RESUMO
OBJECTIVE: To evaluate the prognostic accuracy of qSOFA for predicting in-hospital mortality among patients with suspected infection presenting to the ED of a public tertiary hospital in Brazil. METHODS: We performed a retrospective cohort study of consecutive adult patients (age ≥ 18 years) with suspected infection who presented to an academic tertiary ED in Porto Alegre (Southern Brazil) during an 18-month period. The qSOFA was calculated by using information collected at triage and patients were followed throughout hospitalization for the primary outcome of in-hospital mortality. Sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratios with corresponding 95% CIs were calculated for the qSOFA and qSOFA65. RESULTS: A total of 7523 ED visits of patients with suspected infection in which an intravenous antibiotic was administered within 24 h were included, which resulted in 908 in-hospital deaths (12.1%). There were 690 (9.2%) patients whose triage qSOFA was ≥2 points. When such cutoff was used, the sensitivity for in-hospital death was 24.6% (95% CI 21.8 to 27.4%) and the specificity was 92.9% (95% CI 92.3% to 93.5%). The sensitivity increased to 67.4% (95% CI 64.2% to 70.3%) when a cutoff of ≥1 was tested, but the specificity decreased to 55.3% (95% CI 54.1% to 56.5%). Using a cutoff of ≥2, the qSOFA65 had a sensitivity of 51.0% (95% CI 47.7% to 54.3%) and a specificity of 75.7% (95% CI 74.6% to 76.7%). CONCLUSIONS: The qSOFA score yielded very low sensitivity in predicting in-hospital mortality. Emergency physicians or ED triage nurses in low-to-middle income countries should not be using qSOFA or qSOFA65 as "rule-out" screening tools in the initial evaluation of patients with suspected infection.