Prehospital prediction of clinical course in patients with suspected sepsis: a prospective cohort study

BACKGROUND: Early prediction of the clinical course may aid clinical decisions for patients with suspected sepsis. We examined how variables observed by ambulance personnel predicted the clinical course in patients with suspected sepsis. METHODS: We prospectively collected prehospital data from patients with suspected sepsis in the ambulance service and supplemented the data with in-hospital findings. Prehospital data were used to calculate outputs of quick Sequential Organ Failure Assessment (qSOFA) and National Early Warning Score-2 (NEWS2). Outcomes were (1) sepsis severity after hospital admission, measured by Sequential Organ Failure Assessment (SOFA) score; (2) development of septic shock, defined by need for intravenous fluid resuscitation and pressors; and (3) 30-day all-cause mortality. Factors that predicted the three outcomes were identified with regression analysis. We performed receiver operating characteristic curve analysis to find the discriminatory abilities of prehospital scores. RESULTS: We included 398 patients with a median age of 76 years and a 30-day all-cause mortality of 10.1%. Oxygen saturation, systolic blood pressure and qSOFA were associated with sepsis severity and septic shock; Glasgow Coma Score and male sex with sepsis severity and mortality; age with sepsis severity; ashen or marbled skin and focus of infection with septic shock; cyanosis, nursing home resident and comorbidity with mortality. No variable predicted all three outcomes. Area under the curve (AUC) for septic shock was higher for qSOFA than for NEWS2 (p=0.033). AUC for septic shock was higher for males than for females if NEWS2 score was used (p=0.049), indicating that NEWS2 is better at predicting shock among males compared with females. CONCLUSIONS: While there may be a role for the use of prehospital scoring systems in patients who are critically ill on presentation, established clinical decision aids must be used with caution and predictive information is lost if the tools are applied alone.