Captured Is Not Counted: A Reflection From the RISE Risk Adjustment Forum

On Thursday morning at the RISE Risk Adjustment Forum, I joined our CTO, Yubin Park, and Susan Waterman of Baylor Scott & White on stage for Leading Edge AI Innovation in Risk Adjustment. The angle we brought to that stage was the leg of risk adjustment that is often missing and least talked about across the industry: what happens to a diagnosis after it's captured during an encounter between a provider and a member, on its way to CMS to be reflected in payment.

After the session, Yubin posted a few observations from the exhibit floor, including this one:

"Surprisingly, we were the only booth focused on encounter data submission."

He added that everyone we spoke with agreed submission matters, and that it has been overlooked lately across the industry. I agree. Looking back at the agenda and through the sessions, I was struck by how well the rest of the RISE program set up our conversation. Put together, the forum covered the whole of risk adjustment.

Three Legs, Equal Weight

In my experience, risk adjustment stands on three legs:

1. Documentation and capture: prospective suspecting, point-of-care prompts, CDI, chart retrieval, and coding
2. Submission to payment: getting each service, with its associated diagnoses, to CMS, accepted, into the risk score, and reflected in the capitation payment
3. Validation: RADV and the other audits that come later from CMS, OIG, and others to confirm the medical record supports what was paid

The second leg is the necessary middleware. It connects what was documented to what CMS uses for the capitation payment, and that payment to what auditors later validate.

All three are equally important. A plan needs all of them, woven together, to receive Fair, Accurate, Complete, and Timely (FACT) risk adjustment payments. Strong capture with weak submission leaves supported conditions out of the payment. Strong submission without supporting documentation fails validation. And validation can only check what the first two legs produced.

The forum, together with the exhibitor booths, placed its greatest emphasis on the first and third legs. On the first leg, there were sessions on prospective CDI, accuracy before the audit, and the risk adjustment workflow, plus spotlights on predicting uncoded HCCs and surfacing gaps at the point of care. On the third, RISE put CMS and OIG in the same room as the plans they oversee: a CMS session on RADV's expansion from 60 audited plans per payment year to more than 500, an OIG session on chart reviews and in-home health risk assessments, and sessions on audit-ready operating models and RADV operations. With RADV expanding and enforcement real, plans needed to hear that directly.

On the exhibit floor, sponsors showed what's possible in chart retrieval, coding, and audit readiness. Our booth and our panel brought the middle leg into that picture.

The prospective CDI session laid out a strong loop for the first leg:

Data → Signal → Evaluation → Document → Retrieval → Coding/Validation → Learn

The middle leg picks up exactly where that loop ends: when the encounter data record, with its diagnoses, leaves the building.

The Middle Leg

Two years ago, I wrote this on LinkedIn:

"Plans do not get paid on risk adjustment data in their systems nor in provider systems. But they get paid for data that is captured, submitted, accepted, filtered, allowed, added, modeled and then paid."

We came back to that idea during our panel last week. Most of what the industry talks about lives in the first word of that sentence. A documented, coded, validated diagnosis has only been captured. To count, it still has to be:

1. Submitted: on an MA encounter data record that passes pre-submission edits and reaches CMS before the deadline
2. Accepted: by CMS's Encounter Data System, which the MAO-002 report shows, line by line
3. Filtered: through CMS's risk adjustment filtering, which checks whether the record and its diagnoses are eligible for risk adjustment
4. Allowed: flagged on the MAO-004 as allowed for risk adjustment once filtering is done
5. Added: to the member's diagnosis profile, and not later removed by a delete, void, or replacement
6. Modeled: mapped to an HCC in the payment year's model, then run through hierarchies, interactions, normalization, and the coding intensity adjustment
7. Paid: combined with the member's demographic and other factors into a risk score, and reflected in the monthly capitation payment that actually lands in the plan's bank account

Each of those steps is a place where a perfectly good code can stop. The encounter is rejected for a provider identifier. The service line fails filtering. The chart review never links to an accepted encounter. The correction is ready one week after the final deadline.

None of those failures is a coding error. Every one of them costs the same as a coding miss: the condition does not count.

Submission Is Where Compliance Happens, Too

The CMS and OIG sessions made clear that regulators are focused on plans submitting too much: unsupported diagnoses, one-way chart reviews, and settlements in the hundreds of millions.

That's exactly where the third leg and the middle leg meet. A delete is a submission, too. When a plan finds a diagnosis the record doesn't support, the correction only counts if it reaches CMS, is accepted, and lands before the deadline.

So the submission pipeline is where both directions of risk adjustment become real. It's where a supported diagnosis gets folded into payment, and where an unsupported one gets removed. If a plan can't see what happened to its data after it left, it can't prove either.

Here are a few questions I'd ask any risk adjustment team after last week:

• Of the HCCs your coders captured for the last payment year, what share appears on the MAO-004?
• How many rejected encounters carrying risk-adjusting diagnoses are open right now, and who owns them?
• When a review finds a code to delete, how do you confirm the delete was accepted before the deadline?
• Can finance, coding, and encounter operations look at one member and agree on which diagnoses CMS actually used?

What Risk Adjustment Actually Is

So what is risk adjustment? My answer is the one Yubin wrote after the forum. Risk adjustment isn't just coding. It's payment integrity: getting the data right so each plan receives Fair, Accurate, Complete, and Timely (FACT) payment for the members it serves.

That takes all three legs, carrying equal weight. Documentation and capture establish what is true about the member. Submission makes sure that truth reaches CMS correctly and is reflected in payment. Validation confirms it held up. Take any one away and the payment is no longer FACT.

The industry has invested heavily in the first and third legs, and rightly so. The opportunity now is to give the middle leg the same attention, and to weave all three together so that what's documented, what's submitted, and what's validated tell the same story: the member's story.

AI will make capture faster and audits sharper. At Falcon Health, Falcon Scope uses AI anchored on the middle leg, the one that has fallen between coding, IT, and finance for too long. From there, it strengthens the other two: showing capture teams which documented diagnoses never reached payment, and giving validation teams a clear record of what was submitted, accepted, and paid.

Every file. Every member. Every dollar. Reconciled.

Thank You, RISE

None of these thoughts would have come together without the forum itself. Thank you to the RISE team for producing a forum that brought CMS, OIG, health plans, providers, and vendors into one place for three days of candid conversation. Thank you for building a program where all three legs had a place, and for having Falcon on the stage and on the exhibit floor to bring the middle leg into the conversation. Thank you to Susan Waterman for sharing the panel with us, and to everyone who stopped by booth 305 to talk. We left Orlando hopeful, and we're already looking forward to the next one.

See Falcon in Action

If your team wants to know what happened to every diagnosis after it left your building, we'd like to show you what that looks like for your plan.

Schedule a conversation with Falcon Health

This week, we're in San Diego for NHCAA's 2026 Annual Training Conference. If you're around, stop by booth 520 and say hello.