Every risk-adjustment program I review asks the same question: are the diagnoses we submitted supported? It is the right question. It is also only half of one.
The other half almost nobody asks: which patients never made it into the analysis at all? Because a patient who is invisible to your data is invisible to your risk score, your care management, and your quality reporting simultaneously.
1. The two failure modes, and why only one gets attention
Over-capture triggers an audit. Under-capture just quietly costs you money and outcomes.
Submitting an unsupported diagnosis is a compliance failure with a name, a regulator, and a penalty. Missing a real one is an invisible failure with no alarm attached — no one sends a letter about the condition you never documented.
That asymmetry in consequences produces an asymmetry in attention. Organizations build entire programs around defending what they captured and almost nothing around finding what they missed — even though the second failure harms patients directly.
2. Who falls out of the denominator
The patients missing from your analysis are rarely random. They are the ones who did not have a qualifying face-to-face encounter in the data-collection period — the homebound, the transportation-limited, the disengaged, the newly attributed, the ones who see specialists outside your network.
There is a cruel logic to it: the patients hardest to reach are the ones least likely to generate the encounter that documents how sick they are. Your data then reports them as low-risk, and your care management deprioritizes them accordingly.
3. This is the same trap as cost-trained risk models
Absence of documentation gets read as absence of disease.
It is worth naming the pattern, because it recurs. A model trained on cost mistakes low spending for low need. A risk score built only on submitted encounters mistakes low documentation for low illness. Same error, different mechanism.
In both cases the system concludes that the patient it has failed to serve is the patient who does not need serving. Under a total-cost-of-care contract, that patient reappears later — in the emergency department, at full price.
4. A bidirectional review is the professional standard
This is where the audit discipline actually points. A defensible risk-adjustment program reviews in both directions: removing diagnoses the record does not support, and identifying conditions the record clearly supports that were never submitted.
One-way review — adding codes but never deleting them — is precisely the pattern federal oversight has flagged as a red flag. A program that only adds looks like revenue maximization. A program that corrects in both directions looks like what it should be: accuracy. And note the compliance obligation runs with it — unsupported codes identified must be addressed, not quietly left in place.
5. How to actually audit your denominator
Practical steps. Identify attributed patients with no qualifying encounter in the period — that list is your blind spot, and it is usually longer than leadership expects. Cross-check patients with historically documented chronic conditions that did not recur this year, and ask whether the condition resolved or the visit simply never happened. Look for patients whose pharmacy data implies conditions their diagnosis data never captured; a patient filling insulin with no diabetes documentation is telling you something about your process, not their health.
Then treat the output as a care problem before a coding one: these people need to be seen, not merely coded.
Final Thoughts
The risk-adjustment field has become very good at defending the numerator and largely indifferent to the denominator. That imbalance is a compliance artifact, not a clinical one.
As someone trained as a physician who later became a certified medical auditor, I have come to think the denominator question is the more revealing one. It tells you who your system is failing to reach — and under any risk-bearing contract, the patient your data cannot see is the patient who eventually costs you the most.
If your risk-adjustment work only asks whether your submitted diagnoses are supported, you are auditing half the problem. At HealtheNomics I help organizations run bidirectional reviews — defending what you captured and finding the patients your data never saw.
Explore the services: https://healthenomics.com/services-2/
Request a strategy conversation: https://healthenomics.com/contact-us/
Connect on LinkedIn: https://www.linkedin.com/in/muhammad-ayoub-ashraf/
Website: https://www.drayoubashraf.com
Watch on YouTube: https://www.youtube.com/@HealtheNomics
