Conditions that were confirmed and submitted but never counted come back as fresh suspects. How to measure what your coders are working twice.
Ask a coder who has worked the same population for three years which conditions they recognize. They will name them. Same members, same diagnoses, same supporting documentation, opened and confirmed and signed again, cycle after cycle.
Some of that is normal. Chronic conditions require annual capture, so a recurring condition on a recurring member is the system working. What is not normal, and what nobody is counting, is the subset where last year’s capture was confirmed, coded, and submitted, and never reached the risk score at all. That condition did not recur. It failed, silently, and came back as a suspect because nothing in the process knew the difference.
Every hour spent there was paid for twice and produced nothing either time. It is the only line item in a risk adjustment budget that is pure loss, and most organizations have never put a number on it.
How a Confirmed Capture Becomes Next Year’s Suspect
The path is short and every step of it looks successful from the inside.
A coder opens a suspected condition, finds the supporting documentation, confirms it, and signs. Internal capture reporting records it. The diagnosis goes onto an encounter and the file is submitted. Then one of three things happens that nobody downstream sees: the encounter is not accepted, or the diagnosis is not treated as eligible for risk adjustment, or a later void removes it.
None of those return an error to the coding team. There is no queue of failed captures, no flag on the member, no note on the chart. The condition simply is not in the risk score.
Next cycle, the suspect logic runs. It looks at prior-year HCCs and current-year claims, sees a chronic condition that is not reported this year, and surfaces it. Correctly, by its own rules. The coder opens it, finds the same documentation they found last year, confirms it again, and signs. The organization has now paid twice for one determination, and if the underlying failure was structural rather than random, it will pay a third time next year.
The condition was never a documentation problem, never a coding problem, and never a clinical problem. It was a submission outcome that no one connected back to the work that produced it.
Why Nothing in Internal Reporting Shows It
Rework is invisible for a specific and fixable reason: both cycles record a success.
Capture reporting counts work performed. It answers how many conditions the coding team confirmed, which is a real and useful number and has nothing to say about whether those conditions counted. A condition worked twice and accepted neither time appears in reporting as two captures. Productivity looks fine. The coders were busy and the charts were real.
The only place the truth lives is in the CMS response files. MAO-002 shows whether an encounter was accepted. MAO-004 shows which submitted diagnoses were treated as eligible for risk adjustment, including those that were rejected or deleted. Neither is a coding report, which is exactly why they usually sit with a different team and are read for a different purpose, if they are read at all.
So the failure and the record of the work that produced it live in two systems, owned by two functions, joined by nobody. The suspect list is rebuilt each year from clinical and claims data that has no knowledge of either.
Want to see how much of your worklist is repeat work?
Measure It on One Population
This is a finite analysis on data you already hold, and it produces a number that is hard to argue with.
Take a prior cycle that has finished reconciling. Pull the conditions your coders confirmed and signed. Check which of them CMS recognized, against MAO-004 and the MOR rather than against internal capture reporting. Take the ones that were not recognized, and check how many appeared again on the following year’s suspect list and were worked again.
That count, multiplied by your average time per condition, is coder capacity you purchased twice. Recovering it requires no additional headcount, no additional identification, and no change to how anyone documents.
The second cut is more useful still. Group those failures by reason. If they cluster around a provider, a claim type, an encounter path, or a specific diagnosis, the rework is structural and will repeat indefinitely until the cause is addressed. Random failures cost you the hours. Structural ones cost you the hours every year, compounding for as long as the pattern goes unexamined.
What the Fix Actually Requires
The fix is not better coding and not better identification. Both of those worked correctly in every example above.
It requires that a condition keep its identity from the chart through submission and back, so that when CMS responds, the response can be matched to the determination that produced the diagnosis. With that link, a failed condition returns to the coding queue labeled for what it is: a known, evidenced, already-confirmed condition that needs a corrected encounter, not a fresh chart review. The coder does not repeat the determination. Somebody fixes the submission.
Without that link, the only mechanism available is the suspect list, which will keep surfacing the condition as though it had never been worked, because from its point of view it never was.
This is the difference between a coding tool that reports accuracy and a platform that can tell you which of its own outputs survived.
How Invent Health Closes It
Invent Health runs risk adjustment as one connected loop rather than three tools that integrate, and it starts with analytics, not coding.
Risk Analytics decides what is worth working. Suspect logic runs across historical, lab, pharmacy, comorbidity, and chart-derived signals, drawing on EMR data through CCD and FHIR. Conditions are ranked by how much they move the score and how likely they are to close, and internal capture is validated against CMS-recognized HCCs in the MOR with RAF progression tracked across payment cycles in the MMR.
The Coder Workbench confirms the condition is real. AI-assisted coding reads the clinical documentation, shows the supporting evidence, and recommends ICD-10-CM codes for a certified coder to confirm and sign. Invent Health’s NLP detects that supporting evidence with 85 percent or higher accuracy out of the box, and the Coder Workbench averages 95 percent coding accuracy. It is a two-pass, coder-in-the-loop model, and every code stays linked to the chart for audit-defensible lineage.
Encounter Submissions makes sure it counts, validating 837P, 837I, and DME for Medicare EDPS and Edge Server XML for ACA, flagging missing and non-risk-eligible diagnoses before a file leaves, then reconciling MAO-002 and MAO-004 for Medicare and the Edge Server reports for ACA. A diagnosis that did not count returns to the worklist as a known condition needing a corrected encounter rather than reappearing as a fresh suspect.
That last sentence is the entire point of the three products sharing a platform. The link that makes it possible cannot be added afterward by a report, because the identity of the condition has to survive every step for the response to be matchable at the end.
Why This Gets More Expensive in 2027
Rework has been survivable because a retrospective sweep could recover the condition eventually, at a cost nobody attributed to the original failure.
The CY 2027 Rate Announcement ends unlinked chart reviews beginning with the 2027 payment year, so a chart review diagnosis has to tie back to an encounter CMS accepted, and the July HPMS memo on CY 2027 risk adjustment implementation sets out what has to be in place. A condition whose encounter failed no longer has a chart review path to payment, which means the repeat work is not just wasted, it is wasted on something that now cannot be recovered late.
V28 removed a large share of diagnosis codes from risk-adjustment mapping and re-based values, so fewer conditions carry payment and the hours spent repeating work on one of them are hours not spent on the smaller set that still pays.
Run It on a Closed Cycle
Pull the conditions your coders confirmed in a finished cycle, check which ones CMS recognized, and count how many of the rest came back and got worked again. That number is the capacity you already bought.
Frequently asked questions
What is coder rework in risk adjustment?
Coder rework is time spent re-confirming a condition that was already confirmed, coded, and submitted in a prior cycle but never reached the risk score. Because nothing flags the failure, the condition reappears on the suspect list and is worked again from the beginning.
Why do the same HCC conditions reappear as suspects every year?
Chronic conditions legitimately require annual capture. The problem is the subset where last year’s capture was confirmed and submitted but not accepted, filtered as not risk-eligible, or later voided. Suspect logic cannot distinguish those from conditions that were never captured.
Why does internal capture reporting not show rework?
Because capture reporting counts work performed rather than diagnoses CMS recognized. A condition worked twice and accepted neither time appears as two captures, so productivity and volume both look normal while the underlying determination produced no revenue.
How do you measure how much coder time goes to rework?
Take a reconciled cycle, pull the conditions coders confirmed, check which CMS recognized against MAO-004 and the MOR, then count how many unrecognized conditions appeared on the next suspect list and were worked again. Multiply by average time per condition.
What is the difference between MAO-002 and MAO-004?
MAO-002 reports whether a submitted encounter was accepted or rejected. MAO-004 reports which diagnoses on encounters were treated as eligible for risk adjustment, including those rejected or deleted. Read together they show whether a coded diagnosis actually counted.
Is repeat suspecting always a problem?
No. Chronic conditions must be captured each year, so recurrence is expected. The problem is the portion of recurrence caused by a prior submission failure rather than by the annual capture requirement, which is invisible without CMS response reconciliation.
How can you tell whether rework is random or structural?
Group the failed conditions by provider, claim type, encounter path, and diagnosis. Clustering indicates a structural cause that will repeat every cycle until it is addressed. Scattered failures cost the hours once; structural ones cost them indefinitely.
What should happen when a submitted diagnosis is not accepted?
It should return to the worklist identified as a known, already-confirmed condition needing a corrected encounter, not as a new suspect requiring a fresh chart review. That requires the condition to keep its identity from chart through submission and back.
Why can a reporting layer not solve rework on its own?
Because matching a CMS response to the determination that produced it requires the condition to be identifiable at every step. If that identity is not preserved through coding and submission, no report added later can reliably reconnect the two.
Does the 2027 chart review change affect rework?
Yes. Starting with the 2027 payment year, chart review diagnoses must tie back to an accepted encounter, so a condition whose encounter failed has no late path to payment. Repeat work on it is not just duplicated, it is spent on something that can no longer be recovered retrospectively.
How does rework affect coder capacity planning?
It consumes hours that appear productive in every internal measure, so capacity models built on capture counts overstate real output. Recovering those hours requires no additional headcount and no additional identification spend.
How does Invent Health prevent coder rework?
Analytics, coding, and encounter submission run on one platform, so each diagnosis stays linked from chart through coder validation and submission. When MAO-002 and MAO-004 are reconciled, a diagnosis that did not count returns as a correction rather than as a fresh suspect.
