Improved sepsis surveillance using a fully automated electronic health record-based algorithm compared to diagnostic coding.

BACKGROUND: Accurate diagnostic coding of sepsis is essential for surveillance, resource allocation, and health policy planning. Studies assessing the usability of claims-based data (ICD-10 codes) compared to clinical criteria for sepsis surveillance are needed. OBJECTIVES: To assess the concordance between ICD-10 diagnostic coding of sepsis and classification using a previously validated fully automated electronic health record (EHR)-based algorithm applying Sepsis-3 criteria. METHODS: We conducted an observational study, including adult in-hospital admissions during 2016-2024 at Karolinska University Hospital and 2022-2024 at three hospitals in Region Västerbotten, Sweden. The Sepsis-3 algorithm identified suspected infection combined with an increase in Sequential Organ Failure Assessment (SOFA) score ≥2 points. Sepsis ICD-10 codes were categorised as explicit or specific (R65.1, R57.2). Concordance was evaluated using descriptive and kappa statistics, and time trends were analysed. RESULTS: Among 174,343 admissions during the common study period (2022-2024), the Sepsis-3 algorithm classified sepsis in 11.4% of admissions at Karolinska and 6.2% in Västerbotten, compared to 2.5% and 1.2% with explicit codes and 1.4% and 0.6% with specific codes, respectively. Concordance of diagnostic codes with the Sepsis-3 algorithm was low (kappa 0.21 and 0.14 for explicit and specific codes at Karolinska; 0.15 and 0.09 in Västerbotten). During the study period 2016-24 at Karolinska, introduction of an automated SOFA calculator in the EHR system was associated with increased usage of specific sepsis codes but remained far below algorithm-based classification. CONCLUSION: Automated EHR-based algorithms allow for data-driven sepsis surveillance and may support more reliable diagnostic coding practices.