Classification versus Prediction of Mortality Risk using the SIRS and qSOFA Scores in Patients with Infection Transported by Paramedics

Abstract

Objective: Identifying patients with sepsis in the prehospital setting is an important opportunity to increase timely care. When assessing clinical tools designed for paramedic sepsis identification, predicted risk may provide more useful information to support decision-making, compared to traditional estimates of classification accuracy (i.e., sensitivity and specificity). We sought to contrast classification accuracy versus predicted risk of a modified version of the Systemic Inflammatory Response Syndrome score (i.e., excluding white blood cell measure which is often unavailable to paramedics; mSIRS) and quick Sepsis Related Organ Failure Assessment (qSOFA) for determining mortality risk among patients with infection transported by paramedics. Methods: A one-year cohort of patients with infections transported to the Emergency Department (ED) by paramedics was linked to in-hospital administrative databases. Scores were calculated using the first reported vital sign measure for each patient. We calculated sensitivity and specificity of mSIRS and qSOFA for classifying hospital mortality at different score thresholds, and estimated discrimination (using the C-statistic) and calibration (using calibration curves). Regression models for predicting hospital mortality were constructed using the mSIRS or qSOFA scores for each patient as the predictor. Results: A total of 10,409 patients with infection who were transported by paramedics were successfully linked, with an overall mortality rate of 9.2%. The mSIRS score had higher sensitivity estimates than qSOFA for classifying hospital mortality at all thresholds (mSIRS ≥ 1: 0.83 vs. qSOFA≥ 1: 0.80, mSIRS ≥ 3: 0.11 vs. qSOFA ≥ 3: 0.08), but the qSOFA score had better discrimination (C-statistic qSOFA: 0.72 vs. mSIRS: 0.63) and calibration. The risk of hospital mortality predicted by the mSIRS score ranged from 8.0 to 19% across score values, whereas the risk predicted by the qSOFA score ranged from 10 to 51%. Conclusion: Assessing the predicted risk for the mSIRS and qSOFA scores instead of classification accuracy reveals that the qSOFA score provides more information to clinicians about a patient’s mortality risk, supporting its use in clinical decision-making.