Empirical antimicrobial prescribing for pyelonephritis in patients discharged from 15 US Emergency Departments: an opportunity for improvement

Abstract

Introduction

Urinary tract infections (UTIs) are one of the most commonly treated infections in the emergency department (ED), resulting in over 3 million ED visits and 400 000 hospital admissions annually in the USA.^1,2 UTIs account for over 15% of outpatient antibiotic prescriptions.² The IDSA/ESCMID guidelines for acute uncomplicated cystitis and pyelonephritis in women recommend specific antimicrobial therapy based on local resistance rates.¹ For the treatment of outpatient uncomplicated pyelonephritis, empirical treatment with a fluoroquinolone (e.g. ciprofloxacin) is recommended.¹ If the prevalence of Escherichia coli resistance to fluoroquinolones exceeds 10%, a one-time dose of a long-acting parenteral agent (e.g. ceftriaxone or an aminoglycoside) is also recommended as adjunctive therapy. It is unclear if emergency medicine providers use this approach before discharging patients. Additionally, it is unknown if providing a one-time dose of a long-acting parenteral agent increases the likelihood of providing initial active antimicrobial treatment.

Our previous study found significant rates of E. coli resistance among a nationally representative cohort of patients presenting to the ED with a UTI.³ Overall, 22.1% of E. coli isolates were resistant to fluoroquinolones. Among patients with pyelonephritis, 25% of E. coli isolates were resistant to fluoroquinolones.³ This finding has implications for empirical therapy and may influence the likelihood of initial active treatment and time to cure. The purpose of this study is to characterize the empirical treatment of pyelonephritis in a nationally representative sample of ED patients and to identify patient- and treatment-specific factors associated with receiving initial inactive antimicrobial treatment.

Methods

This was a multicentre, retrospective observational cohort study at 15 hospital EDs across the USA. Sites were participants in the Emergency Medicine PHARMacotherapy Research NETwork (EMPHARM-NET). International Classification of Diseases (ICD-10) diagnosis codes N30.00, N30.01 and N39.0, corresponding to a primary diagnosis of uncomplicated or complicated pyelonephritis, were used to identify patients between 1 January 2018 and 31 December 2020.

Patients ≥18 years of age were included if they had a primary or secondary diagnosis of UTI in the ED. It is standard practice in the ED for urine cultures to be obtained prior to antimicrobial administration. The following exclusion criteria were applied: pregnancy; suspected or confirmed acute bacterial prostatitis; lack of urine culture or negative urine culture; orchitis, epididymitis or chronic bacterial prostatitis as determined by history and/or physical examination; gross haematuria requiring intervention other than administration of antibiotics for UTI; or removal or exchange of a urinary catheter or urinary tract surgery within 7 days prior to ED presentation. For this analysis, patients with isolated cystitis or admitted to the hospital were also excluded.

Data collection

All data variables were defined a priori and were available for abstraction from the electronic medical record (EMR). Data collected included patient demographics, signs and symptoms of UTI (e.g. fever, dysuria, flank pain, frequency/urgency), ED disposition and urine microbiological results and susceptibilities. Data were abstracted at each site by a trained data abstractor and each site’s principal investigator did a random sample audit to ensure quality of data collection and documentation. All sites obtained institutional review board approval. The study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.⁴ We used the REDCap^© electronic-data capture tool (Version 6.18.1, 2019; Vanderbilt University, Nashville, TN, USA) hosted by the University of Iowa to manage all study data.

Definitions

Pyelonephritis was defined as the presence of the following systemic symptoms in addition to urinary symptoms: flank pain; chills; fever (temperature >38°C); costovertebral-angle tenderness; and nausea or vomiting.^1,5 Patients were further classified as having complicated pyelonephritis if the following factors were present: male sex; pre-existing genitourinary tract anatomical abnormalities; or current immunocompromised status. Pre-existing urinary tract anatomical abnormalities included a history of kidney stones, urinary obstruction, neurogenic bladder and renal insufficiency.⁶ Immunocompromised status included active cancer, chronic systemic corticosteroid use, current use of immunosuppressants, and HIV disease. Patients who did not meet criteria for complicated pyelonephritis were classified as uncomplicated infections.

A positive urine culture was defined as a specimen with ≥10⁴ cfu/mL bacteria isolated with no more than two organisms isolated in the urine culture.^3,7–11 If a urine culture grew >2 organisms, it was considered to be contaminated and the patient was excluded.^3,7,11 Additionally, the following pathogens were considered to be contaminants: Lactobacillus; non-saprophyticus CoNS; Corynebacterium species; or β-haemolytic streptococci.⁸

Endpoints

The primary endpoint was the proportion of patients who received initial inactive empirical treatment based on urine culture susceptibilities.⁵ Additionally, we aimed to determine predictive factors of inactive therapy, based on resistance to both IV and/or oral antimicrobials received in the ED or upon discharge from the ED. Factors associated with initial inactive treatment were assessed in patients with pyelonephritis discharged from the ED; these patient- and treatment-specific factors included: exposure to IV or oral antibiotics 90 days prior to the ED visit; haemodialysis dependence; urinary tract abnormality (e.g. long-term urinary catheter, nephrolithiasis, renal transplant, neurogenic bladder, nephrostomy tubes); residence in a long-term care facility; age; gender; infection type (cystitis versus pyelonephritis); history of MDR pathogen (blood or urine culture site) in the prior 12 months; and receipt of a long-acting agent (ceftriaxone, aminoglycoside or ertapenem) in the ED for the index UTI episode.^8–10 MDR pathogens included the following: ESBL-producing Enterobacterales; carbapenem-resistant Enterobacterales (CRE); MRSA; VRE; Stenotrophomonas maltophilia; Pseudomonas aeruginosa; Acinetobacter spp.; AmpC-β-lactamase-producing bacteria; or resistance to antibiotics from three different classes.

Statistical analysis

Descriptive characteristics were used to characterize patient demographics and antimicrobial therapy. Nominal data were compared by using the chi-squared or Fisher exact test as appropriate. Continuous data were compared by using Student’s t-test or the Mann−Whitney U-test. To identify independent predictors of initial inactive treatment, a forward stepwise logistic regression was conducted. Variables were selected a priori based on previous literature^7,12 All analyses were conducted using SAS v9.4 (Cary, NC, USA). Inferential tests were two-tailed, and P < 0.05 was considered statistically significant.

Results

We evaluated 3714 patients for inclusion, of which 1257 were excluded due to admission to the hospital, leaving 2457 patients with a diagnosis of UTI discharged from the ED. Of these, 391 patients had a diagnosis of pyelonephritis, of which 223 (57%) had a positive urine culture that grew no more than two uropathogens at ≥10⁴ cfu/mL and were included. The urine culture was negative (i.e. no growth or contamination) in 168 patients (Table 1). Median patient age was 50.1 years (IQR 35.9–66.8) and they were mostly female (78.3%; Table 1). The most common comorbid conditions were hypertension (35.5%) and diabetes (20.7%). UTI was the chief complaint for ED visit in most patients (70.8%). The most common presenting symptoms were flank pain (61.9%), dysuria (43.5%) and frequent/urgent urinary symptoms (30.7%). The most frequently isolated uropathogens were E. coli (70.4%), followed by Klebsiella pneumoniae (9.9%), Enterococcus spp. (5.4%) and Proteus spp. (4.3%).

Table 2 describes the antibiotics administered in the ED and prescribed at discharge in patients with pyelonephritis and a positive urine culture. Overall, 40.4% of patients received an IV antibiotic in the ED. The most common IV antibiotic administered was ceftriaxone (prescribed to 35% of patients with pyelonephritis, comprising 86.7% of IV antibiotics), followed by ciprofloxacin (prescribed to 3.1% of patients; 7.8% of IV antibiotics). Most patients (97.8%) received an oral antibiotic upon discharge from the ED. The most frequently prescribed oral antibiotics at discharge were as follows: cefalexin (31.8%); ciprofloxacin (14.3%); cefdinir (13.5%) and sulfamethoxazole/trimethoprim (12.9%).

Discharge antibiotics and patterns of resistance in all positive culture isolates are displayed in Table 3. Of patients prescribed oral antibiotics upon discharge from the ED, 13.2% of isolates were resistant to the antibiotic prescribed. Of those patients with resistant isolates, 41.4% received an IV antibiotic prior to discharge from the ED and 58.6% did not receive an IV antibiotic. The overall rate of the receipt of inactive treatment across all isolates was 10.6%. Inactive therapy was highest in patients discharged from the ED on cefpodoxime (40%), cefuroxime (23.1%), amoxicillin/clavulanate (25%), ciprofloxacin (15.6%) and sulfamethoxazole/trimethoprim (14.3%). Only three patients who were discharged after receipt of IV and oral antibiotics grew isolates resistant to both antibiotics. Each of these patients received IV ceftriaxone followed by cefdinir, cefpodoxime or ciprofloxacin and grew ESBL E. coli.

The cumulative susceptibility profile for all uropathogens isolated are listed in Table S1 (available as Supplementary data at JAC Online). Any antimicrobial resistance was observed in 54.9% of isolates. The antimicrobials with the highest rates of resistance overall were ampicillin (34.1%), ampicillin/sulbactam (29.1%), sulfamethoxazole/trimethoprim (22%) and cefazolin (17%). The rate of fluoroquinolone resistance was 13.9%. Of the guideline-recommended IV antimicrobials for empirical treatment of pyelonephritis, aminoglycosides had the lowest resistance rates (amikacin 0%, gentamicin 4.9% and tobramycin 4.5%). Also of note, resistance to ceftriaxone was 6.3%.

Table 4 presents the results of the univariate and bivariate analysis for predictors of initial inactive therapy. In the unadjusted analysis, receipt of a long-acting IV antibiotic significantly reduced the odds of inactive therapy (unadjusted OR 0.24, 95% CI 0.07–0.83). After adjustment in a multivariable analysis, long-acting IV antibiotic was the only factor predictive of inactive therapy (OR 0.23, 95% CI 0.07–0.83).

Discussion

We found that receipt of long-acting IV antibiotics decreased the odds of receiving inactive antibiotic therapy. Inactive treatment occurred in 10.3% of patients overall and was the highest in patients who received cefpodoxime, cefuroxime, amoxicillin/clavulanate, ciprofloxacin and sulfamethoxazole/trimethoprim. Only three patients (4.3%) who received both long-acting IV antibiotics and oral antibiotics grew pathogens resistant to both antibiotics. This finding both reinforces the IDSA/ESCMID guideline recommendations and underscores the importance of IV antibiotic administration in the ED upon suspicion of pyelonephritis.¹

We found that only 40.4% of patients discharged from the ED with pyelonephritis received IV antibiotics prior to discharge, the most common of which was IV ceftriaxone, which comprised 86% of IV antibiotics. Relative to most oral options, ceftriaxone is more likely to be active against the most common pathogens that cause pyelonephritis.¹³

The IDSA/ESCMID guidelines for acute uncomplicated cystitis and pyelonephritis in women recommend specific antimicrobial therapy based on local resistance rates.¹ Presently no treatment guidelines direct therapy in complicated pyelonephritis. The IDSA/ESCMID guidelines recommend oral ciprofloxacin for pyelonephritis where the prevalence of resistance of community uropathogens to fluoroquinolones is not >10%. Where prevalence of fluoroquinolone resistance exceeds 10%, an initial one-time IV dose of a long-acting parenteral antimicrobial (ceftriaxone or an aminoglycoside) is recommended.¹ Sulfamethoxazole/trimethoprim following an IV dose of long-acting antibiotic is an alternative. Our fluoroquinolone resistance rate was >10% among the 15 centres, suggesting patients should have received an IV dose of long-acting antibiotics.³ Overall, less than half of patients with pyelonephritis received guideline-concordant therapy, with only 19% of patients receiving a fluoroquinolone and 13% of patients receiving sulphamethoxazole/trimethoprim.¹

The IDSA/ESCMID guidelines also state that β-lactam antibiotics are less effective for pyelonephritis and should not be used.¹ Interestingly, the most common class of oral antibiotic was β-lactams (45%), with cefalexin being the most frequently prescribed. Furthermore, 2% of patients with complicated pyelonephritis and 15.5% of uncomplicated pyelonephritis were treated with nitrofurantoin, which does not provide adequate penetration to the bladder and upper genitourinary tract and thus should not be used for this indication.¹⁴

While antimicrobial stewardship programmes have historically focused on inpatient treatment, our finding of significant guideline-discordant treatment for UTIs highlights the important role that stewardship should play in patients being discharged from the ED.¹⁵ Common ED antimicrobial stewardship interventions include culture follow-up programmes (which typically target antimicrobial escalation or change based on microbiological data and rarely include de-escalation), formulary restriction, collaboration with emergency medicine pharmacists and ED-specific antibiograms.^16,17 A recent quasi-experiment found the use of urinary-specific ED antibiograms and associated guidelines was associated with improved rates of guideline-concordant therapy, although the impact of this intervention on the rate of initial active therapy was not assessed.¹⁸ These findings are encouraging and we advocate for similar approaches to be implemented in EDs across the USA, given our study demonstrates substantial need for improvement in UTI prescribing practices and the initiation of active therapy.

Limitations

Although our study leveraged a large patient database that should be generalizable across the USA, there are several limitations to mention. First, the retrospective and observational design may lead to inaccurate documentation. We intentionally selected variables for analysis that are likely to be accurately reported and are easy to abstract; however, there may have been slight variation between centres in recording of data. Furthermore, misclassification bias may be a risk as we were not able to characterize antimicrobial appropriateness in all patients, given we weren’t assessing patient-specific factors like medication allergies, renal function, pregnancy and other factors. Also, while we have 15 sites included in our study representing various geographical areas across the USA, our regions in the south and Pacific coast are under-represented in our network, thus these findings may be difficult to generalize to those geographical areas. Finally, we were unable to perform subgroup analysis in patients with pyelonephritis due to a limited sample.

Conclusions

In our prospective, multicentre observational study of patients presenting to the ED with pyelonephritis, administration of long-acting IV antibiotics was associated with a decrease in the odds of receiving inactive empirical treatment upon discharge. IV antibiotics were underutilized, given to only 40.4% of patients. This, in addition to the high rates of guideline-discordant treatment, highlights the need for focused improvements on antimicrobial prescribing for ED patients diagnosed with UTI.

Acknowledgements

We thank: Keith Burrell, University of Iowa; Devin Spolsdoff, MS, University of Iowa; Joann Huynh, PharmD Candidate (2022), University of Illinois at Chicago College of Pharmacy; Kevin Johns, PharmD Candidate (2022), University of Illinois at Chicago College of Pharmacy; Danielle Garza, PharmD; Monica Frauhiger, PharmD; Lyudmila Garbovsky, PharmD, BCPS, BCCCP, Hospital of the University of Pennsylvania, Philadelphia, PA; David Gajdosik, PharmD, BCPS, Hospital of the University of Pennsylvania, Philadelphia, PA; Rachel Lam, Stritch School of Medicine, Loyola University Chicago; and Mollie Shutter, Stritch School of Medicine, Loyola University Chicago.

Funding

Spero Therapeutics provided an unrestricted research grant via its investigator-initiated research programme. The REDCap database was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number UL1TR002537. This study was partially funded by a research grant from the Mayo Midwest Pharmacy Research Committee.

Transparency declarations

The following authors received funding by Spero Therapeutics who provided an unrestricted research grant via its investigator-initiated research programme: M.A.R., B.A.F., P.V., S.J.H., G.S., D.Z., G.T.H., M.C., C.N.T., M.J., T.F. and D.A.T. C.S.B. was partially funded by a research grant from the Mayo Midwest Pharmacy Research Committee. All other authors have no conflicts of interest to declare.

Supplementary data

Table S1 is available as Supplementary data at JAC Online.

References