Improving the Performance of Clinical Decision Support for Early Detection of Sepsis: A Retrospective Observational Cohort Study.

Sepsis remains a significant global health problem. It is a life-threatening, but poorly defined and recognized condition. Early recognition and intervention are essential to optimize patient outcomes. Automated clinical decision support systems (CDS) may be particularly beneficial for early detection of sepsis. The aim of this study was to use retrospective data to develop and evaluate seven revised versions of an electronic sepsis alert rule to assess their performance in detecting sepsis cases and patient deterioration (in-hospital mortality or ICU admission). Four revised options had higher sensitivity but lower specificity than the original rule. After discussion with clinical experts, two revised options with the highest sensitivity were selected. Further analysis on the number of alerts and time intervals between alerts and patient outcomes was conducted to decide the option to be implemented. This study has provided a data-driven approach to improve the CDS on early detection of sepsis.