One of the Largest Regional Provider-Owned Health Plans in Its State Surfaces 90K+ HCC/RxHCC Suspects with RAAPID
510K eligible members. 257K targeted. One prospective review built on incomplete data.
Challenge
From a large population to a reviewable one
- The health plan shared data on 874K members, about 40M pages of charts.
- 510K of them were eligible for prospective review.
- 03. RAAPID matched charts to claims. Only 257K members had both.
Many eligible members had missing charts or limited claims history. Identifying emerging conditions meant working across these data gaps, not just reviewing the charts that were available
Solution Overview
RAAPID scored every targeted member using charts, claims, labs, and pharmacy data, including members with no charts. Document AI read the unstructured charts, including tables, forms, handwritten notes, and multi-column layouts, and identified dates of service and document types. Neuro-Symbolic AI combined neural models with clinical Knowledge Graphs and coding rules so that every suspect links to its supporting evidence and the reasoning can be traced. Suspects went to the physician before the visit. Nothing was coded unless the provider confirmed the condition during a real encounter.
Key Outcomes
- 90K+ HCC/RxHCC suspects identified
- 61K members with suspect conditions (1 in 4 reviewed)
- ~$75M potential value
- ~30X potential ROI, net of project cost
- Every suspect linked to traceable clinical evidence
Potential value is based on the plan’s average values of $2,500 per HCC and $600 per RxHCC, at 50% provider confirmation. Potential value, not realized revenue.
Client Testimonial
“We needed suspecting emerging conditions in front of our providers during the encounter, with the evidence right there. RAAPID reviewed our population, and every suspect came back linked to the chart evidence behind it, with a multi-million dollar potential benefit for our plan.”
— Director, Risk Adjustment, Regional Provider-Owned Health Plan