The three main risk adjustment models used in U.S. healthcare are the CMS Hierarchical Condition Categories (CMS-HCC) model for Medicare, the HHS Hierarchical Condition Categories (HHS-HCC) model for ACA marketplace plans, and the Chronic Illness and Disability Payment System (CDPS) for Medicaid. Each model serves a different population, but they all do the same basic thing: predict how much a person’s healthcare will cost based on their age, sex, and medical conditions, then adjust payments to insurers accordingly.
Risk adjustment exists to keep the insurance market fair. Without it, plans that enroll sicker patients would lose money while plans with healthier members would profit, creating a strong incentive to avoid covering people who actually need care. These models remove that incentive by transferring money from plans with lower-risk enrollees to plans with higher-risk enrollees.
CMS-HCC: The Medicare Model
The CMS-HCC model is the oldest and most widely discussed of the three. It’s used to set payments for Medicare Advantage plans, the private insurance option available to people 65 and older and those with qualifying disabilities. The model works prospectively, meaning it uses diagnoses from the current year to predict costs in the following year. This is an important distinction because it means the payment amount is set before a patient’s care actually happens.
CMS calculates each person’s risk score by adding together “relative factors,” which are numerical weights assigned to demographic variables (age and sex) and each diagnosed condition category. The score starts at zero, and each applicable factor gets added on. A person with diabetes and heart failure, for example, would have the weights for both conditions added to their demographic baseline. The result is a raw risk score that represents how costly that person is expected to be relative to an average Medicare beneficiary.
The “hierarchical” part of the name refers to how the model handles related conditions of different severity. If someone has both a mild and severe form of the same type of illness, only the most severe version counts. This prevents double-counting within a disease category while still capturing the full range of a person’s unrelated conditions.
CMS recently completed a major update to this model. Version 28 (V28) replaced the older Version 24, phased in over three years: one-third V28 in 2024, two-thirds in 2025, and full implementation in 2026. The update uses more recent data and a modernized mapping of diagnosis codes to condition categories. According to MedPAC, the independent body that advises Congress on Medicare, coding intensity in Medicare Advantage is projected to increase payments to plans by about 4 percent even after applying a coding adjustment, which remains an ongoing policy concern.
HHS-HCC: The ACA Marketplace Model
The HHS-HCC model was built specifically for the Affordable Care Act’s individual and small group insurance markets. It covers a fundamentally different population than CMS-HCC: primarily working-age adults and their families, most of whom are under 65. Like its Medicare counterpart, it uses demographics and diagnoses to generate a risk score that reflects how costly a person is expected to be for their plan.
CMS used the Medicare model as a starting point when developing HHS-HCC but had to make three significant adaptations. First, the HHS model is concurrent rather than prospective. It uses diagnoses from the same year it’s predicting costs for, which gives it stronger predictive accuracy. Concurrent models generally do a better job of matching payments to actual costs, though they also create greater incentives for aggressive coding since a diagnosis recorded today directly affects this year’s payment.
Second, the population is different. Medicare data comes from people 65 and older or those with disabilities. Marketplace enrollees tend to be younger and have different patterns of illness and healthcare use. Third, the HHS model accounts for prescription drug spending alongside medical costs, whereas the original CMS-HCC model was designed to predict medical spending only. This matters because drug costs make up a significant share of total spending for many conditions common in a younger, commercially insured population.
In practice, the HHS-HCC model drives a transfer payment system. At the end of each year, plans whose enrollees were healthier than average (lower risk scores) pay into a pool, and plans whose enrollees were sicker than average (higher risk scores) receive money from that pool. This is what allows insurers to offer coverage to everyone at community-rated premiums without being financially penalized for attracting people with serious health needs.
CDPS: The Medicaid Model
The Chronic Illness and Disability Payment System takes a different approach tailored to Medicaid’s unique population. Developed by researchers at the University of California, San Diego, CDPS is a diagnostic classification system that state Medicaid programs can use to adjust capitated payments to managed care organizations. Capitated payment means the insurer receives a fixed amount per enrollee per month, regardless of what care that person actually uses.
CDPS was designed to address the wide diversity of diagnoses and illness burden among Medicaid beneficiaries, particularly two groups: people who qualify through disability and those who qualify through low income (historically called TANF or AFDC recipients). These populations have very different healthcare needs and cost profiles. A flat per-person payment would drastically overpay for healthy enrollees and underpay for those with complex chronic conditions or disabilities, giving managed care plans a financial reason to avoid the sickest patients.
The model groups diagnoses into categories and assigns payment weights that states can apply when setting managed care rates. Because Medicaid is administered at the state level, CDPS functions more as a tool states can adopt and customize rather than a single federally mandated formula. Some states also use pharmacy data alongside diagnostic data to improve their risk adjustment, since prescription records can reveal conditions that might not appear in claims data if a patient hasn’t had a recent office visit.
How Risk Scores Are Calculated
All three models follow a similar additive logic. Each person starts with a baseline score of zero. The model then adds numerical weights for demographic factors like age and sex, plus additional weights for each qualifying condition category. If you’re a 72-year-old man with diabetes and chronic kidney disease, your risk score would be the sum of the weight for your age-sex group, the diabetes category weight, and the kidney disease category weight. The final number represents your predicted cost relative to the average person in that program.
A risk score of 1.0 means average expected cost. A score of 1.5 means the person is predicted to cost 50 percent more than average. A score of 0.7 means 30 percent less. Insurers receive higher payments for members with higher scores and lower payments for those with lower scores. This is why accurate documentation of diagnoses matters so much: every condition that gets recorded and coded affects the risk score, which directly affects how much money flows to the plan.
Key Differences Between the Models
- Population served: CMS-HCC covers Medicare (primarily 65+), HHS-HCC covers ACA marketplace plans (primarily under 65), and CDPS covers Medicaid (low-income and disabled populations of all ages).
- Timing of prediction: CMS-HCC is prospective, using this year’s diagnoses to predict next year’s costs. HHS-HCC is concurrent, using the same year’s diagnoses to match that year’s costs. CDPS varies by state implementation.
- Spending scope: CMS-HCC originally predicted medical costs only. HHS-HCC includes both medical and prescription drug costs. CDPS can incorporate pharmacy data depending on the state.
- Administration: CMS-HCC and HHS-HCC are federally standardized. CDPS provides a framework that states adapt to their own Medicaid programs.
Together, these three models cover the vast majority of Americans enrolled in government-sponsored or government-regulated health insurance. They represent different solutions to the same problem: making sure health plans are paid fairly for the actual health risk of their members, so that covering sick people is financially sustainable rather than something insurers try to avoid.