A Sepsis Treatment Algorithm to Improve Early Antibiotic De-escalation While Maintaining Adequacy of Coverage (Early-IDEAS): A Prospective Observational Study

ABSTRACT Background Empiric antibiotic treatment selection should provide adequate coverage for potential pathogens while minimizing unnecessary broad-spectrum antibiotic use. We sought to pilot a rule- and model-based early sepsis treatment algorithm (Early-IDEAS) to make optimal individualized antibiotic recommendations. Methods The Early-IDEAS decision support algorithm was derived from previous Gram-negative and Gram-positive prediction rules and models. The Gram-negative prediction consists of multiple parametric regression models which predict the likelihood of susceptibility for each commonly used antibiotic for Gram-negative pathogens, based on epidemiologic predictors and prior culture results and recommends the narrowest spectrum agent that exceeds a predefined threshold of adequate coverage. The Gram-positive rules direct the addition or cessation of vancomycin based on prior culture results. We applied the algorithm to prospectively identified consecutive adults within 24-hours of suspected sepsis. The primary outcome was the proportion of patients for whom the algorithm recommended de-escalation of the primary antibiotic regimen. Secondary outcomes included: (1) the proportion of patients for whom escalation was recommended; (2) the number of recommended de-escalation steps along a pre-specified antibiotic cascade; and (3) the adequacy of therapy in the subset of patients with culture-confirmed infection. Results We screened 578 patients, of whom 107 eligible patients with sepsis were included. The Early-IDEAS treatment recommendation was informed by Gram-negative models in 76 (71%) of patients, Gram-positive rules in 66 (61.7%), and local guidelines in 27 (25%). Antibiotic de-escalation was recommended by the algorithm in almost half of all patients (n=50, 47%), no treatment change was recommended in 48 patients (45%), and escalation was recommended in 9 patients (8%). Amongst the patients where de-escalation was recommended, the median number of steps down the a priori antibiotic treatment cascade was 2. Among the 17 patients with relevant culture-positive blood stream infection, the clinician prescribed regimen provided adequate coverage in 14 (82%) and the algorithm recommendation would have provided adequate coverage in 13 (76%), p=1. Among the 25 patients with positive relevant (non-blood) cultures, the clinician prescribed regimen provided adequate coverage in 22 (88%) and the algorithm recommendation would have provided adequate coverage in 21 (84%), p=1. Conclusions An individualized decision support algorithm in early sepsis could lead to substantial antibiotic de-escalation without compromising adequate antibiotic coverage.