Methods

This single-center retrospective cohort included non-pregnant adults aged ≥18 years receiving ≥48 hours of active antibiotics for urinary isolates of S. marcescens, M. morganii, or Providencia spp. between October 2019 and June 2024. Patients were identified from a microbiology isolate repository using a first-isolate-per-patient-per-year approach similar to that used for antibiogram preparation. The institutional antibiogram was used to identify organisms of interest. All eligible patients identified from available repository records during the study period were included as a convenience sample, and no predetermined sample-size target or a priori sample-size calculation was used. Exclusions were antibiotic treatment for other infectious syndromes, including bacteremia; polymicrobial cultures requiring broader Gram-negative or additional Gram-positive coverage; and target isolates intermediate or resistant to CRO or piperacillin–tazobactam (TZP). Other polymicrobial cultures were retained without organism-number or colony-count thresholds. The study protocol was approved by the Institutional Review Board of Beth Israel Deaconess Medical Center (BIDMC; 2024P000842) and was conducted in accordance with the principles of the Declaration of Helsinki. Individual informed consent was waived due to the retrospective nature of the study. Data were obtained through manual review of the electronic medical record and entered and managed using REDCap hosted at BIDMC.

Group 1 (AmpC-vulnerable β-lactams) comprised initial treatment with CRO, cefpodoxime (CPD), and TZP. Group 2 (AmpC-stable β-lactams or non-β-lactam agents) comprised FEP, meropenem (MEM), TMP/SMX and FQ. The index regimen was the first active agent administered for ≥48 hours, and brief empiric exposures of less than 48 hours preceding the index regimen did not determine assignment. For example, one dose of FEP followed by at least 48 hours of CRO would have resulted in assignment to Group 1. Subsequent regimens were summarized separately and did not change assignment.

Patients documented as having asymptomatic bacteriuria were retained if they received antibiotics at the treating providers’ discretion. Urinary and nonurinary symptoms were classified based on provider documentation with overlapping categories; nonspecific presentations were not independently adjudicated to determine whether symptoms were attributable to a UTI. Complicated UTI was classified using historical criteria, including male sex, urinary catheter or nephrostomy tube, urinary obstruction, structural or functional urinary tract abnormalities, diabetes, urinary retention, and immunocompromise (Table 1). ^{Reference Gupta, Hooton and Naber9–Reference Johansen, Botto and Cek12} These criteria differ from the 2025 IDSA approach, which emphasizes infection extending beyond the bladder with systemic illness or upper tract findings and includes catheter-associated UTIs. Male sex, underlying urologic abnormalities, diabetes, or immunocompromise alone no longer establishes complicated UTI. However, patients with complicated UTIs may still have underlying conditions such as having an indwelling urinary catheter, neurogenic bladder, urinary obstruction, or urinary retention. ^{Reference Trautner, Cortés-Penfield and Gupta13} The historical classification was therefore retained because available documentation did not permit reliable retrospective assessment of symptom attribution, tenderness location, and catheter-associated symptoms.

Table 1.

Baseline characteristics

Table 1. Long description

A table comparing baseline characteristics of patients with AmpC-vulnerable and AmpC-stable beta-lactams. The table has 25 rows and 5 columns. Column headers are Group 1, Group 2, and Total cohort. Group 1 is further divided into AmpC-vulnerable beta-lactams with 79 patients and Group 2 is further divided into AmpC-stable beta-lactams/non-beta-lactams with 33 patients. Row labels include Male sex, Age, Level of care, Hospital length of stay, Clinical signs/symptoms, Historical complicated UTI classification, and Isolated organisms of interest. Each row provides specific data points for each group and the total cohort. Notable trends include the higher percentage of male sex in Group 2 and the higher median age in Group 1. The table also shows differences in clinical signs/symptoms, historical complicated UTI classification, and isolated organisms of interest between the two groups.

^a Group 1: CRO/CPD/TZP (n = 79). Group 2: FEP/MEM (n = 11) or FQ/TMP-SMX (n = 22), pooled n = 33. These initial-treatment groups do not imply combination therapy or clinical equivalence.

^b Other symptoms include falls, syncope, weakness, fatigue, penile pain, purulent suprapubic-catheter discharge, malfunctioning nephrostomy tube, and elevated CRP.

^c Recorded asymptomatic bacteriuria as determined by the treating providers.

^d Other factors: ileal conduit, preheart transplant, and spinal surgery.

The primary outcome was microbiological recurrence, defined as re-isolation of the same organism in urine culture within 28 days after treatment completion. Secondary outcomes included recurrence of the same isolate within 365 days, antibiotic changes, and treatment duration. Recurrence with retreatment at 28 days was assessed post hoc, however, recurrence symptoms were unavailable. Changes from initial antibiotics were classified as escalation or de-escalation/oral stepdown using study-specific conventions based on antimicrobial spectrum. Intravenous (IV) to oral (PO) switches were included in either category, depending on the agents involved. For example, CRO to CPD and FEP to TMP/SMX were classified as de-escalation, whereas CRO to a fluoroquinolone was classified as escalation because of the broader spectrum of the latter.

Counts, percentages, medians, and interquartile ranges were calculated, and two-sided Fisher’s exact tests were used for outcome comparisons (Python 3.12.13; SciPy 1.17.0). Post hoc analyses excluded asymptomatic bacteriuria, initial non-β-lactam comparators, both, or any documented TMP/SMX or FQ exposure (Table S2). A two-sided P value < .05 was considered statistically significant. No adjustments for multiple comparisons were applied. ChatGPT/Codex assisted with analysis code (OpenAI; https://chatgpt.com; 7 September 2026; GPT-6 Astra).