Transforming healthcare risk adjustment with AI-powered NLP

The challenge: Vendor dependence and risk exposure in risk adjustment

A leading healthcare insurance provider with over $60 billion in assets faced a growing challenge in its risk adjustment process. Heavy reliance on third-party vendors for medical record coding led to data quality issues, increased costs, and compliance risks.

With vendor costs escalating and turnaround times impacting revenue cycles, the payer recognized the urgent need for an in-house solution. The challenge was compounded by the complexity of medical coding, requiring sophisticated natural language processing to accurately extract and categorize diagnoses from unstructured clinical notes. The organization needed a solution that could match or exceed vendor accuracy while providing greater control, flexibility, and cost-effectiveness.

Our AI-powered NLP solution: From vendor dependence to in-house excellence

Inspire11 partnered with the client to embark on an ambitious 8-week proof-of-concept project, designed to validate the feasibility of a proprietary NLP solution. Our approach was methodical yet agile:

Over eight weeks, we led a focused initiative to transform the risk adjustment process with AI. In the first two weeks, we conducted deep-dive workshops with clinical coders, IT leaders, and business stakeholders to map the end-to-end workflow. These sessions revealed key pain points, current vendor dependencies, and a critical insight: 80% of coding efforts focused on a narrow set of high-value diagnosis codes. This allowed us to target our proof of concept for maximum impact.

From weeks three to six, we rapidly built a functional NLP engine using transformer models fine-tuned for medical language. The solution featured custom entity recognition for ICD-10 codes, context-aware logic for negation and temporality, confidence scoring to flag ambiguous cases, and integration APIs for seamless EMR connectivity. Weekly demos with end-users ensured alignment and continuous refinement. We validated performance by processing over 10,000 historical records and benchmarking against vendor-coded data.

In the final phase, we developed a clear business case showing strong ROI. We compared vendor vs. in-house costs, projected accuracy-driven revenue lift, outlined compliance safeguards, and presented a phased implementation plan. A dedicated change management approach ensured coders would be equipped and empowered for the transition.