HEDIS® Risk-Adjusted Utilization Tables: New Measures, Shared Table Updates and FAQs

The Risk Adjusted Utilization (RAU) Tables and HEDIS^® MY 2026 Volume 2 Risk Adjusted Utilization Tables User Manual were released on March 31. These resources—available through the NCQA store—provide the logic and inputs for calculating the risk adjustment determination and weighting used in measures within the Risk Adjusted Utilization domain.

Why Risk Adjustment Matters

Individual health outcomes are shaped by underlying risk factors, which can distort comparisons between health plans if we do not properly account for them. Risk adjustment ensures that performance comparisons reflect differences in care delivery, not differences in the distribution of members’ health status (i.e., case mix). Risk adjustment allows for “apples to apples” comparison between health plans.

Essentially, risk adjustment asks: How would performance compare if all organizations had the same patient population?

There are multiple risk adjustment methods. NCQA uses statistical models for our HEDIS measures to predict outcomes by considering factors such as:

• Age and gender.
• Comorbidities.
• Procedure subtypes.
• Discharge conditions.

These models are the source of the risk weights found in NCQA’s RAU tables. Health plans use the tables to calculate an expected event rate, which is then compared to the observed event rate using an observed-to-expected ratio. The observed-to-expected ratio reflects risk-adjusted performance and shows whether a plan performed better or worse than expected based on its unique case mix.

When interpreting measure results, calibrate the ratio by dividing the individual organization ratio or national percentiles by the national average ratio. A calibrated ratio of <1.0 indicates better than expected performance, while a calibrated ratio of >1.0 indicates worse than expected performance.

For example, for the Plan All-Cause Readmission measure, a plan with a calibrated ratio of 0.8 may be successful at achieving fewer readmissions than expected, given its patient population.

New for MY 2026: Four Risk‑Adjusted Utilization Measures

NCQA has added four new RAU measures for HEDIS MY 2026:

• Acute Hospitalizations Following Outpatient Orthopedic Surgery (HFO).
• Acute Hospitalizations Following Outpatient General Surgery (HFG).
• Acute Hospitalizations Following Outpatient Colonoscopy (HFC).
• Acute Hospitalizations Following Outpatient Urologic Surgery (HFU).

These measures evaluate the risk-adjusted ratio of observed-to-expected unplanned acute hospitalizations (inpatient and observation stays) for any diagnosis within 15 days of an outpatient surgical procedure, for persons 65 years of age and older. Each measure focuses on a targeted outpatient surgical procedure.

Risk-Adjusted Tables Overview

NCQA publishes two types of RAU tables:

• Shared Tables: Provides the logic for mapping diagnosis codes into clinical categories and applies across risk-adjusted measures.
• Measure-Specific Tables: Provides measure-specific risk weights used to calculate expected values. There are 10 measure-specific tables—one for each risk-adjusted measure. Some measures report multiple product lines and each product line has its own set of weights.

Note: Measures in the Medicare product line have different sets of risk weights for enrollees ages 65+ and enrollees under 65.

Updates to the Risk-Adjusted Utilization Tables

The HEDIS MY 2026 RAU Shared Tables introduce a new table.

New: Table Proc-Mapping

A new tab titled “Table Proc-Mapping” was added to the Shared Table to support identification of procedure subtypes used in risk adjustment weights for three of the four new RAU measures (HFG, HFO and HFU). This table maps CPT codes to Clinical Classifications Software (CCS) procedure subtypes.

The risk adjustment model identifies all CPT codes associated with each outpatient surgery episode date. Each CPT code is assigned to a procedure subtype using Table Proc-Mapping. Only CPT codes in the denominator value set are included when assigning CPT codes to procedure subtypes. For example, in the HFU measure, only map the CPT codes in the Urologic Surgery Value Set. All associated CCS codes are captured for each episode. CPT codes that cannot be mapped to a CCS category are excluded.

Example

An outpatient surgery episode includes CPT codes 10160, 11762 and 15934:

• CPT 10160 maps to CCS 170 (Excision of skin lesion).
• CPT 11762 maps to CCS 175 (Other OR therapeutic procedures on skin/breast).
• CPT 15934 maps to CCS 170 (Excision of skin lesion).

Final procedure subtypes: CCS 170 and CCS 175 (with CCS 170 counted once). These CCS codes are used as risk weight variables in the risk adjustment calculation.

Note: The HFC measure currently does not assign CCS categories because colonoscopies only fall under one CCS category.

Conclusion

The MY 2026 Risk Adjusted Utilization Tables introduce new measures, enhanced mapping tools and substantive model updates designed to improve fairness and accuracy in health plan comparisons. By refining how underlying patient risk is captured, NCQA strengthens the reliability of HEDIS reporting—ensuring results reflect clinical performance, not population differences.

If you have any questions regarding the measures or ordering the RAU tables, submit a question to NCQA staff through My NCQA.

Frequently Asked Questions (FAQs)

Why did the risk weights change in MY 2024?

Risk weights are refreshed every 3–4 years to keep pace with changes in healthcare data patterns. The risk adjustment models are generated from past cross-sections of utilization data and are used to predict outcomes in future measurement years. As utilization patterns, coding practices, care management trends and population characteristics change, older models become less predictive.

NCQA also periodically re-estimates the models based on more contemporary data, allowing the variables included in the models and their associated weights to reflect changes to underlying relationships between the risk adjustment variables (e.g., age, gender, comorbidities as recorded in claims) and the outcomes (e.g., hospital readmissions). Re-estimating the models supports both measure reliability and validity.

NCQA derives many of the clinical conditions used in risk models from the CMS Hierarchical Condition Category (HCC) risk adjustment methodology. These risk models are also updated regularly. The Shared Tables include a tab summarizing changes for that year.

When the weights and models are re-estimated, new data is incorporated, which can reveal changes in the relationships among different variables.

Why might a condition that appears to be more severe be assigned a lower HCC risk weight than a related condition?

Several statistical and population-based factors can cause this:

• Multicollinearity (or sometimes just collinearity): There is a correlation among HCCs; people with a “severe” level condition might be more likely than people with a “moderate” level of the same condition to have other HCCs that absorb some of the excess risk associated with the condition.
• Outlier exclusion: People with a severe level of a condition may be considerably more likely to have enough hospitalizations to reach the outlier threshold and thus be excluded from the denominator entirely.
• Compositional effect: Those who remain could be unusually unlikely to experience an event, which can be thought of as a compositional effect.

Any of these dynamics could result in the “moderate” or “mild” level of a condition having a higher risk weight than the “severe” level of the condition.

What models are used for the RAU measures?

NCQA employs statistical prediction models to estimate expected event rates for each measure outcome. To obtain the risk weights, statistical relationships between the potential risk adjustors and the outcomes are assessed using generalized linear models:

• Logistic regression is used to estimate model coefficients and values are summed across a plan population for measures with outcomes based on proportions (i.e., each denominator unit can only have one instance of the outcome).
• Logistic + Poisson regressions are used to estimate model coefficients for measures with outcomes based on rates (i.e., each denominator unit can have many instances of the outcome).

The expected rates derived from the models are compared to observed performance to generate risk-adjusted performance assessments (observed-to-expected ratios). NCQA fits these separately for each utilization measure to produce risk weights.

Can you give more details about the statistical models you use?

For the Plan All-Cause Readmissions (PCR) measure and the Hospitalization Following Discharge From a Skilled Nursing Facility (HFS) measure, NCQA uses penalized logistic regression to predict whether an index hospitalization will result in a readmission.

For the other risk-adjusted measures, NCQA uses penalized logistic regression to predict whether the denominator member would have any numerator event (versus none) and then penalized Poisson regression to predict the number of numerator events, among those who have at least one.

Each measure accounts for a combination of risk weight variables:

• Age and gender.
• Comorbidities (HCCs).
• Procedure type.
• Discharge conditions.
• Surgeries.
• Observation stay discharge.
• COVID discharge.
• Medication.

Note: Not every measure or product line has every type of variable.

In addition, the risk-adjustment models consider interactions using the “combination” HCCs, which are specified in the Shared Tables, as some combinations present a greater amount of risk when observed together.

The models address effect modification by estimating separate sets of risk weights for different populations (e.g., Medicaid, Medicare age 18–64, Medicare age 65+).

Example

Considering the PCR measure, the model specifies that the log odds of a hospital readmission within 30 days of an index hospital discharge are a linear combination of a set of indicators:

• Age and gender combinations (of which each denominator unit belongs to exactly one; all combinations are shown in the risk weight tables).
• Comorbidities observed via diagnosis codes in claims in the year prior to the index hospital discharge (shown in the Shared Tables with the HCC labels).
• Conditions primarily associated with the index hospital stay itself (these have the “discharge CC” label in the Shared Tables and/or risk weight tables).
• Whether the index hospital stay was associated with a surgery.
• Whether the index hospital stay was an observation stay.
• Whether the index hospital stay had a principal discharge diagnosis of COVID-19 (for Medicare 65+ only).

Not all possible predictors are in each population’s set of risk weights, which means that for some populations, some of the risk weights are zero.

HEDIS^® is a registered trademark of the National Committee for Quality Assurance (NCQA).