Characteristics and predictors of in-hospital death or transfer to intensive care unit in patients with suspected bacterial infection and without fulfulling the sepsis criteria on admission

Aim
To describe baseline characteristics and to examine predictors of serious outcomes among emergency department (ED) patients with suspected bacterial infections without meeting the sepsis criteria on admission to the ED.
Methods
A prospective observational study of patients with suspected bacterial infection admitted to the emergency department during 1.10.2017–31.03.2018. A National Early Warning Score (NEWS2) ≥5 within the first four hours in the ED was assumed to be compatible with sepsis with a high risk for serious outcomes defined as the composite endpoint (primary outcome) of in-hospital death or transfer to the ICU. Patients achieving the primary outcome were grouped according to fulfillment of the NEWS2 ≥5 criteria and compared to baseline characteristics. We used logistic regression analysis to estimate unadjusted and adjusted odds for the composite endpoint per unit increase of different covariates among patients with either a NEWS2 score < 5 (NEWS2÷) or a NEWS2 score ≥5 (NEWS2+).
Results
A total of 2,055 with a median age of 73 years were included. A total of 198(9.6%) achieved the composite endpoint, 59(29.8%) NEWS2÷ patients and 139(70.2%) NEWS2+ patients, respectively. Compared to NEWS2+ patients, we found that NEWS2÷ patients were older, more likely to be female, and suffer from diabetes. The proportion with Sequential Organ Failure Assessment Score (SOFA) ≥2 and new-onset atrial fibrillation on admission was lower in the NEWS2÷ group and they also had lower values of leucocytes, bilirubin, lactate, and glucose. Urinary tract infections were more commonly the source of infection among NEWS2÷ patients. The median time to antibiotic treatment was lower in the NEWS2÷ group (4.2 vs 6.7 hours). The Do-not-attempt-cardiopulmonary- resuscitation order (DNACPR) was registered on admission among 16.9% and 27% of NEWS2÷ and NEWS2+ patients, respectively.
The final regression model for NEWS2÷ patients showed that diabetes (OR 2.23;1.23-4.0%), a SOFA score of ≥2 (OR 2.57;1.37-4.79), and a DNACPR order on admission (OR 3.70;1.75-7.79) were predictive variables for the composite endpoint (goodness-of-fit test p-value 0.291; AUROC for the model 0.72). A final regression model for NEWS2+ patients revealed that a SOFA score ≥2 (OR 2.93;1.68-5.13), hypothermia (OR 2.48;1.30-4.75), and a DNACPR order on admission were predictive variables for the composite endpoint (goodness-of-fit test p value 0.62, AUROC for the model 0.70).
Conclusion
Almost one-third of patients with suspected bacterial infections and serious outcomes during hospitalization did not meet the NEWS2 sepsis criteria. Our study has identified some factors with independent predictive values for the development of serious outcomes which should be tested in future prediction models.