Panel
4,000 lives
AI Capability · HCC Risk Adjustment AI
HCC risk adjustment AI for Medicare Advantage and ACO programs.
HCC risk adjustment AI reads patient charts, identifies clinically supported HCC categories, flags suspected diagnoses missing from claims, and supports the recapture documentation process. Done well, AI-supported HCC closes 20-40% more gap-closure opportunities than manual chart review alone. Done poorly, it generates suspect lists nobody trusts.
The V28 scoreboard, in six numbers.
Every figure below is either published CMS V28 model data or an anonymized benchmark from our own client book. No hypotheticals.
Avg RAF change, V28 transition
-9.3%
Same patient panels, re-scored under V28.
Medical groups we audit
ICD-10 codes re-weighted
2,300
218 codes removed from the risk model entirely.
CMS-HCC V28 final rule
Plan revenue per 0.1 RAF
$1,040
Per attributed life, per year.
Panel-level math
Avg RAF lift in 90 days
+0.14
Well-implemented mid-sized group engagements.
Anonymized client benchmark
Top-quartile recapture rate
86-88%
Industry baseline sits at 62-72%.
Anonymized client benchmark
More gap closures vs manual
20-40%
AI-supported review vs chart review alone.
Well-implemented programs
Inside the product
The chart-review pipeline, end to end.
From raw chart to RADV-defensible code: OCR extraction with per-document confidence, V28 hierarchy resolution, MEAT evidence on every suspect, and the RAF math your CFO actually asks for.
● aisuite.asprcm.com/hcc-pipeline
Chart intake · OCR extraction
STAGE 1
Progress note · office visit
PDF · 6 pages · EHR export
Extraction confidence98.2%
Discharge summary
Fax · scanned · 11 pages
Extraction confidence91.4% · verify
Lab panel · metabolic
CCD · structured feed
Extraction confidence99.1%
Suspect conditions · MEAT evidence
STAGE 2
CKD stage 4 · Patient A. (anon)
N18.4 → HCC 327 · 0.514
96%
MEAT
Evidence · progress note p.4eGFR trending down across two draws; nephrology follow-up ordered this encounter.
Confirm code →
DM2 w/ neuropathy · Patient B. (anon)
E11.42 → HCC 37 · 0.166
71%
MEAT
Evidence · medication listGabapentin active; assessment implied but not explicit in the note body.
Query CDI →
V28 hierarchy resolution
LIVE
Pays · most severe in family
E11.22 · DM2 w/ CKD
→ HCC 18 + HCC 138
RAF +0.724
Suppressed by hierarchy
E11.42 · DM2 w/ neuropathy
→ HCC 37 · 0.302
Superseded · does not stack
Removed in V28
E11.9 · DM2 w/o complication
→ No HCC
RAF 0.000
87 hierarchies enforced · only the most severe condition in each family pays
·
RAF current
1.04
→
RAF target
1.18
=
Lift
+0.14
×
Per 0.1 RAF / life
$1,040
=
Annual panel impact
$5.8M
EXTRACTION AUDIT
Model
claude-haiku-4-5
Latency
182 ms
Prompt ver
v28.4.2
Source pages
Logged per snippet
CODE-LEVEL AUDIT
MEAT evidence
Attached
Coder confirm
Required
Model version
Traceable
Audit trail
● RADV-ready
Illustrative console · anonymized demo data · coefficients per CMS-HCC V28
Chart in, defensible code out. Six stages.
Every stage leaves an audit artifact. That is the difference between a suspect list and a program that survives RADV.
01 · INGEST
Chart intake
EHR exports, faxed records, CCD feeds, labs, med lists, discharge summaries.
02 · EXTRACT
OCR + NLP
Per-document extraction confidence. Low-confidence pages routed to human verify.
03 · MAP
V28 + hierarchy
V28 condition map applied; hierarchies resolve which code in each family pays.
04 · EVIDENCE
MEAT check
Monitored, Evaluated, Assessed, Treated signals extracted; gaps prompt CDI queries.
05 · CONFIRM
Coder validation
A certified coder confirms or rejects every suggestion. No auto-submitted codes.
06 · DEFEND
RADV log + recapture
Per-code audit trail written; unconfirmed suspects feed the recapture campaign.
How hcc risk adjustment ai works in revenue cycle.
HCC risk adjustment under CMS-HCC V28 is harder than V24. Average RAF scores dropped 9.3% in V28 transition. The economic pressure on accurate documentation is more acute than ever. AI does not replace coder judgment, but it surfaces opportunities at scale that no human team can cover.
How HCC risk adjustment AI actually works
HCC AI platforms ingest patient charts (notes, problem lists, lab results, medication lists, imaging reports) and surface suspected HCC categories with supporting evidence. The platform compares suspected HCCs against the patient's claims history and flags chronic conditions that were captured prior year but not yet recaptured current year, plus new conditions not yet coded. Documentation gap reports go to clinical for recapture during the patient's next visit.
Where it works well
Large Medicare Advantage panels (5,000+ attributed beneficiaries) see the most lift. ACO REACH and MSSP populations benefit comparably. FQHC populations with high Medicaid-Medicare dual coverage see significant gap-closure opportunities because chart documentation often runs ahead of claims-level capture.
Where it struggles
Specialty-only practices (orthopedics, cardiology, etc.) see less benefit because the documentation universe is narrower. Pediatric practices have minimal HCC exposure. Hospital-employed practices with shared documentation often need EHR access to inpatient notes that the platform cannot reach without integration work.
V28 specifics
CMS-HCC V28 changes which conditions map to HCCs and reweights the coefficients. Several diabetes complications dropped to lower payment categories. Chronic kidney disease coefficients changed. Major depression and anxiety lost HCC weight. AI platforms updated for V28 are essential; legacy V24-only platforms now produce stale lists that overstate the opportunity. Verify V28 compliance before signing any HCC AI contract.
How ASP-RCM is structured differently
We do not just produce suspect lists. Our HCC AI workflow includes documentation review, recapture campaign management, and post-recapture validation, so identified opportunities actually become coded encounters. Our HCC AI dashboard tracks RAF lift, capture rate, and revenue impact in one view. Pure-play HCC AI vendors hand you a list and step aside. We close the loop.
What changes for your team.
Same coders, same clinicians, different day. The AI does the reading and ranking; your people keep the judgment calls.
Before · manual chart review
- Coders read every page of every chart hunting for conditions, highest-value and lowest-value alike.
- Recapture runs at the industry baseline of 62-72%. Prior-year conditions quietly fall off the claim.
- Suspect lists come from V24-era logic and overstate the opportunity under V28.
- MEAT gaps surface at audit time, after the chart is closed and the clinician has moved on.
- RADV prep is a scramble to reconstruct evidence for codes captured months earlier.
After · AI-assisted pipeline
- Coders validate a ranked queue, highest RAF impact first, evidence snippet attached to each suspect.
- Top-quartile clients close recapture at 86-88%, driven by pre-visit prep and AWV outreach.
- The V28 condition map and 87 hierarchies run native, so attention sits where the RAF actually lives.
- MEAT prompts fire before the chart closes, while the clinician can still fix the note.
- Every confirmed code carries its own RADV-ready audit trail from day one.
Go deeper
Frequently asked questions: hcc risk adjustment ai.
How HCC risk adjustment AI actually works
HCC AI platforms ingest patient charts (notes, problem lists, lab results, medication lists, imaging reports) and surface suspected HCC categories with supporting evidence. The platform compares suspected HCCs against the patient's claims history and flags chronic conditions that were captured prior year but not yet recaptured current year, plus new conditions not yet coded. Documentation gap reports go to clinical for recapture during the patient's next visit.
Where it works well
Large Medicare Advantage panels (5,000+ attributed beneficiaries) see the most lift. ACO REACH and MSSP populations benefit comparably. FQHC populations with high Medicaid-Medicare dual coverage see significant gap-closure opportunities because chart documentation often runs ahead of claims-level capture.
Where it struggles
Specialty-only practices (orthopedics, cardiology, etc.) see less benefit because the documentation universe is narrower. Pediatric practices have minimal HCC exposure. Hospital-employed practices with shared documentation often need EHR access to inpatient notes that the platform cannot reach without integration work.
V28 specifics
CMS-HCC V28 changes which conditions map to HCCs and reweights the coefficients. Several diabetes complications dropped to lower payment categories. Chronic kidney disease coefficients changed. Major depression and anxiety lost HCC weight. AI platforms updated for V28 are essential; legacy V24-only platforms now produce stale lists that overstate the opportunity. Verify V28 compliance before signing any HCC AI contract.
Does ASP-RCM offer hcc risk adjustment ai?
Yes. ASP-RCM Solutions delivers hcc risk adjustment ai as part of a full revenue cycle service, with senior partners on every account and a BHCOE channel partnership in the ABA segment. Request a free 30-day RCM audit.