WellSpan Bringing AI Tools to Clinical Documentation Integrity
The Centers for Medicare and Medicaid Services (CMS) recently announced it would ramp up Risk Adjustment Data Validation (RADV) audits to confirm that medical records support diagnoses used for payment. Calling this announcement the “elephant in the room,” Julie Platt, M.D., M.B.A., medical director for clinical transformation at WellSpan Health, recently detailed the work that York, Pa.-based WellSpan has been doing with Premier Inc.’s Stanson Health business on adding AI-based tools to enhance clinical documentation integrity.
Speaking during a recent webinar, Platt noted that with the new RADV rule, financial penalties are no longer limited to the audit sample. They're extrapolated across the entire Medicare Advantage, population. “Documentation integrity is no longer a compliance detail,” she said. “It's a major financial and operational lever.”
WellSpan is an integrated healthcare system based in central Pennsylvania. It has more than 23,000 employees and more than 250 care locations, seven acute-care hospitals, two specialty hospitals and seven ambulatory surgery centers. It also has its own behavioral health organization and home health agency.
WellSpan’s value-based care portfolio includes a large ACO, a co-owned Medicare Advantage plan, an employee health plan, and a diverse mix of commercial and government contracts. Approximately 28% of its system revenue is tied to value-based care arrangements.
Platt said the tool WellSpan is working on with Premier won't fix any past documentation errors, but it will improve Hierarchical Condition Categories (HCC) accuracy going forward, and support sustainable, high-quality documentation at scale.
She said the tool they are working on is not just about smarter coding. It's about smarter collaboration. “It's designed to reduce friction, improve provider engagement and align documentation with clinical truth,” Platt said. “In today's regulatory environment, we can't afford ambiguity. Clear, accurate documentation isn't just best practice, it's essential for compliance, financial sustainability and patient trust.”
Unlike many healthcare systems, Platt explained, WellSpan does not have a dominant presence in Medicare Advantage that it owns. “This has shaped a unique approach to clinical documentation integrity. With a clear focus on doing the right thing for patients and providers, our strategy prioritizes clinical accuracy and provider engagement,” she said. “HCC risk capture is framed not as a coding exercise, but as a reflection of the true patient complexity. This matters because our care management workflows are directly informed by the risk scores, enabling more personalized, proactive care.”
She then explained why AI-based tools are beginning to transform this space. “When we began designing an ambulatory clinical documentation integrity program in 2022, we needed a solution that was people-centered, that could integrate seamlessly in our existing workflows, support our teams in improving diagnostic accuracy from the very start of the patient visit to the post-visit coding,” she said. “We explored the market for technical solutions. Epic didn't offer what we needed, and other platforms fell short. Meanwhile, ChatGPT was emerging as a powerful tool for summarizing and synthesizing data — capabilities that we thought were highly relevant to our goals. Given our strong relationship with Stanson and Premier, we began collaborating in 2022 to explore a new path forward to leverage AI to streamline point-of-care diagnostic accuracy while accelerating post-visit coder efficiency.”
In March Premier’s CodingCare solution was granted official Toolbox status by Epic. “That's a rare distinction currently held by just 26 programs across the country,” she said. “I’ll be honest, I did not initially grasp how significant that designation was, but what it means is that Epic has recognized this solution as both unique and valuable, and they are actively investing resources to support its deeper integration.”
'Major step forward'
The progress is already visible, she said. “In just the past few weeks, we've seen exciting new developments, most notably the rollout of AI-driven text highlighting in the pre-visit workflow to support suggested diagnoses. This is a major step forward in making documentation smarter, faster and more aligned with the clinical story.”
Platt said the core goal with this work is simple but powerful — to use AI to surface the right diagnosis at the right time with the clarity and the context so that providers can act with confidence. “This isn't about more alerts. It’s about better ones grounded in clinical logic, delivered within the workflow and designed to reduce friction while improving the accuracy of HCC capture and documentation from the very beginning of the encounter,” she said.
This includes targeted problem list cleanup, ensuring that only accurate, validated conditions flow into the Epic-native HCC alerts. During the visit, anything that the provider saw would be reviewed by the nurse. They could remove inaccurate diagnoses and add substantiated ones supported by relevant clinical data, Platt added. “That approach builds trust in the tool and enhances provider efficiency. Then in the post-visit space, our vision was to include real-time AI MEAT [Monitoring, Evaluating, Assessing/Addressing, and Treating] detection within the clinical note so that these insights were served up to the coders, who could then confirm the diagnosis or send a query to the provider if clarification was needed. If our providers did not respond within the time that they were allotted, that HCC was automatically removed before claim submission.”
Platt says this model strikes a balance between accuracy, efficiency and accountability, while reducing friction across the care team. “In summary, we wanted to leverage AI in all portions of the workflow, and we have found the most value so far in the post-visit workspace. We anticipate the pre-visit and provider portion of the workflow will improve quickly when AI summaries are operational in the fall.”
