How Aetna Uses Gen AI to Automate HEDIS Medical Chart Reviews

Aetna has deployed a generative AI-driven document intelligence platform that reduced the manual effort required for annual medical record reviews by 65%, according to Nathan Frank, Aetna’s chief digital and technology officer. The platform, known as AI Medical Chart Review, automates the extraction of clinically relevant data from millions of unstructured medical records to identify gaps in patient care.

The technology specifically targets the Healthcare Effectiveness Data and Information Set (HEDIS), a series of performance measures for the managed care industry. Developed by the nonprofit National Committee for Quality Assurance (NCQA) in 1991, HEDIS measures track whether patients receive necessary preventative care and chronic disease management, such as cancer screenings, immunizations, and blood sugar tests for diabetics.

Scaling Medical Record Analysis with Generative AI

Large managed care providers like Aetna review more than 10 million medical records annually to close care gaps. Frank stated that industry benchmarks for large providers suggest a manual review process would require approximately 50,000 work weeks per year, which is equivalent to nearly 1,000 full-time employees. He noted that a team of 50 reviewers would take more than 20 years to complete a single annual review using purely manual methods.

The AI Medical Chart Review platform utilizes cloud services and generative AI to process these records. The system ingests unstructured documents, including physical clinical documentation and charts containing handwritten notes, to identify diagnosis codes, medication records, lab results, and visit documentation. According to Frank, the platform uses large language models to decipher charts and build correlations to identify high-value codes.

Once the data is extracted, the platform generates a prioritized list of records based on the strength of the clinical evidence and the likelihood of closing a care gap. These prioritized records are then passed to trained medical coders for human review and validation.

Development Timeline and Operational Impact

Aetna developed the platform through an iterative process involving engineers and subject matter experts. Frank said the team ideated the platform and designed a proof of concept that processed millions of records in two weeks within a six-month window. The company has since processed 14 million documents using the system.

The reduction in manual labor has allowed Aetna to shift staff focus toward quality control and verifying the accuracy of the automated reviews. Frank stated that the platform has streamlined workflows and increased gap closure rates, which leads to higher reimbursement and improved Star Ratings.

“Something that might have taken weeks or months we can now do in days.”

Nathan Frank, chief digital and technology officer at Aetna

AI Governance and Implementation Strategy

Aetna’s approach to the build focused on a cloud-native model with an emphasis on elasticity and cost optimization. Frank attributed the speed of the rollout to the use of small teams and a reduction in bureaucracy, noting that business subject matter experts worked in the same physical or virtual rooms as the engineers.

To manage the risks associated with generative AI, the company implemented an AI governance model to ensure the technology is used with appropriate guardrails. Frank said that security, compliance, and responsible AI use were core principles from the start of the project. Due to these innovations, the AI Medical Chart Review platform received a CIO 100 Award for IT innovation.