Abstract
Background:
Antimicrobial resistance (AMR) is a growing problem worldwide, with differences in regional resistance patterns partially driven by local variance in antibiotic stewardship. Trauma patients transferring from Mexico have more AMR than those injured in the United States; we hypothesized a similar pattern would be present for burn patients.
Patients and Methods:
The registry of an American Burn Association (ABA)-verified burn center was queried for all admissions for burn injury January 2015 through December 2019 with hospital length-of-stay (LOS) longer than seven days. Patients were divided into two groups based upon burn location: United States (USA) or Mexico (MEX). All bacterial infections were analyzed.
Results:
A total of 73 MEX and 826 USA patients were included. Patients had a similar mean age (40.4 years MEX vs. 42.2 USA) and gender distribution (69.6% male vs. 64.4%). The MEX patients had larger median percent total body surface area burned (%TBSA; 11.1% vs. 4.3%; p ≤ 0.001) and longer hospital LOS (18.0 vs. 13.0 days; p = 0.028). The MEX patients more often had respiratory infections (16.4% vs. 7.4%; p = 0.046), whereas rates of other infections were similar. The MEX patients had higher rates of any resistant organism (47.2% of organisms MEX vs. 28.1% USA; p = 0.013), and were more likely to have resistant infections on univariable analysis; however, on multivariable analysis country of burn was no longer significant.
Conclusions:
Antimicrobial resistance is more common in burn patients initially burned in Mexico than those burned in the United States, but location was not a predictor of resistance compared to other traditional burn-related factors. Continuing to monitor for AMR regardless of country of burn remains critical.
In the United States, up to 10,000 people die from a burn-related infection every year. 1 Burn patients are highly susceptible to infection because of a suppressed immune response, loss of the protective skin barrier, and systemic inflammatory dysregulation. 2 Because of the many medical advances made in burn resuscitation, wound care, and critical care, infection now poses the most significant threat to life after burn injury. Sepsis is the most common cause of death in burn patients, and at least half of all deaths in burn victims can be attributable to infection.3,4 The proportion of deaths from infection has increased over time because of both a decrease in death from other causes as well as an increase in the incidence of multi-drug–resistant (MDR) organisms. In addition, because of prolonged hospital stays and frequent invasive procedures, burn patients are particularly at risk for MDR nosocomial infections. 5 The microbial colonization profile of patients changes over time, and is related to both patient population and location, making it imperative that the trends of microbial cultures and antibiotic resistance be identified and tracked. 6
Multi-drug–resistant organisms are also increasingly common worldwide, with regional variance in bacterial type and resistance patterns.7–9 Of particular concern are resistant gram-negative organisms, which are increasingly prevalent in Latin America and account for a higher percentage of infections there than in the United States or Canada.10,11 Patients initially treated in higher-prevalence areas are at risk of active MDR infections, and screening protocols for transfer patients have been established in some centers.8,12 Hospitals along the United States-Mexico border have demonstrated increasing resistance patterns, raising concern for transfer of MDR organisms across the border. 13 We have previously shown increased rates of gram-negative resistance in trauma patients transferred in after injury in Mexico, leading to a change in empiric antibiotic therapy for these patients to avoid delays in adequate treatment. 14 We hypothesized that patients transferred to a U.S. burn center after being burned in Mexico will have a higher incidence of infections due to antibiotic-resistant organisms than those treated in the United States alone, and may require alternate empiric antibiotic therapy while awaiting culture results.
Patients and Methods
Data source, study setting, and participants
The registry of an American Burn Association (ABA)-verified adult and pediatric burn center was retrospectively queried for all admissions for burn injury from January 2015 to December 2019. Patients of all ages were included if they were admitted to the burn service and had burn as a cause of injury. Patients were excluded if they had a hospital length of stay (LOS) shorter than seven days or admission for a non-burn etiology (Stevens-Johnson syndrome, toxic epidermal necrolysis, road rash, necrotizing fasciitis, etc.). Patients presenting with both burns and traumatic injuries were included if they were transferred to the burn service once their traumatic injuries were addressed. Patients were divided into two groups based upon country of burn occurrence: inside the United States (USA) or in Mexico (MEX). Cultures were reviewed to remove duplicates and for confirmation of clinical infection. This study was approved by the local Institutional Review Board with waiver of consent. All data collection and storage occurred in compliance with Health Insurance Portability and Accountability Act of 1996 rules and regulations.
Data collection
Data were collected from the burn registry and electronic health record, including demographics, comorbid conditions, burn mechanism, burn size (percent total body surface area burned [%TBSA]), inhalation injury (standardly diagnosed via bronchoscopy), operations and procedures, culture results, antibiotics used, duration of antibiotics, and outcomes, including mortality. Culture results were reviewed to remove duplicates and classify results as infections versus colonization as described below. For the purposes of this study, only bacterial infections were included, whereas fungal and yeast culture results were excluded. Burn wound infections, graft site infections, and donor site infections were identified via chart review, with an infection counted as present if physician clinical suspicion led to initiation of antibiotics for treatment of the wound site and cultures grew one or more dominant bacteria. For urinary tract infections, cultures were designated as infections only if >100,000 colonies/mL were present on final culture. For respiratory infections, cultures were characterized as ventilator-associated pneumonia (VAP), non-ventilator-associated pneumonia, or no infection. Ventilator-associated pneumonia was identified only if the 2016 National Trauma Data Standard VAP guidelines were met, 15 which also correlates with the definition of VAP in the ABA data dictionary. Non-ventilator-associated pneumonia was counted if VAP criteria were not met but pneumonia was clinically diagnosed and treated by the burn team at the time of care. For blood stream infections, blood culture results were individually reviewed if positive for coagulase-negative Staphylococcus or if only one of multiple culture bottles yielded positive results. Final determination of infection was determined by consensus of multiple authors.
Culture results were also reviewed for both individual antibiotic resistance as well as specific resistance patterns. Organisms were labeled as having a resistance pattern if such a pattern was identified on the final microbiology report based on standard laboratory procedures; this was the case for all instances of methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococcus faecium (VRE). Resistance was inferred for other organisms if specific criteria were met as follows. For extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae, either identification on the final microbiology report as an “ESBL” organism, or resistance to one or more third-generation cephalosporin or aztreonam by Clinical and Laboratory Standards Institute (CLSI)- or Phoenix-defined minimum inhibitory concentration 90 (MIC90) breakpoints. For MDR Pseudomonas aeruginosa (MDR Pseudomonas), resistance to one or more agent in three or more of the following classes by CLSI- or Phoenix-defined MIC90 breakpoints: monobactams, antipseudomonal cephalosporins, β-lactam/β-lactamase inhibitor combinations, carbapenems, fluoroquinolones, aminoglycosides, or lipopeptides. For MDR Acinetobacter baumannii (MDR Acinetobacter), resistance to one or more agent in three or more of the following classes by CLSI- or Phoenix-defined MIC90 breakpoints: β-lactam/β-lactamase inhibitor combinations, cephalosporins, carbapenems, fluoroquinolones, aminoglycosides, tetracyclines, anti-folate, or lipopeptides. For carbapenem-resistant Enterobacteriaceae (CRE), non-susceptibility to one or more of the following carbapenems, according to CLSI- or Phoenix-defined MIC90 breakpoints: imipenem, meropenem, doripenem, or ertapenem.
Outcomes
The primary outcome was presence of infections and their identified causative organisms in MEX patients compared to USA patients. Secondary outcomes included prevalence of antibiotic resistance, specific resistance patterns, infection subtypes, and additional risk factors for resistant infection.
Statistical analysis
Statistical analysis was performed using SPSS Statistics, version 24 (IBM Corp, Armonk, NY). In bivariable analysis, we used the Pearson χ 2 test for categorical variables. For independent variables that did not meet the assumptions for the χ 2 test (i.e., <5 observations), the Fisher exact test was used to test the association. Continuous variables were compared using the independent samples Student t-test if they met normality assumptions, and with the Mann-Whitney U test if they did not. Variables were considered for inclusion in the logistic regression model if their bivariable p value was <0.2. Model reduction was performed using backward stepwise regression with criteria for entry set at p < 0.05 and criteria to remove from the model at p > 0.1. Variables that did not meet p value criteria were kept in the model if they were deemed to be important confounders or if their exclusion caused a change in model performance.
Results
A total of 899 patients met criteria for inclusion, of whom 73 were injured in Mexico and 826 in the United States. The MEX and USA groups were similar in age (40.4 ± 2.7 standard error [SE] years vs. 42.2 ± 0.8 SE, respectively; p = 0.518), sex (69.9% male vs. 64.4%; p = 0.350), and comorbidities (Table 1). Flame burns were the most common in each group, but MEX patients had a larger median %TBSA (11.1% [4–19] vs. 4.3% [2–9.6]; p < 0.001). Rates of inhalation injury were not statistically different (11.0% MEX vs. 6.3% USA; p = 0.365). There was also no difference in time from burn to admission to our facility (median 1 day [interquartile range {IQR}, 0–3] MEX vs. 0 days [IQR, 0–3] USA; p = 0.629). The MEX patients had longer hospital LOS (median 18 days [IQR, 13–26] vs. 13 days [IQR, 10–19]; p < 0.001), intensive care unit (ICU) LOS, and more ventilator days, but no difference in mortality (1.4% vs. 2.7%; p = 0.384).
Demographics and Outcomes by Country
Values are number (%) or median [25th–75th percentile] unless otherwise specified.
SE = standard error; %TBSA = percent total body surface area burned; LOS = length of stay; ICU = intensive care unit.
Overall infections, by country
The MEX patients were more likely to have any cultures sent (49.3% of patients vs. 32.7%; p = 0.004) and had a higher rate of respiratory infections (16.4% vs. 7.4%; p = 0.046), than USA patients (Table 2). There was no difference in blood stream infections, urinary tract infections, or body site infections between patient groups. There was also no difference in the percentage of patients with any positive culture (23.3% MEX vs. 17.1% USA; p = 0.324). Time from admission to first positive culture was not significantly different between groups (median 6 days [1–10] MEX vs. 6 days [2–10] USA; p = 0.498).
Culture Results, Resistance Patterns, and Antibiotic Use by Country
ESBL = extended-spectrum β-lactamase; MRSA = methicillin-resistant Staphylococcus aureus; VRE = vancomycin-resistant Enterococcus; MDR = multi-drug–resistant; CRE = carbapenem-resistant Enterobacteriaceae; MRSE = methicillin-resistant Staphylococcus epidermidis; IQR = interquartile range.
Organisms and resistance patterns, by country
When organisms grown were assessed, there was a significant difference in the distribution of organisms identified by country, as well as in rates of resistance (Table 2). The MEX patients had more resistant organisms overall (47.2% of organisms resistant MEX vs. 28.1% USA; p = 0.013). Similar trends were seen both for gram-negative organisms (43.9% resistant MEX vs. 24.8% USA; p = 0.009), and gram-positive organisms (51.5% resistant MEX vs. 33.5% USA; p = 0.045). Looking at specific resistance patterns, rates of MDR Pseudomonas/Acinetobacter organisms (primarily MDR Acinetobacter) were higher in MEX patients (14.9% of Pseudomonas/Acinetobacter MDR in MEX patients vs. 6.3% USA; p = 0.048), and there were high rates of MRSA (20.3% of all Staphylococcus aureus in MEX patients vs. 10.9% USA; p = 0.059). Conversely, patients from the United States had the only identified cases of both MRSE and carbapenem-resistant Enterobacteriaceae. There was no difference by country in rates of ESBL organisms or VRE.
Antibiotic use by country
The MEX patients had both more antibiotic days (median, 8 days [IQR, 2–15] vs. 5 [IQR, 1–10] USA; p = 0.007), and more antibiotic-free days (median, 11 days [IQR, 7–16] vs. 8 days [IQR, 5–12] USA; p = 0.011), given their longer hospital LOS (Table 2). The MEX patients were taking antibiotics a larger percentage of hospital days (median, 46.2% vs. 36.4%; p = 0.416) than USA patients, but this difference was not significant.
Risk factors for infection
Bivariable analysis was performed to assess risk factors for any resistant infection (Table 3). Multiple risk factors were identified, including burn in Mexico (p = 0.015), increasing age (p = 0.005), burn type (p = 0.004), transfer from another facility (p = 0.040), inhalation injury, %TBSA burn, hospital and ICU LOS, ventilator days, and number of operations performed (all p < 0.001).
Bivariable Analysis, Risk Factors for Any Resistant Infection
%TBSA = percent total body surface area burned; LOS = length of stay; ICU = intensive care unit.
On multivariable analysis assessing risk factors for both any infection and resistant infection, %TBSA, inhalation injury, age, and history of alcoholism remained predictors (Table 4 and 5). Diabetes mellitus and drug dependence were also risk factors for any infection (Table 5). Burn in Mexico was no longer an independent risk factor in the presence of these other predictors.
Multivariable Analysis, Risk Factors for Resistant Infection
Other included variables (not significant): transferred from another facility, burn type, diabetes mellitus.
OR = odds ratio (logistic regression); CI = confidence interval; %TBSA = percent total body surface area burned.
Multivariable Analysis, Risk Factors for Any Infection
Other included variables (not significant): transferred from another facility, burn type, current smoker, hypertension, psychiatric illness, obesity.
OR = odds ratio (logistic regression); CI = confidence interval; %TBSA = percent total body surface area burned.
Discussion
With the increase of antibiotic resistance globally, identifying at-risk patients may guide screening and treatment protocols. Previously, cross-border infection patterns have been examined in trauma patients, finding that patients initially treated in Mexico were more likely to harbor resistant organisms when infection was present. Our study is the first to examine antibiotic resistance patterns among burn patients transferred across a major international border and is novel in the literature to date. We found that patients burned in Mexico then cared for at a U.S. burn center experienced a higher incidence of infection due to resistant organisms, including both gram-negative and gram-positive organisms; however, on multivariable analysis, burn in Mexico was not an independent risk factor for either infection or resistant infection compared with other patient factors.
Exposure to the Mexican healthcare system for the same acute illness is thought to increase risk for resistant infections; this was a proposed mechanism for the resistance previously reported in our trauma population, who had a median time of two days from injury to admission. However, this same delay was not seen in our burn patients; as such we suspect that the burn patients transferred from Mexico arrived colonized with community-acquired organisms rather than more resistant hospital-acquired organisms. This may partially account for the fact that country of burn was no longer an independent risk factor for infection on multivariate analysis, though the key question remains: do burn patients transferred from Mexico harbor skin microflora or a microbiome that predisposes them to resistant infection? It is also possible that the availability of antibiotics in Mexico may increase the risk of colonization with antibiotic-resistant organisms even in the community setting, as antibiotics are often available over-the-counter with less emphasis on antibiotic stewardship and evidence-based prescribing.
Our results suggest that even if patients transferred from Mexico do harbor more resistant flora, country of burn occurrence alone is not sufficient to explain the infection patterns seen in our population; rather, other factors known to increase risk of infection appear to be more clinically significant than country of burn. Although we have changed the empiric antibiotic regimen often used for our trauma patients injured in Mexico based on the increase in resistant gram-negative organisms seen in that population, this evidence indicates that a similar change is not needed for our burn patients. It also raises the question of whether time spent in Mexico between injury and transfer to the United States, or more specifically time in contact with the Mexican healthcare system, is a more appropriate risk factor for resistance than injury in Mexico alone; similar to increased risk of and screening for MRSA often performed in patients transferred from U.S. healthcare settings.
The MEX burn patients had larger %TBSA burns and a longer hospital LOS than USA patients. Larger burns, prolonged periods of inadequate cutaneous coverage, and multiple surgical debridements all predispose patients to longer periods of compromised immunity and place them at higher risk for infection.16–18 As such, it is unsurprising that %TBSA remained predictive on multivariable analysis for resistant infection. Given the larger size of burns seen in MEX patients, %TBSA is likely a contributor to the difference seen with this population. The longer hospital length of stay for MEX patients likely contributed as well, as resistant organisms acquired in our hospital are likely to dominate infections later in a patient's stay. Although MEX patients had longer hospital stays and larger %TBSA burns in conjunction with higher rates of antimicrobial resistance, the mortality rate was not different between the two groups (1.4% MEX vs. 2.7% USA). This is consistent with prior studies indicating infection with multidrug-resistant organisms in burn patients is not independently associated with mortality.19,20
The MEX patients also had a different pattern of both organisms grown and of organism resistance than USA patients. Previous studies have noted that the gram-positive pathogen Staphylococcus aureus, and gram-negative organisms such as Pseudomonas and Acinetobacter baumannii, are the most common bacteria isolated among burn wound infections. 21 In our study Staphylococcus aureus was common in both MEX and USA patients though higher in the MEX group, whereas we unexpectedly had much higher rates of Acinetobacter in the MEX group, with very little found in USA patients. Conversely, rates of Pseudomonas were slightly higher in the USA population.
In addition to overall differences in organisms between the country cohorts, different patterns of antibiotic resistance were identified between the two groups, with MDR Pseudomonas/Acinetobacter organisms (primarily MDR Acinetobacter) more common in MEX patients, whereas both carbapenem-resistant organisms and methicillin-resistant Staphylococcus epidermidis were found only in USA patients. Interestingly, MDR Acinetobacter was found only in MEX patients, whereas MDR Pseudomonas was only found in USA patients; the reasons for this are unclear. One prior study by Munier et al. 22 reported that higher rates of MDR Acinetobacter correlated with higher patient Abbreviated Burn Severity Index (ABSI) and Simplified Acute Physiology (SAPS) II scores, indicating these infections may be tied to more severe burns and possibly longer ICU stays. Given that our MEX group on average had larger %TBSA and longer stays, this may explain why Acinetobacter was more common in the MEX group. It is evident that Acinetobacter carries intrinsic mechanisms conferring frequent resistance to commonly used antibiotics, increasing the prevalence of multi-drug resistance; consistent with the high rate found in MEX patients.23,24 Still, it is unclear why Pseudomonas resistance was conversely present only in the USA population, indicating that either our study could be underpowered or another risk factor that was not analyzed may explain the difference.
Type of infection differed between the groups only for respiratory infections, which were more common in MEX burn patients. This increase in respiratory infections may be explained by the higher number of ventilator days seen on average in MEX patients, likely the result of more severe burns as previously discussed. It is well known that both severe burns and inhalation injuries are a risk factor for VAP, as are intubation and prolonged ventilation. 25 Ventilator-associated pneumonia is the most common nosocomial infection in the ICU with up to 27% to 44% of mechanically ventilated trauma patients developing VAP over the course of their stay in the ICU. 26 Because burns induce systemic inflammation and immunosuppression, the larger average burn size identified in MEX burn patients may be a contributor to the difference seen in respiratory infections. Additionally, higher average ventilator days in MEX patients compared with USA patients further suggest a higher burden of more critically ill patients in the MEX group. Although inhalation injury rates were not statistically different between the two groups, there was a trend towards a higher rate in MEX patients (11.0% vs. 6.3%), and inhalation injury was a risk factor for both any infection and resistant infection, suggesting it may play a role in the difference in respiratory infections observed.
Several limitations are acknowledged. First, this study is retrospective, and although country of burn injury precludes randomization, prospective data could be beneficial in further elucidating the significance of country of burn, as well as in collecting more granular data or sending cultures on admission to assess for pre-existing colonization with resistant organisms. Our study population also skewed toward USA patients, thus it is possible that the sample of MEX patients was not adequate to detect the relevance of country of burn on multivariate analysis; however, a post hoc power analysis calculated a power of 72.5% for the primary outcome of patients with any resistant organism, so we feel this risk is low. Additionally, patients transferred from Mexico to our burn unit often arrive without records of care provided prior to transfer, leaving charts incomplete. The limited availability of comprehensive primary care in Mexico also increases the risk of undiagnosed comorbidities. Specifically, although the prevalence of comorbid conditions such as diabetes mellitus, drug dependence, and alcoholism were similar between our MEX and USA cohorts, the prevalence may be underestimated in patients from Mexico. Finally, although number of operations was examined in our analysis, the extent of surgical management such as surface area of excision, time to grafting, type of graft, and their effects on wound infections were not analyzed in this study as the small number of MEX patients makes these data underpowered for robust analysis.
Conclusions
This study is the first to compare infection rates and antibiotic resistance patterns between patients initially burned in Mexico versus the United States. Although burn patients transferred from Mexico demonstrated a higher incidence of infections from resistant organisms, the country in which the burn occurred was not an independent risk factor for either infection or resistant infection on multivariate analysis. Instead, other patient factors such as burn size, inhalation injury, age, and history of alcoholism were identified as stronger predictors of resistant infection than country of burn occurrence. Although patients burned in Mexico may have an increased risk for infection with resistant organisms, this is surpassed by traditional burn-related factors.
Footnotes
Authors' Contributions
Study design: Dunbar, Santorelli, Haines, Box, Lee, Costantini, Doucet, Berndtson. Data collection: Dunbar, Santorelli, Strait, Smith, Berndtson. Data analysis: Dunbar, Smith, Berndtson. Data interpretation: Dunbar, Santorelli, Marshall, Haines, Box, Lee, Berndtson. Writing—original draft: Dunbar, Marshall, Berndtson. Writing—review and editing: Dunbar, Santorelli, Marshall, Haines, Box, Lee, Strait, Costantini, Doucet, Berndtson. Software: Smith. Resources: Smith. Conceptualization: Berndtson. Supervision: Berndtson.
Funding Information
There was no specific funding source for this project.
Author Disclosure Statement
All authors declare they have no conflicts of interest.