a b s t r a c t

Background: Pediatric burns account for 120,000 emergency department visits and 10,000 hospitalizations annu- ally. The American Burn Association has guidelines regarding referrals to Burn centers; however there is variation in burn center distribution. We hypothesized that disparity in access would be related to burn center access. Methods: Using weighted discharge data from the Nationwide Inpatient Sample 2001-2011, we identified pedi- atric patients with International Classification of Diseases-9th Revision codes for burns that also met American Burn Association criteria. Key characteristics were compared between pediatric patients treated at burn centers and those that were not.

Results: Of 54,529 patients meeting criteria, 82.0% (n = 44,632) were treated at burn centers. Patients treated at burn centers were younger (5.6 versus 6.7 years old; p b 0.0001) and more likely to have burn injuries on mul- tiple body regions (88% versus 12%; p b 0.0001). In urban areas, 84% of care was provided at burn centers versus 0% in rural areas (p b 0.0001), a difference attributable to the lack of burn centers in rural areas. Both length of stay and number of procedures were significantly higher for patients treated at burn centers (7.3 versus 4.4 days, p b 0.0001 and 2.3 versus 1.1 procedures, p b 0.0001; respectively). There were no significant differences in mortality (0.7% versus 0.8%, p = 0.692).

Conclusion: The majority of children who met criteria were treated at burn centers. There was no significant dif- ference between geographical regions. Of those who were treated at burn centers, more severe injury patterns were noted, but there was no significant mortality difference. Further study of optimal referral of pediatric burn patients is needed.

(C) 2019

Introduction

Each year in the United States (U.S.), pediatric burns comprise ap- proximately 120,000 Emergency Department (ED) visits and 10,000 hospitalizations [1,2] Over the past several decades, advances in critical care, fluid resuscitation, surgical technique, skin substitutes, and nutri- tion have led to significant decreases in burn-related mortality in the United States [3]. Many of these advances stem from the advent of spe- cialty burn centers developed during World War II [4]. These centers provide a single source of care for the entire spectrum of burn treat- ment, including initial assessment and triage, resuscitation, wound cov- erage, surgical care, and rehabilitation. Furthermore, the American Burn

? This paper was presented as a poster at the 2016 American Association for the Surgery of Trauma National Conference, Honolulu, HI.

* Corresponding author at: University of California San Francisco, East Bay Department of Surgery, 1411 East 31^st Street, Oakland, CA 94602, United States of America.

E-mail address: [email protected] (C. Ewbank).

Association (ABA), founded in 1976, has developed national guidelines for burn center practices, and has put forth well-accepted criteria delin- eating which burns require specialized treatment at a burn center [5].

Burn centers require a high level of expertise and resources. Because of this, they are generally situated in urban areas, which results in nearly one-third of the United States population living N2 h by ground trans- port from a burn center [6]. Additionally, despite the success of special- ized burn center care in decreasing mortality and morbidity, burn centers have decreased in number over the past few decades. From 1976 to 2011, the number of U.S. burn centers decreased from 180 to 123 [7]. As such, a large portion of burned patients initially present to emergency departments not affiliated with a burn center [8]. Although early transfer to a burn center has been shown to reduce complications and mortality for patients meeting ABA criteria [9,10], many patients are never transferred [11-13].

Pediatric burns comprise a disproportionate percentage of all burn injuries in the United States, with Younger children affected most fre- quently [14]. Twenty-eight percent of all burns requiring hospitalization

https://doi.org/10.1016/j.ajem.2019.158401

0735-6757/(C) 2019

from 2001 to 2011 occurred in children [2]. Due to differences in burn type, cause, and physiological response, pediatric patients have worse outcomes after a severe Burn injury than adults [15]. Delays in resusci- tation, wound care, and grafting drive morbidity and mortality in se- verely burned children [3]. These factors have led to guidelines recommending a low threshold for transferring pediatric patients to qualified burn centers. To date, few studies have looked at burn center utilization among pediatric inpatients meeting ABA criteria for burn center referral [1].

The goal of this study was to conduct a population-wide study of pe- diatric patients meeting ABA criteria and thus recommended for care at a burn center, and to describe their access to care and the characteristics associated with treatment at a burn center or non-burn center. We hy- pothesized that under-referral of severely burned children to burn cen- ters would be associated with lack of access to a burn center rather than severity of burn and that children treated at non-burn centers would have worse outcomes relative to their injuries.

Methods

Study design and population

This was retrospective cohort study of pediatric burn patients meet- ing American Burn Association (ABA) guidelines for treatment at a burn center. This study used weighted inpatient admissions data from the Nationwide Inpatient Sample (NIS), Healthcare Cost and Utilization Pro- ject (HCUP), Agency for Healthcare Research and Quality (2001-2011). This database samples 20% of hospitals nationally, with sampling strata defined by census region, hospital ownership, urban-rural location, teaching status, and bed size [16]. A total of 103,425 patients ages 1 to 17 years old from 47 states were identified using the International Clas- sification of Diseases-9th Revision (ICD-9) discharge codes for burn in- juries (940.0-949.5). Infants b1 year have different patterns of burns that warrant special considerations and were thus excluded [17,18]. In- clusion and exclusion criteria are summarized in Fig. 1.

The NIS reports American Hospital Association (AHA) identification numbers, which can be linked with institutional information. We used a linkage file available for fiscal years 2008 and 2011 that designates burn center status for 31 of 47 states in the NIS. Burn center status is self-identified under this designation, and should not be confused

Fig. 1. Summary of Inclusion/exclusion criteria for this study. Data are from the Nationwide Inpatient Sample dataset, years 2001-2011. Patients ages 1-17 years, ICD-9 for burn injury (940.0-949.5). ICD-9 = International Classification of Diseases-9th Revision. AHA = American Hospital Association. ISS = Injury Severity Score. ABA = American Burn Association.

with ABA verification. Patients from the following 16 states without a linkable hospital identification number were excluded: Georgia, Idaho, Indiana, Kansas, Louisiana, Michigan, Mississippi, Nebraska, New Mexico, Ohio, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, and Wyoming (n = 31,147 (30.1%)). Additionally, Montana, North Dakota, and New Hampshire did not have burn centers in their state, and patients from these states were excluded from the study (n

= 1501 (1.5%)). Our final dataset included patients from 28 states (Fig. 2). Patient records from facilities missing burn center designation status were also excluded (n = 635 (0.6%)). Lastly, patients who were transferred out from the initial hospital of admission (n = 1426 (1.4%)) were excluded from the study to avoid potential duplicate discharges.

ABA criteria variable coding

The ABA guidelines recommend that certain burns be evaluated at regional burn centers [5] (Table 1). A categorical variable was created to describe patients meeting one or more of the listed ABA criteria. We obtained the diagnoses of total body surface area burned, body re- gion, and burn thickness from ICD-9, as well as E codes where pertinent. The International Classification of Diseases Programs for Injury Catego- rization (ICDPIC) were used to calculate an Injury Severity Score (ISS) from ICD-9 Diagnosis codes [19]. Of the 68,715 potential patients for our cohort, 54,529 met one or more ABA criteria and were included in our study.

Because patients with concomitant major trauma (ISS N 15) require trauma center stabilization before treatment at a burn center [20-22], they were excluded from this study (n = 398 (0.4%)). The total study population included 54,529 children.

Demographic and clinical variables

Variables of age, sex, race/ethnicity, primary payer, mechanism of in- jury, burn body location, total body surface area, burn depth, hospital length of stay, number of burn-related procedures performed, discharge disposition, hospital region, and hospital location were compared for patients treated at burn and non-burn centers.

Age was categorized within traditional age groupings defined as 1-4 years, 5-8 years, 9-12 years, and 13-17 years old. Race/ethnicity was defined as White, Black, Hispanic, and other (Asian, Native American, or other). Primary payer was defined as Public insurance (of which 99.8% were covered by Medicaid and 0.2% by Medicare), pri- vate insurance, self-pay, and other. Mechanism of injury (E code) in- cluded hot object/scald, fire/flame, electric (including Lightning injury), chemical, and other (when the mechanism of injury code did not match any of the aforementioned categories). The region of the body burned was determined through ICD-9 codes (e.g. “head, neck, and face” and “truck, back, and genitalia”), and was not mutually exclu- sive. We created an additional variable, “multiple burn regions” for pa- tients with burns to two or more discrete regions. Burn procedures were identified using the ICD-9 Clinical Classifications Software Services and Procedures Procedure Categories for burn wound debridement (ICD-9 CSS 169). Hospital region (Northeast, South, Midwest, and West) and hospital location (urban versus rural) were defined by the United States Census Bureau [23]. Discharge disposition was defined as routine, home health care, died in hospital, and other (including against medical advice, to law enforcement, and discharged alive to un- known destination).

Outcome measures

The primary outcome of interest was access and treatment at a burn center versus a non-burn center. Secondary outcomes of interest were length of stay, number of burn procedures performed, and discharge disposition.

Fig. 2. United States map showing number of weighted discharges per hospital for both non-burn centers (dark blue) and burn centers (red). The size of the each circle is proportional to number of weighted discharges for each hospital. Light blue states were included in the analysis; grey states did not have a hospital linkage variable in NIS and thus were not able to be linked, and were not included.

Theory/calculation

Univariate analysis of independent variables in the weighted sample was determined with an adjusted Wald test with an alpha set at 0.05. Univariate analysis compared patients meeting ABA criteria who were treated at burn versus non-burn centers. Patients with missing data for variables related to burn center referral were excluded from the re- gression analysis. Of note, Total body surface area (TBSA) Diagnostic codes were not reported for nearly half of patients meeting ABA criteria (n = 10,007). The exception to this exclusion strategy was the race cat- egory. Due to wide variability in collection of data on race in hospitals, and even entire states, patients with missing race data were included in a separate “missing” category for analysis (i.e. missing indicator method [24]).

A multivariable logistic regression was used to analyze lack of refer- ral to a burn center (factors associated with non-burn center care). We controlled for age group, gender, race/ethnicity, and hospital region. Burn location on body, multiple body regions burned, and having mul- tiple ABA criteria were included in the model as proxies of burn severity. Patients with incomplete data were excluded from the multivariable analysis in the same manner as in the univariate analysis (n = 12,795 (23.5%)), resulting in a multivariate model with 41,734 patients. Model goodness of fit was calculated using the Hosmer-Lemeshow test with p = 0.97. In the Hosmer-Lemeshow test, a significant p– value provides evidence of a poor fit, and thus, larger p-values indicate a better fit model [25]. All statistical analyses were performed using Stata 13 software [26]. The follow study was exempt from institutional review board given de-identified data.

Results

Of pediatric patients meeting ABA criteria for treatment at a burn center 44,632 (82%) were treated at a burn center, and 9897 (18%) were treated at a non-burn center. A summary of the number of patients meeting ABA criteria is provided in Table 1.

Demographic characteristics of patients treated at burn and non- burn centers are summarized in Table 2. Patients at burn centers were younger than those at non-burn centers (5.6 +- 0.2 years old versus 6.7 +- 0.2, p b 0.0001). Of patients treated at burn centers, 26,433

(59%) patients were aged 1-4 years old, compared to 5091 (51%) pa- tients at non-burn centers. Gender distribution of patients treated at burn and non-burn centers was 3703 (37%) and 16,128 (36%) female patients at burn and non-burn centers, respectively (p = 0.49). Patient race/ethnicity was also not significantly associated with treatment cen- ter type, although data on race/ethnicity were noted to be recorded in- consistently in the NIS and were unavailable for 23% of patients. Primary payer (e.g. public, private, self-pay, and other) did not vary significantly between burn and non-burn center patients. The largest payer in both cohorts was public insurance followed by private insurance. Self-pay was relatively infrequent in both groups.

This study includes a total of 84 burn centers and 598 non-burn cen- ters (Fig. 2). All 84 burn centers were located in urban settings. Despite geographical variation in burn center distribution [6,13], there were no significant differences in the proportion of patients treated at burn and non-burn centers in each United States region (Northeast, Midwest, South, West). The average volume of pediatric burn patients admitted per year was 44 at the burn centers compared to 1.4 at the non-burn centers. Children were treated at 429 urban and 169 rural non-burn centers. Of note, 88% percent of all non-burn center patients were treated in urban hospitals, despite ready access to burn centers in most of the urban areas sampled.

The characteristics of the burns treated at burn versus non-burn cen- ters were significantly different (Table 3). While the entire study popu- lation was defined as patients meeting ABA criteria for burn center treatment, patients treated at a burn center met a greater number of ABA referral criteria than those treated at a non-burn center (1.53 +- 0.03 versus 1.37 +- 0.02, p b 0.0001). Mechanism of injury varied be- tween the two treatment center types. Patients with scald injuries were more likely to be treated at burn centers than non-burn centers (23,579 (53%) versus 3234 (33%) patients, p b 0.0001). Both center types were equally likely to treat flame injuries (16% for both burn

Table 1

American Burn Association burn center referral criteria (1); associated ICD-9 codes and numbers of patients. Per ABA recommendations, burn injuries meeting criteria 1-7, 9, and 10 warrant transfer to a burn center. Burns meeting criteria 8 warrant trauma center stabilization prior to burn center treatment (2). These criteria are defined through ICD-9 codes, E codes, and V codes in the National Inpatient Sample dataset, listed in the table above. ISS is calculated from ICD-9 codes pertaining to trauma and ranges on a scale of 0 to 75.

American Burn Association criteria Variable coding Patients (n)

partial thickness burns N10%

total body surface area (TBSA). ICD-9 948.1-948.99 5306

ICD-9 940-940.9, 941-941.59,

(3358 patients, 34%). Total body surface area (TBSA) data were evalu- ated, and the only significant factor was whether a patient had greater or b40% involvement. Burn centers treated 38,878 (89%) and non- burn centers treated 4924 (11%) of patients with b40% TBSA (p b 0.0001). Among patients with N40% TBSA, 639 (89%) were treated at burn centers and 81 (11%) were treated at non-burn centers. Presence of burns for all distinct body regions were significantly greater at burn centers due to increased prevalence of multiple regions burned.

Hospital course and discharge disposition also varied significantly between pediatric burn patients meeting ABA criteria at burn and non-burn centers (Table 4). Burn center patients had a significantly lon- ger length of hospital stay (7.3 +- 0.7 versus 4.4 +- 0.2 days, p b 0.0001),

Burns that involve the face,

hands, feet, genitalia, perineum, or major joints.

Third degree burns in any age group.

electrical burns, including

942.X5, 943.X2, 943.X5,

944-944.58, 945.X2, 945.X3, 945.

X5, 947.4 47,196

ICD-9 941.3-941.59,

942.3-942.59, 943.3-943.59,

944.3-944.58, 945.3-945.59,

946.3-946.5, 948.X1-9,

949.3-949.5 20,334

higher frequency of burn procedures performed (19,507 (44%) versus 2192 (22%) patients, p b 0.0001), and higher average number of burn procedures per hospitalization (2.3 +- 0.2 versus 1.1 +- 0.1 procedures, p b 0.0001). There were no significant differences in in-hospital mortal- ity (0.7% versus 0.8%, p = 0.692) or discharge disposition between the treatment center types.

A multivariate model was used to predict factors associated with

lightning injury. E925.X, E907 176

chemical burns. E924.1 1111

ICD-9506.9, 508.2, 518.5, 518.81,

518.84, 947.1, 987.9, E890.2,

under-referral to a burn center (i.e. non-burn center treatment of a pa- tient meeting ABA criteria), summarized in Table 3. Gender, race/eth- nicity, payer, and hospital region were not significantly associated

Inhalation injury.

Burn injury in patients with preexisting medical disorders that could complicate management, prolong recovery, or affect mortality.

Any patient with burns and concomitant trauma (such as fractures) in which the burn injury poses the greatest risk of morbidity or mortality. In such cases, if the trauma poses the greater immediate risk, the patient may be initially stabilized in a trauma center before being transferred to a Burn unit. physician judgment will be necessary in such situations and should be in concert with the regional medical control plan

E891.2 445

AHRQ comorbidity elements except depression and psychiatric illness, which are included in

criteria 10 7324

with treatment center type. Children in older age categories, and those

with burns to the head, neck, and face had increased odds of under- referral (OR 1.52, 95% CI: 1.14-12.03). Patients with multiple body re- gions burned (OR: 0.43, 95% CI: 0.34-0.53) and children meeting multi- ple ABA criteria for burn center treatment were more likely to go to a burn center; in other words they had decreased odds of under-referral (OR 0.77, 95% CI: 0.64-0.93).

Discussion

In our analysis, we demonstrated that the majority of children that met American Burn Association criteria for referral to a burn center re- ceived inpatient treatment at a burn center, independent of resource limitations as reflected by regional access and urban setting. Klein et al. found a regional difference in proximity to a burn center by ground and air transport, and determined that nearly 80% of Americans live within 2 h of a burn center by ground or rotary air transport [6]. This

and Triage protocols. ISS b 15 398?

Burned children in hospitals without qualified personnel or equipment for the care of

children. NA 0

Burn injury in patients who will

require special social, emotional, AHRQ comorbidity elements for

or rehabilitative intervention. depression and psychiatric illness 676

Total patients meeting any criteria 54,529^?

* Patients with severe trauma were excluded from the study.

^? Not a sum of Criteria 1-10, as some records may meet more than one criterion.

and non-burn centers, p = 0.25). Electrical and chemical burns were significantly less likely to be treated at burn centers than non-burn cen- ters (551 (1%) and 680 (2%) versus 205 (2%) and 407 (4%), respectively p= 0.0058 and p b 0.0001). Mechanism of injury was missing or unclas- sifiable for 12,485 (28%) of patients at burn centers versus 4463 (45%) of patients at non-burn centers. There was no significant difference in presence of inhalation injury for burn and non-burn center patients (3% for both, p = 0.56).

Body region burned also varied significantly between burn center and non-burn center patients. Among patients at burn centers, 27,537 (62%) had burns on multiple distinct body regions, compared to 3868 (39%) non-burn center patients (p b 0.0001). On average, patients at burn and non-burn centers had 2.03 +- 0.04 and 1.59 +- 0.03 distinct body regions burned, respectively (p b 0.0001). Burn centers most com- monly treated trunk, back, and genitalia burns (20,390 patients, 46%). Non-burn centers most commonly treated head, neck, and face burns

may explain the compensatory referral patterns across regions noted in this study.

Of those who were referred to a burn center, we found that they gen- erally appeared sicker, with increased numbers of children with multi- ple regions burned, more commonly including the trunk, back, and genitalia. Transfer to a pediatric burn center also resulted in longer length of stay and increased procedures, which taken together suggest a higher level of acuity among these patients. We found that the body region and severity of burns were associated with the location of care. Younger, more severely burned children were more likely to receive burn center treatment. Other characteristics associated with referral to a burn center included: 2 or more ABA criteria met, scald burn, and N40% TBSA. Patients with scald injuries were more frequently cared for in burn centers, perhaps reflecting more extensive injuries after body submersion in hot liquids, or even non-accidental trauma. No data were available on non-accidental trauma as an Underlying etiology. Interestingly, patients with chemical and electrical burns as well as patients with burns on the face were more likely to be treated at a non-burn center. Electrical burns can be deceptive, with the degree of soft tissue damage easily underappreciated. For example, oral electrical burns are most common in young children (e.g. they bite an electrical cord). These injuries can lead to devastating contractures around the mouth, compromising oral competency. Early specialized therapy is es- sential to optimizing recovery, and may include adjuncts such as custom garments, hand therapy, and oral splints. This referral pattern raises the possibility that there is under recognition of burns that requires special-

ized interventions to prevent late adverse sequelae [27,28].

Table 2

Demographics of pediatric burn patients meeting one or more American Burn Association criteria for treatment at a burn center. Column percentages are expressed in parentheses; row percentages are expressed in brackets.

	Non-burn center	Burn center	p value
	(n = 9897)	(n = 44,632)
Patients >=1 ABA criteria, no. (column %) [row %]	9897 (100) [18]	44,632 (100) [82]
Female, no. (column %) [row %]	3703 (37) [16]	16,218 (36) [84]	0.4938
Age, avg +- stdev	6.7 +- 0.2	5.6 +- 0.2	b0.0001
Age group, no. (column %) [row %]
1-4 years old	5091 (51) [16]	26,433 (59) [84]	0.0001
5-8 years old	1209 (12) [17]	5731 (13) [83]	0.5381
9-12 years old	1116 (11) [19]	4647 (10) [81]	0.3470
13-17 years old	2487 (25) [24]	7819 (18) [76]	0.0001
Race/ethnicity, no. (column %) [row %]
White	3825 (39) [19]	16,023 (36) [81]	0.5128
Black	1760 (18) [20]	7157 (16) [80]	0.5883
Hispanic	1509 (15) [16]	7737 (17) [84]	0.4533
Other	686 (7) [16]	3539 (8) [84]	0.4245
Missing	2117 (21) [17]	10,176 (23) [83]
Primary payer, no. (column %) [row %] Public insurance?	4749 (48) [18]	21,547 (48) [82]	0.9256
Private insurance	4242 (43) [19]	18,516 (41) [81]	0.7674
Self-pay	537 (5) [19]	2264 (5) [81]	0.7802
Other	364 (4) [14]	2168 (5) [86]	0.2892
Missing	5 (0) [4]	137 (0) [96]
Hospital region, no. (column %) [row %]
Northeast	3074 (31) [22]	10,943 (25) [78]	0.4440
Midwest	1777 (18) [17]	8713 (20) [83]	0.8018
South	2571 (26) [18]	11,423 (26) [82]	0.9669
West	2480 (25) [15]	13,551 (30) [85]	0.5528
Hospital location, no. (column %) [row %]			b0.0001
Urban	8685 (88) [16]	44,632 (100) [84]
Rural	1217 (12) [100]	0 (0) [0]

* Public payer is 99.8% Medicaid, 0.2% Medicare.

Some studies suggest that burn center volume is inversely propor- tional to mortality [11,12]. However, we found that the average burn center treats 44 children per year versus 1.4 children treated per year at non-burn centers, raising concerns that there is inadequate volume to maintain skills in non-burn centers, although the NIS does not include patients treated and discharged from the emergency department or an outpatient clinic. Interestingly, the vast majority of patients sampled in this study were in urban areas, but many were treated at non-burn centers. The specific city in the urban category is not included in the NIS, so we could not determine what proportion of this urban sample was in a city or state served by a burn center. However, this could ex- plain some portion of the decreasED referral to burn centers in urban areas. Zonies et al. studied factors that contribute to burn center referral, and found that patients of all ages in states with burn centers were less likely to be treated in non-burn centers (RR, 0.52; 95% CI, 0.51 to 0.54). Conversely, they found that those on Medicaid (0.77, 95% CI, 0.69 to 0.85) were less likely to be treated in verified burn centers [13]. Other studies that include adults found that patients without insurance or with Workman’s Compensation were more likely to be cared for at a burn center, while patients with Medicare or Medicaid were more likely to be cared for at a non-burn center [10,13]. Our analysis, similar to a previous study of payer mix limited to pediatric patients [1], did not find a difference between public and private payer, suggesting that burn centers were just as likely to treat children with public insurance as non-burn centers. While our findings may be reflective of different criteria for children versus adults to obtain emergency Medicaid, further investigation of the impact of insurance status on pediatric burn care is merited.

When patients were not transferred to burn centers, our analysis showed no significant difference in disposition outcomes, including mortality. Doud et al. similarly used clinical and outcome data (includ- ing extent of injuries, length of stay, and mortality differences) from a statewide dataset to infer that pediatric burn patients treated at burn centers were more severely burned [1]. In their statewide analysis of pe- diatric burn admissions in North Carolina, Palmieri et al. showed no

difference in mortality between children treated at burn versus non- burn centers, although both Doud and Palmieri reported that patients treated at burn centers and verified burn centers sustained more signif- icant injuries than those at non-burn centers [1,29]. Since mortality is such a rare event (b1%), even with a large dataset our analysis should be interpreted with caution, and we cannot draw any firm conclusions about the role the type of facility has on mortality. Given these discrep- ancies and the complex and varied presentation, more data are needed to assess long-term physiological and psychological morbidity [9] as it relates to triage patterns and severity of burns in pediatric patients. It would also be beneficial to better understand the capacity for both inpa- tient and outpatient intervention at non-burn centers. We cannot con- clude from these data what strategies are being utilized to offload the burn centers and to what extent these are scalable or appropriate.

5.1. Limitations

Our analysis has limitations. Identification of burn centers by linking hospital identification number to the AHA database excluded 31,147 possible patients from 16 states without a linkable hospital identifica- tion number. Not all states provide a hospital identifier that can link the NIS data to the AHA, a phenomenon which varies by year and in- cludes patients from major centers in large states such as Texas and Ohio. Notably, HCUP cautions against drawing interstate comparisons from NIS data, as the sampling frame is tied to census division, not state, and states may have variable representation in the sample from year to year [16]. This study identified burn centers based on self- reported burn designation by the AHA. This is a much broader definition than the ABA criteria and includes burn centers without verification. Further, unlike the ABA, the AHA does not differentiate between adult and pediatric burn centers. Thus, burn center status in this investigation should be understood in the broader context of facilities that treat burns. This could impact our results by diluting the higher performance of verified burn centers with a larger pool of non-verified centers which may not have equivalent outcomes. However, it does represent a

Table 3

injury characteristics and outcome measures for pediatric burn patients meeting ABA criteria for treatment at a burn center. Column percentages are expressed in parentheses; row per- centages are expressed in brackets.

	Non-burn center	Burn center	p value
	(n = 9897)	(n = 44,632)
Number of ABA criteria met, avg +- stdev	1.37 +- 0.02	1.53 +- 0.03	b0.0001
>=2 ABA criteria met, no. (column %) [row %]?	2942 (30) [14]	17,691 (40) [86]	b0.0001
Mechanism of injury, no. (column %) [row %]
Hot object/scald	3234 (33) [12]	23,579 (53) [88]	b0.0001
Fire/flame	1589 (16) [18]	7336 (16) [82]	0.2482
Electric?	205 (2) [27]	551 (1) [73]	0.0058
Chemical	407 (4) [37]	680 (2) [63]	b0.0001
Missing or unclassifiable	4463 (45) [26]	12,485 (28) [74]
Inhalation injury, no. (column %) [row %]	293 (3) [17]	1471 (3) [83]	0.5595
Body area involved by region, no. (column %) [row %]
Head, neck, and face	3358 (34) [15]	18,823 (42) [85]	b0.0001
Trunk, back, and genitalia	2969 (30) [13]	20,390 (46) [87]	b0.0001
Upper limb, excluding wrist and hand	2649 (27) [12]	18,644 (42) [88]	b0.0001
Wrist and hand	3112 (31) [16]	16,300 (37) [84]	0.0045
Lower limb	3115 (31) [17]	15,694 (35) [83]	0.0639
Multiple regions?	3868 (39) [12]	27,537 (62) [88]	b0.0001
Missing	323 (3) [53]	284 (1) [47]
Number of regions burned?, avg +- stdev	1.59 +- 0.03	2.03 +- 0.04	b0.0001
Total body surface area burned, no. (column %) [row %]
b40% TBSA	4924 (50) [11]	38,878 (88) [89]	b0.0001
N40% TBSA	81 (1) [11]	639 (1) [89]	0.9970
Missing	4892 (49) [49]	5115 (11) [51]
Third degree (full-thickness) burn present, no. (column %) [row %]	1961 (20) [10]	16,965 (38) [90]	0.1624
Length of stay, avg days +- stdev	4.4 +- 0.2	7.3 +- 0.7	b0.0001
Burn procedure performed, no. (column %) [row %]?	2192 (22) [10]	19,507 (44) [90]	b0.0001
Number of procedures, avg +- stdev	1.1 +- 0.1	2.3 +- 0.2	b0.0001
Discharge disposition, no. (column %) [row %]
Routine	8780 (89) [18]	38,964 (87) [82]	0.631
Home health care	978 (10) [16]	5300 (12) [84]	0.54
Died in hospital	78 (b1) [20]	305 (b1) [80]	0.6917
Other?	40 (0) [38]	56 (0) [62]	0.1012
Missing	21 (0) [75]	7 (0) [25]

TBSA = Total Body Surface Area burned.

* Includes burns secondary to lightning injury.

^? Body regions were divided into the subcategories under the variable “body area involved” (Ex: “Head, neck, and face” represents one body region). “Multiple regions” is defined as the presence of any combination of more than one of these regions for a given record.

^? Inpatient burn procedures were identified using the ICD-9 Clinical Classifications Software Services and Procedures Procedure Categories for burn wound debridement (ICD-9 CSS 169).

^? Includes discharges against medical advice, to rehabilitation, and to law enforcement.

pragmatic sample of how hospitals are functioning. Although NIS has been used before to characterize patient populations and referral pat- terns, important variables, including race/ethnicity, and variables re- quired to classify the severity of a burn, including total body surface area, burn depth, and mechanism are incomplete. Race/ethnicity has been problematic in NIS data and these variables are frequently missing. Our analysis included otherwise complete patient records with missing race data in order to detect a significant relationship, but improved reporting would allow detection of more subtle contributions of this variable in future studies. Burn specific variables are needed to adjust for severity, particularly burn extent and depth. While the National Burn Repository (NBR) dataset, maintained by the ABA, does provide burn-specific injury characteristics for each record, these data are lim- ited to verified burn centers and other burn centers that choose to re- port, and data from non-burn centers are not available [30]. More information about TBSA and burn depth as it relates to mechanism of in- jury may have more clearly differentiated patients and allowed for a better understanding of burn center referral patterns. Burn area calcula- tions were based on a compressed scale of ICD-9 codes by age group, and therefore more granular statistical analysis was not possible using the available data. Although these limitations significantly decrease the sample size and granularity of cases reported, such databases are the only way to look at the state of burn care across the country, and rea- sonable conclusions about transfer patterns can be inferred and used to improve large scale care delivery.

Additionally, the NIS is limited to inpatient data, so we could not

study referral patterns that involve ambulatory treatment of more

minor injuries. Current trends suggest a drive towards the ambulatory treatment of pediatric burns when feasible, and studying appropriate referral patterns in that setting could be worthwhile, especially consid- ering the relative abundance of more minor injuries [31,32]. This dataset also limits our ability to evaluate the decision to admit the child, and we are not able to assess the appropriateness of outpatient management of the child with minor burns discharged from the emer- gency department or Outpatient setting. Similarly, the NIS does not con- tain granular outcomes data on longer horizons such as functional recovery or scarring/contractures. Investigating how referral patterns translate into long-term functional and patient-reported outcomes could further assist in understanding the value of specialized burn care. Lastly, this study is also limited by the geographical information available in the dataset. In addition to limitations of included states, pa- tient specific geographic information was unavailable. Distance from

the injury site to the hospital of admission is not known.

Conclusion

The results of this study suggest that the majority of children meet- ing ABA criteria are treated at a burn center. The children who are not referred to burn centers appear to be less severely burned. As mortality is a rare event, longitudinal comparative analyses are needed to deter- mine the true impact of treatment center type on long-term functional outcomes. Such studies should evaluate differences in measures such as burn-specific staff training, recordkeeping, and transfer patterns, and could also include developmental data (cognitive, physical, emotional,

Table 4

Multivariate model of non-burn center use (under-referral) in pediatric patients with burns meeting ABA burn center referral criteria (n = 41,734). Patients with records miss- ing data are excluded (n = 12,795). Hosmer-Lemeshow goodness of fit performed (p = 0.97); larger p-values indicate better fit model.

References

1. Doud AN, Swanson JM, Ladd MR, et al. Referral patterns in pediatric burn patients. Am Surg 2014;80(9):836-40.
2. Web-based injury statistics query and reporting system (WISQARS). [Internet]. Available from: . http://www.cdc.gov/injury/wisqars, Accessed date: 7 February

   Characteristic Odds ratio 95% confidence interval

   Lower Upper

   Age group 13-17 years old	Reference
   1-4 years old	0.61	0.46	0.80	0.000
   5-8 years old	0.66	0.48	0.92	0.014
   9-12 years old	0.77	0.58	1.03	0.080
   Female	1.07	0.97	1.22	0.169
   Race/ethnicity
   White	Reference
   Black	1.07	0.71	1.60	0.741
   Hispanic	0.90	0.63	1.30	0.587
   Other	0.81	0.49	1.35	0.417
   Payer Public?	Reference
   Private	0.96	0.68	1.36	0.828
   Other	0.97	0.66	1.42	0.877
   Hospital region
   Northeast	Reference
   Midwest	0.46	0.16	1.37	0.163
   South	0.92	0.33	2.58	0.870
   West	0.71	0.24	2.15	0.545
   Body region
   Lower limb	Reference
   Head, neck, and face	1.52	1.14	2.03	0.004
   Trunk, back, and genitalia	1.22	0.95	1.57	0.125
   Upper limb, excluding wrist and hand	1.19	0.95	1.49	0.134
   Wrist and hand	1.04	0.83	1.33	0.705
   Multiple body regions burned	0.43	0.34	0.53	0.000
   >=2 ABA criteria met	0.77	0.64	0.93	0.005

   * Public payer is 99.8% Medicaid, 0.2% Medicare.

   p-Value

   2019.

   Wolf SE, Rose JK, Desai MH, et al. Mortality determinants in massive pediatric burns. An analysis of 103 children with N or = 80% TBSA burns (N or = 70% full-thickness). Ann Surg 1997;225(5):554-65 [discussion 65-9].

3. Brigham PA, Dimick AR. The evolution of burn care facilities in the United States. J Burn Care Res 2008;29(1):248-565.
4. Burn Center Referral Criteria Chicago, IL American Burn Association. Available from . http://www.ameriburn.org/BurnCenterReferralCriteria.pdf.
5. Klein MB, Kramer CB, Nelson J, et al. Geographic access to burn center hospitals. JAMA 2009;302(16):1774-81.
6. Ortiz-Pujols SM, Thompson K, Sheldon GF, et al. Burn care: are there sufficient pro- viders and facilities? Bull Am Coll Surg 2011;96(11):33-7.
7. Johnson SA, Shi J, Groner JI, et al. Inter-facility transfer of pediatric burn patients from U.S. Emergency Departments. Burns 2016;42(7):1413-22.
8. Sheridan R, Weber J, Prelack K, et al. Early burn center transfer shortens the length of hospitalization and reduces complications in children with serious burn injuries. J Burn Care Rehabil 1999;20(5):347-50.
9. Klein MB, Nathens AB, Heimbach DM, et al. An outcome analysis of patients trans- ferred to a regional burn center: transfer status does not impact survival. Burns 2006;32(8):940-5.
10. Holmes JH, Carter JE, Neff LP, et al., The effectiveness of regionalized burn care: an analysis of 6,873 burn admissions in North Carolina from 2000 to 2007. J Am Coll Surg 2011;212(4):487-93, 93.e1-6; [discussion 93-5].
11. Carter JE, Neff LP, Holmes JH. Adherence to burn center referral criteria: are patients appropriately being referred? J Burn Care Res 2010;31(1):26-30.
12. Zonies D, Mack C, Kramer B, et al. Verified centers, nonverified centers, or other fa- cilities: a national analysis of burn patient treatment location. J Am Coll Surg 2010; 210(3):299-305.
13. D’Souza AL, Nelson NG, McKenzie LB. Pediatric burn injuries treated in US emer- gency departments between 1990 and 2006. Pediatrics 2009;124(5):1424-30.
14. Thombs BD, Singh VA, Milner SM. Children under 4 years are at greater risk of mor- tality following acute burn injury: evidence from a national sample of 12,902 pedi- atric admissions. Shock 2006;26(4):348-52.
15. Nationwide Inpatient Sample (NIS). In: AfHRa Quality, editor. Healthcare cost and utilization project (HCUP); 2001-2011 Rockville, MD.
16. Burlinson CE, Wood FM, Rea SM. Patterns of burn injury in the preambulatory infant.

    Burns 2009;35(1):118-22.

    Ojo P, Palmer J, Garvey R, et al. Pattern of burns in child abuse. Am Surg 2007;73(3):

    social outcomes) from the outpatient setting after burn patients are discharged. Due to the relative infrequency of burn injuries in sparsely populated areas of the country, development of new, rural burn centers is not likely to occur, nor would it be likely to improve outcomes based on the existing literature. Despite geographic limitations, children in this study with the most severe burn injuries were treated at burn cen- ters without significant Regional variation in early outcomes. Further re- search should explore reasons for not referring pediatric patients to burn centers when they meet ABA criteria. It would also be beneficial to study long-term outcomes of children treated at burn centers versus those treated at non-burn centers, including consideration of those pa- tients who are followed in a burn center after discharge, regardless of initial site of care.

    Funding

    This research did not receive any specific grant from funding agen- cies in the public, commercial, or not-for-profit sectors.

    Declaration of competing interest

    None.

    Acknowledgements

    Not applicable.

    253-5.

    Clark DE, Osler TM, Hahn DR. ICDPIC: Stata module to provide methods for translat- ing International Classification of Diseases (Ninth Revision) diagnosis codes into standard injury categories and/or scores Statistical Software Components ; 2010.

17. Hawkins A, Maclennan PA, McGwin G, et al. The impact of combined trauma and burns on patient mortality. J Trauma 2005;58(2):284-8.
18. Rosenkranz KM, Sheridan R. Management of the burned trauma patient: balancing conflicting priorities. Burns 2002;28(7):665-9.
19. Santaniello JM, Luchette FA, Esposito TJ, et al. Ten year experience of burn, trauma,

    and combined burn/trauma injuries comparing outcomes. J Trauma 2004;57(4): 696-700.

    Urban and rural classification. United Status Census Bureau [Internet]. Available from . https://www.census.gov/geo/reference/ua/urban-rural-2010.html; 2010.

20. Zhou X, editor. Applied missing data analysis in the health sciences. Hoboken, New Jersey: Wiley; 2014 [230 p].
21. Hosmer D, Lemesbow S. Goodness of Fit tests for the multiple logistic regression model. Communications in statistics-Theory and Methods 1980;9(10):1043-69.
22. StataCorp. Stata statistical software. Release, vol. 13. College Station, TX: StataCorp LP; 2013.
23. Umstattd LA, Chang CW. Pediatric oral electrical burns: incidence of emergency de- partment visits in the United States, 1997-2012. Otolaryngol Head Neck Surg 2016; 155(1):94-8. https://doi.org/10.1177/0194599816640477 Jul. [Epub 2016 Apr 5].
24. Pontini A, Reho F, Giatsidis G, Bacci C, Azzena B, Tiengo C. Multidisciplinary care in severe pediatric electrical oral burn. Burns 2015;41(3):e41-6. https://doi.org/10. 1016/j.burns.2014.12.006 May. [Epub 2015 Feb 21].
25. Palmieri TL, Taylor S, Lawless M, et al. Burn center volume makes a difference for burned children. Pediatr Crit Care Med 2015;16(4):319-24.
26. American Burn Assiociation. National Burn Repository. Chicago, IL. [Internet]. Avail- able from . https://ababurnportal.ebresearch.com/; 2017.
27. Abali AE. Cost-effective outpatient burn-care for minor burns. Burns J Int Soc Burn Inj 2015;41(3):639-40 May.
28. Brown M, Coffee T, Adenuga P, Yowler CJ. Outcomes of outpatient management of pediatric burns. J Burn Care Res Off Publ Am Burn Assoc 2014;35(5):388-94 Oct.