Design and evaluation of an automated real-time SOFA score calculator in an electronic health record system, for early sepsis diagnosis

BackgroundIn patients with infections, the Sequential Organ Failure Assessment (SOFA) score should be used to diagnose sepsis. However, manual SOFA calculation is time consuming. Thus, we developed an automated real-time SOFA score application that has been fully implemented into the major electronic health record (EHR) system (TakeCare) in Stockholm. We aimed to describe the method and evaluate its performance for detection of sepsis. MethodWe developed an automated SOFA application that presents a total score (SOFATotal) over time and a baseline score (SOFAChronic) based on outpatient data prior to hospital admission. We evaluated its performance on 583 sepsis alert patients in an emergency department, including 472 with sepsis according to manually calculated SOFA (SOFAManual). SOFATotal at 30 minutes and 3 hours, and quick-SOFA (qSOFA) were compared with SOFAManual for detection of sepsis. The acute SOFA score (SOFAAcute) was calculated by subtracting SOFAChronic from SOFATotal. ResultsSOFATotal at 3 hours showed moderate-good agreement with SOFAManual (R{superscript 2} = 0.76). Areas under receiver operating characteristic curves for sepsis detection were 0.83 at 30 minutes, 0.94 at 3 hours for SOFATotal, and 0.67 for qSOFA. Among patients with SOFAChronic [≥]1 (21% of cases), SOFATotal of [≥]2 was observed in 90% of cases (mortality 13.5%), although SOFAAcute of [≥]2 was noted in 63% (mortality 16.7%). ConclusionThe automated SOFA score demonstrated effectiveness in early sepsis detection, outperforming qSOFA, but adjustments for chronic baseline scores were necessary to avoid over-diagnosis. Further studies on broader patient populations would be useful to validate its clinical use.