Neonatal Escherichia coli Bloodstream Infections
Neonatal Escherichia coli Bloodstream Infections
We identified all infants with E. coli BSI discharged from 1 of 77 U.S. neonatal intensive care units (NICUs) managed by the Pediatrix Medical Group in 2012. Data were obtained from an electronic medical record that prospectively captures information from notes generated by clinicians on all infants cared for by the Pediatrix Medical Group. Data are extracted, de-identified and stored in the Pediatrix Clinical Data Warehouse. Information stored for infants on a daily basis includes maternal history, demographics, medications ordered, laboratory results, microbiology results, diagnoses and procedures. Antimicrobial susceptibility data for microbiologic isolates are reported as the presence or absence of resistance to several common antibiotics. Comprehensive antimicrobial resistance patterns are not captured in the database.
We identified all blood cultures positive for E. coli during the first 120 days of life. All positive cultures obtained within 21 days of each other were analyzed as single infectious episodes. We assumed the same antimicrobial resistance for all positive blood cultures obtained within a single BSI episode. We defined the duration of bacteremia as the difference between the day of the first and last positive blood cultures within a single episode. We defined an early-onset BSI as an initial positive blood culture within the first 3 postnatal days and a late-onset BSI as an initial positive blood culture after the first 3 postnatal days.
We defined empiric antimicrobial therapy as any antimicrobial administered on the day of the first positive blood culture for each episode of bacteremia. For ampicillin-resistant E. coli isolates, we categorized the number of effective empiric antimicrobials received on the day of first positive blood culture as none or ≥1. We considered the following antimicrobials effective empiric therapy against ampicillin-resistant E. coli: beta-lactam/lactamase inhibitor combinations, cephalosporins, antipseudomonal penicillins, aztreonam, carbapenems, aminoglycosides, fluoroquinolones and trimethoprim/sulfamethoxazole. We defined inotropic support as exposure to any inotrope (dobutamine, dopamine, epinephrine, milrinone/amrinone, norepinephrine or phenylephrine), supplemental oxygen requirement as any fraction of inspired oxygen >21% and mechanical ventilation as any need for invasive mechanical ventilation on the day of the first positive blood culture for each episode of bacteremia. We categorized E. coli isolates as ampicillin-susceptible or ampicillin-resistant as reported by each site.
The primary outcome of our study was mortality within 30 days of the first positive blood culture. Secondary outcomes were duration of BSI, mortality within 7 days of the first positive blood culture and mortality at hospital discharge. We used standard summary statistics including counts (percentages) and medians (interquartile ranges) to describe the categorical and continuous study variables, respectively. We compared study variables across groups using Wilcoxon rank sum and χ tests of association where appropriate. We used multivariable logistic regression to compare outcomes between ampicillin-susceptible and ampicillin-resistant E. coli BSI. The following covariates were included in a full model: E. coli ampicillin-susceptible versus ampicillin-resistant, birth weight, male gender, gestational age (GA), small-for-GA status (SGA), early-onset versus late-onset BSI and inotropic, ventilator or oxygen support on the day of the first positive culture. Reduced models were then constructed by removing prespecified groups of covariates and comparing the reduced model to the full model using likelihood ratio tests. The most parsimonious model that fit the data well was retained and contained the following covariates in addition to ampicillin resistance: GA, SGA, early-onset versus late-onset BSI, and inotropic, ventilator, or oxygen support on the day of the first positive culture. Lastly, we accounted for the clustering of data by fitting separate random intercepts for each NICU. We used standard graphing techniques and statistical tests to evaluate the assumptions of our models. To evaluate the impact of effective empiric antimicrobial therapy, 30-day mortality was compared between infants with ampicillin-resistant E. coli BSI receiving 0 or ≥1 empiric antimicrobial agents known to be effective against ampicillin-resistant isolates. Multivariable logistic regression was used to compare 30-day mortality for infants with ampicillin- resistant E. coli BSI receiving 0 versus ≥1 empiric antimicrobial agents active against ampicillin-resistant E. coli. Covariates incorporated into the model included GA, SGA, early-onset versus late-onset BSI and inotropic, ventilator or oxygen support on the day of the first positive culture. Finally, median duration of BSI for infants with ampicillin-resistant E. coli isolates receiving 0 versus ≥1 effective empiric antimicrobial agents was compared. We used Stata 13.0 (College Station, TX) to conduct all analyses and considered a P < 0.05 statistically significant. The Duke University Health System Institutional Review Board determined that the study met the definition of research not involving human subjects and was, therefore, exempt.
Methods
We identified all infants with E. coli BSI discharged from 1 of 77 U.S. neonatal intensive care units (NICUs) managed by the Pediatrix Medical Group in 2012. Data were obtained from an electronic medical record that prospectively captures information from notes generated by clinicians on all infants cared for by the Pediatrix Medical Group. Data are extracted, de-identified and stored in the Pediatrix Clinical Data Warehouse. Information stored for infants on a daily basis includes maternal history, demographics, medications ordered, laboratory results, microbiology results, diagnoses and procedures. Antimicrobial susceptibility data for microbiologic isolates are reported as the presence or absence of resistance to several common antibiotics. Comprehensive antimicrobial resistance patterns are not captured in the database.
We identified all blood cultures positive for E. coli during the first 120 days of life. All positive cultures obtained within 21 days of each other were analyzed as single infectious episodes. We assumed the same antimicrobial resistance for all positive blood cultures obtained within a single BSI episode. We defined the duration of bacteremia as the difference between the day of the first and last positive blood cultures within a single episode. We defined an early-onset BSI as an initial positive blood culture within the first 3 postnatal days and a late-onset BSI as an initial positive blood culture after the first 3 postnatal days.
We defined empiric antimicrobial therapy as any antimicrobial administered on the day of the first positive blood culture for each episode of bacteremia. For ampicillin-resistant E. coli isolates, we categorized the number of effective empiric antimicrobials received on the day of first positive blood culture as none or ≥1. We considered the following antimicrobials effective empiric therapy against ampicillin-resistant E. coli: beta-lactam/lactamase inhibitor combinations, cephalosporins, antipseudomonal penicillins, aztreonam, carbapenems, aminoglycosides, fluoroquinolones and trimethoprim/sulfamethoxazole. We defined inotropic support as exposure to any inotrope (dobutamine, dopamine, epinephrine, milrinone/amrinone, norepinephrine or phenylephrine), supplemental oxygen requirement as any fraction of inspired oxygen >21% and mechanical ventilation as any need for invasive mechanical ventilation on the day of the first positive blood culture for each episode of bacteremia. We categorized E. coli isolates as ampicillin-susceptible or ampicillin-resistant as reported by each site.
The primary outcome of our study was mortality within 30 days of the first positive blood culture. Secondary outcomes were duration of BSI, mortality within 7 days of the first positive blood culture and mortality at hospital discharge. We used standard summary statistics including counts (percentages) and medians (interquartile ranges) to describe the categorical and continuous study variables, respectively. We compared study variables across groups using Wilcoxon rank sum and χ tests of association where appropriate. We used multivariable logistic regression to compare outcomes between ampicillin-susceptible and ampicillin-resistant E. coli BSI. The following covariates were included in a full model: E. coli ampicillin-susceptible versus ampicillin-resistant, birth weight, male gender, gestational age (GA), small-for-GA status (SGA), early-onset versus late-onset BSI and inotropic, ventilator or oxygen support on the day of the first positive culture. Reduced models were then constructed by removing prespecified groups of covariates and comparing the reduced model to the full model using likelihood ratio tests. The most parsimonious model that fit the data well was retained and contained the following covariates in addition to ampicillin resistance: GA, SGA, early-onset versus late-onset BSI, and inotropic, ventilator, or oxygen support on the day of the first positive culture. Lastly, we accounted for the clustering of data by fitting separate random intercepts for each NICU. We used standard graphing techniques and statistical tests to evaluate the assumptions of our models. To evaluate the impact of effective empiric antimicrobial therapy, 30-day mortality was compared between infants with ampicillin-resistant E. coli BSI receiving 0 or ≥1 empiric antimicrobial agents known to be effective against ampicillin-resistant isolates. Multivariable logistic regression was used to compare 30-day mortality for infants with ampicillin- resistant E. coli BSI receiving 0 versus ≥1 empiric antimicrobial agents active against ampicillin-resistant E. coli. Covariates incorporated into the model included GA, SGA, early-onset versus late-onset BSI and inotropic, ventilator or oxygen support on the day of the first positive culture. Finally, median duration of BSI for infants with ampicillin-resistant E. coli isolates receiving 0 versus ≥1 effective empiric antimicrobial agents was compared. We used Stata 13.0 (College Station, TX) to conduct all analyses and considered a P < 0.05 statistically significant. The Duke University Health System Institutional Review Board determined that the study met the definition of research not involving human subjects and was, therefore, exempt.
