This dataset is extracted from an ambulance quality registry of patients with suspected sepsis managed by the ambulance department of the University hospital of Northern Norway. Data was collected from patients with suspected sepsis who were given pre-hospital intravenous antibiotics and transported to hospital by the University of Northern Norways ambulance service from May 2018 to August 2022. The dataset was extracted to conduct a study on the ability of data available in the prehospital setting to predict the clinical trajectory for patients with suspected sepsis. The dataset contains demographic data and clinical data from both prehospital and in-hospital setting.
Background:
Prediction of the clinical trajectory for patients with suspected sepsis aid clinical decisions. Early intravenous antibiotic therapy is important for a subgroup of the patients and it was applied in the ambulance service we studied. Our aim was to compare how parameters from the patient history and clinical findings available in the prehospital setting can identify patients with the most severe prognosis. In addition, we studied how the National Early Warning Score 2 (NEWS2) and quick Sepsis-Related Organ Failure Assessment (qSOFA)-scores could predict outcomes.
Methods:
We prospectively collected prehospital data from patients with suspected sepsis in the ambulance service and supplemented the data retrospectively with in-hospital findings. The primary outcomes were 1) sepsis diagnosis after admission, 2) development of septic shock and 3) 30-day all-cause mortality. Admission SOFA score was used as a proxy for sepsis, and need for intravenous fluid resuscitation and pressor defined patients with septic shock. We used regression analysis to identify factors that predicted the three outcomes and performed receiver operating characteristic curve analysis of qSOFA and NEWS2’s ability to predict the same outcomes.
Results:
We included 398 patients with a median age of 76 years and a 30-day all-cause mortality of 10.1 %. Different prehospital variables were associated with the three different outcomes. In the unadjusted regression analysis NEWS2-score, oxygen saturation, Glasgow Coma Score, and sex were associated with all outcomes. With cut-off points for the scoring systems set by Youden’s J statistic, the difference in sensitivity and specificity between NEWS2 and qSOFA was not clinically significant. Some of the differences between qSOFA and NEWS2 were explained by the cut-off values and not by inherent properties in the scoring systems. Optimal cut-off values for NEWS2 differed between sexes for prediction of sepsis and septic shock, but this must be verified in future studies. In the final regression model, neither NEWS2 nor qSOFA contributed to prediction of all three outcomes.
Conclusions:
Our findings suggest that septic shock is better predicted from prehospital variables than sepsis, and that sepsis is more easily predicted than 30-day mortality. Predictive information is lost if scoring systems are applied alone.
