Early diagnosis, antibiotics and supportive therapy are essential in sepsis. The diagnostic value of clinical and laboratory variables were evaluated in a prospective observational study.
A cohort of 404 adult patients admitted to the Department of Infectious Diseases from the emergency room (ER) for suspected severe infection was studied. A bacterial infection requiring antibiotic treatment was diagnosed in 306 patients (pneumonia 130 patients, urinary tract infection 80 patients, skin/soft tissue 43 patients, other bacterial infections 53 patients). Nonbacterial infections or noninfectious conditions were diagnosed in 82 patients. Significant bacteremia was detected in 68 patients (most common isolates: pneumococci 19, Escherichia coli 18, Staphylococcus aureus eight, β-haemolytic streptococci seven). Physiological variables recorded were temperature, heart rate, blood pressure, respiratory rate (RR), oxygen saturation, urine output, cerebral status. Laboratory variables were C-reactive protein (CRP), lactate, bicarbonate, creatinine, urea, hemoglobin (Hb), white blood cells (WBC), neutrophils, platelets, International Normalized Ratio, D-dimer, albumin, bilirubin, pro-calcitonin (PCT), IL-6 and LPS binding protein (LBP).
The value of each variable in identifying patients with bacteremic sepsis or bacterial infections requiring antibiotics was evaluated. In a univariate analysis, PCT, IL-6, LBP, CRP, bilirubin and maximum RR during the first 4 hours (RRmax 0–4 h) was significantly associated with bacteremia with P < 0.001 and CRP, PCT, IL-6, LBP, WBC, neutrophils, RRmax 0–4 h and Hb was associated with a bacterial infection with P < 0.001. In a multivariate logistic regression, PCT, RRmax 0–4 h, bilirubin and CRP each contributed significantly to the accurate prediction of bacteremia. To predict a bacterial infection, CRP, WBC, Hb and RRmax 0–4 h contributed significantly. If patients with pneumonia were excluded, the RR still contributed significantly to the prediction of bacteremia.
The studied patients have a high level of suspicion of a serious infection. The patients without bacterial infections often have other inflammatory processes, sometimes mimicking sepsis. This is indeed a challenging population for a diagnostic variable to prove its value but also one where it would be most needed. The results show that the RR is the best discriminatory physiological variable and that PCT is best fitted to predicting bacteremia whereas CRP best predicts bacterial infections.