Study Design
We conducted a multicentre retrospective cohort study including patients with first positive concurrent blood and urine cultures for E. coli from January 1, 2016 to December 31, 2016 within Fraser Health, which serves 1.8 million people in British Columbia, Canada. The study was approved by Fraser Health Research Ethics Board (FHREB File #: 2017-015).
Patients with E. coli positive blood and urine cultures were identified from an electronic microbiology database. Patients were included in this study if they were 18 years or older and had both blood and urine cultures positive for E. coli within 72 hours of each other. Patients were only included if they received empiric intravenous beta-lactam therapy and followed by step down to oral beta-lactam or oral fluoroquinolone for treatment completion. Timing and agent of step down therapy was at the discretion of the treating physician. Patients who did not survive for at least 72 hours after first positive blood culture were excluded.
Patients were also excluded for the following reasons: non-urinary source of E. coli bacteremia as documented by the treating physician; treatment for a concomitant infection; presence of complicated pyelonephritis identified within first 7 days of therapy (emphysematous pyelonephritis, renal parenchymal abscess, perinephric abscess, renal papillary necrosis); prostatitis; extended spectrum beta-lactamase producing, AmpC, or carbapenem-resistant E. coli; HIV with CD4 <200; neutropenia with ANC <0.5; or renal transplant.
Demographic, clinical and outcome data were acquired from electronic medical records. Charlson co-morbidity index and Pitt bacteremia score data were collected at baseline. Sepsis and septic shock was defined using the Sepsis-3 international consensus definitions [18]. Infections were classified as community-acquired healthcare-associated infections if any of the following criteria were met: Living in a long-term care facility or nursing home, receiving outpatient intravenous therapy within the past 30 days, attending a hospital or hemodialysis clinic within the past 30 days or being admitted for more than 2 days within the past 90 days. A full list of definitions can be found in Additional file 1.
Outcome Measures
The primary outcome was clinical cure defined by meeting all of the following criteria: resolution or improvement of symptoms during treatment, no recurrence of symptoms or signs of urinary tract infection within 30 days, and no discontinuation or change of treatment because of worsened or persistent symptoms or occurrence of adverse events.
Secondary outcomes included length of hospitalization following first positive blood culture, all-cause mortality within 30 days of first positive blood culture, and C. difficile positive stool sample within 30 days of first positive blood culture.
Microbiology Methods
Blood and urine cultures were processed by a centralized clinical microbiology laboratory located within Fraser Health. Antibiotic susceptibility was determined by the Vitek 2 System (bioMérieux, Laval, Quebec). Clinical and Laboratory Standards Institute breakpoints were used to determine antimicrobial susceptibility breakpoints (2016 M100 26th edition).
Statistical Analysis
Categorical variables were analyzed using Pearson chi-squared or Fisher’s exact test. Continuous variables were analyzed using Mann Whitney U test. Statistical significance for all tests was set at a p-value of less than 0.05.
Association between antibiotic therapy (oral beta-lactam or oral fluoroquinolone) and clinical cure was modelled using logistic regression and expressed as odds ratios (ORs) and 95% confidence intervals (CI). Additional adjustment for covariates with a P value of <0.20 on univariate analysis were included in the adjusted logistic regression model.
Further analysis using propensity score was performed to account for potential residual confounding. We constructed a propensity score for receiving antibiotic therapy (oral beta-lactam or oral fluoroquinolone) as the dependent variable using logistic regression modelling. Covariates included in generating the propensity score included age, sex, diabetes mellitus, chronic kidney disease, heart conditions, immune compromise, recurrent UTI risk factors (history of recurrent UTI, anatomical or functional urinary tract abnormalities, or nephrolithiasis), Pitt bacteremia score, sepsis, beta-lactam allergy, and duration of intravenous therapy. A final analysis using logistic regression to assess for association between antibiotic therapy and clinical cure was performed using propensity score as a single covariate. Statistical analyses were performed using STATA IC version 13.1 and RStudio version 1.2.503.