Actuarial Notes

Risk Adjustment in State Medicaid Programs

By Ross Winkelman and Rob Damler · Society of Actuaries · 2008

Study guide: Exam GH-301

An article in the January 2008 issue (no. 57) of the Society of Actuaries Health Section's Health Watch on the most important considerations in implementing risk adjustment within a Medicaid Managed Care program. It draws on a guide published by the University of Maryland, Baltimore County and Actuarial Research Corporation, and sets out nine steps for implementation.

On the syllabus

  • Exam GH-301 — Risk Adjustment, outcomes 3a and 3b; the whole article.

Introduction

  • Risk adjustment, if done properly, allows managed care organizations (MCOs) to compete on how efficiently they deliver care and negotiate provider reimbursement rather than on how well they enroll the healthiest individuals.
  • Claims-based risk adjustment systems were first developed in the late 1980s; before them, rates were based primarily on age, gender, geographic region and other demographic characteristics, which have much lower predictive power.
  • The tools in use in Medicaid Managed Care capitation rate setting are CDPS, MedicaidRx, ACGs, CRxGs and DxCGs; each assigns members to demographic and morbidity categories with a risk weight.
  • A table lists states using or planning to use risk adjustment by system, individual or aggregate approach, and prospective or concurrent model.
  • Nine steps are listed: choose the system, the data, the eligibility groups, prospective or concurrent, individual or aggregate, whether to customize risk weights, the criteria for including individuals, the claims criteria, and the phase-in and risk corridors.

Two Important Definitions

  • The experience period is the data collection period, usually 12 months, usually preceding the period for which rates are paid.
  • The rate period is the period for which rates are paid, usually 12 months, with usually three to nine months between the two periods to allow claims to complete.

Choosing a Risk Adjustment System

  • The goal in Medicaid managed care is to capture overall relative risk at the MCO level, not the individual level, so among the top systems the data used and the ability to customize matter more than published individual-level accuracy.

Data to be Used

  • All risk adjustment tools use eligibility data (age, gender and eligibility category) because it is high quality and not subject to gaming.
  • Three broad categories of additional data: diagnosis data from inpatient admissions, diagnosis data from outpatient services, and pharmacy data.
  • Florida uses a pharmacy-based risk adjuster (MedicaidRx) initially, then a diagnosis-based model (CDPS) as MCO encounter data improves.

Which Eligibility Groups to Risk Adjust

  • Two considerations: how much health status varies among beneficiaries in the rate category, and whether the system captures that variation.
  • The SSI population shows significant variation; the TANF population shows less variation and high turnover; the SOBRA population consists of pregnant women who fail the TANF eligibility standards.

Prospective or Concurrent/Retrospective

  • Prospective risk adjustment uses experience period data to estimate morbidity for a future period; concurrent risk adjustment estimates morbidity during that same period and is more accurate.
  • Because retroactive rate adjustments are generally not favored by states or MCOs, most states use a prospective model.

Individual Versus Aggregate Risk Factor Calculation

  • Individual: risk scores for individuals calculated during the experience period follow beneficiaries, and an MCO's factor is the weighted average of the scores of those enrolled during the rating period.
  • Aggregate: the average risk score of enrollees during the experience period is assumed to represent the average risk of enrollees during the rating period.
  • The two approaches treat new and existing enrollees differently, as laid out in a table.

Customization of Risk Weights

  • Customization is often necessary because of benefit carve-outs, data coding differences, regional practice and utilization patterns, regional cost differences, differences in eligibility categories, and the desire to include individuals with limited exposure.
  • Birth and delivery "kick" payments, and mental health and pharmacy carve-outs, require customization.

Criteria for Including Individuals

  • Many states require at least six months of eligibility exposure in the experience period; pharmacy-based models require fewer months.

Phase-in and Risk Corridors

  • Phase-in is the portion of differences in risk adjustment applied to the MCO's capitation rate; risk corridors limit the upward or downward adjustment to revenue.
  • Risk corridors can make payments asymmetric and so make the system not budget neutral.

Other Considerations

  • Budget neutrality; timing of updates; data testing and validation; HIPAA considerations and controls; newborns.

Conclusion

  • Risk adjustment is an important tool to align incentives between health plans and state Medicaid managed care programs, and all stakeholders need to work collaboratively and openly share data and implementation decisions.

Sources