Maximize RAF & close risk gaps without manual chart-chasing

CMS-HCC v22/v24/v28, ESRD-HCC v21/v24, RxHCC v08, and ACA/HHS with plan-specific overrides. ICD-10-CM/PCS vocab alignment for explainability.

Encounters/notes, problems, meds, labs, claims, eligibility; we operate on any subset and improve as sources expand.

Tunable; typical precision/recall 0.90–0.95 with explainability and MEAT validation gates. Benchmarks are validated by model version before rollout.

Yes—via FHIR (Problems/Tasks/Communication), native APIs where available, and OpenEMR care-management flags on the patient banner. Epic: Live. athenahealth: Live. OpenEMR: Live. eClinicalWorks: Pilot. NextGen/Meditech: Roadmap → Q2'26.

<2 weeks for standard FHIR + SFTP feeds; enterprise SSO and VPC peering commonly add 1–2 weeks. P50: 12 days. P90: 28 days.

Yes—evidence packets, lineage, and binder exports reduce RADV exposure and prep time. Audit binder prep: 3.5 weeks → 4.5 days (-76%). RADV discrepancies (sampled): -41%.

By member-month or chart volume; unlimited internal users on Growth/Scale.

Yes—Scale supports private VPC deploys and customer-managed keys.

In-workflow suggestions cite source text + MEAT and respect provider specialty & recency rules. Confidence thresholds + specialty priors + 90-day suppressions cut query volume −28%.

Every suggested HCC must have Monitor, Evaluate, Assess, or Treat support before it can be accepted. If MEAT is not met, reviewers reject or hold the HCC with an insufficient-documentation reason and route CDI follow-up.

Yes—optional ingestion improves suspecting and risk stratification. Z-codes + external indices supported.

Yes. When ESRD status flows in from the EHR, such as dialysis codes N18.6 and Z99.2 in OpenEMR, new RAF jobs default to the ESRD-HCC v24 model and log the segment change for audit.

Versioned policies, champion–challenger, and drift monitoring with rollback. Monthly challenger runs; promotion if AUPRC +1.5 pts with no ↑ in abrasion. Drift alerts if precision drops >1.5 pts week-over-week.

If a job stays in PROCESSING for more than 5 minutes, reviewers can open logs, wait 10 minutes, then use Cancel and Retry. Multiple stuck jobs should be filed as a High-priority support ticket because they usually indicate a coefficient cache, coefficient file, database, or concurrency-cap issue.

Gap closure, RAF uplift, review time per chart, and audit packet completeness. Typical first-month improvements: Gap closure +31.8%, First-pass acceptance 96.2%.

Yes. Clinicians can accept, modify, or reject suggested HCCs in review. Their answer wins, the RAF score updates to match, and the decision is logged for audit.

Graceful degradation with evidence sufficiency score; low-sufficiency items gated for manual review.

Yes — CSV/Parquet via UI/API; S3 drops; Binder PDFs per payer template.

Template library (CMS, state, major MA payers); custom templates in ≤10 business days.