Given that risk adjustment is fundamental to these programs, they continue to demand considerable attention for ensuring coding accurately and appropriate rate levels.
Although varying based on the application, risk adjustment in health insurance is essentially a process that incorporates health conditions, demographic profiles, and plan-level information to estimate healthcare costs. In other words, risk adjustment attempts to capture the risk level of a health plan’s population, and it is a crucial risk mitigation component in programs seeking to ensure adequate coverage and access to care in the healthcare marketplace.
The MA and ACA populations, both subject to risk adjustment, are significant segments of the U.S. health benefits coverage ecosystem and continue to expand. MA enrollment has more than tripled in the past 16 years, increasing MA penetration to a majority of the Medicare-eligible population. In 2024, 32.8 million individuals were enrolled in a MA plan, representing over half (54%) of eligible Medicare beneficiaries, as shown in Figure 1 and Figure 2.1 The ACA individual health insurance marketplace has also experienced growing enrollment, totaling 21.4 million enrollees as of 2024, as shown in Figure 3.2
Figure 1: Medicare Advantage penetration 2007-2024
Figure 2: Medicare Advantage enrollment 2007-2024
Figure 3: ACA individual marketplace enrollment 2014-2024
The CMS-HCC and HHS-HCC models
How the models work, and their impact on risk scores
In this paper, we focus on the MA and ACA populations, despite the existence of different risk adjusters in other markets, e.g., Chronic Illness and Payment System (CDPS), Diagnostic Cost Group (DCG). The current risk adjustment models used in the MA and ACA markets, created and managed by the Centers for Medicare and Medicaid Services (CMS) and the U.S. Department of Health and Human Services (HHS), are the CMS-HCC model and the HHS-HCC model, respectively. While different in many respects, both models leverage a hierarchical condition category (HCC) classification system, which is a grouping methodology to assign hierarchies or families of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes to a single HCC code.3 In other words, the main purpose of the HCC systems is to accurately predict or measure plan-level risk by categorizing medical diagnoses into clinical conditions or groups of related conditions with varying degrees of severity or risk. We summarize a few key differences between the two models in Figure 4.
Figure 4: Differences between CMS-HCC and HHS-HCC models
| CMS-HCC | HHS-HCC | |
|---|---|---|
| Target population | Medicare-eligible, over age 65 and/or disabled. | Individual and small group ACA population, primarily under age 65. |
| Purpose | Developed by CMS for risk adjustment reimbursement of the Medicare Advantage (Part C) program. Reimbursement is not dependent on other plans’ risk scores. | Developed by HHS to determine payment transfers among participating plans. These transfers sum to zero within a risk pool separately for individual, catastrophic, and small group plans. |
| Coverage | Predicts relative medical spending only. CMS also developed the RxHCC model for risk adjustment of the Medicare prescription drug (Part D) population. |
Predicts the relative sum of medical and drug plan paid amounts (up to $1 million per member).4 |
| Data used for development | Medical diagnoses and costs only from Medicare fee-for-service (FFS) claims. | Both medical diagnoses and prescription drug data from historical ACA population. |
| Time scope | Prospective—based on data from the base year to predict costs for the following year. | Concurrent—based on data from the current benefit year. |
Figure 5: ICD-10-CM codes vs. HCCS (CY2024 CMS-HCC)
According to the latest report for the calendar year (CY) 2024 CMS-HCC risk model adjustments, and as shown in Figure 5, out of 73,926 ICD-10-CM codes, 7,770 are risk-adjustable and included in the finalized HCC model, categorized into 115 HCCs.5
Ultimately, the HCCs are utilized in conjunction with demographic and plan-level information to develop a risk adjustment factor (RAF) that impacts plan risk scores and, subsequently, adjusts the payments to health plans. It is important to note that not all ICD-10-CM diagnoses are linked to an HCC that affects risk scores and, therefore, not all health conditions are risk adjustable: 90.5% of ICD-10-CM codes are not included in the final payment model. Further, even for risk adjustment-eligible diagnoses, not all performed services qualify for inclusion in the risk adjustment models. Each HCC contributes differently to a member’s RAF, and some HCCs interact with or override the contributions of other HCCs.
Within MA, the CMS-HCC risk scores directly impact plan revenue. For ACA plans, the HHS-HCC risk scores determine payment transfers among participating plans.
Successful risk adjustment and payment reimbursement requires Medicare Advantage organizations (MAOs) and ACA payers to ensure that physicians properly code their patients’ conditions. The HHS Office of Inspector General (OIG) released an analytics tool kit in December 2023 providing a methodology to MAOs to identify diagnosis codes submitted to CMS for use in CMS’s risk adjustment program that are at high risk for improper coding. This tool kit compiles groups of diagnosis codes OIG audits suggest are at high risk for incorrect specification and provides explanations and methodologies to help plans identify these groups in their own populations. According to the OIG, for the several audits performed so far approximately 70% of submitted risk adjustment-eligible diagnosis codes from this relatively small set of high-risk codes were not supported in the associated medical records, increasing up to 90% for some codes.6 For certain groups such as acute stroke and breast cancer codes, the OIG found that upwards of 96% of submitted codes contained errors. However, several audits released by the OIG were disputed by some major payers.7 At a minimum, MAOs have an opportunity to ensure proper coding and confirm medical record support for the uncommon and unique set of circumstances that the OIG identified as being at high risk for improper reporting. ACA plans may also consider applying similar methodologies for their populations.
Both the MA and ACA programs enforce a risk adjustment data validation (RADV) process to assess how well the diagnoses submitted for risk adjustment are supported in the medical records. It is the health plan’s responsibility to ensure overall coding accuracy—both adding diagnosis codes for conditions not initially captured and deleting codes not supported. Through the ACA RADV process, HHS will adjust issuer risk scores in specific cases based on the error rate identified during data validation, which leads to adjustments to payments and charges across the risk pool. MA RADV audits can also lead to adjustments in payments to individual MA plans. In addition to direct financial impacts, significant noncompliance could affect a plan’s reputation and its ability to enroll new members.
HCC prevalence
The Centers for Disease Control and Prevention (CDC) analyzed results from the Behavioral Risk Factor Surveillance System (BRFSS), which is an annual randomized state-based telephone survey of noninstitutionalized adults (age greater than 18). It found that, in 2019, over 53.8% (39.8 million) of adults between ages 18 and 34 reported having at least one chronic condition, as well as 22.3% of adults reporting having more than one chronic condition.8
The most common reported chronic conditions were:
- Obesity (25.5%)
- Depression (21.3%)
- High blood pressure (10.7%)
- High cholesterol (9.8%)
- Asthma (9.2%)
The National Center for Chronic Disease Prevention and Health Promotion (NCCDPHP), a division of the CDC, has a Chronic Disease Indicators (CDI) tool, which provides national and state data on chronic diseases.9
Chronic Conditions Data Warehouse provides national CMS Medicare research data for both Medicare and Medicaid, which includes Medicare chronic condition prevalence.10
Resources like these, which provide high-level health condition prevalence statistics, can be useful references to determine whether there are gaps within a specific population regarding health conditions for the purpose of risk adjustment. However, more sophisticated approaches should be employed to refine the estimated HCC gaps within a specific population.
HCC gaps and risk opportunity identification
A key component of risk adjustment for payers and providers is accurate characterization of the presence of each diagnosis for each member. An inaccurate characterization creates inefficiencies in risk identification, potentially resulting in improper revenue capture. To avoid this, it is crucial to identify areas of improvement for HCC and risk identification and make cost-effective efforts to ensure that medical records and the diagnoses coded accurately reflect the conditions that exist. Outlined here are some of the sources to identify existing but uncaptured (non-coded or ineligibly coded) conditions:
- Rejected claims
- Denied claims
- Unqualified claims
- Chronic conditions
- Related diagnoses, procedures, drugs, and lab results
- Member self-reported conditions
- Predictive analytics or episode grouper
Payers should consider the likelihood that each uncaptured condition exists based on the identification source. To increase efficiency, conditions that have already been captured, as well as conditions lower in the hierarchy than those that have already been captured, can be filtered out.
Rejected claims
A rejected claim is one that has not been processed by the health plan due to errors or missing information. These claims need to be corrected and resubmitted for any risk adjustment consideration.
Denied claims
Medical claims can be denied by the health plan for many reasons,11 including:
- Lack of approved prior authorization
- Procedures or services not covered
- Services rendered at an out-of-network provider
- Delays and denials due to coordination of benefits (COB)
- Disputes over medical necessity
Regardless of the reason for denial, these claims are ineligible for risk adjustment in the HHS-HCC model even if they otherwise would be performed under different circumstances. If risk-adjustable conditions are identified in denied claims and these conditions have not been accounted for in other qualifying claims in the risk adjustment process, it could signify uncaptured risk linked to the member.
However, for the CMS-HCC model, MAOs are instructed by CMS to submit encounter data records (EDRs) for each service or item covered by the plan and provided to an enrollee, regardless of payment status of the claim. Because an EDR is a record of a service or item covered by the plan and provided to an enrollee while enrolled in that plan, the MAO’s final adjudication status of a claim from a provider should not affect whether that encounter is submitted.12
Unqualified claims
Although a patient can have a risk-adjustable condition and the claim is covered by the payer, it is possible that a claim can be rendered ineligible for MA risk adjustment due to an unqualified type of bill (TOB) or Current Procedural Terminology (CPT)/Healthcare Common Procedure Coding System (HCPCS) code.13 Inpatient encounters will be filtered based on TOB only, outpatient encounters will be filtered using a combination of TOB and CPT/HCPCS codes, and professional encounters will be filtered based on CPT/HCPCS codes only.
The ACA risk adjustment model methodology also outlines the criteria for determining services and providers that are allowable for risk adjustment, focusing on the presence of acceptable CPT/HCPCS codes and specific facility bill type codes.14
Telehealth utilization has undoubtedly grown over recent years. According to a survey performed by the National Center for Health Statistics (NCHS), telemedicine usage has increased among physicians to 86.5% in 2021, compared to 15.4% in 2019.15 Although telemedicine is becoming an increasingly viable option for receiving medical services, the CMS-HCC model requires a telehealth visit to be both audio and video together with other requirements in order for any diagnoses to be properly captured for risk adjustment purposes. However, the HHS-HCC model currently allows diagnoses to be captured for risk adjustment from an audio-only visit.16 MA plans could look for clues within these unqualified audio-only telemedicine visits for uncaptured HCCs.
Chronic conditions
Current risk adjustment rules stipulate that an individual’s HCCs are valid only for the calendar year in which the encounter they are associated with is incurred.17 Risk adjustment relies on the consistent and accurate redocumentation of a patient’s conditions every year. Although a patient with a chronic condition can have high utilization, costs, and clinical risk, the risk score will not accurately characterize the patient’s risk to the plan if the condition is not recaptured.
Provider documentation is crucial for supporting a patient’s risk profile. Patients with a persisting chronic condition should have their condition reported each year.
Related diagnoses, procedures, drugs, and lab results
Certain diagnoses can indicate a high likelihood of having other conditions that are risk-adjustable. For example, diabetes with chronic complications could be correlated with heart diseases and kidney disorders.
Procedure codes can be used to help identify specific uncaptured risk-adjustable health conditions. One example is the procedure of insulin administration without a diagnosis of diabetes.
Members can also be taking prescription drugs that are related to a risk-adjustable condition but not already covered by the risk adjustment model. For instance, members taking antidepressants may be experiencing a depressive disorder, a condition that is subject to risk adjustment. If this condition is not captured in the member's medical claims, it will not contribute to the final risk score.
If health plans have access to lab test results, those results can also indicate an undocumented health condition. For example, high blood glucose levels from hemoglobin A1c (HbA1c) labs could imply diabetes or prediabetes, which could then be confirmed by a physician. This is important not only for the patient's health management but also for accurate risk adjustment.
Member self-reported conditions
Members can self-report their health conditions through various sources, including:
- Health risk assessments (HRAs) and any other questionnaires, which provide information about a member's health status, lifestyle, and other behaviors.
- Member surveys conducted by the plans or a third-party organization to gather information about members’ health conditions.
- Member portals where members can input and update their health information.
- Phone or in-person interviews conducted by the plans with members to gather information about their health conditions.
- Care management (CM) programs, where participating members self-report health conditions as part of the CM process.
Any risk-adjustable self-reported health conditions that have not yet been captured through appropriate encounters and medical diagnoses could present an opportunity to increase coding accuracy.
Predictive analytics or episode grouper
Predictive modeling and episode groupers can assist health plans in identifying members with certain risk-adjustable conditions or those progressing into higher-risk conditions. These tools use historical data and algorithms to predict future outcomes and group-related healthcare events, which can help in early detection and management of potential health risks.
Discussion
There are many approaches health plans can employ to identify errors and omissions related to risk-adjustable conditions. It may be in the interest of health plans to validate the accuracy of each data source and balance available code capture resources with the likelihood of condition existence and accurate documentation. When coming up with strategies to close these risk gaps, one way is to consider both the value of a condition in terms of revenue dollars and the likelihood that a patient actually has a suspected condition. Health plans can consider creating an index to combine these two factors to help prioritize the operational efforts needed to address these gaps. Medical record review, member outreach, and provider education are some of the common code capture operations. The choice of the optimal operational method can be dependent on the way the uncaptured conditions were identified and the potential for coded conditions to be deleted due to lack of supporting documentation in the medical record as well as on budget and opportunity cost. It is also important to identify and rectify improperly coded conditions during these operations to ensure that the final risk adjustment results reflect the risk of the covered population.
1 Freed, M. et al. (August 8, 2024). Medicare Advantage in 2024: Enrollment Update and Key Trends. KFF. Retrieved December 20, 2024, from https://www.kff.org/medicare/issue-brief/medicare-advantage-in-2024-enrollment-update-and-key-trends/.
2 KFF (2024). Marketplace Enrollment, 2014-2024. Retrieved December 20, 2024, from https://www.kff.org/affordable-care-act/state-indicator/marketplace-enrollment/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D.
3 We will ignore HHS-HCC model prescription drug categories (RxCs) in our discussions unless expressly noted, as they are not the focus of this paper.
4 CMS. Patient Protection and Affordable Care Act, HHS Notice of Benefit and Payment Parameters for 2025; Updating Section 1332 Waiver Public Notice Procedures; Medicaid; Consumer Operated and Oriented Plan (CO-OP) Program; and Basic Health Program. Retrieved December 20, 2024, from https://www.cms.gov/files/document/cms-9895-f-patient-protection-final.pdf.
5 CMS (March 31, 2023). Fact Sheet: 2024 Medicare Advantage and Part D Rate Announcement. Retrieved December 20, 2024, from https://www.cms.gov/newsroom/fact-sheets/fact-sheet-2024-medicare-advantage-and-part-d-rate-announcement.
6 Frontz, A. (December 14, 2023). Toolkit to Help Decrease Improper Payments in Medicare Advantage Through the Identification of High-Risk Diagnosis Codes. HHS Office of Inspector General Retrieved December 20, 2024, from https://oig.hhs.gov/documents/audit/9174/A-07-23-01213-Complete%20Report.pdf.
7 Minemyer, P. (September 26, 2024). OIG calls for Humana, Aetna subsidiary to repay millions in MA overpayments. Fierce Healthcare. Retrieved December 20, 2024, from https://www.fiercehealthcare.com/regulatory/oig-calls-humana-aetna-subsidiary-repay-millions-ma-overpayments.
8 CDC (July 29, 2022). Chronic Conditions Among Adults Aged 18-34 Years — United States, 2019. Retrieved December 20, 2024, from https://www.cdc.gov/mmwr/volumes/71/wr/mm7130a3.htm.
9 CDC. Chronic Disease Indicators (CDI). Retrieved December 20, 2024, from https://www.cdc.gov/cdi.
10 Chronic Conditions Data Warehouse. Medicare Tables and Reports. Retrieved December 20, 2024, from https://www2.ccwdata.org/web/guest/medicare-tables-reports.
11 Poland, L. & Harihara, S. (April 25, 2022). Claims Denials: A Step-by-Step Approach to Resolution. Journal of Ahima. Retrieved December 20, 2024, from https://journal.ahima.org/page/claims-denials-a-step-by-step-approach-to-resolution.
12 CMS. CSSC Operations: General Submission Questions, 4. Retrieved December 20, 2024, from https://www.csscoperations.com/internet/csscw3.nsf/DID/U3NFLZ5K9V.
13 HHS (December 22, 2015). Final Encounter Data Diagnosis Filtering Logic. Retrieved December 20, 2024, from https://www.hhs.gov/guidance/document/final-encounter-data-diagnosis-filtering-logic.
14 CMS (April 10,2024). Final HHS-Developed Risk Adjustment Model Algorithm “Do It Yourself (DIY)” Software Instructions for the 2023 Benefit Year, April 10, 2024, Update Retrieved December 20, 2024, from https://www.cms.gov/files/document/cy2023-diy-instructions-04102024.pdf.
15 Myrick, K.L. et al. (February 2024). Telemedicine Use Among Physicians by Physician Specialty: United States, 2021. NCHS Data Brief. Retrieved December 20, 2024, from https://www.cdc.gov/nchs/data/databriefs/db493.pdf.
16 CMS (May 13, 2024). Risk Adjustment Telehealth and Audio-only Services FAQ. Retrieved December 20, 2024, from https://www.cms.gov/files/document/telehealth-faq-2024-update-51324-clean.pdf.
17 Martin, E. (April 1, 2019). Realize the Value of HCC Coding. AAPC. Retrieved December 20, 2024, from https://www.aapc.com/blog/46375-realize-the-value-of-hcc-coding.