Measuring the impact of reduced antibiotic use in hospital settings using electronic health records

Thesis

Hospital inpatient antibiotic use has increased year-on-year in the NHS despite considerable investments in stewardship activity. Although hospitals have financial incentives to reduce overuse, current antibiotic reduction targets are modest and hospital stewardship leads have expressed significant concerns about the safety of achieving antimicrobial reduction targets, due to fears of under-treatment leading to patient harm. To reassure practitioners and the public that substantial reductions in antibiotic use could be safe, this thesis begins by exploring associations between variation in hospital-level antibiotic prescribing and patient-level clinical outcomes in 36 million general medicine admissions in England. I use this analysis to estimate what system-wide reductions might, at least in theory, be safely achieved if higher-using hospitals could replicate the prescribing practices of low-using hospitals. I then use patient-level antibiotic and outcome data to investigate the causal effect of using shorter versus longer antibiotic courses by treating time-varying antibiotic use as sequence of non-randomised nested “trials”. Although observational analyses cannot prove shorter antibiotic courses are safe due to the potential for residual confounding, finding no evidence of harm can help to justify the evaluation of antibiotic reduction strategies in randomised experiments for unbiased inference. To inform its adoption in other acute care settings, the thesis then addresses two key questions emerging from the recently completed ARK-Hospital cluster-randomised trial, which tested a multifaceted behaviour change intervention to reduce antibiotic overuse in general medicine. Since the intervention effect varied widely by site, I examined the role of 26 contextual factors as potential mediators of impact, to better understand sources of this between-site variation. Finally, although the intervention had a modest impact on antibiotic prescribing trends overall, trends in non-ARK sites are unknown, and I therefore compared ward-level prescribing trends in ARK vs non-ARK sites to understand whether the intervention had a greater (or lesser) impact than suggested by the trial-site specific analysis alone. The thesis discusses the analytic challenges of using routine electronic databases to compare antibiotic treatment durations and evaluate antimicrobial stewardship initiatives, and highlights the implications of specific data quality and accessibility issues for investigators and policy makers. Overall, the thesis demonstrates the need for high-quality health-record-linked e-prescribing systems to support robust assessments of the impact of antibiotic reduction strategies.

Authors: Budgell, EP

• DOI: 10.5287/ora-g2mrmdxzj
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: antimicrobial stewardship; antibiotics; electronic health records; mortality; cluster-randomised controlled trial; secondary care
• Subjects: Epidemiology