Geographic Variation in Outcomes After Burn Injury

Abstract

Background:

Geography is an important yet underexplored factor that may influence the care and outcomes of burn survivors. This study aims to examine the impact of geography on physical and psychosocial function after burn injury.

Methods:

Data from the Burn Model Systems National Database (1997–2015) were analyzed. Individuals 18 years and older who were alive at discharge were included. Physical and psychosocial functions were assessed at 6, 12, and 24 months postinjury using the following patient-reported outcome measures: Community Integration Questionnaire, Physical Composite Scale and Mental Composite Scale of the 12-Item Short Form Health Survey, Satisfaction with Appearance Scale, and Satisfaction with Life Scale. Descriptive statistics were generated for demographic and medical data, and mixed regression models were used to assess the impact of geography on long-term outcomes.

Results:

The study included 469 burn survivors from the Centers for Medicare and Medicaid Services regions 10, 31 from region 8, 477 from region 6, 267 from region 3, and 41 from region 1. Participants differed significantly by region in terms of race/ethnicity, burn size, burn etiology, and acute care length of stay (P < 0.001). In adjusted mixed model regression analyses, scores of all 5 evaluated outcome measures were found to differ significantly by region (P < 0.05).

Conclusions:

Several long-term physical and psychosocial outcomes of burn survivors vary significantly by region. This variation is not completely explained by differences in population characteristics. Understanding these geographical differences may improve care for burn survivors and inform future policy and resource allocation.

He cost, utilization, and outcomes of health care vary widely across the United States.^1–3 Over the past few decades, many have debated the importance of geography in accounting for these differences.^4–6 A 2013 study by the Institute of Medicine found that the highest spending regions (top 10%) in the United States spent 1.7 times more than the lowest spending regions (bottom 10%).⁶ Residents of the higher spending regions received up to 60% more care than those in lower spending regions.^1,4 Furthermore, according to a recent study of 22 million inpatient admissions, geographic variability in outcomes exceeds variability in costs and is not completely explained by population, comorbidity, and health system characteristics.⁷

Geography is an important factor when considering the outcomes of specific patient populations. For example, mortality after penetrating trauma differs by region, with patients in the Northeast having greater chances of survival than those living in other regions of the United States.⁸ There is also significant geographic variation in mortality and spending after traumatic brain injury.⁹ The western United States has lower mortality rates and higher costs after traumatic brain injuries, whereas the southeast United States has the highest mortality and mixed costs.⁹ In addition, the likelihood of employment for adults with spinal cord injuries varies with geography; those living in more urban areas and areas of higher socioeconomic status (SES) are more likely to find jobs.¹⁰ Beyond these studies, research investigating the relationship between traumatic injury, geography, and long-term outcomes is limited.

Researchers have started to evaluate the impact of geography on burn populations. The existing literature focuses primarily on geographic variation in access to care. As of 2017, there were 134 self-identified burn centers in the United States.¹¹ Of these, 66 met the rigorous standards for “organizational structure, injury prevention and education, qualifications and training of personnel, facilities, and resources” required for burn center verification by the American Burn Association (ABA) and the American College of Surgeons.^11,12 Verified burn centers are located in 31 states,¹² and proximity to these centers varies greatly across the United States.¹³ Access by ground and air transport to both self-identified and verified burn centers is highest in the Northeast and lowest in the South.¹³ The number of beds for burn patients also varies by region, with the highest number of beds in the Midwest and the lowest number in the South.¹³ Delayed transfer to a verified burn center has been associated with longer lengths of stay and higher rates of infectious complications in pediatric burn patients.¹⁴ This observed variation in access and likely also in transport times may have important implications for burn survivors.

Despite this previous work in burn patients, literature focusing on the influence of geography on long-term outcomes is limited. With improved mortality rates, there is greater appreciation that survival alone is not enough.¹⁵ Long-term outcomes and quality of life after burn injury are becoming increasingly important to both patients and providers. Given the current lack of data on the influence of geography on long-term outcomes, this study aims to examine regional differences in physical and psychosocial function after burn injury as measured by several patient-reported outcome measures.

METHODS

Data Source

A retrospective review of the prospectively collected Burn Model System (BMS) National Database between 1997 and 2015 was performed. Six burn centers in the United States have contributed to this database over the lifespan of the project (1994 to present).¹⁶ This study used data from 5 of these 6 centers: the Northwest Regional Burn Model System in Seattle, WA (1994 to present), the North Texas Burn Rehabilitation Model System in Dallas, TX (1994 to present), the University of Texas Medical Branch Shriners Hospitals for Children—Galveston Model System in Galveston, TX (1997 to present), the Boston-Harvard Burn Injury Model System in Boston, MA (2012 to present), and the Johns Hopkins Burn Model System in Baltimore, MD (1997–2012). Participants 18 years and older who were alive at the time of acute care discharge and had follow-up data at 6, 12, or 24 months postinjury were included. Data collection was chosen to end in 2015 to maximize the amount of 24-month follow-up data of the most recently enrolled participants.

Modifications were made to the inclusion criteria over time. Details of the BMS National Database inclusion criteria, data collection process, and data collection sites can be found at http://burndata.washington.edu/. For this period, the BMS National Database included data on survivors who met the ABA criteria for a severely burned person, namely, those with 1 or more of the following:

• Deep 2nd and 3rd degree burns greater than 10% total body surface area (TBSA) in patients older than 50 years

• Deep 2nd and 3rd degree burns greater than 20% TBSA

• Deep 2nd and 3rd degree burns with serious threat of functional or cosmetic threat that involve face, hands, feet, genitalia, perineum, or major joints

• Third degree burns greater than 5% TBSA

• Deep electrical burns including lightning injury

• Burn injury with inhalation injury

• Circumferential burns of the extremity or chest

The BMS National Database uses Research Electronic Data Capture tools hosted at the BMS National Data and Statistical Center at the University of Washington. Research Electronic Data Capture is a secure, web-based application designed to support data capture for research studies.¹⁷

Variables

Demographic and clinical variables were collected through self-report or medical record abstraction at discharge and included the following: age, sex, race/ethnicity, preinjury employment status, burn etiology, burn size, acute care length of stay, presence of inhalation injury, and government insurance status.

In preliminary analyses, a revised breakdown of the ABA regional map was used to divide the study population into 3 geographic groups: Northeast, South, and West. However, these groups were defined broadly, and additional granularity was needed to better identify geographic differences. The Centers for Medicare and Medicaid Services (CMS) regions¹⁸ were subsequently chosen to assess geographic variation in the study population. Regions with fewer than five participants were not included in this study. Region assignment was based on residence at the time of burn.

Outcomes

Physical and psychosocial functions were assessed at 6, 12, and 24 months postinjury using the following patient-reported outcome measures:

1. Community Integration Questionnaire (CIQ) — Social Integration: The CIQ provides a measure of an individual’s level of community integration. The overall 15-item questionnaire is divided into 3 subscales including: (1) questions about home activities, (2) questions about social integration, and (3) questions about educational, vocational, or other productive activities outside the home. The social integration subscale, used in the current study, consists of 6 questions, with scores ranging from 0 to 12. Higher scores on this subscale indicate greater social integration.¹⁹ The CIQ has previously been validated in the adult burn population.²⁰

2. The 12-Item Short Form Health Survey (SF-12): The validated SF-12 survey was created as a shorter version of the 36-Item Short Form Health Survey to measure health status and well-being.²¹ The SF-12 includes 2 subscores: the Physical Component Scale (PCS) and the Mental Component Scale (MCS). Using a t score transformation, PCS and MCS scores are standardized to a mean of 50 and standard deviation of 10, with a maximum of 100, based on a US population.²² Scores greater than 50 represent above average health status.

3. Satisfaction with Appearance (SWAP): The 14-item SWAP Scale is a valid and reliable tool used to determine satisfaction with appearance in the burn population.²³ Participants are asked to rate how well each item describes their thoughts and feelings about their postburn appearance. Each item is rated on a 7-point scale, with 1 indicating strongly disagree and 7 indicating strongly agree. Total SWAP scores are calculated by subtracting 1 from each item score and totaling scores for all 14 items. Total scores range from 0 to 84, with higher scores indicating greater dissatisfaction with appearance and body image postinjury.

4. Satisfaction with Life (SWL): The SWL Scale is a reliable and valid measure of self-perceived quality of life.²⁴ Psychometric evaluation in the spinal cord injury, traumatic brain injury, and burn injury populations has shown the instrument to be useful in evaluating trauma outcomes.²⁵ The SWL Scale includes 5 items, each scored on a 7-point scale, with 1 indicating strongly disagree and 7 indicating strongly agree. Total scores range from 5 to 35; higher scores represent greater life satisfaction.

Data Analysis

Descriptive statistics were generated for demographic and medical data. Mean outcome scores for each region were calculated at 6, 12, and 24 months postinjury. Omnibus t tests were used to determine the overall importance of geography as a predictor of long-term outcomes. Because of concerns about multiple testing, a Bonferroni correction to the α level of 0.05 was used, resulting in P values of <0.005 for demographic and medical data and 0.003 for unadjusted outcome scores for statistical significance.

Mixed regression models were used to assess the impact of geography on outcomes, controlling for demographic and clinical factors. This methodology was chosen because mixed models allow for estimation of individual and population trends over time, as well as the use of uneven follow-up intervals and time-dependent and time-independent covariates.²⁶ In addition, mixed models use all available data on each subject and are unaffected by randomly missing data.²⁶ This approach also avoids problems with multiple comparisons that would be encountered when using separate regressions. A mixed model was created for each standardized outcome measure (CIQ, SF-12 PCS, SF-12 MCS, SWAP, and SWL) and included the following clinically important variables: age, sex, TBSA burned, burn etiology, race/ethnicity, length of hospital stay, presence of inhalation injury, government insurance status, employment status at time of injury, time since injury, and time since the beginning of the BMS National Database.

Standardized coefficients were used to facilitate comparisons across each of the 5 outcome measures. Region 10 was chosen to be the reference group for all mixed models. If the interaction term (time since injury by region) was not significant, it was removed from the model and the model was recalculated. A P value of 0.05 was considered significant for the mixed regression models. For outcome variables that were found to be significantly different by region, raw scores were calculated to determine if these statistical differences were clinically significant. In nonburn US populations, minimal clinically important differences have been found to range from 3.29²⁷ to 8.8²⁸ for the SF-12 PCS and from 3.77²⁷ to 9.3²⁸ for the SF-12 MCS. Based on author consensus, values within these ranges were considered clinically significant for the SF-12 PCS and SF-12 MCS for the purposes of this article. Minimal clinically important differences have not been established for the CIQ, SWAP, or SWL scales. All analyses were performed using STATA version 15.0 (College Station, TX).

To examine potential changes in overall demographics and outcomes over the course of the approximately 20-year study period, 2 subsets of the data (one from 1994 to 1999 and one from 2010 to 2015) were compared using nonparametric t tests (Wilcoxon-Mann-Whitney tests) and χ² tests to assess for differences in demographic, clinical, and patient-reported outcome data. The significance level was adjusted using Bonferroni method (P = 0.0023).

RESULTS

The study included participants from 5 CMS regions: 41 from region 1 (CT, ME, MA, NH, RI, VT), 267 from region 3 (DE, DC, MD, PA, VA, WV), 477 from region 6 (AR, LA, NM, OK, TX), 31 from region 8 (CO, MT, ND, SD, UT, WY), and 469 from region 10 (AK, ID, OR, WA) (Fig. 1). Participants differed significantly by region in terms of race/ethnicity, burn size, burn etiology, and acute care length of stay (P < 0.001; Table 1). Burn size ranged from 12% to 24.3% TBSA, and acute care length of stay ranged from 20.5 to 34.9 days. The proportion of White, non-Hispanic participants ranged from 58.1% to 79.1%, and the percentage of participants who sustained fire/flame injuries ranged from 34.2% to 71.0%. Although not statistically significantly different, the percentage of male participants ranged from 65.5% to more than 80%, and the percentage of participants who were employed before their injury ranged from 64% to 77.4%. The percentage of individuals receiving government-sponsored health insurance (Medicare or Medicaid) ranged from 16.6% to 25.9%.

FIGURE 1.

Number of study participants by CMS Region.

TABLE 1.

Demographic and Medical Characteristics of the Study Population

Variable	Region 10 (AK, ID, OR, WA)	Region 8 (CO, MT, ND, SD, UT, WY)	Region 6 (AR, LA, NM, OK, TX)	Region 3 (DE, DC, MD, PA, VA, WV)	Region 1 (CT, ME, MA, NH, RI, VT)	P
No. subjects	469	31	477	267	41
Age, mean (SD)*	44.2 (14.6)	41.6 (15.7)	42.5 (15.1)	42.3 (15.1)	49.5 (15.7)	0.0179
Male, % (n)†	72.9 (342)	80.7 (25)	75.5 (360)	65.5 (175)	73.2 (30)	0.046
TBSA burned, mean (SD)*	18.4 (15.8)	24.3 (16.8)	24.1 (16.8)	13.9 (14.7)	12.5 (13.8)	<0.001
Length of stay, mean (SD)*	32.1 (25.6)	34.9 (20.1)	29.4 (25.3)	21.2 (21.0)	20.5 (17.2)	<0.001
Preinjury employment status, % (n)†
Working	64.0 (299)	77.4 (24)	74.6 (356)	67.4 (180)	65.9 (27)	0.007
Not working	36.0 (168)	22.6 (7)	25.4 (121)	32.6 (87)	34.2 (14)
Race/ethnicity, % (n)‡
White, non-Hispanic	79.1 (371)	90.3 (28)	58.1 (277)	61.5 (163)	92.7 (38)	<0.001
Non-White (Black, Hispanic, other)	20.9 (98)	9.7 (3)	41.9 (200)	38.5 (102)	7.3 (3)
Burn etiology, % (n)†
Fire/flame	61.6 (289)	71.0 (22)	66.9 (319)	50.9 (136)	34.2 (14)	<0.001
Other	38.4 (180)	29.0 (9)	33.1 (158)	49.1 (131)	65.9 (27)
Insurance, % (n)†
Medicare/Medicaid	25.7 (96)	25.9 (7)	18.5 (68)	16.6 (32)	33.3 (9)	0.025
Other	74.3 (278)	74.1 (20)	81.5 (300)	83.4 (161)	66.7 (18)
Inhalation injury, % (n)‡	11.4 (53)	16.1 (5)	10.3 (49)	14.2 (37)	12.2 (5)	0.479

Omnibus t tests were used to identify any differences in demographic and clinical factors by geographical region. One-way analysis of variance, χ², and Fisher exact tests were used for omnibus determination of difference between groups. The significance level was adjusted with the Bonferroni method to P = 0.005.

^*One-way analysis of variance test.

^†χ² test.

^‡Fisher exact test.

In unadjusted analyses, SF-12 PCS scores differed significantly by geographical region at 6 and 12 months postinjury (P < 0.003). The SWAP scores also differed significantly by region at 6 and 24 months postinjury (P < 0.003; Table 2). The CIQ, SF-12 MCS, and SWL scores did not differ by region at any follow-up time point.

TABLE 2.

Unadjusted Outcome Scores by Region at 6, 12, and 24 Months Postinjury

Outcome	Region 10	Region 8	Region 6	Region 3	Region 1	P
CIQ
6 mo	8.0 (2.5)	8.4 (2.6)	7.5 (2.5)	7.9 (2.5)	8.3 (2.2)	0.0124
12 mo	8.0 (2.5)	8.8 (2.2)	7.5 (2.4)	7.9 (2.4)	7.8 (2.5)	0.0045
24 mo	8.1 (2.5)	8.7 (2.3)	7.6 (2.5)	7.8 (2.4)	7.4 (2.2)	0.0155
SF-12 PCS
6 mo	45.3 (10.9)	46.1 (11.2)	40.9 (10.5)	43.4 (11.2)	43.3 (12.4)	<0.001
12 mo	46.7 (11.1)	48.2 (11.1)	43.2 (10.9)	44.4 (11.2)	45.3 (11.3)	<0.001
24 mo	47.5 (10.8)	49.7 (11.1)	44.9 (10.8)	45.9 (10.4)	46.1 (10.2)	0.0149
SF-12 MCS
6 mo	48.6 (11.6)	50.0 (7.9)	46.6 (11.9)	46.2 (13.8)	48.6 (11.7)	0.0425
12 mo	48.5 (11.8)	53.0 (6.4)	46.4 (12.8)	47.3 (12.4)	47.3 (9.9)	0.025
24 mo	49.1 (11.7)	51.2 (10.9)	46.5 (12.4)	47.1 (11.5)	41.4 (12.7)	0.0072
SWAP
6 mo	28.0 (17.3)	25.0 (21.7)	32.4 (18.7)	26.1 (19.0)	26.9 (17.6)	0.0025
12 mo	27.8 (17.9)	22.1 (17.0)	32.0 (19.8)	25.9 (18.9)	31.3 (20.2)	0.0039
24 mo	26.1 (17.8)	16.4 (11.7)	31.8 (19.7)	27.0 (18.6)	40.5 (10.6)	<0.001
SWL
6 mo	20.8 (8.4)	22.9 (8.3)	20.8 (8.5)	19.9 (8.2)	21.5 (8.4)	0.3955
12 mo	21.5 (8.6)	23.7 (7.7)	20.7 (8.8)	19.7 (8.3)	20.9 (8.1)	0.0852
24 mo	22.1 (8.4)	24.8 (7.4)	21.3 (8.8)	21.0 (7.9)	19.4 (8.3)	0.1654

All scores are presented as mean (SD). Omnibus t test was used to determine the importance of geographical region as a predictor of long-term outcomes. The significance level was adjusted with the Bonferroni method to P = 0.003.

In adjusted mixed model regression analyses, CIQ, SF-12 PCS, SF-12 MCS, SWAP, and SWL scores were found to differ significantly by region (P < 0.05; Table 3). Compared with those from region 10, burn survivors from regions 3 and 6 had significantly lower PCS scores (raw scores, 41.41 vs 46.4 [P < 0.001] and 41.63 vs 46.4 [P = 0.009], respectively). These differences in PCS scores were also clinically significant. Burn survivors from region 6 had significantly lower MCS (raw scores, 45.79 vs 48.7; P = 0.02) and CIQ (raw scores, 7.33 vs 8.0; P = 0.002) scores compared with those from region 10. Differences in MCS scores were clinically significant. Burn survivors from region 1 reported significantly poorer SWAP scores (raw scores, 36.24 vs 27.4; P = 0.022) and survivors from region 3 had significantly lower SWL scores (raw scores, 19.47 vs 21.4; P = 0.029) compared with those from region 10. Furthermore, the interaction term (time since injury by region) was significant for the SF-12 PCS scale: time since injury was associated with improved PCS scores for burn survivors in region 3 compared with those in region 10 (P = 0.002).

TABLE 3.

Mixed Models Examining Geography as a Predictor of Long-term Outcomes

Outcome	Standardized Coefficient	P
CIQ
Region 8	0.23	0.153
Region 6	−0.19	0.002
Region 3	−0.05	0.477
Region 1	−0.06	0.686
SF-12 PCS
Region 8	0.14	0.593
Region 6	−0.27	<0.001
Region 3	−0.29	0.009
Region 1	−0.13	0.193
SF-12 MCS
Region 8	0.15	0.347
Region 6	−0.14	0.020
Region 3	−0.07	0.331
Region 1	−0.31	0.059
SWAP
Region 8	−0.32	0.095
Region 6	0.07	0.325
Region 3	−0.15	0.125
Region 1	0.38	0.022
SWL
Region 8	0.24	0.174
Region 6	−0.05	0.438
Region 3	−0.18	0.029
Region 1	−0.23	0.170

CMS region 10 was chosen to be the reference group for all mixed models. Each model controlled for age, sex, TBSA burned, burn etiology, time since injury, length of hospital stay, presence of inhalation injury, government insurance status, employment status at time of injury, race/ethnicity, and time since the beginning of the BMS National Database. A P value of 0.05 was used for statistical significance.

When examining differences in subsets of participants enrolled in the study from 1994 to 1999 and 2010 to 2015, age (40.8 vs 45.2 years; P = 0.0003), race/ethnicity (76.3% vs 64.5% White-non-Hispanic; P = 0.002), and insurance (11.5% vs 28.9% Medicare/Medicaid; P < 0.0001) were significantly different between groups. All other demographic and clinical factors as well as all examined outcomes did not exhibit significant differences.

DISCUSSION

Although previous work has investigated the relationship between geography and access to burn care,¹³ little attention has been given to the impact of geography on long-term physical and psychosocial functioning after burn injury. To the best of our knowledge, this is the first study to evaluate geographic differences in long-term outcomes of burn survivors.

This population of burn survivors differed significantly by region in terms of race/ethnicity, burn size, burn etiology, and acute care length of stay. Given these findings, it would seem reasonable to attribute the observed variation in outcomes to differences in patient populations. However, even after controlling for these and other factors in mixed model regression analyses, significant variation in physical and psychosocial outcomes remained. Clinically significant variation in 2 of the outcomes of interest, the SF-12 PCS and SF-12 MCS, was identified. Based upon utilization and cost data from the Veterans Health Administration, it has been determined that those scoring 10 points (or one standard deviation) lower on the PCS require an additional US $1462 per patient per year.²⁹ Those scoring 10 points lower on the MCS require US $864 more per patient per year.²⁹ Thus, this geographic variation in physical and mental health outcomes of burn survivors may have important implications for overall costs and resource allocation.

Research in other fields has begun to demonstrate similar geographic variation in long-term outcomes. Employment, a key area of social recovery, has been shown to vary by geography for individuals with spinal cord injury.¹⁰ In addition, a multicenter study of the health status of individuals with chronic obstructive pulmonary disease found statistically and clinically significant regional variation in physical and emotional functioning, as measured by the 36-Item Short Form Health Survey,³⁰ a precursor to the SF-12. Other studies have found significant regional variations in the health status and functioning of Medicare beneficiaires³¹ and veterans³² that were independent of comorbid conditions and lifestyle characteristics. Beyond these, existing literature primarily focuses on geographic variation in the cost and utilization of health care.^6,7 The present study is one of few to begin exploring regional differences in long-term outcomes in the United States.

The unexplained variation in outcomes observed in this study suggests that other unmeasured factors influence long-term outcomes after burn injury. One such factor may be burn practice variation. Through publication of its Practice Guidelines for Burn Care, the ABA has established a set of recommendations for the treatment of burn survivors.³³ Additional guidelines have been proposed based on expert consensus and supplemental data on quality of care.^34,35 Despite these standardization efforts, variations in processes of burn care are prevalent and often reflect clinician’s personal preferences and training.³⁶ A recent study of the Shriners Hospitals for Children burn care system found substantial variation of process indicators between Shriners hospitals; although all centers fulfilled requirements for the American College of Surgeons Committee on Trauma Burn Center Verification Program, 89% of process indicators differed significantly across sites, and adherence to these process indicators ranged from 34% to 93%.³⁷ Although the relationship between process indicators and outcomes has not yet been determined, practice variation may influence burn recovery, and its impact on geographical differences in long-term burn outcomes represents a key area for future research.

As discussed earlier, access to verified burn centers, as well as the number of available beds for burn patients, varies significantly by geographical region.¹³ Those living in rural areas may experience more difficulty accessing specialized burn care than their urban counterparts. Research in other US populations has shown that rural individuals report less utilization of and more unmet needs for medical care than those living in urban areas.^38,39 Furthermore, patients in rural areas are less likely to receive specialized care.³⁹ With respect to specialized burn care, mortality rates at verified burn centers are comparable with rates at nonverified centers⁴⁰; however, literature evaluating outcomes and quality of burn care by region and verification status is limited. Research from other disciplines has shown that trauma care provided at level I trauma centers is associated with better outcomes compared with care at nondesignated trauma centers.⁴¹ The observed differences in access to specialized burn care may also be related to regional variation in SES. Patients of low SES face significant barriers to accessing health care, including lack of insurance coverage and unaffordable costs.⁴² Individuals living in rural areas are more likely to have lower incomes⁴³ and to be uninsured,⁴⁴ and thus, urbanicity may also impact health outcomes in this way. Numerous studies have shown that low SES is a strong predictor of increased disease risk, morbidity, and mortality.^45,46 Within the burn population, the incidence and severity of injury are higher among low SES individuals.⁴⁷ Furthermore, SES may be a surrogate marker of other factors that influence long-term outcomes, including education, transportation, social and community support, diet, and physical activity, among others.^48,49 The present study used race/ethnicity, employment status, and government insurance status as proxy measures of SES. Race/ethnicity was found to differ significantly by region, with region 6 having the largest percentage of non-White burn survivors. In mixed regression models, region 6 demonstrated worse CIQ, SF-12 PCS, and SF-12 MCS scores compared with region 10. Further investigation is needed to determine the relationship between geography, SES, and long-term outcomes after burn injury.

Several other factors may also influence the relationship between geography and long-term burn outcomes. Variation in the availability and use of post–acute care services may help explain some of the observed geographical differences in outcomes for burn survivors. Previous research has shown that the use of inpatient rehabilitation after hospitalization for burn injury varies significantly by state.⁵⁰ The present study was not able to evaluate the impact of post–acute care on geographic variation in burn outcomes; further research is needed. Contextual factors, such as local availability of resources, culture, economy, and communication, may also influence differences in outcomes for burn survivors; however, literature in these areas is extremely limited.

Study Limitations

Several study limitations must be acknowledged. Despite using CMS groupings, geographic regions were defined broadly, with 1 or 2 BMS centers contributing to each group. This study included participants from 18 different states, and several BMS centers contributed data to more than 1 region. Previously published literature has shown the BMS database to be demographically similar to the National Burn Registry; however, we recognize that these participants may not be entirely representative of all burn survivors in their respective geographical region.⁵¹ Furthermore, this study did not include participants from CMS regions 2 (NJ, NY, Puerto Rico, Virgin Islands), 4 (AL, FL, GA, KY, MS, NC, SC, TN), 5 (IL, IN, MI, MN, OH, WI), 7 (IA, KS, MO, NE), or 9 (AZ, CA, HI, NV, Pacific Territories). As per the BMS National Database eligibility criteria, participants included in this study were treated at a BMS center, all of which are verified burn centers. Thus, this study may not be representative of those burn survivors who were treated at nonverified centers. Given the limited number of participating sites (5), it was difficult to distinguish the effects of geography versus practice variation. Future studies would benefit from the inclusion of practice indications as well as additional burn centers from numerous states. In the mixed regression models, region 10 was chosen to be the comparison group. Further between-region analysis was limited by the sample sizes of several regions, namely, those of regions 1 and 8 (41 and 31, respectively); a larger number of individuals from these regions would be needed to detect any statistically significant differences in population characteristics and outcomes. In addition, the minimal clinically important differences for several of the measures used in this study were based on nonburn populations, because they have not been defined in the adult burn population. All follow-up data were collected through self-report and are subject to reporting bias. Lastly, given the 20-year duration of the study, there is a potential for a drift in characteristics of the population over time. However, this was assessed by comparing individuals enrolled during the first and last 5 years of the study, and all variables, with the exception of age, race/ethnicity, and insurance, were not significantly different. Despite these limitations, the impact of geography on long-term outcomes of burn survivors is an underexplored area. The BMS National Database is one of few longitudinal databases of burn recovery and contains unique data on multidimensional, long-term outcomes of burn survivors. This study represents an important first step toward better understanding geographical differences in burn outcomes.

CONCLUSIONS

Several long-term physical and psychosocial outcomes of burn survivors vary significantly by region. This variation is not completely explained by differences in population characteristics. Additional research is needed to better understand geographic trends in burn outcomes; however, this study represents a critical first step in identifying long-term care needs of US burn survivors. The results of this study may have important policy implications, because noted differences in burn outcomes may help guide resource allocation.