The Use of Different Sepsis Risk Stratification Tools on the Wards and in Emergency Departments Uncovers Different Mortality Risks: Results of the Three Welsh National Multicenter Point-Prevalence Studies

OBJECTIVES:. To compare the performance of Sequential Organ Failure Assessment, systemic inflammatory response syndrome, Red Flag Sepsis, and National Institute of Clinical Excellence sepsis risk stratification tools in the identification of patients at greatest risk of mortality from sepsis in nonintensive care environments. DESIGN:. Secondary analysis of three annual 24-hour point-prevalence study periods. SETTING:. The general wards and emergency departments of 14 acute hospitals across Wales. Studies were conducted on the third Wednesday of October in 2017, 2018, and 2019. PATIENTS:. We screened all patients presenting to the emergency department and on the general wards. MEASUREMENTS AND MAIN RESULTS:. We recruited 1,271 patients, of which 724 (56.9%) had systemic inflammatory response syndrome greater than or equal to 2, 679 (53.4%) had Sequential Organ Failure Assessment greater than or equal to 2, and 977 (76.9%) had Red Flag Sepsis. When stratified according to National Institute of Clinical Excellence guidelines, 450 patients (35.4%) were in the “High risk” category in comparison with 665 (52.3%) in “Moderate to High risk” and 156 (12.3%) in “Low risk” category. In a planned sensitivity analysis, we found that none of the tools accurately predicted mortality at 90 days, and Sequential Organ Failure Assessment and National Institute of Clinical Excellence tools showed only moderate discriminatory power for mortality at 7 and 14 days. Furthermore, we could not find any significant correlation with any of the tools at any of the mortality time points. CONCLUSIONS:. Our data suggest that the sepsis risk stratification tools currently utilized in emergency departments and on the general wards do not predict mortality adequately. This is illustrated by the disparity in mortality risk of the populations captured by each instrument, as well as the weak concordance between them. We propose that future studies on the development of sepsis identification tools should focus on identifying predicator values of both the short- and long-term outcomes of sepsis.