AI Features
Every AI feature in Carelytics is anchored to the documentation in the chart. Coding Review cites the chart text it found. Pre-Sign QA flags missing CMS Conditions of Participation items. SOAP draft writes from the structured form data the clinician already entered.
We don't recommend changes purely to maximize reimbursement. OIG-safe by design.
What AI does in Carelytics
Drafts Subjective / Objective / Assessment / Plan paragraphs from the structured form data the clinician already entered. Clinician reviews, edits, signs.
Scans the form for CMS Conditions of Participation issues before the clinician signs. If issues, a modal lists them and the clinician fixes or overrides.
Medicare-compliant PDGM coding accuracy check. Flags RTP risk, specificity gaps, and doc-vs-code mismatches with cited documentation evidence.
Smart copy from the patient's prior visit, skipping vitals and visit-specific stamps. Banner appears automatically when prior visits exist for the same form template.
Why this matters
The HHS Office of Inspector General has been explicit: AI tools that recommend coding changes purely to maximize reimbursement are a fraud risk. Carelytics's AI is built to live on the right side of that line — and to keep your agency on it too.
Every AI finding cites the chart text it found. The clinician sees what the AI saw. There's no opaque "score increased."
Coding Review surfaces downcoding gaps as well as upcoding risks. We're checking accuracy, not optimizing for revenue.
AI never auto-applies coding changes. The clinician (or coder) reviews each finding, accepts or rejects, and signs.
Every AI run is logged with the prompt category, input fingerprint, and which findings were accepted. Reviewable per OASIS.
Enterprise tenants can customize prompts to match their coding manual. We never tune the prompt to "lift case mix."
Every AI feature falls back to rule-based logic when AI is unconfigured. A new agency that hasn't set up an AI key still gets the workflow.
AI POC Generation
§21 Orders/Interventions and §22 Goals are drafted from the patient's actual OASIS data, primary diagnosis, fall risk, hospitalization risk, and visit frequency. Not the same three paragraphs every other agency in the country uses.
