Your retrospective HCC coding program captured 12,000 incremental HCCs last year. Leadership celebrates the revenue impact. You present at executive meetings about program success.
Then someone asks: “How many HCCs did your retrospective program remove or downgrade?”
You don’t track that metric. Your program only adds HCCs, never removes them.
That’s a problem. The add-only pattern is now a red flag for CMS auditors, DOJ investigators, and your own compliance team. Here’s why delete rates matter and what healthy retrospective programs look like.
The Two-Way Coding Principle
Retrospective coding should improve documentation accuracy in both directions. Sometimes improving accuracy means adding HCCs that were undercoded. Sometimes it means removing HCCs that were overcoded or unsupported.
If your retrospective program only finds additions, one of two things is true: (1) Your prospective and concurrent coding is so conservative that it never overcodes (unlikely), or (2) Your retrospective program isn’t actually focused on accuracy, it’s focused on revenue (likely).
CMS and DOJ investigations consistently flag add-only retrospective programs as evidence of revenue-focused rather than accuracy-focused coding.
The fix requires measuring and reporting delete rates alongside add rates. A healthy retrospective program should remove 5-15% as many HCCs as it adds.
The Unsupported Diagnosis Problem
Your retrospective coder reviews a chart where the prospective coder assigned diabetes with nephropathy. The retrospective coder sees “diabetes with nephropathy” documented but finds no GFR values, no nephrologist involvement, no renal medication management, and no MEAT criteria supporting nephropathy.
What should the retrospective coder do?
In add-only programs, the coder leaves it alone. “Someone already coded it, not my job to question it.”
In compliance-first programs, the coder removes the nephropathy diagnosis or downgrades to diabetes without complications because documentation doesn’t support the severity coded.
Organizations with add-only programs are perpetuating overcoding. They capture undercoding but ignore overcoding.
The fix requires explicit guidance to retrospective coders: your job is to ensure coding accuracy, which includes removing or downgrading diagnoses that aren’t adequately supported.
The Coding Drift Detection
Retrospective programs that track delete rates can identify systematic coding problems prospectively.
If retrospective coders are consistently removing CHF codes that prospective coders assigned, that reveals a systematic CHF coding error in the prospective program.
If retrospective coders are consistently downgrading CKD staging, that reveals prospective coders are applying staging criteria incorrectly.
Add-only programs never discover these patterns. They add new HCCs prospective coding missed but don’t track what prospective coding got wrong.
Organizations with two-way coding use delete data to improve prospective coding through targeted retraining.
The Provider Feedback Value
When retrospective coding removes an HCC, that creates opportunity for provider education.
Dr. Martinez documented “diabetic retinopathy” but retrospective review found no ophthalmology exam, no fundoscopic findings, no evidence of retinopathy. The retrospective coder removes the diagnosis.
Sharing this with Dr. Martinez helps improve future documentation: “When documenting diabetic retinopathy, please include supporting exam findings or ophthalmology consultation notes.”
Add-only programs only tell providers about documentation they missed. Two-way programs also educate providers about documentation they overcoded, creating more balanced feedback.
The Audit Defense Position
When CMS conducts RADV audits, they scrutinize your retrospective coding program.
If your retrospective program shows 12,000 adds and 0 deletes, CMS asks: “Was this program designed to find documentation gaps (accuracy-focused) or to find revenue opportunities (revenue-focused)?”
If your retrospective program shows 12,000 adds and 1,200 deletes, you can demonstrate: “Our program improves coding accuracy bidirectionally. We found undercoding and overcoding. Here’s our documentation showing we prioritize accuracy over revenue.”
The delete rate becomes evidence of program intent.
The Realistic Delete Rate Expectations
Organizations implementing two-way retrospective coding ask: “What’s a normal delete rate?”
It depends on your prospective coding quality. If prospective coding is conservative and rarely overcodes, delete rates will be low (2-5%). If prospective coding is aggressive or error-prone, delete rates will be higher (10-20%).
Most organizations fall in the 5-10% delete rate range when they implement genuine two-way coding.
Delete rates below 2% suggest coders aren’t seriously evaluating overcoding. Delete rates above 20% suggest major prospective coding problems.
The Coder Incentive Alignment
Many retrospective programs pay vendors or internal coders based on HCCs captured. This creates financial incentive to add HCCs and disincentive to delete them.
A coder paid per HCC captured will find lots of additions and zero deletions. That’s rational economic behavior given the incentive structure.
Organizations implementing two-way coding need to adjust compensation models. Pay for coding accuracy, not HCC volume. Measure and reward coders who appropriately add AND appropriately delete.
What Actually Works
Building compliance-first retrospective programs requires tracking and valuing delete activity.
Measure delete rates alongside add rates (target 5-15% delete rate relative to adds). Provide explicit guidance that retrospective coders should remove unsupported HCCs. Use delete patterns to identify and fix prospective coding errors. Leverage delete examples for provider education. Present delete rates as evidence of accuracy focus during audits. Align coder incentives to reward accuracy, not volume.
If your retrospective HCC coding program has zero delete activity, you don’t have a coding accuracy program. You have a revenue optimization program. That distinction matters enormously during audits and investigations.
The organizations successfully defending retrospective programs during scrutiny are the ones that can show bidirectional coding focused on accuracy. Add-only programs can’t make that defense credibly.
