Development and evaluation of a hybrid algorithm for identifying hospitalized patients with cirrhosis

Background: Electronic medical record (EMR)–based quality improvement (QI) tools for cirrhosis care require accurate patient identification. Combining administrative and EMR data may enhance cirrhosis identification. This study aims to validate the Alberta code set (a hybrid code set using both administrative and EMR data) for identifying patients with cirrhosis. Methods: Twelve high-performing ICD-10 codes (Alberta code set) were evaluated using a cohort of 719 chart review–confirmed cirrhosis patients. Validation was performed in an independent cohort of 913 consecutively admitted patients at four Albertan hospitals (two tertiary and two non-urban). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were obtained with a 95% confidence interval. Two other code sets (Shearer and SoLiDaRity-10) were also validated using administrative-only, EMR-only, or hybrid data. Results: Using administrative data alone, the Alberta code set showed a sensitivity of 78.9%, specificity of 97.4%, PPV of 80.4%, and NPV of 97.2%. With hybrid data, sensitivity improved to 87.2% and NPV to 98.2%, while specificity (96.5%) and PPV (77.2%) remained similar. Urban sites showed higher sensitivity (86.2% and 91.7%) than non-urban sites (70% and 82.1%), likely attributed to coding practice variability. The Shearer and SoLiDaRity-10 code sets also demonstrated high sensitivity (86.2% and 83.5% respectively) and similar specificity (96.6 and 97.1 respectively) when using hybrid data rather than either admin data or EMR data alone. Conclusions: A hybrid administrative and EMR-based approach effectively identifies in-patient cirrhosis cases across health care settings and holds promise to support QI and research for cirrhosis patients.