Free Medical Billing Audit

Guide

What Is HCC Coding? Risk Adjustment for Medical Practices

What is HCC coding? Hierarchical Condition Category (HCC) coding is the process of reporting a patient’s chronic and serious conditions with ICD-10-CM diagnosis codes that map into risk adjustment categories. Payers such as Medicare Advantage plans use those categories, plus age, sex, and enrollment status, to produce a risk score that predicts how costly a patient is likely to be. For a practice, the work comes down to one habit: document and code every relevant condition, to its full specificity, at a face-to-face or qualifying telehealth visit, every calendar year.

People working at desks in an open office
In this article
  1. What HCC coding means in plain terms
  2. Which risk adjustment models exist
  3. Why a practice cares about HCC coding
  4. Medicare Advantage patients in your panel
  5. Value-based and shared-savings contracts
  6. Practices that manage chronic disease
  7. Which encounters and providers count for risk adjustment
  8. Documentation standards that make a diagnosis defensible
  9. Visit documentation checklist
  10. Specificity: how the same condition can be coded three ways
  11. Annual recapture: why chronic conditions must be re-documented
  12. Compliance risks: RADV audits and over-coding
  13. Patterns that create exposure
  14. The opposite problem: under-documentation
  15. A practical HCC workflow for a practice
  16. Writing a compliant provider query
  17. Worked hypothetical: a missed condition across two visits
  18. What goes wrong and how to fix it
  19. How to measure whether the program works
  20. What Mednex handles and what the practice handles
  21. Questions practice owners ask before starting
  22. Do we need a certified risk adjustment coder?
  23. Should coding follow the score?
  24. What about commercial and Medicaid patients?
  25. Next steps for your practice
Key takeaways
  • HCC coding does not change how a visit is paid on your fee schedule. It changes the risk score attached to the patient, which drives plan payments and some value-based contract benchmarks.
  • A diagnosis only counts if the medical record supports it. Coding a condition that is not documented is a compliance problem, not a revenue tactic.
  • Most HCC categories reset each calendar year, so chronic conditions must be assessed and documented again every year.
  • Specificity is the main lever: “diabetes” and “type 2 diabetes with diabetic chronic kidney disease” are different codes with different risk meaning.
  • Models change. Confirm the current model version and code mapping with CMS rather than relying on a printed cheat sheet.

What HCC coding means in plain terms

Every ICD-10-CM diagnosis code you submit on a claim can be sorted by a risk adjustment model into a condition category. The model groups clinically related diagnoses into an HCC, and the word “hierarchical” means that when a patient has several conditions in the same family, the most severe one counts and the lesser ones are dropped. A patient with a diabetes code that indicates complications, for example, is scored on that category and not also on a plain diabetes category in the same family.

The model then adds each category’s weight (called a coefficient) to demographic factors such as age, sex, whether the person is also on Medicaid, and whether they qualified for Medicare through disability. The sum is the risk adjustment factor, usually called the RAF score. A score near the average for the Medicare population means the patient is expected to cost about an average amount. A higher score means a higher expected cost.

Three practical points follow from this:

  • The code is the input. The model never reads your note. It reads the diagnosis codes that came off a submitted encounter. If the note says one thing and the code says another, the code is what scores, and the note is what an auditor reads.
  • Not every code counts. Only a subset of ICD-10-CM codes map to a payment category. Many common diagnoses do not.
  • Not every encounter counts. The service type, the provider type, and the setting all determine whether a diagnosis is eligible.

HCC coding is a coding and documentation discipline, and it sits next to normal professional billing, not in place of it. If you want the broader context of how coding feeds claims, the medical billing process step by step guide shows where diagnosis coding fits in the claim lifecycle.

Which risk adjustment models exist

“HCC” is not a single list. Several federal models use the same idea, and they apply to different programs. Knowing which one your payer uses prevents a lot of confusion.

Model Program it supports What it predicts
CMS-HCC Medicare Advantage (Part C) payments to plans Medical costs of enrollees, based on diagnoses from the prior year
CMS-HCC ESRD models Medicare Advantage payment for people with end-stage renal disease Costs for dialysis and transplant populations, which are scored separately
RxHCC Medicare Part D plan payments Prescription drug costs
HHS-HCC ACA individual and small group marketplace risk adjustment Relative plan liability across insurers, not a per-visit payment

The version matters. CMS announced that for payment year 2026 it is calculating Medicare Advantage risk scores using only the 2024 CMS-HCC model, after a three-year phase-in that began in 2024. That model regrouped many conditions and changed which diagnosis codes lead to a payment category compared with the earlier model. Industry shorthand often calls the older model V24 and the newer one V28. Because CMS publishes new mapping files and model software on its risk adjustment page, always pull the current mapping rather than trusting a table from a vendor or an old training slide.

Why a practice cares about HCC coding

A practice that bills fee-for-service Medicare and takes no risk contracts may feel distant from all this. Three situations bring it close.

Medicare Advantage patients in your panel

Medicare Advantage plans receive a monthly payment per enrollee that is adjusted by risk score. The plan depends on diagnoses submitted from provider encounters. Your claims are the raw material for that submission, so plans often ask for chart documentation, run coding education programs, or send provider scorecards. Even if you are paid your usual contracted rate for a visit, your documentation shapes how accurately the plan is paid for the patient.

Value-based and shared-savings contracts

Many shared savings arrangements, capitation deals, and quality programs use risk scores to set benchmarks or to adjust performance. If your patients look healthier on paper than they are, the benchmark you are measured against can be set too low, and a practice that provides good care can look like it overspends. The reverse is a compliance risk. Check each contract for the model it names and how the score is calculated.

Practices that manage chronic disease

Primary care, cardiology, nephrology, endocrinology, and similar practices diagnose and manage the conditions that most often map to categories. These practices carry most of the documentation load. For a view specific to a family medicine or internal medicine setting, see the primary care billing services page and the internal medicine billing services page.

For a view of how specific your current diagnosis coding is, a Free Medical Billing Audit is a low-effort place to start.

Which encounters and providers count for risk adjustment

This is where many programs go wrong. A diagnosis does not count just because it appears on a claim. It must come from an acceptable source.

In general terms, CMS-HCC risk adjustment draws diagnoses from three kinds of sources: hospital inpatient stays, hospital outpatient services, and professional services from physicians and other acceptable provider types. The service must be a face-to-face encounter. Telehealth visits can qualify when they use real-time audio and video, and CMS has published guidance on diagnoses from telehealth services for risk adjustment, so check that document for the current rules on audio-only visits and modifiers. CMS also keeps lists of eligible CPT and HCPCS codes and of acceptable physician specialty types on its risk adjustment page.

Source of diagnosis Generally eligible? What to check
Office visit with an acceptable provider type Yes, when the CPT or HCPCS code is on the eligible list Face-to-face, provider credentials and signature, date of service
Real-time audio and video telehealth visit Generally yes Current CMS telehealth guidance, platform used, modifier and place of service on the claim
Audio-only visit Check current CMS guidance before relying on it Payer policy and the model year
Diagnostic radiology and lab-only services Generally not an acceptable source The diagnosis should come from the treating provider’s encounter
Problem list or history page without an assessment Not by itself The condition needs to be evaluated at an eligible visit
Superbill or coding sheet checkbox with no note support No The signed note must support the code

If you are unsure about a provider type or a code, confirm with your Medicare Administrative Contractor or the Medicare Advantage plan. Rules for commercial and Medicaid risk programs differ again, so treat the table as a CMS-HCC orientation and not a universal rule.

Documentation standards that make a diagnosis defensible

A reviewer deciding whether a diagnosis is supported looks for evidence that the provider actually considered the condition at that visit. Many organizations teach this with the acronym MEAT, which stands for Monitor, Evaluate, Assess or Address, and Treat. MEAT is an industry teaching device, not a CMS regulation, but it matches how auditors read a note. At least one element should be visible for each condition you report.

Element What it looks like in a note Weak example
Monitor Symptoms, vital signs, home readings, disease progression noted “Diabetes” in the history only
Evaluate Exam findings, lab or test results reviewed with the value cited “Labs reviewed” with no result or condition named
Assess or Address Provider’s clinical judgment: stable, worsening, improved, at goal A list of conditions with no comment
Treat Medication, referral, counseling, therapy, orders A medication list that is not linked to a diagnosis

Visit documentation checklist

  1. The visit is face-to-face or qualifying telehealth, and the note says which.
  2. The note carries the patient name, date of service, and provider name with credentials.
  3. The note is signed or electronically authenticated by the treating provider, with the date.
  4. Each chronic condition is named in the assessment, with status and plan, not only in the history.
  5. Linked conditions are stated as linked when the provider believes they are linked, for example a kidney problem caused by diabetes.
  6. The diagnosis codes on the claim match the diagnoses documented in the note.
  7. Past conditions that no longer exist are not coded as active. Status codes (for example amputation status) are used when they are correct.

Coders cannot assume a link between conditions. In ICD-10-CM, some combinations are presumed linked by the classification (the “with” convention), and others need the provider to state the relationship. Read the official guidelines for the code range in question. The provider must make the clinical call, and the coder must not fill in what the note does not say.

Specificity: how the same condition can be coded three ways

The most common loss in risk adjustment is vague documentation. A provider may know the patient has serious kidney disease from diabetes, but write “DM, CKD.” A coder then reports an unspecified diabetes code and an unspecified kidney code, and the record loses clinical detail.

Clinical reality Vague documentation Specific documentation Code family
Type 2 diabetes with diabetic chronic kidney disease, stage 4 “DM2, CKD” “Type 2 diabetes with diabetic CKD, stage 4” E11.22 with N18.4
Chronic systolic heart failure “CHF” “Chronic systolic heart failure, compensated” I50.22 versus I50.9
Recurrent major depression, moderate “Depression” “Major depressive disorder, recurrent, moderate” F33.1 versus a nonspecific depression code
COPD with acute exacerbation “COPD flare” “COPD with acute exacerbation” J44.1 versus J44.9

Whether a given code maps to a payment category depends on the model year, so use the codes above as examples of specificity differences, not as a statement of what pays. Look up each code in the current CMS mapping file for the model your payer uses.

Specificity does not mean inflation. A provider should document what is clinically true. If the patient’s kidney function is stage 3, the note says stage 3, even if stage 4 would score higher. This point matters for audits, and it is also the reason clinical education beats any coding trick.

Annual recapture: why chronic conditions must be re-documented

Under the CMS-HCC approach, diagnoses collected during one calendar year predict payment for the next. The categories do not carry forward on their own. If a patient has an eligible diagnosis documented in 2025, it supports the 2026 payment year. If nobody documents it again in 2026, the condition drops out of the score for 2027, even though the patient still has it.

This is called recapture, and it is a workflow problem, not a coding problem. The conditions are known. The gap is that the patient did not have an eligible visit where the provider assessed each one. Practices that run recapture well usually do the following:

  1. Build a suspected and prior-year condition list from last year’s claims, so each provider can see which chronic conditions were reported before.
  2. Schedule the annual wellness or chronic care visit early in the year, not in December.
  3. Prompt the provider at the visit to evaluate each listed condition: still present, resolved, or never confirmed.
  4. Review open gaps in the autumn and call patients who have not had an eligible visit.
  5. Remove conditions that do not hold up. A prior-year code is a prompt for clinical review, not an instruction to code it again.
A prior-year list is a question, not an answer

A suspect list says “this was coded before.” It does not say “code it again.” The provider must evaluate the patient in front of them and document the finding. If the condition resolved, the correct outcome is no code and a corrected problem list.

Compliance risks: RADV audits and over-coding

Because risk scores move payments, risk adjustment diagnoses are audited. CMS runs Risk Adjustment Data Validation (RADV) audits for Medicare Advantage. In plain terms, RADV checks whether the diagnoses a plan submitted are supported by the enrollee’s medical record, and CMS can recover overpayments when they are not. Audits occur after the final risk adjustment data submission deadline for the contract year. The plan is the audited party, but the evidence that decides the outcome is your documentation, so plans may request records from you.

Federal enforcement has also reached the provider side in some cases, including False Claims Act matters tied to unsupported diagnoses. A practice does not have to be a plan to face questions if it submits diagnoses without support.

Patterns that create exposure

  • Coding from a problem list or an old note with no assessment at the visit.
  • Coding a condition as active when the note describes it as history or resolved.
  • Selecting an unsupported complication or severity to reach a higher category.
  • Using a diagnosis from a source that is not eligible, such as a lab-only service.
  • Adding diagnoses after the visit through a retrospective chart review without the treating provider’s documentation to support them. Check the payer’s rules and CMS guidance on this before any retrospective program.
  • Pressure on providers through productivity scorecards that reward code counts instead of accuracy.

The opposite problem: under-documentation

Compliance cuts both ways. Under-documenting makes a patient look healthier than they are, which can shrink the resources a plan or a value-based arrangement allocates to their care. The target is accurate, complete documentation, not the highest possible score.

A practical HCC workflow for a practice

A workable program sits across the front desk, clinical staff, providers, and billing. The table below shows who normally does what.

Stage Task Usual owner
Pre-visit Prepare a prior-year condition list and flag open gaps Coder or practice manager
Scheduling Book eligible visit types for patients with open gaps Front desk or virtual assistant
Visit Evaluate and document each chronic condition with status and plan Provider
Coding Select specific ICD-10-CM codes that the note supports Certified coder
Query Ask the provider to clarify vague or conflicting documentation Coder to provider
Claim Submit with correct diagnosis pointers and visit type Billing
Review Sample charts for accuracy and report findings Compliance or outside reviewer

Writing a compliant provider query

A query asks the provider to clarify. It must not lead. A leading query names the answer the practice would like. A compliant query presents the clinical facts from the record and offers reasonable options, including “unable to determine” or “other.”

Hypothetical example, compliant query: “The note lists CKD and type 2 diabetes. The most recent eGFR in the chart is shown below. Please document the stage of CKD, and whether the provider considers the CKD related to diabetes, due to another cause, or unable to determine.”

Hypothetical example, leading query: “Please confirm diabetic CKD stage 4.” This tells the provider what to write and creates audit risk.

Worked hypothetical: a missed condition across two visits

Hypothetical example. The numbers below are round illustrations, not real scores or payments. A Medicare Advantage patient is seen twice in one year.

  1. Visit one, spring. Note says: “Diabetes, CKD. Continue current medications.” The coder reports a nonspecific diabetes code and an unspecified CKD code. The assessment names no stage, and no link is stated.
  2. Visit two, autumn. The provider reviews labs, writes “Type 2 diabetes with diabetic chronic kidney disease, stage 3b, eGFR trending down, continue ACE inhibitor, refer to nephrology.” The coder reports the specific diabetes-with-CKD combination and the stage 3b code.

Only visit two supports the specific picture. If the patient never came back in the autumn, the record would carry only the vague codes for the year. Suppose, for illustration, the specific coding places the patient in a category that adds 0.2 to the risk score, and the vague coding adds 0.1. The difference of 0.1 is a modeling artifact of the example, but it shows the mechanism: specificity changes the category, and the visit date matters.

Now the audit view. If the plan were audited and asked for evidence, visit two supports the diabetic CKD code because it contains monitoring, evaluation, assessment, and treatment. Visit one would not support the specific code at all. Nothing in the example required a coding trick. It required a provider who wrote what they saw and a patient who returned for a visit.

What goes wrong and how to fix it

Problem Likely cause Fix
Chronic conditions missing from claims Provider assessed them but the coder only captured the chief complaint Train coders to code all documented, addressed conditions, and add a claim scrubber check for missing chronic codes in known patients
Unspecified codes used often Templates and shorthand notes lack stage, type, or severity Update note templates with required specificity fields and give providers short feedback examples
Diagnoses from ineligible visits Capture came from lab, radiology, or non-face-to-face contacts Check visit type and provider eligibility before counting a diagnosis
Conditions coded with no assessment Copy-forward notes and problem lists Audit for copied text. Require current status and plan for each coded condition
Patients with no eligible visit in the year No outreach for open gaps Run a gap list monthly and book visits before the year ends
Provider sees code counts as the goal Incentives reward volume Measure accuracy and documentation quality instead

How to measure whether the program works

Measure accuracy first and volume second. A sound program tracks:

  • Chart accuracy rate. In a sample of charts, the share of reported HCC-relevant diagnoses that the note fully supports. Record the sample size and who reviewed it.
  • Specificity rate. The share of chronic condition codes that are specific and not “unspecified.”
  • Recapture completion. Of the patients with a chronic condition last year, how many had an eligible visit this year where the condition was assessed.
  • Query rate and response time. How many charts needed clarification, and how quickly providers answered.
  • Visit completion for open gaps. How many patients on the gap list were scheduled and seen.

Compare these against your own prior quarters. Public benchmarks depend on patient mix, payer mix, and specialty, so a comparison with another practice is rarely fair. Use sampling that a compliance officer would accept, and keep the method steady over time so trends mean something.

What Mednex handles and what the practice handles

Mednex Group works on the billing and coding side of the revenue cycle. HCC documentation is clinical, so responsibilities split clearly. The scope below describes process and responsibility, and the exact arrangement is set in the engagement agreement.

Mednex can support The practice keeps
Coding review of sampled encounters for specificity and note support All clinical judgments and the content of every note
Flagging unspecified or conflicting diagnoses for provider query Answering queries based on the clinical picture
Claim preparation so diagnoses and pointers match the documentation Provider signatures and attestations
Reporting on denial and rejection patterns tied to diagnosis coding Compliance policy and decisions about risk contracts
Scheduling and patient follow-up support through virtual assistants, where engaged Whether a patient receives a visit, and what is medically appropriate

Mednex cannot guarantee a risk score, a payment, or the outcome of an audit, and does not make clinical documentation decisions. Details such as how records are accessed and shared depend on the arrangement and are confirmed in writing. The broader service set is described on the medical billing page and the revenue cycle management page.

Questions practice owners ask before starting

Do we need a certified risk adjustment coder?

Not by law for most practices, but accuracy improves when someone trained in ICD-10-CM guidelines and risk adjustment rules reviews charts. Credentials exist for risk adjustment coding. Whether to hire, contract, or train depends on your volume and contracts.

Should coding follow the score?

No. The note comes first, the code follows the note, and the score follows the code. If a team starts from the score it wants, the process is inverted and the audit exposure grows.

What about commercial and Medicaid patients?

Some commercial plans and Medicaid programs use their own risk models. The same documentation habits help, but the eligible encounter rules and model versions differ. Ask each plan which model it uses.

Denials tied to diagnosis coding often show up in the same reports that reveal documentation weakness. The common medical billing errors guide lists the patterns, and the why medical claims are denied article explains how to read the remittance.

Next steps for your practice

A reasonable first step is a small look-back. Pull a sample of recent encounters for patients with chronic conditions, compare each note to the diagnoses on the claim, and count how many used unspecified codes or lacked an assessment. That gives you a baseline without changing any workflow.

If you want an outside view of coding and claim patterns, the Free Medical Billing Audit reviews a de-identified or secure-portal sample of recent claims, denial reasons, and A/R aging. You supply report exports and never send patient records through the website form. You can also reach the team through the contact page or read more practical guides in the Mednex blog.

General information only

This guide is educational. It is not legal, coding, compliance or clinical advice. Rules differ by payer, plan, state and contract, so confirm against the payer's current policy before acting.

Frequently asked questions

What is HCC coding in simple terms?

HCC coding means reporting a patient's chronic and serious conditions with ICD-10-CM diagnosis codes that a risk adjustment model sorts into condition categories. The model combines those categories with age, sex, and enrollment status to produce a risk score that predicts expected cost. The score affects what plans are paid and how some value-based benchmarks are set. It does not change your fee schedule payment for a visit. The code must always be supported by the signed note. Think of it as documentation that tells the plan how sick a population is, so payments can match expected need. Accurate, specific coding is useful, and inflated or unsupported coding is a compliance problem, not a revenue strategy.

What does HCC stand for and what is a RAF score?

HCC stands for Hierarchical Condition Category. Hierarchical means that when a patient has several related conditions, the most severe one counts and the milder ones in the same family are dropped. The RAF, or risk adjustment factor, is the total score built from demographic factors plus the weight of each category. A score near the population average predicts average cost, and a higher score predicts higher cost. Each model version uses its own weights. Because models are updated and recalibrated, a diagnosis that carried a certain weight in one year may carry a different weight later. Check the current CMS model version rather than relying on a number you remember.

Which visits can be used to report HCC diagnoses?

For CMS-HCC, diagnoses generally come from hospital inpatient, hospital outpatient, and professional services by acceptable provider types, and the visit must be face-to-face. Real-time audio and video telehealth can qualify, but CMS publishes specific guidance, including on audio-only visits, so check the current document. Lab-only and diagnostic radiology services are generally not acceptable sources. A problem list without an assessment does not support a code by itself. If you are unsure whether a provider type or code qualifies, confirm with the Medicare Advantage plan. A practical habit is to confirm at scheduling that chronic condition patients have a qualifying visit booked, rather than discovering in December that a face-to-face encounter never happened.

Why do chronic conditions have to be documented every year?

Under the CMS-HCC approach, diagnoses collected in one calendar year predict payment for the following year, and the categories do not carry forward automatically. If a condition is assessed in 2025 but not again in 2026, it falls out of the score for 2027 even though the patient still has it. Practices manage this with an annual visit schedule, a prior-year condition list that prompts clinical review, and a monthly gap report. For patients who are hard to schedule, consider whether an eligible telehealth visit is allowed under current CMS and plan rules, and document the modality clearly in the note.

What is a RADV audit and how does it reach my practice?

RADV stands for Risk Adjustment Data Validation. CMS uses it to check whether diagnoses submitted by Medicare Advantage organizations are supported by enrollees' medical records, and it may recover overpayments if they are not. The plan is the audited party, but the evidence is your documentation, so plans can request charts from you. Complete, signed, dated notes that show assessment and plan are your best protection. Keep records organized by date of service so they can be produced quickly. Practices should keep a clear process for responding to plan record requests, with a named contact and a tracking log, since late or incomplete responses create problems for both the plan and the practice.

What is MEAT and is it required?

MEAT stands for Monitor, Evaluate, Assess or Address, and Treat. It is a teaching device many coding programs use to test whether a note supports a diagnosis, and it is not a CMS regulation. A condition mentioned only in the history, with no sign that the provider considered it at that visit, usually fails the test. A condition with a stated status, a reviewed lab value, and a treatment plan passes. Use it as a checklist for provider education, not as a legal standard. Plans and auditors may phrase their expectations differently, so follow the specific contract and payer guidance you operate under.

Can a coder add diagnoses that the provider did not write?

No. A coder translates documented diagnoses into codes and may not infer or add a condition the note does not state. If documentation is vague or conflicting, the coder sends a compliant query that presents the clinical facts and offers neutral options, including unable to determine. A leading query that names the answer creates audit risk. Retrospective chart review programs should be checked against payer and CMS rules first. Ask your compliance lead to approve the query template. If a retrospective review program is used, keep it limited to documentation that already exists in the chart and make sure any query is compliant, non-leading, and answered by the treating provider.

Does HCC coding apply to small practices without risk contracts?

It still matters if you see Medicare Advantage patients, because plans use your claims and records for their risk submissions and may request documentation or audit your charts. Practices in shared savings or capitation arrangements feel it more directly. A solo practice with only fee-for-service patients has less exposure, but good specificity and annual assessment of chronic conditions are sound habits for any practice. Plans may also send scorecards or request charts, so documentation habits matter either way. Ask each Medicare Advantage plan in your network what programs they offer, such as scorecards or documentation support, and whether they have specific requests, then read the contract terms before agreeing to any arrangement.

How is HCC accuracy measured inside a practice?

Review a sample of charts and compare each reported diagnosis to the note. Track the share fully supported, the share of chronic condition codes that are specific, how many patients with a chronic condition had an eligible visit this year, and how fast providers answer queries. Record the sample size and reviewer, and compare against your own earlier quarters, not another practice, because patient and payer mix differ. Keep the sampling method steady so trends mean something. Keep results in a simple table by provider and quarter, share them in plain language, and focus education on the most common specific gaps rather than on individual blame.

Can Mednex guarantee a higher risk score or audit outcome?

No. Risk scores depend on clinical documentation and patient conditions, and audit outcomes are decided by the payer or CMS. Mednex can review coding for specificity and note support, flag items for provider query, and align claims with documentation. Clinical judgments and note content stay with the practice. Details of scope and access are confirmed in the engagement agreement. Mednex also cannot make clinical documentation decisions or predict payer behavior. The practice remains responsible for clinical accuracy and signatures. Mednex supports the coding and workflow side, and any decision about a diagnosis stays with the treating provider.

Sources

Want this reviewed against your own claims?

The free medical billing audit looks at a sample of your recent claims, denial reasons and A/R aging. You get a written summary and a walkthrough call.

Do not include patient names, insurance IDs or clinical details. If records are needed later, we set up a secure way to share them.