Integrated Clinical Decision Support for Empiric Antibiotic Selection in Sepsis: A Protocol For A Cluster Randomized Cross-Over Trial (IDEAS-CRXO)

IntroductionAs antibiotic resistant organisms and infections continue to proliferate globally, it becomes increasingly difficult to select empiric antibiotic therapy, particularly in patients who stand to benefit from early adequate treatment. The inappropriate treatment of suspected infection, including sepsis, can be both too narrow and too broad. There is a need to optimize and tailor selection of antibiotic therapy to each patient, such that those at risk for infections due to antibiotic resistant organisms receive broader therapy, and those that are not at risk receive narrower antibiotic therapy. By helping clinicians select the right antibiotic for each individual patient we can potentially reduce unnecessarily broad antibiotic prescribing while preserving (and potentially improving) adequacy of treatment. Individualized clinical prediction models and decision support interventions are promising approaches that can meet these needs by better defining patient risk for antibiotic resistant or susceptible infections, but rigorous prospective evaluations are needed to confirm their utility. MethodsWe propose a two-period two-sequence non-blinded cross-sectional cluster randomized cross-over trial of an individualized antibiotic prescribing decision support intervention, administered by antibiotic stewardship pharmacists, for providers treating hospitalized patients with suspected infection including sepsis. The trial is taking place at 3 hospitals in Ontario, Canada, with clusters defined as medical, surgical, and critical care services. Eligible patients are adults with suspected infection characterized by initiation of a broad-spectrum antibiotic and recent collection of blood cultures. This decision support intervention will be based on a combination of proven decision heuristics (for Gram-positive organisms), modelled predicted susceptibilities (for Gram-negative organisms), and well-defined syndromes, that are individualized to the patient. Our primary outcome will be whether or not patients are de-escalated from their initial empiric regimen at 48 hours. The target sample size is 18 clusters with anticipated data from 1,440 patients, designed to achieve 80% power to detect an absolute increase in de-escalation at 48 hours of 7.5%. Secondary outcomes include adequacy of therapy, process measures, and safety. AnalysisThe primary and secondary outcomes will be analyzed at the patient-level using generalized linear mixed effects regression with fixed effects for treatment and period to account for the cross-over design, as well as fixed effects for the stratification factors and a priori patient risk factors to improve power and efficiency. Cluster and cluster-period random effects will be included to account for within-period and between-period intracluster correlation and the Kenward-Roger correction will be applied to minimize small sample bias. The primary effect estimate will be presented as a risk ratio with 95% confidence interval. Ethics and DisseminationThis trial has research ethics board approval at all participating sites and adheres to the Ottawa Statement ethics guidelines. The study is registered with clinicaltrials.gov, and the results will be published open access in a peer-reviewed journal. RegistrationClinicaltrials.gov, NCT06103500. STRENGTHS AND LIMITATIONSO_LIIDEAS-CRXO is a cluster randomized cross-over trial of a pharmacist administered decision support intervention for antibiotic prescribing in patients with suspected infection. C_LIO_LIThe intervention is based upon a generalizable algorithm incorporating Gram-positive heuristics, Gram-negative modelled susceptibility, and local well-defined syndrome treatment guidelines. C_LIO_LIThe primary outcome is a relevant measure for antibiotic stewardship - antibiotic de-escalation from the initial empiric index regimen at 48 hours from receipt of antibiotics. Secondary outcomes are focused on other clinical, process, and safety measures. C_LIO_LIThis rigorous cluster randomized cross-over trial will be conducted with 18 clusters (across three institutions) which will achieve 80% power in a two-period two-sequence design to detect a clinically important absolute difference of 7.5% in the primary outcome at a 5% two-sided significance level. C_LIO_LIThis study could be impacted by treatment contamination, however the cluster design will limit this potential source of bias. C_LI