Developing Computable Quality Indicators for Knee Osteoarthritis Using Structured and Free-Text EMR Data

Background: To develop and evaluate a rule-based pipeline that generates computable quality indicators (QIs) for outpatient knee osteoarthritis (OA) using structured and free-text electronic medical record (EMR) data. Methods: We conducted a retrospective observational study using seven years (2018–2024) of outpatient EMRs from a tertiary hospital in China. Knee OA encounters were identified using a computable phenotype integrating structured codes and free-text terms. Five guideline-aligned process QIs were operationalized with transparent rules applied to structured fields and curated text lexicons. Indicator extraction, distributions, temporal trends, and correlations were assessed. Results: A total of 11,833 encounters met criteria. All five QIs were computable across all encounters. The mean total quality score was 51.3 (SD 17.2). Structured-data indicators showed higher and more stable performance, whereas text-dependent indicators particularly exercise or rehabilitation advice had extremely low extractability (mean 0.27/10). Indicator correlations were low, indicating distinct care domains. Scores showed modest temporal variation with stable pipeline performance. Discussion : Rule-based computable QIs enabled scalable, interpretable measurement of knee OA care and revealed persistent gaps in guideline-recommended knee OA management. Findings highlight structural limitations in EMR documentation that constrain visibility of high-value care processes. Conclusion: Integrating structured and unstructured EMR data enables reproducible, guideline-based quality assessment for chronic OA care and offers a foundation for continuous quality monitoring and improvement.