induced Demand Factors
  • A.2 Stakeholder Feedback on CSR Induced Demand Factors
    • A.2.1 Risk Term Analysis
    • A.2.2 Rating Term Analysis
  • A.3 Potential Future Changes to CSR Induced Demand Factors
    • A.3.1 Risk Term Update Options
    • A.3.2 Rating Term Update Options
  • A.4 Next Steps
  • The appendix describes the factors used in the state payment transfer formula, stakeholder feedback on them, and options likely to be implemented no earlier than the 2024 benefit year.

Appendix B List of Acute HCCs Used in Enrollment Duration Factors Analysis

Appendix C HCCs in the 2021 Adult Model Severity Illness Indicator Variable and HCCs Selected for HCC Interacted Counts Variables for the Adult and Child Models

Sources