(117) Accelerating quality gap closure with artificial intelligence: Clinical natural language processing and large language model-driven extraction of HEDIS measures from structured and unstructured medical records

Background: : Improving patient health outcomes is a core objective of value-based care models, which reward healthcare organizations for delivering high-quality, efficient care. A critical component of achieving these outcomes is the timely identification and closure of quality gaps. However, variability in clinical documentation formats, especially CCDs, CCDAs, and PDFs, significantly hinders timely HEDIS measure abstraction and substantially increases the manual effort required to collect and validate clinical evidence. Artificial intelligence (AI) solutions, particularly Clinical NLP and Large Language Models (LLMs), are reshaping this landscape by enabling large scale accurate extraction of quality indicators from diverse clinical data formats - both structured and unstructured.

Objective: : To design and validate an AI-based solution leveraging Clinical NLP and LLMs for rapid, accurate extraction of HEDIS measures from heterogeneous clinical data sources, thereby enabling providers and payers to accelerate quality gap closure, enhance reporting workflows, and advance population health initiatives.

Methods: : A modular AI framework was developed, integrating targeted Clinical NLP pipelines with context-aware extraction algorithms powered by LLMs. The system was trained and tested on a diverse dataset of CCDs, CCDAs, and unstructured PDFs sourced from multiple EHRs. A comparative analysis was conducted against manual review processes, focusing on key outcomes such as extraction precision, time to gap closure, and a positive financial impact resulting from improvements in operational efficiency.

Results: : The deployment of the AI solution for Depression Screenings, Mammograms, and Colonoscopies led to a 45% reduction in the time required to identify open gaps, achieved 92% extraction accuracy, and resulted in a 2.5× increase in chart review throughput. Providers reported increased confidence in documentation workflows, while payers observed accelerated reporting cycles and a positive impact on Star Ratings trajectory.

Conclusions: : AI-powered HEDIS measure extraction from both structured and unstructured clinical records presents a significant opportunity for health plans and provider groups to enhance quality performance, maximize incentive revenues, and optimize resource utilization. Clinical NLP and LLMs are emerging as key enablers of scalable, data-driven care management strategies within value-based care frameworks.

Sponsorship:: Datalink Software LLC