Abstract
BACKGROUND: The herd effects of antimicrobial interventions used to prevent ICU-acquired infections are unknown. The objective here was to estimate these herd effects within a single three-tiered cluster randomized trial (CRT) emulated using ventilator-associated pneumonia (VAP) data from randomized concurrent control trials (RCCTs) abstracted within Cochrane reviews. METHODS: Control and intervention group data derived from 13 Cochrane reviews of 72 RCCTs of antibiotic (Tier 3) and antiseptic (Tier 2) decontamination versus 109 RCCTs of various non-decontamination (Tier 1, serving as benchmark) VAP prevention interventions were arranged as a three-tiered CRT. The direct and indirect (herd) effects of Tiers 2 and 3 each versus Tier 1 interventions were obtained using estimators derived in meta-regression models. RESULTS: Benchmark (Tier 1) VAP incidences derived for control and intervention groups from non-decontamination RCCTs were 23.3 (95% CI: 20.6-26.1; n = 111) and 19.2 (95% CI: 16.8-21.8; n = 112), respectively. The mean VAP incidences for antibiotic and antiseptic decontamination control groups were 5% to 15% higher than the control group benchmark. The direct effects of antibiotic and antiseptic interventions versus Tier 1 benchmarks (ORs) were 0.77 (95% CI: 0.55-1.09) and 0.97 (95% CI: 0.71-1.33) whereas the indirect effects were 2.17 (95% CI: 1.56-3.03) and 1.38 (95% CI: 1.0-1.91), respectively. CONCLUSIONS: Indirect (herd) effects from antimicrobial interventions, although inapparent within individual RCCTs, are strong. These effects on control group VAP incidences, which spuriously conflate the appearance of benefit, constitute herd peril.