Acute cystitis accounts for a significant proportion of ambulatory care visits every year in the United States. Empiric antibiotic selection varies widely among providers, even those working within the same health system. Current Infectious Diseases Society of America guidelines emphasize the use of local susceptibility data for determining initial antimicrobial therapy. Reference Gupta, Hooton and Naber1 Previous findings suggest that site-specific antibiograms (ie, urine, blood, etc) may result in improved empirical therapy. Reference Marsh, Mundy, Holter and Johnson2 We collated a local urinary antibiogram to promote recommended empiric antibiotic therapies for the treatment of cystitis. A clinical decision support system (CDSS) and order set were nested within the electronic medical record to encourage guideline-concordant prescribing. Using a quasi-experimental time series analysis, we assessed the impact of this intervention on β-lactam and fluoroquinolone prescribing rates for outpatient acute cystitis.
Methods
We performed a quasi-experimental, interrupted time series analysis to evaluate the effect of a local antibiogram-guided order set on antibiotic class selection for the treatment of outpatient urinary tract infections. Patients were treated in the Durham Veterans’ Affairs Health Care System (DVAHCS) from April 2016 through October 2019. Encounters were identified retrospectively using the following criteria: outpatient visits with appropriate International Classification of Disease, Ninth Revision (ICD-9) or ICD-10 documentation for acute cystitis, a collected urine culture, and an antibiotic prescription filled within 24 hours of visit through the DVAHCS pharmacy. Prescribing data were associated with the appropriate encounter. In total, 5,517 prescriptions were analyzed.
The study consisted of a preintervention phase (April 2016 through July 2018), an intervention period (August 2018), and a postintervention phase (September 2018 through October 2019). During the preintervention phase, a local urinary antibiogram, stratified by admission status, was created from all urine cultures across DVAHCS to inform empiric antibiotic choices. Escherichia coli was the most commonly isolated pathogen (n = 531 isolates, 41% of all outpatient urine cultures), with the following susceptibility profile: cefazolin 91%, ciprofloxacin 71%, and trimethoprim-sulfamethoxazole 77%. We then developed a CDSS highlighting nitrofurantoin and β-lactams (cephalexin, amoxicillin-clavulanate) as preferred agents over fluoroquinolones for the treatment of uncomplicated acute cystitis. In August 2018, the CDSS was integrated within the facility-wide computerized patient record system (CPRS).
The primary outcomes measured were changes in monthly proportions of antibiotic classes prescribed for cystitis. Classes were defined as fluoroquinolones (ciprofloxacin and levofloxacin), β-lactams (cephalexin, cefuroxime, amoxicillin, and amoxicillin-clavulanate), nitrofurantoin, and trimethoprim-sulfamethoxazole. Median monthly proportions and interquartile ranges were compared by unpaired Wilcoxon rank-sum test. Antibiotic concordance was compared in a subset of prescriptions with cultures using Fisher’s exact test. To control for unmeasured factors, we performed a segmented interrupted time series (ITS) analysis using a Poisson generalized linear regression model to assess changes in antibiotic prescribing associated with the intervention. All statistical analyses were performed using R version 3.6.3 software (R Foundation for Statistical Computing, Vienna, Austria).
Results
Patient age, race, and gender were similar before and after the intervention (Supplementary Table 1 online). Prior to implementation, monthly fluoroquinolone prescriptions accounted for a median of 45.0% of all antibiotics prescribed for outpatient treatment of acute cystitis (IQR, 41.0–49.0). After the intervention, fluoroquinolone prescriptions comprised a median of 32.0% (IQR, 28.9–35.1) of all antibiotics prescribed for cystitis monthly (P < .001). Conversely, prescriptions for β-lactams increased from a monthly median of 14.0% (IQR, 11.5%–16.5%) of all chosen antibiotics preintervention to a median of 24.5% of antibiotic prescriptions (IQR, 21.6%–27.4%; P < .001). Nitrofurantoin and trimethoprim-sulfamethoxazole prescribing were unaltered by implementation of the acute cystitis order set. In the ITS, CDSS implementation resulted in a −20.7% level change (95% CI, −33.8% to −7.5%; P = .002) and −1.4% change in slope (95% CI, −3.0% to 0.2%; P = .09) in fluoroquinolone prescribing for cystitis. The ITS demonstrated a 28.5% level change (95% CI, 15.5% to 41.7%; P < .001) and 1.2% change in slope (95% CI, −0.3% to 2.8%; P = .13) with regard to β-lactam prescriptions (Fig. 1). In a subset of patients treated in the emergency department, microbiological concordance was unchanged (79.1% vs 77.1%; P = .6514) (Supplementary Table 2 online).
Fig. 1. Monthly antibiotic selection as a proportion of all antibiotics prescribed for cystitis. The data points represent the observed proportions and the solid line represents the Poisson regression model.
Discussion
Prior to our intervention, fluoroquinolones accounted for the plurality of antibiotics prescribed for acute cystitis, despite an unfavorable susceptibility profile. CDSS implementation resulted in significant changes in prescribing patterns of fluoroquinolones and β-lactams for treatment of acute cystitis, resulting in greater concordance with the local urinary antibiogram. As nitrofurantoin and trimethoprim-sulfamethoxazole prescribing remained constant, we hypothesize that β-lactam therapies replaced fluoroquinolones for the treatment of uncomplicated cystitis. This intervention most likely succeeded by leveraging order entry, allowing providers to easily select guideline-recommended antibiotics, while simultaneously discouraging fluoroquinolone use.
Our results are consistent with those of prior studies demonstrating a positive effect of CDSS on antibiotic prescribing. Reference Hecker, Fox and Son3,Reference Patros, Sabol, Paniagua and Lans4 To our knowledge, however, our study is the first to assess the impact of this intervention across the entire ambulatory network of a healthcare system.
This study has several limitations: (1) It was conducted at a single institution with a predominantly male cohort, these findings lack generalizability to other settings. (2) The provider-level data regarding utilization of the CDSS were not available, preventing a direct linkage between CDSS usage and antibiotic choice. (3) The use of ICD codes may have over- or underestimated the true frequency of acute cystitis associated with antibiotic prescriptions; however, any discrepancy likely remained constant throughout the study period, and the observed trends are not likely related to this limitation.
In summary, CDSS combined with local urine antibiograms, even without prescriber education or audit and feedback, can be an effective tool for antimicrobial stewardship.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2020.357
Acknowledgments
We would like to recognize all the support from information technology, infection prevention, microbiology, nursing, and leadership at the Durham Veterans’ Affairs Medical Center while implementing this initiative.
Financial support
No financial support was provided relevant to this article.
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
