Prediction of bloodstream infection using triage variables in the emergency department: retrospective derivation and validation cohort.

BACKGROUND

Blood cultures are frequently performed in febrile emergency department (ED) patients despite low yield and risks of contamination. Existing prediction tools are not usable at triage. The objective was to derive and internally validate a score for predicting bloodstream infection using variables assessable immediately after ED triage.

METHODS

This observational retrospective cohort study was conducted in the two EDs of a University Hospital between 1 January and 31 December 2021. Adult patients for whom a blood culture was collected during ED visit were included. The primary endpoint was bloodstream infection. A multivariate analysis using backward stepwise logistic regression was performed to identify risk factors for bloodstream infection and derive the aGe, cOmplaint, Temperature and Hemodynamic for bloodstream Infection Classification (GOTHIC) score. Diagnostic accuracy of the score was evaluated in the derivation and validation cohorts.

MAIN

RESULTS

6740 visits were analysed, including 4493 patients in the derivation cohort and 2247 patients in the validation cohort. The prevalence of bloodstream infection was 11% (n=512) and 10% (n=234) in each cohort, respectively. The GOTHIC score included seven factors associated with bloodstream infection: age ≥75 years (OR=2.90; +1 point), tachycardia >90 beats per minute (OR=1.51; +1 point), systolic blood pressure <100 mm Hg (OR=1.90; +1 point), diastolic blood pressure <60 mmvHg (OR=1.41; +1 point), fever >38°C (OR=2.14; +1 point), isolated fever as chief complaint at triage (OR=1.62; +1 point) and protective chief complaint (dyspnoea, COVID-19 symptoms or vaso-occlusive crisis) (OR=0.47; -1 point).

CONCLUSION

We propose a simple risk score for bloodstream infection that can be easily calculated using seven variables assessable at triage.