Automating Quality to Accelerate Transformational Improvements in Care (AQUATIC)

Abstract Current methods for evaluating quality measures are time-consuming, costly, and often delayed, limiting the effectiveness of timely quality improvement (QI) initiatives. Typically, quality performance assessments are restricted to small patient samples evaluated months after discharge, resulting in inconsistent and statistically underpowered data. This lag hampers the ability to use insights for immediate feedback and improvement. To overcome these limitations, we have developed QuaLLM, an innovative automated toolset that allows real-time abstraction and evaluation of quality measures. QuaLLM enables key stakeholders to assess quality and the effectiveness of improvement-oriented changes in near real-time, with greater efficiency. Our approach leverages Fast Healthcare Interoperability Resources (FHIR) to enable seamless electronic health record (EHR) data exchange and generative artificial intelligence (GenAI), specifically large language models (LLMs), to automate clinical quality reporting. The proposed work aims to develop a foundational model for quality measure abstraction that is easily adaptable to new measures. We will prospectively validate QuaLLM against existing reports generated within a large academic health system using a variety of infection related measures that are reported to CMS. This proposal seeks to revolutionize healthcare QI by replacing the traditionally manual, resource-intensive abstraction process with a real-time, AI-driven solution. The successful implementation of this project could lead to enhanced patient outcomes through timely insights into quality performance, substantial reductions in healthcare costs, and the establishment of a more agile and responsive learning healthcare system.