Where HCC Recapture Programs Quietly Lose Accuracy
Every measurement year, a familiar pattern shows up in risk adjustment programs: leadership reviews the prior year's RAF trend, sees a dip, and asks what changed. Usually, nothing dramatic changed. The accuracy eroded slowly, in three predictable places that don't show up until the aggregate numbers move.
The first is the gap between suspected and confirmed conditions. Prospective chart prep is supposed to hand providers a concise list of likely undocumented chronic conditions ahead of a visit. In practice, that list often arrives too close to the appointment, buried in the EHR, or formatted in a way that doesn't fit how a provider actually works through a visit. A suspect list a provider doesn't see is functionally the same as no suspect list at all — and recapture rates reflect that quietly, one missed condition at a time.
The second is retrospective review that stops at "was it coded," instead of asking "was it coded correctly and specifically." A condition can be technically captured and still miss the correct HCC category — diabetes without documented complications coded as if complications were present, or vice versa. These aren't dramatic errors individually, but they compound across a population and they're exactly the kind of finding that turns into a RADV audit problem later.
The third, and the one most programs underweight, is documentation specificity that would not hold up if it were actually audited. It's possible to build a recapture program that looks accurate on a coding accuracy scorecard while still carrying real audit risk, because the coding matched what was written, but what was written wasn't specific enough to support the code under formal review standards.
The fix for all three is the same discipline, applied earlier: build prospective review that's genuinely usable at the point of care, run retrospective review against RADV-level documentation standards rather than presence-of-code standards, and close the loop with providers on documentation patterns rather than only sending more queries. Programs that do this don't just protect their RAF score — they build a recapture process that's defensible if a payer or CMS ever looks closely.