Evaluating effects of antibiotic treatment on resistance and patient outcomes using linked microbiology and patient electronic health record data

Thesis

There is substantial pressure on medical services to reduce antibiotic prescribing, aimed at reducing the spread of antimicrobial resistance. Reducing prescribing in the acute hospital setting provides unique challenges, with antibiotic decisions made in the first few hours of admission with uncertainty over both diagnosis and severity of illness. The potential benefit to the patient must be balanced with the risks of side-effects and of future resistant infections to both the patient and the population as a whole.

I first review the evidence currently available to the acute clinician to assist in this decision. I then conduct a small study of antibiotic use in the acute medical setting, demonstrating that whilst substantially reducing antibiotic initiation and prescribing is possible, this is at the cost of more admissions and longer length-of-stay. I highlight the need for larger studies to meaningfully assess clinical outcomes, including resistant infections.

I then perform two studies investigating the accuracy of electronic health record data. The first assesses to what degree diagnostic codes can be used to estimate Charlson comorbidities to enable risk-adjusted studies of clinical outcome, concluding that diagnostic coding is relatively robust. The second assesses the degree to which diagnostic codes can be used to identify infections and their bacterial cause, finding that careful data curation is required, and significant limitations exist in using coded data to study causative organisms.

I use results from these studies to inform the subsequent design of a study of patients with Escherichia coli and Klebsiella pneumoniae bloodstream infections using linked electronic health record data and laboratory microbiology results. I find that previous broad-spectrum beta-lactams and trimethoprim exposure are associated with beta-lactam resistance in these infections at the level of an individual patient, but no evidence of association between duration of antibiotic exposure and development of resistance.

This work supports current efforts to limit use of broad-spectrum antibiotics, particularly beta-lactams, and to encourage the clinician, if an antibiotic is required, to try and use narrow spectrum antibiotics. It also showcases the power of electronic health record data, if used carefully, to progress the current knowledge in identifying rational prescribing strategies to reduce development of antimicrobial resistance.

Authors: Fawcett, N

• DOI: 10.5287/ora-nz5aovbgj
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Subjects: prescribing; clinical medicine; general medicine; antibiotic; antibiotic resistance; electronic health record data; diagnostic coding