2.2. Data Collection
Data were collected manually from the electronic medical record during individual chart review and compared between those who did and did not develop an infection with a DTp. Demographic data collected included: age, race, sex, comorbidities, and the presence of hospital-acquired infection risk factors (HAI RF) on admission (i.e., intravenous access, history of chemotherapy, positive urine drug screen or reported social history, resident in a nursing home or long-term acute care hospital, admission to the hospital in the last 90 days, or end-stage renal disease requiring dialysis). The collected burn injury characteristics included: mechanism of injury, % total body surface area burned (TBSA), % partial and full thickness injury, and presence and grade of inhalation injury. Patient outcome data included: total length of stay (if patient survived index hospitalization) or days to mortality. The surgical course was noted with data collected on number of surgical interventions, days to first burn wound excision, and days to final grafting procedure.
Culture results were collected, and susceptibility results were analyzed. Organisms in this study could have been isolated from any source including skin swabs, tissue cultures following surgical excision, bronchoalveolar lavage, sputum, tracheal aspirate, urine, bone, or blood. Due to the retrospective nature of the study and resource limitations, specific genetic data for resistance mechanisms are not routine at the study center and were not available for classification. Organisms were considered to be DTp if they fell into one or more of the following categories: MDR (non-susceptibility to at least one agent in three or more antimicrobial categories), XDR (non-susceptibility to at least one agent in all but two or fewer antimicrobial categories), vancomycin-resistant, confirmed or presumed ESBL-production [ceftriaxone minimum inhibitory concentration (MIC) ≥ 2 μg/mL], confirmed or presumed AmpC-production (resistance to ceftriaxone, cefotaxime, and ceftazidime), carbapenem-resistant (confirmed carbapenemase producers or resistant to at least one carbapenem), or methicillin-resistant Staphylococcus aureus (MRSA), CRAB, Stenotrophomonas spp., or DTR-Pseudomonas (not susceptible to at least one antibiotic in at least three antibiotic classes for which P. aeruginosa susceptibility is generally expected: penicillins, cephalosporins, fluoroquinolones, aminoglycosides, and carbapenems). Susceptible pathogens were defined as strains whose MIC were interpreted to be susceptible to a given antibiotic. Non-susceptible pathogens were defined as strains whose MICs were interpreted to be resistant or intermediate to a given antibiotic. An organism’s intrinsic resistance was taken into consideration and the organism was not classified as resistant if intrinsically resistant (e.g., Enterococcus casseliflavus was not considered to be vancomycin-resistant). Organisms not meeting any of the above criteria were classified as non-DTp. Information on culture sources and length of stay prior to culture obtainment were also collected.
Hierarchical data were collected according to patient, pathogen, and date of infection. Exposures of topical and systemic antimicrobials were tracked over time. Exposure was recorded at the pathogen level and was defined as exposure to any patient site (i.e., exposure at the patient level) prior to each culture obtainment. For example, a single patient could have had an infection present early in the hospital stay that may have had different (likely less) exposures than pathogens that were treated a month later in the same stay. Topical agents analyzed in this study included: mafenide, bacitracin, silver sulfadiazine, mupirocin, Dakin’s, silver nitrate solution, hypochlorous acid, other topicals, solid silver dressings, as well as chlorohexidine gluconate (CHG) and nasal decolonization. Exposures to systemic antimicrobial classes were also collected, including: non-pseudomonal beta-lactams (i.e., cefazolin, ceftriaxone, ampicillin, or ampicillin-sulbactam), anti-pseudomonal beta-lactams (i.e., piperacillin-tazobactam, cefepime), carbapenems (i.e., meropenem, ertapenem, or imipenem-cilastain), fluoroquinolones (i.e., ciprofloxacin or levofloxacin), anti-MRSA agents (i.e., vancomycin, daptomycin, or linezolid), aminoglycosides (i.e., amikacin, gentamicin, or tobramycin), extended-spectrum beta-lactam beta-lactamase inhibitors (i.e., ceftazidime-avibactam, meropenem-vaborbactam, or ceftolozane-tazobactam), metronidazole, tetracyclines (i.e., doxycycline or minocycline), sulfamethoxazole-trimethoprim, and antifungals (i.e., fluconazole, micafungin, or isavuconazole).
During this study period, the institution’s microbiology laboratory utilized the Brucker MALDI Biotyper for the identification of bacteria and yeast. This system uses mass spectrometry to determine the proteomic fingerprints of microorganisms and compares these to the research-use-only database for microbial identification. The BD Phoenix™ automated system was used as a backup method for identification, and was the primary system used for antimicrobial susceptibility testing. The identification of microbes was based on the results of 45 chromogenic and fluorogenic substrates. ESBL production identified in isolates of E. coli, K. pneumoniae, and K. oxytoca was based on differential responses to third-generation cephalosporins in the presence and absence of the beta-lactamase inhibitor clavulanic acid. Carbapenem resistance for other organisms was determined by resistance to either meropenem or ertapenem, which were the representative carbapenems on the antimicrobial susceptibility testing panel. The BD Phoenix system uses a Carbapenemase-producing Organism Detect Panel. Specific genetic testing for resistance at our institution requires samples to be sent out to consulting laboratories, and must be requested. The rules for antibiotic reporting and interpretation for MIC values from the BD Phoenix™ system were based on United States Food and Drug Administration-cleared interpretations built within the automated system. Additional or reflex-sensitivity testing that may have been reported was done iteratively or collaboratively according to intramural antimicrobial stewardship guidance and adhered to Clinical & Laboratory Standards Institute breakpoints using the Kirby–Bauer disk diffusion method. Disk diffusion was rare and utilized only in cases of ceftazidime-avibactam, meropenem-vaborbactam, or ceftolozane-tazobactam susceptibilities for multidrug-resistant Acinetobacter spp., Enterobacterales spp., or Pseudomonas spp. Quality control for Phoenix identification and MICs followed the package insert for the Phoenix products. Kirby–Bauer interpretation, reporting and quality control followed CLSI recommendations.
2.3. Statistical Analysis
Demographic data and injury characteristics were reported using descriptive statistics. Nominal data were reported as n (%) and compared using Chi-square or Fisher’s exact tests, depending on expected counts. The normality of continuous data was analyzed visually and statistically via the Shapiro–Wilk and Kolmogorov–Smirnov tests. Non-parametric data were reported as median (25th, 75th percentiles), while parametric data were reported as mean ± standard deviation. To test the primary hypothesis, exposures were first subjected to univariable and then multivariable logistic regression with a manual backward elimination of exposures. Variables were considered significant if p < 0.05; however, all variables with p < 0.1 during univariable analysis were considered in multivariable modeling and covariates were manually included. Considering potential covariates (i.e., demographics, injury characteristics, and surgical course) as determined by a literature review, multivariable analysis, and Akaike information criterion, TBSA, flame injury, and full-thickness injury produced the most predictive model for the predicted presence of DTp. However, once exposures were entered into the model, the presence of full-thickness injury continually fell out of the model, while the other two covariates remained strongly associated with DTp. It was determined that TBSA and flame injury would need to be included as covariates. The exposure model was verified via repeated measures generalized linear mixed modeling as date of culture within patient as the repeated measure. No differences were observed between the models, so the most parsimonious model was chosen to model exposures. Although the a priori plan for the multicenter study was a Cox-proportional hazard model, survival was analyzed using the Kaplan–Meier method, as the number experiencing the event was few. Cox-proportional hazards will be revisited in the next iteration to better control for covariates that are known to affect mortality after burn injury.