The one-line answer
The advantage that used to come from translating a written measure specification into code is going away. The new advantage is owning a governed, transparent data foundation, normalizing messy source data into a standard shape, and running a reasoning layer that decides what to do about the result. Teams that build for that now will be ready; teams that wait will be retrofitting under deadline.
What is actually changing
NCQA, the body that certifies many of the quality measures Medicare Advantage plans report on, has published a plan to take measurement fully digital. It intends to retire the legacy hybrid method by measurement year 2029 and move to fully digital reporting by 2030, and it now publishes the measure logic itself in two machine-readable standards: FHIR, a data model for clinical information, and CQL, a language for expressing measure logic. The full timeline is on the NCQA blog.
For two decades, a meaningful part of the work was reading a written specification and turning it into running code, then defending that translation when the numbers were questioned. When the certifying body ships the logic in a standard form, that translation step shrinks. The work moves upstream, to the data.
That sentence is easy to nod along to and hard to act on, so it is worth being concrete about what “upstream” means. A measure expressed in CQL asks questions of FHIR resources: is there an encounter of this type in this window, a lab result with this code above this value, a medication dispensed within this many days. It does not ask where the resource came from, whether the code was mapped correctly, or whether the same event exists three times under three local identifiers. Every one of those questions is now yours, and the answer you give is the answer the measure will faithfully compute.
What the hybrid method was doing for you
The hybrid method deserves a fair epitaph, because a lot of quality operations are about to discover how much it was quietly covering.
Under hybrid reporting, a measure could be satisfied from administrative data, and where the administrative data fell short, a sample of medical records was reviewed by hand. That chart retrieval season was expensive and slow, and it is the part everyone is glad to see go. But it was also a correction layer. When a screening was performed but never coded, when a result lived in a scanned document, when a specialist’s note never made it into the claim, a reviewer could find it, and the measure got credit for care that actually happened.
Digital measurement removes the reviewer. That is the point of it. It also removes the correction. In a fully digital world, care that is not represented as a clean, correctly coded resource did not happen as far as the measure is concerned. The care gap that opens is real to the rating even when it is not real to the member, and no one is going to pull the chart to argue otherwise. The plans that have been leaning on retrieval to close the last few points on a measure are the ones with the most to lose, and most of them do not know it yet, because the hybrid rate and the administrative rate have never been shown to them side by side.
Why this matters for Medicare Advantage Star Ratings
CMS Star Ratings decide a large share of a Medicare Advantage plan’s economics, and the cut points move every year. When measurement was slow and manual, a plan often learned where it landed after the measurement year had already closed, with no time left to act. Digital measurement shortens that cycle. Measures can run on real clinical and claims data with less manual chart retrieval, which means a plan can see exposure earlier and still have room to move a bubble measure before the year ends.
The catch is that digital measurement is only as good as the data underneath it. A measure expressed in CQL still needs clean, well-mapped inputs. If your source data is inconsistent, the standard logic will faithfully produce a wrong answer. That is why the foundation matters more, not less, in a digital world.
There is a second-order effect that matters for Stars specifically. As more plans report the same measures on the same logic against the same standards, the spread between plans narrows for reasons that have nothing to do with care. Cut points are set from the distribution of plan performance, so when the distribution tightens, the distance between one star and the next shrinks with it. A rate that comfortably cleared a threshold in the hybrid era can sit on the bubble in the digital one without a single member’s care changing. Digital measurement does not just change how you compute your rate. It changes the field you are being ranked against.
Three capabilities that decide who wins
- Normalizing messy data into a standard shape. Real-world clinical and claims data does not arrive in FHIR. Getting it there, consistently and with documented lineage, is the new core competency. This is governed data architecture, not a one-time conversion.
- A reasoning layer that decides what to do. Knowing a member is out of a measure is not the same as knowing which action moves the rating most per unit of effort. Prioritization by impact, traced to the claim event that caused the gap, is where analytics earns its keep.
- Certification as a trust signal. As measurement standardizes, the differentiator shifts from “we can compute the measure” to “our result holds up to scrutiny.” Certification and transparent lineage become the signal that unlocks enterprise procurement.
What changes for the quality operation
Most of the writing on this transition is aimed at the data team. The quality team’s year changes just as much, and in ways that are easier to miss.
The calendar flattens. Under hybrid reporting the year had a shape: administrative data through the measurement year, retrieval and abstraction after it, results months later. Digital measurement can run continuously, which means the question “where do we stand on this measure” has an answer in March, not next spring. That is a gift, and it is also a demand. A quality leader who can see exposure in March is expected to act on it in March, and the annual improvement plan built around a retrieval season stops making sense.
Outreach gets more precise, and more accountable. When a gap is computed from a FHIR resource rather than a claim summary, the reason for the gap is knowable. A member is open on a screening measure because there is no qualifying procedure resource in the window, and you can see which feeds were checked and what they contained. The outreach call is better informed, and when the member says the screening already happened, the follow-up is a data fix rather than a shrug.
Provider data sharing becomes a quality lever. A large share of the care that closes gaps happens in provider systems the plan does not own. In the hybrid era, chart retrieval was the bridge. In the digital era the bridge is a clinical data feed, and its quality is the quality of your rate. Provider engagement stops being a relationship function and becomes part of the measurement pipeline, with the same need for lineage and validation as any other source.
The audit changes character. A reviewer questioning a hybrid rate asked to see the charts. A reviewer questioning a digital rate will ask to see the resources, the mappings, and the version of the logic that produced the number. If the answer takes a week and three people, you have a problem that no amount of clinical improvement will fix.
What does not change
It is tempting to hear “fully digital” and assume the whole discipline resets. It does not, and the parts that stay put are the parts that still decide ratings.
Cut points still move every year, and the weighting between measure domains is set by CMS, not by the format of the logic. Member experience measures are not computed from FHIR resources at all, and they carry the same weight in your rating whether or not your clinical measures are digital. The bubble measure that decides whether you hold a star is still a bubble measure, and finding it still requires knowing your position, the threshold, and the distance between them. Digital measurement gives you an earlier and more honest view of that position. It does not change the arithmetic of what the position is worth.
And the data problem does not change, it moves. The team that spent years reconciling claims feeds is not out of work. It is now reconciling FHIR resources, and the errors have new names.
A worked example: the outcome measure that regressed
Consider a blood pressure control measure, an outcome measure that depends on a recorded reading below a threshold. Under hybrid reporting, a plan performed well on it, with a meaningful share of its compliant members identified through chart review, because readings were documented in clinic notes and never sent as structured data.
The plan moves to digital reporting. The measure logic is the same in substance; it now runs in CQL against FHIR observation resources. The rate drops. Not because members’ blood pressure got worse, and not because the logic changed, but because a reading in a free-text note is not an observation resource, and the correction layer that used to find it is gone.
The plan now has three possible responses, and only one of them is right. It can treat the drop as a care problem and fund clinical outreach, spending real money on members who are already controlled. It can treat it as a vendor problem and question the measure engine, which computed exactly what it was given. Or it can trace the rate back through the pipeline, find that a category of clinic feeds is arriving without structured vitals, and fix the feed. The third response is the cheapest and the only one that moves the number, and it is only available to a team that can see the lineage from the reported rate to the individual observation.
Questions to put to any vendor, including us
The transition will bring a wave of vendors selling digital readiness. Some of it is real. Here is how to tell.
- Where does normalization to FHIR happen, and who owns the mappings? If the answer is “inside the engine, you do not need to worry about it,” you need to worry about it.
- Can a reported rate be traced to the individual resources that produced it, and can that trace be shown to an auditor without an engineering ticket?
- When the certifying body publishes a new version of a measure’s logic, what changes in your pipeline, and how do you show the effect of the version change separately from the effect of the data?
- What happens to a member whose qualifying care exists only in an unstructured document?
- Which of your results are computed, and which are inferred? Inference has a place in prioritization. It has no place in a reported rate.
A vendor who can answer those five questions plainly is selling infrastructure. One who cannot is selling a dashboard on top of someone else’s assumptions.
What analytics teams should do now
You do not need to rebuild everything this quarter. You do need to stop treating the data foundation as plumbing and start treating it as the product. Three practical moves:
- Inventory where your quality data comes from and how it is transformed today, and write down the lineage. If you cannot trace a number to its source, a digital measure will not save you.
- Pick one high-value measure and model it end to end on real inputs, so you learn where your data breaks before the standard forces the issue.
- Make transparency a requirement, not a nice-to-have. When a regulator, an auditor, or a skeptical Director of Analytics asks how you got a number, the answer should be one click away.
Add a fourth if you report on any measure that currently leans on chart retrieval: compute the administrative-only rate for it now, next to the hybrid rate, and look at the gap. That gap is a preview of what the digital transition will do to your rating if nothing else changes, and it is the most persuasive number you can bring to a budget conversation about data infrastructure.
Common ways to get this wrong
The mistakes are predictable, which means they are avoidable.
Treating the transition as a software purchase. The engine is the easy part, and the certified engines will all compute the same logic. The difference between plans will be entirely in what they feed it.
Assuming standard logic means the same answer. Two plans running identical CQL against different normalization pipelines will produce different rates, and neither will be able to explain the difference unless both kept lineage.
Leaving lineage to the end. Lineage is cheap to capture at every step and nearly impossible to reconstruct afterward. A pipeline built without it will produce rates that are correct and indefensible, which in an audit is the same as wrong.
Waiting for the deadline. The transition window exists so that digital results can be run in parallel with the established method while the differences are still harmless. Every measurement year spent not doing that is a year of comparison data you will never get back.
The ClearStars view
This shift is the reason we built ClearStars the way we did. The Glass Box is not a slogan: every answer carries its lineage, every classification is visible, and every number can be traced to the source. ClearStarsLM is built digital-native and FHIR-first, for exactly the future NCQA is describing, and ClearStarsQI puts Star Ratings exposure in front of the year while there is still time to act. Clear Insights, 5 Star Outcomes. In a digital world, those insights start with the data.
References
Sources are linked inline above and collected in the references list on this page. All claims here trace to NCQA and CMS primary sources.