Platform-reported code agreement
Suggested codes vs. internal reference coding on evaluated encounters. Not a contractual SLA for every specialty.
Reported coding labor cost reduction
Observed in deployments that shifted repetitive chart volume from manual assignment to AI-assisted review.
Preventable coding-denial movement
Applies to coding-integrity denials (mismatch, unbundling, unspecified codes)—not eligibility or authorization.
How we measure
- Reference set: dual-review or customer gold-standard codes on a held-out encounter sample.
- Agreement: exact code match or acceptable equivalent per published coding guidelines for that date of service.
- Exclusion: illegible scans, missing notes, and encounters outside contracted specialties.
- Denials: coding-reason codes only; mix changes when payer mix or CDI program changes.
Official code-set rules remain those of ICD-10-CM and CPT. Benchmarks do not replace CMS or AMA publications.
Frequently asked questions
Are accuracy figures third-party audited?+
Figures on the benchmarks page are HealthCoder-reported from internal evaluation and customer pilots unless a named study is cited. Ask for the latest validation packet under NDA.
What is 99.9% accuracy measuring?+
Platform-reported agreement between suggested codes and a reference coding standard on evaluated encounters—not a guarantee for every specialty or note quality.