Risk adjustment exists to solve a fundamental problem in healthcare: not everyone costs the same to treat, but the system needs to work fairly for both patients and the organizations covering their care. Without it, insurers would have every reason to chase healthy enrollees and avoid sick ones, payments wouldn’t reflect actual patient needs, and the entire insurance market would tilt toward plans that are best at dodging costly patients rather than caring for them.
How Risk Adjustment Works
At its core, risk adjustment uses a person’s age, sex, and medical diagnoses to calculate a risk score, a number that predicts how expensive that person’s healthcare will be relative to the average. Someone with well-controlled high blood pressure gets a lower score than someone managing diabetes with acute complications, who in turn scores lower than someone with multiple organ-system diseases.
The most widely used system is the Hierarchical Condition Categories (HCC) model. Your diagnosis codes from medical claims get sorted into condition categories, which are then arranged in hierarchies so that only the most severe version of a related condition counts. For example, diabetes with acute complications ranks higher than diabetes without complications, and only the higher category factors into the final score. The model then uses these categories alongside demographic information to produce a single number representing your predicted cost of care.
That score drives real money. In Medicare Advantage, plans receive higher payments for members with higher risk scores. In the Affordable Care Act marketplaces, the system transfers funds from plans whose enrollees are healthier than average to plans whose enrollees are sicker than average. The formula averages all individual risk scores within a plan, then uses those plan-level averages to calculate the dollars that move between insurers.
Stopping Cherry-Picking Before It Starts
The ACA requires insurers to charge similar premiums regardless of a person’s health status. That’s a critical consumer protection, but it creates a perverse incentive: if you can’t charge sick people more, the next best strategy is to avoid enrolling them altogether. Plans could design narrow networks that exclude top cancer centers, bury their prescription drug coverage, or make enrollment paperwork harder for people with complex conditions. This behavior is called adverse selection, or more bluntly, cherry-picking.
Risk adjustment neutralizes that incentive. Plans that attract sicker enrollees receive transfer payments from plans with healthier populations, so there’s no financial penalty for covering people with serious health needs. Research on ACA marketplace enrollees with mental illness found that risk adjustment significantly reduced both underpayment to plans serving those populations and overpayment to plans that happened to attract healthier members. The result is that insurers compete on the quality and efficiency of care, not on their ability to screen out expensive patients.
Making Payments Match Patient Needs
Without risk adjustment, every insurer or provider organization would receive roughly the same payment per person, regardless of that person’s health. A flat payment of, say, $15,792 per member (the average in one recent Medicare analysis) works fine for the statistically average enrollee but dramatically overpays for healthy 30-year-olds and dramatically underpays for someone managing heart failure and diabetes simultaneously.
The HCC model cuts that payment prediction error substantially. In a cross-sectional analysis of Medicare spending, the HCC model reduced the mean absolute error in predicting individual healthcare costs by $2,337 per person compared to simply paying everyone the average amount. That’s a meaningful improvement: it means the money flowing to a plan or provider organization more closely tracks the actual cost of caring for its patients. More advanced statistical models have shown the potential to reduce prediction error even further, by $3,690 per person relative to flat payments, though the current HCC system balances accuracy with transparency and simplicity.
Why Documentation Accuracy Matters
Risk scores are only as good as the diagnosis codes that feed them, which means the accuracy of the entire system rests on clinical documentation. Every chronic condition a patient has needs to be recorded during a visit, assessed, and linked to a current treatment plan. If a provider manages a patient’s COPD but doesn’t document it because it wasn’t the reason for that particular appointment, the condition drops out of the risk score calculation.
This is a real and measurable problem. Analysis comparing traditional Medicare and Medicare Advantage found that conditions like vascular disease, depression, substance use disorders, and COPD were among the biggest contributors to coding gaps between the two programs. Conditions requiring frequent visits, like HIV/AIDS and diabetes, showed coding persistence above 90 percent. But conditions that might not come up at every visit, like stable COPD or well-managed paraplegia, often went unrecorded, dragging risk scores below where they should be.
Healthcare organizations invest heavily in coding documentation improvement programs for exactly this reason. These programs train providers on best practices for recording all active diagnoses, using diagnosis codes at the highest appropriate level of specificity, and ensuring the medical record supports every code submitted. The goal isn’t to inflate scores. It’s to make sure the data accurately reflects how sick a patient population actually is.
The Foundation for Value-Based Care
Risk adjustment isn’t just an insurance mechanism. It’s the financial scaffolding underneath nearly every value-based care arrangement in the country. When a provider group accepts a fixed per-patient payment (capitation) or agrees to be accountable for total cost of care, risk adjustment determines the baseline: how much money they receive to care for their specific panel of patients.
CMS describes this directly: in value-based models, you get paid more for patients who have more health problems than for healthy patients who may not need as many services. The pre-payment amount accounts for the health conditions that predict whether a patient will cost more or less than average. Without accurate risk adjustment, a primary care practice that specializes in managing complex chronic disease would be financially penalized compared to one seeing mostly healthy young adults, even if both delivered excellent care. Risk adjustment levels that playing field so providers can focus on outcomes rather than patient selection.
Social Factors and the Limits of Current Models
One active area of development is whether social determinants of health, factors like income, housing stability, and neighborhood resources, should be built into risk adjustment formulas. The logic is straightforward: two patients with the same diagnoses may cost very different amounts to treat if one has stable housing and reliable transportation while the other doesn’t.
Early results are mixed. A 2025 Society of Actuaries study examining Medicaid risk adjustment found that adding community-level social factors as separate variables had limited impact on payment accuracy. The researchers concluded that much of the influence of social determinants was already being captured indirectly through the condition categories themselves, since people facing more social disadvantage tend to accumulate more diagnoses. Community-level data also lacks the granularity needed to distinguish between individuals in the same ZIP code who may have very different circumstances.
Individual-level social determinant data may prove more useful. As of January 2024, a new billing code became effective specifically to pay providers for administering social determinant screenings, which could generate the individual-level information future models need. For now, risk adjustment is best understood as one component of a broader approach to addressing health inequities, not a standalone fix.
Recent Changes to the HCC Model
In 2024, CMS began paying Medicare Advantage organizations under a significantly updated version of the HCC model, version 28, replacing the version 24 model that had been in use. The new model decreased the number of diagnosis codes that map to a payment-relevant condition category while simultaneously increasing the number of condition categories that trigger additional payment. In practical terms, this means the model got more selective about which diagnoses count but more granular in how it distinguishes between levels of disease severity. The HHS Office of Inspector General is actively studying the patterns and payment implications of this transition.