(228) Development of a risk categorization framework to quantify antimicrobial resistance risk in female patients with uncomplicated urinary tract infection

Publication Number: 888

Background: Patients with uncomplicated urinary tract infections (uUTIs) typically receive empiric antibiotic (Abx) treatment, which may not be effective if the uropathogen is resistant due to growing antimicrobial resistance (AMR). Data-driven approaches inform empiric prescribing by evaluating patient-level AMR risk for Abx prescribed for uUTIs.

Objectives: To construct risk categories for resistance to four common Abx classes recommended by prescribing guidelines for uUTI (nitrofurantoin [NFT], trimethoprim/sulfamethoxazole [SXT], beta-lactams [BLs], and fluoroquinolones [FQs]) based on probabilities of resistance derived from predictive models.

Methods: Electronic health record data from Oct 2015–Feb 2020 was used to identify female patients aged ≥12 years, with a uUTI diagnosis confirmed by positive Escherichia coli urine culture, and treatment with ≥1 NFT, SXT, FQ, or BL. Separate predictive models were developed and validated to quantify the probability of resistance, defined as not susceptible (NS; resistant and intermediate), to each Abx class. Three risk categories were developed based on the predicted probability of being NS: patients were “low risk” if their predicted probability of being NS to each Abx class was at or below a predicted probability equivalent to a 20% false negative rate, based on clinical input; “high risk” patients had a predicted probability of being NS to each Abx class higher than the United States national prevalence; “moderate risk” patients had a predicted probability of being NS between the “low risk” and “high risk” categories. Demographic and clinical characteristics were summarized for patients in each risk category for each Abx class.

Results: Among 87,478 eligible patients, the proportion of truly NS patients was 3–10-folds higher among patients categorized as “high risk” versus “low risk” for all Abx classes (NFT: 11.7% vs. 1.8%; FQ: 64.5% vs. 6.5%; SXT: 53.9% vs. 15.3%; BL: 30.5% vs. 9.1%, respectively); the proportion of patients categorized as “high risk” was 3–12-folds higher among truly NS patients versus truly susceptible patients (NFT: 28.9% vs. 7.4%; FQ: 31.1% vs. 2.6%; SXT: 32.5% vs. 8.7%; BL: 36.8% vs. 14.1%, respectively). “High risk” patients versus “moderate” or “low risk” patients across all Abx classes were generally more likely to have a recurrent UTI, prior Abx use, prior AMR, older age (except for patients NS to SXT), and non-White race (except for patients NS to FQs).

Conclusions: Our AMR risk categorization framework provides a useful approach to contextualize the probability of AMR to four common Abx classes used to empirically treat uUTIs. This framework could be used to build a tool to help physicians more effectively manage appropriate empiric Abx treatment options for patients with uUTI.