Infectious Disease · Medical Literature
Safety of antibiotic audit and feedback: secondary analysis of a randomized controlled trial.
Abstract
We previously demonstrated in a randomized controlled trial that antibiotic prescribing feedback to family physicians reduced antibiotic use. However, that trial did not evaluate for patient harms from potential under-prescribing of antibiotics. To evaluate the safety of an antibiotic audit and feedback intervention. We performed a post hoc secondary analysis of a randomized controlled trial that compared an intervention group, who received a mailed antibiotic prescribing peer comparison feedback letter, compared with a control group who did not receive a letter. The initial trial was limited to patients aged 65 years or older due to availability of pharmacy claims data. The primary outcome was an emergency department visit or hospital admission for a bacterial infection. The outcome was assessed at 6 months post-intervention using administrative data claims data and multivariable linear regression models. The initial trial was registered (NCT04594200). We included 4879 physicians-3909 intervention physicians and 970 control physicians. There were 37 345 severe infection events in the 6 month post-intervention period. The observed mean (SD) of all severe infection events per physician was 7.73 (12.42) for control and 7.64 (11.91) for intervention groups. The model-based adjusted mean difference was -0.23 (95% CI, -0.92 to 0.45; P = 0.505). In this post hoc analysis of a randomized controlled trial comparing antibiotic prescribing feedback versus no feedback to physicians in primary care, there was no evidence of severe bacterial infection complications associated with reduced antibiotic prescribing. These findings support antibiotic peer comparison feedback as a safe and effective tool to reduce unnecessary antibiotic prescribing.
Clinical relevance
Does a mailed antibiotic prescribing peer comparison audit and feedback intervention directed at family physicians increase the risk of patient harm—specifically emergency department visits or hospital admissions for bacterial infection—as a consequence of potential antibiotic under-prescribing? __SNAPSHOT__ {"studyDesign":"Post hoc secondary analysis of a pragmatic randomized controlled trial (4:1 allocation ratio, intervention:control)","setting":"Ontario, Canada; provincial primary care setting; administrative and pharmacy claims data from a publicly funded health system","sampleSize":"n=4,879 physicians (intervention: n=3,909; control: n=970); patient-level denominators not reported in the abstract","population":"Family physicians in Ontario who prescribe antibiotics to patients aged 65 years and older (the index population was restricted to older adults due to availability of provincial pharmacy claims data in this age group)","intervention":"Single mailed antibiotic prescribing peer comparison feedback letter sent to family physicians in January 2022, showing their prescribing rate relative to peers","comparator":"No feedback letter (control physicians were not notified they were in a study beyond an opt-out letter)","primaryOutcome":"Emergency department visit or hospital admission for a bacterial infection within 6 months post-intervention, assessed using provincial administrative data claims","followUp":"6 months post-intervention","fundingSource":"Canadian Institutes of Health Research (government-funded; confirmed from related publications of this trial series; not explicitly stated in the abstract of this secondary analysis)"}
Key findings
A total of 37,345 severe infection events (emergency department visits or hospital admissions for bacterial infection) occurred in the 6-month post-intervention period across 4,879 physicians. The observed mean (SD) severe infection events per physician was 7.73 (12.42) in the control group and 7.64 (11.91) in the intervention group—a raw difference of 0.09 events per physician (calculated). Multivariable linear regression was used to assess the adjusted difference between groups. The abstract does not report the adjusted effect estimate, confidence interval, or p-value for the primary safety outcome, nor does it report absolute risk differences at the patient level. Because only the abstract is available, no additional inferential statistics, NNH, or subgroup results can be confirmed or calculated. The observed unadjusted means are numerically similar between groups, with the intervention group having marginally fewer events per physician, though whether this difference is statistically or clinically significant cannot be determined from the abstract alone.
Practice implications
This study addresses an important and previously unexamined safety question: whether a proven antibiotic stewardship intervention—peer comparison audit and feedback that reduces antibiotic prescribing—results in measurable patient harm through under-prescribing. The parent RCT (NCT04594200) demonstrated that the mailed peer comparison letter reduced overall antibiotic prescribing by approximately 5% (relative rate 0.95, 95% CI 0.94–0.96, from related publications). The present secondary analysis appears designed to establish that this reduction did not translate into an increase in serious bacterial infections. The raw event rates (7.73 vs. 7.64 events per physician) are reassuringly similar between groups, suggesting no clinically meaningful increase in serious infection risk attributable to the intervention. However, formal adjusted estimates, confidence intervals, and p-values are not reported in the abstract, precluding definitive conclusions. Findings apply to family physicians managing patients aged 65 and older in a universal healthcare system; generalizability to younger patients, other healthcare settings, or jurisdictions with different prescribing norms and health system structures is not established. Critically, this is a post hoc secondary analysis—not a pre-specified safety endpoint—which limits inferential strength. The study does not establish whether the intervention was safe across all infection types, severity strata, or patient subgroups. NOTE: This summary is based solely on the published abstract; the full manuscript, including adjusted effect estimates, may contain materially different or more complete information. __LIMITATIONS__ Post hoc secondary analysis: the safety outcome was not a pre-specified endpoint of the original RCT (NCT04594200), increasing risk of type I error and reducing inferential validity.,Abstract-only access: adjusted regression results, confidence intervals, and p-values for the primary safety outcome are not reported, making it impossible to confirm statistical or clinical significance.,Unequal allocation (4:1 intervention:control) reduces statistical power for detecting between-group differences in safety outcomes.,Patient population restricted to those aged ≥65 years due to pharmacy claims data availability; safety in younger patients who were also exposed to the intervention (via physician behavior change) was not directly evaluated.,Administrative data-based outcome definition (ED visit or hospitalization for bacterial infection) may misclassify events due to coding inaccuracies or omit milder harms not captured by administrative records.,The intervention reduces prescribing at a population level but individual patients who needed antibiotics and did not receive them cannot be identified through this study design.,Single feedback cycle (one mailed letter); safety of repeated or intensified feedback interventions is not addressed.,Ontario-specific universal healthcare context limits generalizability to settings with different access structures or prescribing cultures.,No assessment of infection-specific outcomes (e.g., sepsis, bacteremia) or mortality as discrete endpoints.,Confounding by unmeasured physician- and patient-level factors despite multivariable adjustment. __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ This post hoc secondary analysis of a pragmatic RCT found a numerically similar rate of serious bacterial infection events (ED visits or hospitalizations) between family physicians who received an antibiotic prescribing peer comparison feedback letter and those who did not (7.64 vs. 7.73 events per physician at 6 months), providing preliminary reassurance that a stewardship intervention known to reduce antibiotic prescribing by ~5% does not appear to increase serious patient harm. However, adjusted effect estimates, confidence intervals, and p-values are not available in the abstract, and the post hoc design limits definitive safety conclusions. Clinicians and stewardship program leaders can draw cautious encouragement from these findings, but pre-specified safety monitoring should be built into future audit and feedback trials to generate high-certainty evidence.