Abstract
Background:
Patients with burn injuries are especially prone to infections; however, the influence of socioeconomic status on infection following burn injury remains poorly understood. This study aimed to evaluate the association between socioeconomic disadvantage and risk of infection among burn patients. We hypothesize that burn patients with more socioeconomic disadvantage have an increased risk of infections.
Patients and Methods:
The burn registry of an American Burn Association-verified burn center was queried for all admissions from 2015 to 2019. Patients admitted for <7 days or with a home address outside California were excluded. Collected data included demographics, burn characteristics, and culture results. Patient home addresses were geocoded, and correlated Area Deprivation Index (ADI) scores were classified into quintiles. Unhoused patients were classified as a separate category. Groups were then compared using univariate and multivariate analysis. The primary outcome was any infection. Resistant infections were analyzed as a secondary outcome.
Results:
Overall, 788 patients were included for analysis. The median ADI state decile was 6 (interquartile range = 5–8); 4.6% (n = 36) of patients were unhoused. 16.6% (n = 131) of patients had an infection during admission, whereas 6.3% (n = 50) had an infection with a resistant organism. Risk factors for having an infection included age (p < 0.001), percent total body surface area burned (p < 0.001), and diabetes mellitus (0.004). ADI quintile was not associated with infection. The unhoused state was an independent risk factor for resistant infection (odds ratio = 3.37, 95% confidence interval = 1.05–9.31, p = 0.027).
Conclusions:
ADI was not associated with an increased risk of infection, but unhoused patients are at increased risk of developing an infection with a resistant organism during admission for a burn injury.
Introduction
Each year, approximately 40,000 patients are hospitalized with burns in the United States. 1 Patients with burn injuries are highly susceptible to healthcare-associated infections (HAIs) due to immune system dysregulation, loss of the cutaneous barrier, and a systemic inflammatory response, which places them at risk for significant complications. 2 Infection is the leading cause of mortality in burn patients, with half of deaths in this population attributed to sepsis.3,4 Furthermore, burn-related infections result in longer lengths of stay, poor wound healing, and increased healthcare costs. 5
Patient-specific risk factors such as age and percent total body surface area (%TBSA) burned are associated with an increased risk of infection and mortality.6,7 Patients living in neighborhoods with high poverty tend to have more extensive burns and higher fatalities.8,9 Socioeconomic inequalities have been shown to significantly impact healthcare outcomes and quality. 10 Disadvantaged patients who are injured often have higher rates of complications and more prolonged hospitalizations compared with those of higher socioeconomic status. 11 However, traditional socioeconomic markers like median income, insurance status, and the federal poverty line struggle to capture disadvantaged communities. 12 The Area Deprivation Index (ADI) was designed to overcome this limitation by incorporating a variety of U.S. census factors such as housing, education, and employment status.
Socioeconomic deprivation, as identified by a higher ADI, has previously been shown to be an independent risk factor for burn incidence, inhalation injury, and burn severity.9,13,14 A combination of ADI and geospatial mapping has previously been utilized to identify neighborhoods at increased risk of pediatric scald burns. 13 Geographic location of injury has also been shown to impact both the organisms present in burn infection and rates of antibiotic resistance. 15 However, the association between ADI and infections after burns has yet to be delineated. Therefore, in this study of burn patients, we investigate the association between patient comorbidities, ADI, and HAI. We hypothesize that burn patients with higher ADI have an increased risk of infectious complications.
Patients and Methods
Study population and design
Patients of all ages with burn injuries treated at a single American Burn Association (ABA) accredited burn center between January 2015 and December 2019 were retrospectively identified from a prospectively maintained burn data registry. The study included patients with a burn injury as their principal diagnosis, including patients with both burn and traumatic injuries if they were transferred to the burn service after their traumatic injuries were resolved. Patients were excluded if they had a hospital length of stay (LOS) of less than seven days, an admission due to a non-burn etiology (i.e., Stevens–Johnson syndrome or necrotizing fasciitis), or a home address outside of California. The institutional review board approved the study at the University of California, San Diego, with a waiver of consent.
The state percentiles of the ADI were calculated based on the the patient’s home address using the Neighborhood Atlas tool. 16 The University of Wisconsin School of Medicine and Public Health most recently modified the original model to create a cross-reference database at the neighborhood level. Multiple healthcare institutions and specialties have used the ADI scale to evaluate the impact of socioeconomic disparities on healthcare outcomes.16–18 The ADI includes 17 factors incorporating education, employment, housing quality, and poverty measures drawn from U.S. Census data. Patients were divided into quintiles based on their state ADI values, with the lowest quantiles signifying the least disadvantaged patients. However, because there is no corresponding ADI for the unhoused state, and a standard ADI cannot be assigned without a home address, these patients were grouped separately. 19 Analysis was performed on these six groups.
Data were collected retrospectively from the burn registry and electronic medical records. They included demographics, comorbid conditions, burn mechanism, %TBSA, procedures, cultures, antibiotics, and outcomes. All culture results were reviewed to remove duplicates. For this study, we only included bacterial infections. Wound infections were identified via chart review, with an infection counted as present if physician clinical suspicion led to the initiation of antibiotics for treatment of a burn wound site, graft site, or donor site, and cultures of the site grew one or more dominant bacteria. Urinary tract infections were characterized by a final urine culture with >100,000 colonies per mL. Respiratory cultures were classified as ventilator-associated pneumonia (VAP), non-ventilator-associated pneumonia, or no infection. VAP criteria were based on the 2016 National Trauma Data Standard VAP Guidelines, 20 which also correlate 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. All blood culture results were individually reviewed if positive for coagulase-negative Staphylococcus or if only one of multiple culture bottles yielded positive results; in these cases, the authors’ consensus determined the final determination of positive blood stream infection.
Based on laboratory sensitivities, culture results were reviewed for specific antibiotic resistance and resistance patterns. If a culture had antibiotic resistance, the infection was classified as resistant. Organisms were labeled as having a resistance pattern if such a pattern was identified on the final microbiology report as well as per organism-specific criteria as we have previously published. 15
Outcomes
The primary outcome was the infection rate, as defined above, compared between the ADI quintiles and unhoused patients. Secondary outcomes include antibiotic-resistant infections, specific resistance patterns, and antibiotic duration.
Statistical analysis
All statistical analysis was performed using R Studio (R Version 4.3.0, Posit, Boston, MA). Categorical data were presented as percentages, whereas continuous data were presented as a median with an interquartile range (IQR). Chi-square testing was used to analyze comparisons of categorical variables. Mann-Whitney U and Kruskal-Wallis testing were used to compare continuous variables of two groups and three or more groups, respectively. ADI was grouped into quintiles, with unhoused patients as a separate category. Multivariable regression was performed to identify risk factors for both infection and resistant infection using patient age, ADI quintile or the unhoused state, %TBSA, diabetes mellitus, and tobacco use history, which were selected for the regression a priori. The unhoused state was also tested separately (independent of ADI). Since ADI and the unhoused state were tested together and separately in regression models for total infections and resistant infections, a false discovery rate (FDR) correction using the Benjamini-Hochberg method was used to report adjusted p values. Performance characteristics of the logistic regression models using the Hosmer-Lemeshow goodness-of-fit test and the area under the receiver operating curve (AUC) were reported. p Values were defined as statistically significant if <0.05.
Results
Demographics, comorbidities, injury profile, and hospital outcomes
Of the 788 burn patients included in the study, 133 (16.9%) were in the bottom ADI quintile, 48 (6.1%) were in the top ADI quintile, and 36 (4.6%) were unhoused (Table 1). The median age was 45 (IQR = 25–60), which was similar between groups (p = 0.513). There was no difference in biologic sex across groups (p = 0.487), with 501 (63.6%) being male. Unhoused patients were more likely to have a diagnosis of alcohol use disorder (p < 0.001). There were some differences in burn mechanism as well, with the unhoused group experiencing more flame injuries and fewer scald injuries than the other groups (p = 0.036 and 0.002, respectively). The %TBSA burned was similar between all groups at a median of 4–7% (p = 0.427), but rates of inhalation injury varied and were overall higher in the more disadvantaged ADI groups and unhoused patients (p = 0.037). Between ADI quintiles, there was no significant difference in mortality, hospital LOS, or ICU LOS. All groups also had no difference between days on antibiotics or antibiotic-free days (Table 2).
Demographics and Comorbidities by Area Deprivation Index Quintile
Higher-Area Deprivation Index quintile = greater socioeconomic deprivation.
IQR = interquartile range; %TBSA = percent total body surface area.
Clinical Outcomes by Area Deprivation Index Quintile
All values are median (IQR) unless otherwise specified.
Higher-Area Deprivation Index quintile = greater socioeconomic deprivation.
LOS = length of stay; ICU = intensive care unit; IQR = interquartile range.
Overall infections by ADI
In total, 257 (32.6%) of patients had a culture of any kind (blood, urine, respiratory, or wound) sent during their hospitalization, and 131 (16.6%) of patients had at least one positive culture (Table 3). There was no significant difference between groups in the rate of positive cultures, though there was a trend toward more positive cultures in unhoused patients (p = 0.076). Time from admission to the first positive culture was not statistically different between groups (p = 0.508). There were, however, differences in the types of infection between the groups, with unhoused patients experiencing more wound infections (p = 0.016) and ADI quintiles 3 and 5 having more respiratory infections than other groups (p = 0.007). When organisms cultured were assessed, there was no significant difference in organisms identified or rates of resistance by groups. The unhoused group had 6 (17%) resistant infections compared with 44 (5.9%) in all other groups, but this did not reach statistical significance (p = 0.128). There was no difference in antibiotic resistance patterns between groups.
Microbiology Culture Results by Area Deprivation Index Quintile
Higher-Area Deprivation Index quintile = greater socioeconomic deprivation.
IQR = interquartile range; ESBL = extended-spectrum β-lactamase; MDR = multidrug-resistant; MRSA = methicillin-resistant Staphylococcus aureus; CRE = carbapenem-resistant Enterobacteriaceae; MRSE = methicillin-resistant Staphylococcus epidermidis; VRE = vancomycin-resistant Enterococcus.
Risk factors for infection
Multivariable logistic regression was performed to identify risk factors for any infection (Table 4). Identified risk factors included age (p < 0.001), %TBSA (p = <0.001), and history of diabetes mellitus (p = 0.004). ADI group and the unhoused state were not risk factors for overall infection. The fit of the regression model using the Hosmer-Lemeshow test demonstrated no evidence of poor fit (p = 0.066), and the AUC was 0.764 (95% confidence interval [CI] 0.673–0.854). We then compared the top three ADI scores (1–3) to the lowest two (9,10) combined with the unhoused state. This identified age (p = 0.018), %TBSA burned (p < 0.001), smoking (p < 0.028), and active alcohol intoxication (p = 0.003) as risk factors for infection; however, the unhoused state and ADI were again not associated with infection.
Multivariable Analysis, Risk Factors for Any Infection
Higher-Area Deprivation Index quintile = greater socioeconomic deprivation.
Adjusted p value after false discovery rate correction using the Benjamini-Hochberg method.
Significant at alpha value of 0.05.
%TBSA = percent total body surface area; ADI = area deprivation index.
On multivariable analysis assessing risk factors for resistant infections, only age (p = 0.005) and %TBSA (<0.001) remained significant. However, when we specifically looked at unhoused versus housed groupings rather than ADI overall, being unhoused was associated with an increased risk of resistant infection (OR = 3.22, 95% CI = 1.02–8.68, p = 0.03); this was supported by a significant FDR-adjusted p value (p = 0.045) (Table 5). The fit of the regression model using the Hosmer-Lemeshow test demonstrated no evidence of poor fit (p = 0.769), and the AUC was 0.765 (95% CI = 0.561–0.969).
Multivariable Analysis, Risk Factors for Resistant Infection
Adjusted p value after false discovery rate correction using the Benjamini-Hochberg method.
Significant at alpha value of 0.05.
%TBSA = percent total body surface area.
Discussion
Infectious complications in burn patients are a significant source of morbidity; identifying high-risk patients may help guide screening protocols or the selection of prophylactic antibiotics. In this study, the overall incidence of infection was 16.6%, and the incidence of a resistant infection was 6.3%. We identified that lower socioeconomic status, as measured by ADI, was not an independent risk factor for infection or resistant infections in burn patients, though the confidence intervals for these analyses remain wide and a small association is not entirely excluded. When we assessed the unhoused group compared with all other patients, we identified that our unhoused population had a higher incidence of resistant infections than housed patients. These findings underscore the importance of patient-specific factors influencing microflora and antibiotic resistance, and identify an opportunity to target and potentially improve care for our unhoused patients. These findings also raise the question of how best to study patient-specific outcomes; ultimately, we found that the patient-specific dichotomy of housed or unhoused was a better prediction of individual infection rates than census tract-level data.
Low socioeconomic status is a known risk factor for adverse health outcomes. Previous research has shown that neighborhood disadvantage is associated with higher rates of pneumonia, methicillin-resistant Staphylococcus aureus (MRSA), and bacteremia.21–23 This is likely due to a variety of different acute and chronic stressors, which place patients at higher risk of disease. More disadvantaged regions may have increased exposure to harmful pathogens and challenge a person’s immune system to recover after infection. 24 Furthermore, more disadvantaged patients have a higher risk of readmission following treatment for sepsis, 25 indicating that patients in higher ADI neighborhoods may lack access to care or resources for follow-up. This trend holds true for surgical patients as well. Higher ADI is associated with an increased risk of requiring emergent colorectal surgery and lower rates of long-term follow-up following vascular trauma.26,27
In our study, ADI was not associated with an increased risk of infection during acute burn hospitalization. This indicates that higher ADI alone may not increase patient risk sufficiently to overcome other risk factors for infection in burn patients. Our burn center treats all patients by a standardized, evidence-based protocol, likely contributing to similar infection rates across ADI groups. Unlike other hospitalized patients, burn patients receive daily wound inspection/care, universal contact precautions, and early surgical debridement, contributing to the substantial improvement in wound-related infections. 28 To the best of our knowledge, this is the first study evaluating the association between ADI and infections in burn patients. Previous studies have shown higher ADI associated with increased %TBSA burns, inhalation injury, and injury severity.9,14 These are similar risk factors for infection, and although our multivariable analysis did not show a significant difference, higher ADI burn patients may be at risk of more severe injuries.
Although ADI was not associated with overall infection, the unhoused state was associated with an increased risk of resistant infections on multivariate analysis and wound infections on univariate analysis. The current understanding of socioeconomic status as a risk factor for colonization with resistant organisms is limited. Antibiotic-resistant infections are a global health concern, with antibiotic resistance rates rising worldwide. 29 A systemic review by Alividza et al. shows a complex relationship between housing, income, education, water quality, and rates of resistant organisms 30 ; however, it remains unclear if unhoused patients harbor skin microflora or a microbiome predisposing them to resistant infections. It is also possible that unhoused patients who frequent emergency rooms more often have increased exposure to antibiotics or to the resistant organisms often found in healthcare settings. Being unhoused is also associated with a strongly suppressed immune response, placing these patients at higher risk of morbidity and mortality from infection. 31 Currently, no data supports systemic antibiotic prophylaxis in burn patients. 32 However, in this vulnerable population, alternative strategies for infection monitoring, empirical antibiotic choice, or even prophylaxis in specific circumstances warrant further investigation.
This study has several limitations to acknowledge, including those inherent to its single-center retrospective design, such as reporting bias, misclassification, and missing data. The burn registry attempts to mitigate this by using trained clinical data extractors, and additional data were collected retrospectively via chart review. Our study population encompasses a diverse population across Southern California; still, it does not include patients hospitalized or managed at other institutions. Therefore, the results may not be generalizable to patients managed outside burn centers or in different geographic regions. Additionally, we did not compare the rate of infection between burn patients and control patients. Our overall sample size was heavily skewed toward housed patients, and thus, the small sample size of unhoused patients increases the risk of a type II error in our results. However, given the single-center design, all patients received the same standard of care in a robust and well-trained burn unit.
Conclusions
The findings of this small single-center retrospective study revealed that socioeconomic status as measured by ADI was not clearly associated with an increased risk of infection among burn patients; however, the unhoused state was a risk factor for both wound infection and resistant infections. Unhoused patients appear to be uniquely different than housed patients when considering risk of infection; this may be due in part to increased exposure to resistant microflora, greater prior antibiotic use inducing resistance, or more compromised immunity, though further research is needed to elucidate specific risk factors. We also noted that the patient-specific comparison of housed versus unhoused was more accurate when predicting infections than census-level data as determined by ADI. Ultimately, our results suggest that unhoused patients may require different empirical antibiotics when infections develop, in addition to protocol-driven, evidence-based burn care.
Footnotes
Authors’ Contributions
S.S. and J.E.S. contributed to the research design, data analysis and interpretation, and the writing of the abstract/article. L.P. contributed to the design and implementation of the research, data acquisition and analysis, and interpretation of the results. J.E.S., L.N.H., J.G.L., E.S., T.W.C., and J.J.D. contributed to the research idea, data interpretation, and critical revision of the article. A.E.B. contributed to the design and implementation of the research, data analysis and interpretation, drafting of the article, and final approval of the version to be published.
Author Disclosure Statement
The authors report no conflicts of interest, financial or otherwise.
Funding Information
This research received no specific grant from funding agencies in the public, commercial, or not-for-profit sectors.