HCC Prospective Coding

Coding & Risk AdjustmentAsync jobs required

Point-of-care HCC coding — only conditions actively addressed in the encounter.

HCCProspectivePoint of careRisk adjustment

About

Real-time HCC coding for the current encounter. Unlike retrospective coding, which sweeps the whole chart, the prospective pipeline only surfaces conditions the provider actively managed in this visit — enforced by an actively_managed governance flag on every code.

Built for pre-visit planning and in-encounter assist, where the question is "what risk-adjustable conditions is this provider addressing right now", not "what was ever documented".

How it works

  1. 1Encounter note upload → OCR (skipped for the /text variant)
  2. 2Pass 1 — prospective diagnosis extraction limited to actively addressed conditions
  3. 3Pass 2 — shortlisting, MEAT validation, HCC V24/V28 enrichment, governance (including actively_managed)

Intended use

  • Routed to when users ask for "prospective HCC" or "point-of-care" coding
  • Pre-visit planning views and encounter-time coding nudges
  • Prospective risk-capture programs that must avoid coding from history alone

Key outputs

  • result.ICD_descriptions[] — same evidence-rich code structure as HCC Retrospective
  • Per code: hcc_v24 / hcc_v28, MEAT evidence with bboxes, governance_checklist
  • governance_flags.actively_managed — the prospective-specific gate

Model comparison

F1 on Gwen's healthcare benchmark for this task — the Gwen pipeline vs the prompt-optimized model alone, with the uplift the pipeline adds, per model.

#ModelGwen pipelineModel onlyUplift
1
Claude Opus 4.8Best
0.967
0.967+0.000
2
GPT-5.5
0.924
0.855+0.069
3
Gemini 3.5 Flash
0.897
0.813+0.085

Endpoints

Try each endpoint with your signed-in session — usage counts toward your monthly budget.

Use synthetic data only. Do not submit real patient records or PHI when testing endpoints.

Limitations & caveats

  • No reference_codes input — prospective coding deliberately ignores externally supplied code lists
  • Conditions documented only in past medical history are excluded by design; use HCC Retrospective for full-chart sweeps
  • Runs 1–3 minutes; the async /jobs flow is mandatory for document uploads