Retrospective chart review has long given Medicare Advantage organizations another opportunity to identify HCC diagnoses after a patient encounter has already closed. If a clinically valid diagnosis surfaced later, it could be submitted through a chart review record even when that record was not linked to a specific encounter data record.
Beginning in CY 2027, that approach loses much of its value.
CMS has finalized the exclusion of diagnoses submitted through most unlinked chart review records, or CRRs, from risk score calculation. A limited exception applies when a beneficiary switches between Medicare Advantage organizations under different parent organizations. MA organizations can continue submitting unlinked chart reviews, but outside that exception, those diagnoses will not be used in CY 2027 risk score calculation. CMS finalized the policy in the 2027 Medicare Advantage and Part D Rate Announcement.
For plans and provider organizations that have relied heavily on retrospective capture, the obvious question is what takes its place.
At the CMS submission level, the answer is encounter data records and linked chart reviews. Operationally, the answer is bigger: risk adjustment work needs to move closer to the patient encounter, while the provider can still review the evidence, assess the condition, and document it appropriately.
That shift makes point-of-care risk adjustment software more important heading into CY 2027.
An unlinked chart reviews contains diagnosis information that is not associated with a specific beneficiary encounter in encounter data. CMS will exclude diagnoses from most of these records when calculating CY 2027 risk scores, although the records themselves can still be submitted.
The scale of the issue helps explain why CMS made the change.
CMS analyzed approximately 88.8 million unlinked chart reviews submitted in 2023 for 2024 payment. Roughly 85% could not be associated with a submitted encounter when CMS attempted to match basic information including the beneficiary, billing provider, and dates of service within three days. CMS said a significant majority of those records should either have been submitted as encounters documenting the underlying services or linked to an existing encounter. The full analysis appears in the CY 2027 Rate Announcement.
For organizations that have treated unlinked chart review as a major source of diagnosis capture, this changes the economics of the workflow. Finding a clinically relevant condition after the visit may still be useful for quality assurance and future care, but finding it late does not solve the encounter-linkage problem.
The unlinked chart reviews exclusion is also not the only diagnosis-source change arriving in CY 2027. CMS is excluding diagnoses from certain audio-only encounters when the encounter contains only service lines with the applicable audio-only modifiers and no other risk-adjustment-eligible service. The two changes are part of a broader move toward encounter-supported diagnosis data, which we cover in 3 Things to Know About Risk Adjustment in 2027.
An unlinked chart review record contains diagnosis information that is not tied to a known beneficiary encounter represented in encounter data.
A linked chart review, by contrast, associates additional diagnosis information with an identifiable encounter data record, or EDR.
The distinction matters because CY 2027 does not eliminate chart review. CMS specifically states that MA organizations can continue to capture clinically valid diagnoses through an EDR or through a linked CRR that associates the diagnosis with a known encounter. CMS explains that distinction in its response to comments on the finalized policy.
An EDR also should not be confused with a traditional paid claim. MA organizations are required to submit encounter data for items and services provided to Medicare enrollees, which means a clinical interaction does not necessarily need to produce a conventional claim before it can be represented in encounter data.
The issue for CY 2027 is not whether chart review happened. It is whether the diagnosis used for risk adjustment can be associated with the appropriate clinical encounter.
There are two answers, and they operate at different levels.
From a CMS submission standpoint, eligible diagnoses can continue to reach risk adjustment through encounter data records and linked CRRs. If additional diagnoses need to be submitted for an encounter, a linked chart review can associate them with the known EDR, provided they otherwise meet risk adjustment requirements.
From an operational standpoint, the replacement is a different way of working.
In a retrospective model, the patient encounter happens first. The chart closes, a coding or chart review team later identifies a missed HCC, and the organization tries to recover the diagnosis after the clinical work has already taken place.
A CY 2027-ready workflow moves more of that work upstream. Patient information is reviewed before or during the visit, relevant HCC opportunities are surfaced with supporting evidence, and the provider evaluates the condition while the encounter is still active. If the diagnosis is supported, it can be documented as part of the visit and move through the encounter-based submission process.
The goal is not to eliminate retrospective review. It is to reduce how often organizations need retrospective review to rescue something that could have been captured correctly the first time.
That is the broader transition from retrospective to prospective risk adjustment workflows.
Retrospective review still has an important role in CY 2027. It can uncover unsupported diagnoses, identify coding errors, expose recurring documentation problems, and show where provider education or workflow changes are needed.
What it cannot do is solve the unlinked chart review problem simply by increasing review volume.
A reviewer may find clinical evidence suggesting that a patient has a condition, but that does not automatically make the diagnosis eligible for risk adjustment without an appropriate encounter and documentation behind it.
The later a gap is found, the fewer options the organization may have. If a chronic condition was relevant to an earlier visit but was never assessed or documented, a retrospective reviewer cannot change what took place during the encounter. Depending on the circumstances, the organization may need additional clinical review, clarification, another qualifying encounter, or may have to accept that the opportunity was missed.
This is one of the reasons health plans still miss HCCs at the point of care. The problem is often not the ability to find information somewhere in the chart. It is getting the right information in front of the provider when there is still an opportunity to use it.
Risk adjustment software should help identify clinically relevant HCC opportunities before or during an eligible encounter, give the provider enough evidence to evaluate each one, and support accurate documentation without turning the visit into a manual coding exercise.
Claims and previous diagnosis lists are useful signals, but they do not contain the entire clinical story.
Modern HCC software should be able to analyze structured and unstructured information in the medical record, including clinical notes, medications, labs, and prior diagnoses. That provides more context for identifying conditions that may deserve attention during the current encounter.
Inferscience HCC Assistant, for example, uses natural language processing to analyze structured and unstructured chart data and applies clinical rules to identify diagnoses that may have been overlooked. Suggestions are displayed in the provider’s EHR workflow for review.
A diagnosis identified months after the visit creates another retrospective task. The same potential condition surfaced while the provider is reviewing the patient can become part of the clinical decision-making already underway.
The provider can determine whether the condition remains active, whether the available evidence supports it, and whether it is appropriate to document during the encounter.
A suspect list without context shifts the research burden back to the clinician.
The software should show why a condition was surfaced so the provider does not have to search through years of chart history just to evaluate a suggestion. The evidence should make review easier, not create another chart-chasing exercise.
Risk adjustment technology should not turn a suspected condition into a documented diagnosis automatically.
The provider remains responsible for deciding whether the condition is clinically supported and appropriate to document. The goal is an accurate record of what the clinician evaluated, not simply another diagnosis in the chart.
Every additional screen, queue, inbox, and disconnected application adds friction.
Point-of-care tools should work within the provider’s existing clinical workflow as closely as possible. That gives organizations a better chance of capturing supported diagnoses correctly during the original encounter instead of creating another downstream process.
For organizations evaluating risk adjustment factor software, RAF lift should not be the only measure of success. It is also worth asking how much valid risk is captured correctly during the original encounter and how much still depends on retrospective intervention. Our guide to proving ROI in risk adjustment beyond chart review looks at that measurement shift in more detail.
Moving risk adjustment upstream does not mean asking providers to perform a manual chart review before every patient visit.
The software can analyze the clinical information already available and identify conditions that may deserve attention. Those conditions are presented as opportunities for review rather than diagnoses that should automatically be added to the chart.
The provider sees the potential HCC alongside the evidence that caused it to be surfaced, determines whether the condition is current and clinically appropriate, and documents supported diagnoses with the necessary specificity.
Because that assessment happens as part of the clinical encounter, the diagnosis can move through the encounter-based submission process rather than depending on an unlinked retrospective CRR later.
This is the same point-of-care approach described in Closing HCC Coding Gaps at the Point of Care: give clinicians the relevant information while the clinical context is still available and reduce the amount of reconstruction required later.
HCC Assistant is designed to move HCC identification closer to the point of care.
The platform analyzes structured and unstructured information in the patient’s chart, applies clinical rules to identify diagnoses that may have been overlooked, and displays suggestions in the EHR. The provider reviews those suggestions, selects the ones that apply, and retains the final decision over what is documented.
The difference is where the risk adjustment work happens.
Instead of discovering a potential condition only after the chart has closed, the provider can see the opportunity while still working with the patient record, review the supporting information, and decide whether the diagnosis is appropriate.
HCC Assistant does not replace CMS encounter data requirements, and it does not replace clinical judgment. It helps reduce operational dependence on identifying valid HCCs only after the encounter and trying to recover them retrospectively.
The point-of-care workflow is only one side of a defensible risk adjustment program. Organizations also need a way to check whether captured diagnoses are supported before submission.
HCC Validator provides concurrent and retrospective HCC validation, including MEAT-based documentation review, identification of coding and documentation deficiencies, and review of higher-risk diagnoses. That gives coding and compliance teams another opportunity to identify unsupported diagnoses and documentation gaps before they become audit findings.
Together, the two workflows address different parts of the same problem: improve first-pass capture during the encounter, then confirm that the resulting diagnosis is supported and defensible.
No. Its job changes.
Organizations will still need retrospective review to validate documentation, identify unsupported diagnoses, monitor coding accuracy, find recurring documentation problems, and prepare for RADV and other audits.
What becomes less viable is using retrospective chart review as the primary safety net for HCC capture.
Under the older model, a condition missed during the visit could often become a recovery project afterward. Under a more encounter-based model, retrospective review should increasingly tell the organization why a diagnosis was missed and what needs to change so it is less likely to be missed again.
That makes retrospective review a tool for validation and workflow improvement rather than simply a revenue-recovery mechanism.
RAF still matters, but RAF alone does not tell an organization whether the underlying workflow is improving.
A stronger measurement framework looks at a small group of connected signals: first-pass HCC capture, unlinked chart review volume, linked versus unlinked diagnosis submissions, provider action on surfaced opportunities, and the amount of retrospective querying or chart chasing still required.
Together, those measures answer a more useful question: How much valid risk are we documenting correctly the first time?
That distinction matters because CMS expects the financial effect of the unlinked chart review exclusion to vary based on how heavily an MA organization has relied on those records. Plans with greater dependence on unlinked chart reviews have more workflow exposure to address.
The diagnoses from 2026 dates of service will form the basis of final CY 2027 risk-adjusted payments, so the encounters that matter are happening now. CMS states this explicitly in the final Rate Announcement.
The practical starting point is to quantify current dependence on unlinked retrospective capture. Identify the HCCs, provider groups, markets, and workflows producing the highest volume of unlinked chart reviews, then look at why those diagnoses were not captured through an encounter in the first place.
Sometimes the relevant condition was buried in the chart. Sometimes the provider did not have the evidence available during the visit. In other cases, a condition remained on a problem list without being evaluated, documentation lacked specificity, or the operating model simply assumed a retrospective team would clean up gaps later.
Each problem points to a different fix.
Timing matters here as well. CMS’s July 2026 risk adjustment implementation guidance confirms that the unlinked chart review and audio-only exclusions begin with the CY 2027 Midyear model run rather than the Initial model run. That means the Midyear update can change payments prospectively for July through December and produce retroactive adjustments for payments made from January through June. Inferscience breaks down those mechanics in CMS Just Published the CY 2027 Risk Adjustment Playbook. The federal HHS Guidance Portal lists CMS’s CY 2027 Risk Adjustment Implementation guidance as final guidance issued in July 2026.
CMS has also made clear that MA organizations have until at least January 31, 2028 to submit data for CY 2027 payment. That submission window should not be mistaken for extra time to fix the clinical workflow. By then, the 2026 patient encounters themselves will already be over.
CY 2027 does not end chart review. It changes what organizations can expect chart review to accomplish.
A risk adjustment program built primarily around finding missed diagnoses after the visit will have fewer options for turning those diagnoses into risk-adjusted payment through unlinked submissions. The more durable approach is to identify the opportunity earlier, put the supporting evidence in front of the provider while the patient record is open, document supported conditions during the encounter, and validate those diagnoses before submission.
That is the real replacement for an unlinked chart review strategy.
HCC Assistant helps surface HCC opportunities and the clinical information behind them within the provider workflow. HCC Validator adds a validation layer to help ensure submitted diagnoses are supported and defensible.
Together, they support the larger shift from retrospective recovery toward accurate first-pass capture.
Contact Inferscience to see how that workflow can work inside your organization.
No. Medicare Advantage organizations may continue to submit unlinked chart reviews. However, diagnoses from most unlinked chart reviews will be excluded from CY 2027 risk score calculation, with an exception for qualifying beneficiaries who switch between MA organizations under different parent organizations. See the CMS CY 2027 Rate Announcement.
Yes. CMS states that clinically valid diagnoses can continue to be captured through encounter data records or linked CRRs that associate diagnoses with a known encounter, provided the diagnoses otherwise meet risk adjustment requirements.
At the submission level, encounter data records and linked CRRs remain available for eligible diagnoses. Operationally, organizations need to move more HCC identification and documentation into prospective, point-of-care workflows so supported diagnoses can be associated with clinical encounters rather than depending on unlinked retrospective recovery.
Risk adjustment software should identify potential HCC opportunities before or during the patient encounter, show the clinical information supporting each suggestion, allow the provider to evaluate the condition, and support accurate documentation within the existing clinical workflow.
The terms are often used for overlapping categories of technology. Rather than focusing on the label, organizations should evaluate whether the software helps clinicians identify and document supported HCCs during qualifying encounters while improving the accuracy and defensibility of RAF capture.
