The perfect blend: AI and human validation in condition capture

Capture a more accurate and complete picture of member health

In the ever-evolving landscape of healthcare, the need for complete and accurate risk adjustment has never been more critical. Health plans rely on risk adjustment to understand the health status of their members. However, traditional methods of risk adjustment programs, which often involve collecting and reviewing medical charts to capture conditions not reported on claims, can be resource-intensive and time-consuming. 

The advent of AI has transformed this landscape. AI can efficiently sift through unstructured data, such as provider chart notes, to help identify clinical indications that may be present but not meet the stringent coding guidelines required for submission. These indications can identify suspected conditions the member may have that would benefit from a provider assessment. This process helps payers capture a more comprehensive and accurate picture of their members' health status, which is crucial for risk adjustment.

At Optum, we are committed to providing solutions that help health plans support high quality care and health outcomes for their members. Our AI-powered tools not only help identify new suspects but also connect the dots between the “tech” and the “touch,” using human intervention and interpretation to drive access, outcomes, efficiency and performance. Information from these AI-enabled reviews can be fed directly into the payer's workflow, helping to provide the most up-to-date and accurate information to inform their assessments. By leveraging AI, we can minimize the burden on providers and maximize the reuse of existing data, helping to improve compliance and accuracy of the information.

While AI plays a crucial role in identifying potential new suspects, it is only a component and not a standalone solution. The output from the AI model is reviewed by experienced medical coders, who use their expertise to verify the legitimacy of the AI-identified conditions. This human validation approach helps to ensure that only valid suspects are presented to the payer or provider, adhering to compliance and quality standards. The net result of this process is significant. On average, our AI model, combined with the verification by medical coders, identifies about 5% net new suspects. It’s important to note that the physician performs the assessment and documents any conditions present for the member.*

Our data-driven approach to analytics innovation is designed to help payers make precise and efficient decisions. By leveraging AI combined with the expertise of medical coders, we are helping payers capture a more accurate and complete picture of their members' health status, ultimately driving improved outcomes for payers, providers and members. As we continue to evolve, we will focus on adopting new technologies to help improve our solutions and meet the changing needs of the market.

Learn about Prospective Solutions for risk adjustment.