ORIGINAL ARTICLE

Relationship Between Patient Characteristics and Number

of Procedures as well as Length of Stay for Patients
Surviving Severe Burn Injuries: Analysis of the American
Burn Association National Burn Repository

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

Eliza Kruger, MHEcon,* Stacey Kowal, MS,* S. Pinar Bilir, MS,* Eileen Han, PhD,* and
Kevin Foster, MD, MBA†

This study establishes important, national benchmarks for burn centers to assess length of stay (LOS) and
number of procedures across patient profiles. We examined the relationship between patient characteristics such
as age and total body surface area (TBSA) burned and number of procedures and LOS in the United States,
using the American Burn Association National Burn Repository (NBR) database version 8.0 (2002–2011).
Among 21,175 surviving burn patients (TBSA > 10–60%), mean age was 33 years, and mean injury size was
19.9% TBSA. Outcomes included the number of debridement, excision, autograft procedures, and LOS.
Independent variables considered were: age (linear, squared, and cubed to account for nonlinearity), TBSA,
TBSAs of partial-thickness and mixed/full-thickness burns, sex, hospital-acquired infection, other infection,
inhalation injury, and diabetes status. Regression methods included a mixed-effects model for LOS and ordinary
least squares for number of procedures. A backward stepwise procedure (P <0.2) was used to select variables.
Number of excision and autografting procedures increased with TBSA; however, this relationship did not hold
for debridement. After adjusting for sex, age, and comorbidities, predicted LOS for adults (18+) was 12.1,
21.7, 32.2, 43.7, and 56.1 days for 10, 20, 30, 40, and 50% TBSA, respectively. Similarly, predicted LOS for
pediatrics (age < 18) was 8.1, 18.8, 33.2, 47.6, and 56.1 days for the same TBSA groups, respectively. While
average estimates for adults (1.12 days) and pediatrics (1.01) are close to the one day/TBSA rule-of-thumb,
consideration of other important patient and burn features in the NBR can better refine predictions for LOS.

Approximately 1% of nonfatal injuries among U.S. civilians patients with burns covering up to 90% of their bodies can
are burn injuries.1 According to recently published estimates, survive with appropriate management strategies.3 While
nearly 500,000 burn victims require medical care annu- these improvements highlight the benefits of innovation in
ally, 40,000 of whom are also hospitalized for burn treat- burn care, there remain opportunities to improve healing
ment.2 Dramatic improvements have been made in burn care and clinical outcomes, thereby reducing patient length of
practices over time, resulting in improved clinical outcomes. stay (LOS) and the economic burden of burn injuries.4, 5
During the 1960s, burn-related mortality was common Increased transparency on resource use and the relationship
for patients with burns of 20% or more of total body sur- between patient and burn characteristics is a fundamental
face area (TBSA) given either the initial injury or down- step in providing a benchmark of real-world care practices.
stream infections and complications.3 Today, the number For example, early excision and autografting to achieve de-
of burn-related deaths has declined by more than 50% and finitive closure are recognized cornerstones of modern burn
therapy.6 Still, there is wide variation in practice, including
assessment of depth, timing of eschar removal by wound de-
bridement/excision, extent of excision performed and the
From the *IQVIA, Falls Church, Virginia and †Arizona Burn Center, Phoenix,
Arizona products and procedures that are used to achieve definitive
closure. Identifying characteristics that drive significant var-
Funding: This work was funded by Biomedical Advanced Research Development
Authority, AVITA Medical. This work was supported under the HHS/ASPR/ iation in the number of these procedures as well as resulting
BARDA Contract no. HHSO100201500028C. patient LOS would help care providers understand how their
Conflict of interest statement. S.K., E.K., P.B. and E.H. are employees of IQVIA practice compares to overall practices treating a similar pa-
and received funding to conduct the research. The publication of study results tient population.
was not contingent on the sponsor’s approval or censorship of the manuscript.
Address correspondence to Stacey Kowal, MSc, IQVIA, 3110 Fairview Park
Although studies have sought to describe treatment trends
Drive, Suite 400, Falls Church, VA 22042. Email: skowal@us.imshealth.com and predictive relationships in U.S. burn care, robust data de-
© The Author(s) 2020. Published by Oxford University Press on behalf of the tailing the predictive relationship between individual patient
American Burn Association. characteristics and burn center practice patterns on patient
This is an Open Access article distributed under the terms of the Creative LOS is limited. For example, one study assessed the relation-
Commons Attribution License (http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted reuse, distribution, and reproduction in any me- ship between burn patient characteristics and operating room
dium, provided the original work is properly cited. visits, number of operations, mechanical ventilation use, and
doi:10.1093/jbcr/iraa040 intensive care unit (ICU) days.7 In this analysis, the authors
1037
 Journal of Burn Care & Research
1038  Kruger et al September/October 2020

included all patients regardless of survival status and injured unique codes were applied in the same surgical intervention
TBSA and did not consider specific types of procedures. A sys- (ie, assumed to represent a unique and single operating room
tematic literature review of publications predicting LOS in visit). Therefore, the maximum count of any individual ICD-9
thermal burns noted that age and percent TBSA of burn code avoids double-counting and thus avoids overestimation
were the strongest predictors of LOS, with percent mixed of the number of autografting procedures. Please note that
depth/full-thickness burns, sex, inhalation injury, number while number of operating room procedures is variable in the
of procedures, and depth of burn as additional significant NBR, less than one-quarter (21%) of our analysis sample has
variables.8 However, many studies cited in the review focused this variable populated. Therefore, within this analysis, we
on smaller TBSA ranges, typically less than 20%.6, 9–14 In addi- assumed that presence of the aforementioned ICD-9 codes
tion, one publication did not differentiate between surviving can be interpreted as a surgical intervention, which we re-
patients and nonsurviving patients,7 which may confound ferred to as a procedure throughout this article.
conclusions. Independent variables were informed by a review of the
To the authors’ knowledge, no published research has published burn literature,7, 8 interviews with burn surgeons
examined the factors that predict the number of specific types and availability of variables in the NBR. These variables in-

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

of inpatient procedures and LOS in surviving burn patients cluded patient characteristics such as age (in years), sex,
in U.S. acute care using real-world data. Therefore, the aim selected comorbidities (diabetes status, hospital-acquired in-
of this study was to examine the relationship between patient fection [HAI], other infection, and inhalation injury), burn
characteristics such as age or TBSA of burn on LOS as well TBSA, and whether the burn was superficial partial-thickness
as numbers of procedures (debridement, excision, and defini- (SPT, defined as patients expected to heal in less than 14 to
tive closure with autografting), using national aggregate data 21 days without an autograft procedure). To account for pos-
from the American Burn Association (ABA) National Burn sible nonlinear relationship between the outcomes and the in-
Repository (NBR).15 Specifically, this article seeks to analyze dependent variables, the squared and cubed forms of age and
and develop national aggregate information on procedures TBSA were also included in the model.
used and resulting LOS outcomes by patient and burn type.
Leveraging this information, we also seek to develop predic- Analyses
tive equations that can be leveraged by the burn community
Descriptive statistics were calculated for all variables, stratified
to benchmark individual burn center trends and to inform
by patient age group (pediatrics: age 0–17 years or adult: age
decisions on the value of new interventions. Furthermore,
18 years or older). For continuous variables, the mean was
such information could be leveraged by innovators, key payers,
reported; for categorical variables, the proportion of patients
and provider stakeholders to conduct economic evaluations of
observed in each category was reported. All analyses were
new burn care interventions.
conducted using Stata Version 15.19
Dependent variables were number of procedures (per the
METHODS ICD-9 codes noted above) and LOS. A backward selection
stepwise process was used to identify independent variables in-
Data and Variables cluded in each analysis. Independent variables were removed
The NBR is a voluntary registry sponsored by the ABA and from the model if the level of significance exceeded 20% (ie,
includes ten years of cumulative data from burn centers, P > 0.2). To predict the number of procedures—debridement,
thus representing the largest resource on epidemiology of excision, and autograft—ordinary least squares (OLS) models
burn injuries for patients admitted to burn centers in North were fitted. Only patients with one or more autografting
America.16 This study used the NBR version 8.0 (2002–2011) procedures were included in the analyses for number of
as it was the most up to date version available for analysis.15 autografting procedures. A mixed-effects linear model was
To avoid confounding factors associated with both mortality selected for LOS to adjust for patient case mix and hospital
and resource use, and to better support hospital financial characteristics and to account for the dependence of out-
planning and comparison between specialist centers, we fo- come variables within hospitals. This regression method was
cused on acute care patients with burns covering 10 to 60% considered the most appropriate based on a review of similar
TBSA.17 Patients with TBSA 61%+ were removed from re- regression analyses of LOS of the literature.20–22
gression analyses to reduce a tail effect where outlier patients Descriptive statistics of mean patient and burn charac-
with high TBSA would skew results, as they commonly have teristics for pediatrics and adults for TBSA 10, 20, 30, and
exceptionally long LOS and intensive resource use.7, 17 We 40% were multiplied by regression coefficients to generate
only included surviving patients to reduce confounding in our predicted results for average number of procedures and
predictive analysis, as many of the same predictors of mor- LOS for a range of patient profiles to allow for benchmark
tality (eg, TBSA and inhalation injury) overlap with important comparisons. Output of the regression could be applied to
predictors of resource use such as LOS.18 any burn patient via use of coefficients applied to the patient’s
Key outcomes of interest were the number of procedures unique burn and demographic characteristics (see Table 2).
identified via ICD-9 code, specifically including nonexcisional To help put the output of the regression in context, we used
debridement (debridement, 86.28), excisional debridement regression output and average patient characteristics from the
(excision, 86.22), and autograft (86.60, 86.61, 86.62, 86.63, NBR to estimate outcomes for TBSA ranges of 10 to 60%.
and 86.69), and NBR-reported LOS. As autograft procedures Please note that results for 60% and over are not reported
were identified by multiple codes, we assumed that multiple given that the last included TBSA band was 50 to 59%.
Journal of Burn Care & Research
Volume 41, Number 5 Kruger et al  1039

RESULTS variables were retained in the model. The Bryk and

Raudenbush23 r2 level 1 of the predicted model was 0.43,
From NBR 2002 to 2011 data, a sample of 21,175 surviving suggesting that the model had an adequate fit given un-
patients, with nonmissing data for dependent and independent derlying data. Examining the impact of TBSA alone, all
variables, was identified. Table 1 provides the description of else being equal among other independent variables, TBSA
the sample (descriptive statistics by TBSA are available in the contributes to some but not all of LOS, with each addi-
Supplementary Appendix). Among the final sample, the av- tional percent TBSA leading to approximately 0.723 more
erage age of surviving burn patients was 32.8 years, approx- days of LOS. Given the coefficients observed across other
imately one-third were female, and the average TSBA was independent variables, other factors also played a key role in
19.9%. Viewing the descriptive data in aggregate without any predicting LOS for a given patient. For example, presence of
regression analyses or adjustments, the mean LOS per TBSA HAI was almost as influential as TBSA for debridement and
percent was 0.84 for pediatrics and 1.09 for adults. Compared could be more important than TBSA for smaller burns when
with adult patients, pediatric patients had a higher propor- predicting number of excisions or LOS. Furthermore, while
tion of females (35% vs 24%), higher proportion of mixed age was not a strong predictor of number of procedures, it

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

depth/full-thickness burn (37% vs 33%), and fewer excision had a notable impact on LOS. While the exact impact of
and autografting procedures (excision, 1.7 vs 2.0; autograft, each independent variable such as TBSA, age, sex, HAI, and
1.7 vs 1.8;). Pediatric patients also had a higher proportion of diabetes, is dependent on each patient’s unique character-
patients with SPT (47% vs 40%). This trend is not unexpected, istics, the difference in magnitudes for the coefficients for
as epidemiology reports of burn injury indicate a high inci- each independent variable shows important variation across
dence of scalds in pediatrics.17 As such, this higher proportion key procedures and LOS.
of SPT burns in pediatrics could be driven by a greater pro- Table 3 presents the model predictions for each outcome.
portion of scalds, which is consistent with the noted lower rate As noted above, mean patient and burn characteristics across
of inhalation injury (6% vs 12%). TBSA ranges from the NBR were leveraged to translate the
Table 2 shows the estimated coefficients for each pa- regression findings into benchmark information for com-
rameter in the regression model for LOS. All independent parison. Details on the patient and burn characteristics for
each TBSA range are reported in Supplementary Appendix.
Table 1. Demographic characteristics of surviving burn While TBSA was a significant variable for debridement, overall
patients number of debridement procedures did not increase with
Pediatrics Adult TBSA for either adults or pediatrics, as the impact of TBSA
was outweighed by other factors. However, the number of ex-
(0–17 (18+ All cision procedures did increase with TBSA. When a definitive
years) years) patients
closure (ie, autografting) was required, the predicted number
Number of patients (n) 5957 15,218 21,175 of autograft procedures was similar for adults and pediatrics,
Mean age at time of burn 6.3 43.2 32.8 with the number of autograft procedures increasing as TBSA
injury (years) burned increased.
Sex After adjusting for sex, age, and comorbidities, predicted
Female (%) 35% 24% 27% LOS for adults (age 18+) was 12.1, 21.7, 32.2, 43.7, and
Male (%) 65% 76% 73% 57.5 days for 10, 20, 30, 40, and 50% TBSA, respectively.
Comorbidities For pediatrics (age < 18), the predicted LOS was 8.1, 18.8,
Inhalation injury 6% 12% 10% 33.2, 47.6, and 56.1 days for 10, 20, 30, 40, and 50% TBSA,
HAI 2% 3% 3% respectively. When considering the impact of all independent
Other infection 2% 4% 4% variables, the average LOS per percent TBSA is estimated
Diabetes 0% 4% 3% at approximately 1.12 and 1.01 days for adults and pediat-
Characteristics of burn rics. For pediatrics, the average LOS days per percent TBSA
Total TBSA (%) 19.8% 19.9% 19.9% increased with TBSA, from 0.81, 0.94, 1.11, and 1.19 days for
Partial thickness TBSA (%) 12.4% 13.3% 13.1% 10, 20, 30, and 40%. For adults, LOS days per percent TBSA
Full-thickness TBSA (%) 7.4% 6.6% 6.8% increased by 1.21, 1.08, 1.07, and 1.09 days for TBSA 10,
Proportion patients SPT (%) 47% 40% 42% 20, 30, and 40%, respectively. Trends for 50% TBSA showed a
Number of procedures continued increase in days per percent TBSA for adults (1.12)
Debridement 0.7 0.7 0.7 but a slight decreasing trend for pediatrics (1.15). However,
Excision 1.7 2.0 1.9 this information should be interpreted with caution given ex-
Autograft 1.7 1.8 1.8 pected confounding in this high TBSA category. Although the
LOS (days) observed LOS in pediatrics is in general lower than adults, the
Average 17.4 22.0 20.7 overall trend of increasing LOS with increasing percent TBSA
Per percent TBSA 0.84 1.09 1.02 of burn were similar. As the percent TBSA burned increases,
the relative impact on LOS also increases to become one of
HAI, hospital acquired infection; LOS, length of stay; SPT, superficial partial- the dominant factors influencing LOS outcomes. Figure 1 fur-
thickness; TBSA, total body surface area.
Descriptive statistics for the final sample are provided above. LOS re-
ther illustrates these findings and compares adjusted estimates
ported above is mean values for the sample and is not adjusted for patient for LOS per percent TBSA to unadjusted mean values from
characteristics. the NBR.
 Journal of Burn Care & Research
1040  Kruger et al September/October 2020

Table 2. Regression model coefficients

Debridement (OLS) Excision (OLS) Autograft (OLS) LOS (mixed effects)

Coefficients Beta SE P Beta SE P Beta SE P Beta SE P

TBSA −0.081 0.020 <0.001 0.031 0.029 0.276 0.068 0.003 <0.001 0.708 0.161 <0.001
TBSA2 0.002 0.001 0.001 0.001 0.001 0.454 0.015 0.006 0.012
TBSA3 0.000 0.000 0.003 0.000 0.000 0.662 0.000 0.000 0.087
TBSA PT 0.009 0.002 <0.001 −0.031 0.003 <0.001 −0.023 0.003 <0.001 −0.565 0.016 <0.001
Age 0.008 0.003 0.007 −0.010 0.005 0.030 −0.275 0.039 <0.001
Age2 0.000 0.000 0.014 0.000 0.000 0.068 0.000 0.000 0.014 0.009 0.001 <0.001
Age3 0.000 0.000 0.005 0.000 0.000 <0.001
Female −0.065 0.032 0.042 0.130 0.068 0.056 1.811 0.262 <0.001
HAI −0.110 0.085 0.196 0.608 0.125 <0.001 11.269 0.757 <0.001

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

Other infection −0.282 0.140 0.045 4.110 0.683 <0.001
Inhalation injury 0.210 0.091 0.021 7.563 0.399 <0.001
Diabetes 0.331 0.079 <0.001 0.564 0.116 <0.001 0.430 0.156 0.006 1.880 0.656 0.004
SPT 0.659 0.030 <0.001 −1.676 0.045 <0.001 −9.169 0.267 <0.001
Constant 1.053 0.156 <0.001 1.872 0.230 <0.001 1.886 0.101 <0.001 11.108 1.465 <0.001
N = 21,175 N = 21,175 N = 12,333* N = 21,175
R2 = 0.03 R2 = 0.13 R2 = 0.04 R2 = 0.43

OLS, ordinary least squares, LOS, length of stay; SPT, superficial partial-thickness; HAI, hospital-acquired infection.
*Sample size is reduced for autografting given requirement that patients in the sample for this concept received an autograft.

Table 3. Predicted number of debridement, excision, auto- (expressed as percent change) from the 1 day per TBSA
graft procedures, and LOS by age group and TBSA common clinical approximation across potential patients with
20% TBSA. Moving away from a weighted average of the
Debridement Excision Autograft NBR population characteristics, we can see how LOS changes
Procedures (n) Procedures (n) Procedures (n) LOS (days)
based on sex (male, female), actual age (0.5 to 17 years for
Adults (18+) pediatrics; 18 to 65 years for adults), burn depth, and pres-
TBSA (%) Burned ence of comorbidities. Considering the estimated LOS with
10% 1.0 1.3 2.3 12.1 the 1 day per TBSA approximation is 20 days for a patient
20% 0.7 1.9 2.8 21.7 with 20% TBSA burned, differences in individual patient char-
30% 0.6 2.5 3.4 32.2 acteristics, such as full-thickness depth of injury, can drive up
40% 0.7 3.1 4.0 43.7 to a 66% shift in LOS, or up to a change in LOS of 13.2 days
50% 0.4 3.9 4.8 57.5 (20 days for rule of thumb compared to 33.2 days).
Pediatrics (0–17)
TBSA (%) Burned
10% 1.0 0.9 2.4 8.1
DISCUSSION
20% 0.6 1.7 2.9 18.8 This study represents the first analysis that develops real-world
30% 0.5 2.5 3.6 33.2 evidence-based predictive equations to explore the relation-
40% 0.5 3.4 4.3 47.6 ship between patient characteristics and LOS as well as three
50% 0.6 3.8 4.6 56.1 specific procedures among surviving burn patients with TBSA
10% or more in the United States. When controlling for typical
Estimates above represent averages for the population with each burn depth,
average patient characteristics as captured in the NBR, we find
with patient characteristics informed by the final analysis sample from the
NBR. Please see Supplementary Appendix for more detail on average patient LOS per percent TBSA is estimated at approximately 1.12 days
characteristics by age group and TBSA range. per percent TBSA for adults and 1.01 for pediatrics, with av-
erage LOS per percent TBSA increasing with TBSA. While
TBSA was found to be a significant predictor of excision and
Although the 1 day per percent TBSA rule of thumb may autograft procedures as well as LOS, it is not the only factor
somewhat approximate LOS, the key benefit of generating a that affects these outcomes. Patient age, sex, comorbidities,
predictive equation from regression analysis is the ability to and burn characteristics beyond TBSA may be as important.
capture the impact of many influential characteristics that Notably, large positive coefficients for HAI, infections, and
work in a multifactorial fashion to predict LOS outcomes. The inhalation injury, as well as a large negative coefficient for SPT
ability of TBSA alone to accurately predict LOS is indeed var- burns can influence predicted LOS. Furthermore, the coeffi-
iable based on underlying patient characteristics. For example, cient for partial thickness is almost as large as the coefficient
when evaluating how LOS may present for an individual for TBSA, indicating that the depth of burn is an important
patient, the range of difference from the 1 day per TBSA feature when predicting LOS. Considering additional burn
rule is more notable. Figure 2 shows the relative difference and patient characteristics, the existing 1 day per TBSA rule
Journal of Burn Care & Research
Volume 41, Number 5 Kruger et al  1041

1.40
1.21 1.19
1.20 1.12 1.15
1.08 1.07 1.11 1.09 1.09

1.00 0.94
0.81 0.84
0.80
Days

0.60

0.40

0.20

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

0.00
Pediatrics

Pediatrics

Pediatrics

Pediatrics

Pediatrics

Pediatrics
Adults

Adults

Adults

Adults

Adults

Adults
TBSA 10% TBSA 20% TBSA 30% TBSA 40% TBSA 50% Unadjusted
Straight Mean
Value

Figure 1. Average predicted LOS days per percent TBSA for adults and pediatrics for surviving patients. Columns represent number of inpatient
days per percent TBSA burned. Adult patients (18 and older) represented in light gray and pediatrics (17 and under) in dark gray. X-axis labels de-
note group of patients based on TBSA burned. Adjusted estimates for LOS per percent TBSA based on regression analysis are also shown alongside
average, unadjusted estimates for patient characteristics in this sample was leveraged to generate average LOS estimate. Columns with gradient fill
(for adults, pediatrics) show the straight mean. Please note that these estimates were derived without rounding.

80%
66% 63%
Relave Difference from 1 Day LOS per TBSA

60% 51% 49%

43%
40% 32% 30% 28%
15% 14% 17%
20% 10%
8%
0% 0% 3%
0%

-6% -6% -9%

-20% -11%

-40%
-36%
-52%
-60%
Average Paent SPT Burn PT Burn FT Burn Younger Age Older Age Female Paent Inhalaon HAI Other infecon Diabec Paent
TBSA 20% (NBR (6 mos; 18 yrs) (17 yrs; 65 yrs) injury
Regression)

Pediatrics Adults
Figure 2. Relative difference from 1 day LOS per TBSA across varying patient characteristics (scenario analysis: adult or pediatric patient with
20% TBSA burned). Columns represent relative difference (percent change) from common rule of thumb baseline of 1 day per TBSA (0% Y-axis
represents 1 day per TBSA rule of thumb). Adults patients (18 and older) represented in light gray and pediatrics (17 and under) in dark gray.
X-axis labels denote scenario tested to show variation across patient characteristics, including: average from the NBR, patients with an SPT burn,
PT or FT burn, Patients at younger and older ages within population range, sex, and presence of key comorbidities (ie, inhalation injury, HAI,
other infection, and diabetes). Numbers above 0% represent increase beyond 1 day per TBSA and negative figures represent lower than 1 day per
TBSA. For example, increase of 51 and 66% for pediatrics and adults, respectively, indicate that LOS was 51 and 61% greater than 1 day per TBSA
for patients with full thickness burns.

of thumb could differ from expected LOS by 60% or more, as TBSA exceeded 1 for all surviving patients (all TBSAs), with
illustrated by the 20% TBSA example described above.13 a low of 1.66 days per percent TBSA for infants aged 12 to
Our findings are consistent with summary descriptive sta- 23 months ranging up to 3.94 days per percent TBSA for
tistics provided in the 2017 ABA NBR report.17 Specifically, adults aged over 80.17 The higher average LOS per percent
summary statistics of all NBR patients (regardless of burn size) TBSA from the NBR sample is likely due to the floor effect of
by age found unadjusted number of hospital days per percent including smaller TBSA burns (ie, inpatient days are greater
 Journal of Burn Care & Research
1042  Kruger et al September/October 2020

than 1, even for small burns). These findings suggest that a burn care who may have other comorbid conditions beyond
more nuanced approach to accurately estimate LOS is needed, those captured in the NBR that impact outcomes.
and that considering patient and burn characteristics (in par- This analysis focused on understanding resource utilization
ticular, age and depth of burn) is needed in addition to TBSA. of a different patient cohort than has been examined previ-
While TBSA is a significant predictor for debridement, ously. Specifically, our patient sample includes large burns,
there is no discernable increase in the number of debride- surviving patients, and focused regression analysis on a TBSA
ment procedures with increasing TBSA. It is expected that range of 10 to 60% to reduce the biasing effect of outliers.
unknown factors or differences in clinical practice may play Additionally, these analyses sought to consider specific types
a larger role in determining the number of nonexcisional of procedures, adding granularity on key intervention and
debridement procedures required. For example, eschar re- resource use detail during an inpatient stay. Finally, this
moval via excision may have been preferred for patients analysis provides a more nuanced estimate of LOS days per
subsequently receiving an autograft, while nonexcisional de- percent TBSA for surviving burn patients, highlighting the
bridement may have been preferred for burns not needing differences between average LOS per percent TBSA between
autografting, diluting the impact of TBSA on overall de- pediatrics and adults and when adjusting for typical patient

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

bridement trends. Furthermore, noting the high number of characteristics.
autografting procedures relative to the number of excision Summary information on national aggregate trends for
procedures, this trend may be impacted by either coding procedures and LOS may be useful for burn surgeons and
practices (as this analysis had to rely in ICD-9 codes given the centers seeking to benchmark resource utilization and center
nature of the dataset) or the potential for excision to be done outcomes. This analysis may also facilitate assessment of new
in tandem with grafting in a single procedure, captured under interventions/medical countermeasures (MCMs) by setting a
autografting and not excision codes. Finally, please note that baseline of resource use associated with current standard of
while the debridement and excision may be used interchange- care in this patient cohort. Accordingly, any new interven-
ably, this analysis assumes that the definitions align with the tion that aims to lessen resource utilization may be evaluated
ICD9 codes available in the NBR. This is a limitation given against this threshold. For example, international real-world
that we can only report on trends for these procedures based studies have examined the impact of type of skin replacement
on coding, which may be subject in inaccuracy. Therefore, a surgery on LOS and surgery time outcomes to assess real-
future area of research could leverage survey data to better world effectiveness.24, 25 Furthermore, the predictive equations
understand the relationship between use of these procedures derived in this study were leveraged in an economic model,
in practice and coding approaches to better place the results the Burn-MCM Effectiveness Assessment Cost Outcomes
of our findings in context. Nexus (BEACON) model, to predict costs, outcomes and the
In a previous regression analysis that included both surviving value of new innovations for burn care patients in the United
and deceased burn patients in its sample,7 the authors found States.26, 27
that size and severity of burn, inhalation injury, and age were This study has several limitations that should be noted when
significant predictors of resource utilization, such as the interpreting and using the results. For models predicting the
number of operative procedures (ie, sum of ICD-9 procedure number of procedures, low r2 values indicate that the model
codes), total operating room visits (ie, multiple procedures specifications describe only a small proportion of the varia-
may be performed in the same operating room visit), LOS tion in the numbers of procedures. This suggests that other
in ICU, and length of time on a ventilator. In our analysis unknown factors, such as variation in hospital practices that
of only surviving burn patients, the above variables were also affect outcomes but are not captured in the dataset, may have
significant predictors of excision, debridement, and autograft an important influence. For example, some burn surgeons
procedures and LOS. In addition, depth of burn and sex were may take a more “wait-and-see” approach to burn wounds
significant predictors in our model, which is consistent with of indeterminate depth, whereas others may excise and auto-
previous literature.7, 8 Overall, our research builds on the body graft the burn wound at the earliest possible opportunity.28 In
of evidence that concludes consideration of burn and patient the former case, the surgeon may nonsurgically debride the
characteristics (beyond just TBSA) supports better predictions wound to assess whether the wound may be treated conserv-
of resource utilization, regardless of the type of patients ulti- atively and may therefore have more debridement procedures
mately included in the analysis (eg, severe burn patients and/ compared to the latter approach.29 Furthermore, related to
or deceased patients). This research aimed to establish initial differences in practices on timing of intervention, the true
predictive equations by looking at surviving burn patients and depth of burn is subjective and potentially unreliable.30
limiting to smaller burns. Indeed, while we report burns of Currently, the NBR does not code depth of burn at multiple
more than 50% TBSA, it should be noted that the influence time points, which limits the ability to capture potentially im-
of confounders in this population may have a more substantial portant changes in burn depth diagnosis as surgeons learn
impact on outcomes than other variables. Therefore, an area more over time. Therefore, while this study uses the best avail-
of future research could be to repeat this analysis across dif- able data from the NBR, this limitation should be noted when
ferent samples of patients (such as survivors, those who died interpreting findings. Another key finding of this research is
late in therapy, large burns) to more formally assess trends in that burn care practices have additional relatively uncontrolled
these populations as well as the shifting importance of predic- influencing factors important in driving these outcomes than
tive variables. Additionally, interpretation of this research, as patient characteristics. However, at present, the NBR does not
well as application of this research in practice, should remain include information attributed to region or individual burn
aware of the challenges of treating many complex patients in centers to allow for formal consideration of these factors in
Journal of Burn Care & Research
Volume 41, Number 5 Kruger et al  1043

regression analysis. Therefore, while the presented results are Furthermore, these model equations permit burn centers to
a foundational step to establish a baseline understanding of evaluate their own performance and highlight any potential
outcomes across key patient characteristics, an important area areas for improving efficiency. These estimates can also indi-
of future research will be to more formally evaluate how in- cate whether a given burn center achieves definitive closure
dividual burn center practices may improve patient resource with shorter LOS and fewer procedures. These predictive
utilization-related outcomes. equations also provide second-order information, as com-
In addition, the mixed-effects model for prediction of LOS parative value for cost of interventions can be evaluated by
exhibited greater variability in outcomes for increasing TBSA feeding the equations into a larger burn economic model.26, 33
(ie, heteroscedasticity). Despite an attempt to mitigate this by Finally, it may be feasible to predict costs and resource utiliza-
transformation of LOS to the logarithmic domain, the issue tion at a population or regional level, according to patient mix
largely remained and, further, predictive bias was introduced and expected interventions, supporting a higher level under-
during back transformation.31 This mixed-effect specification standing of the anticipated impact of potential changes or new
reflects the greater variability in outcomes observed in the interventions in burn care.
treatment of larger burns, wherein compounding clinical is-

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

sues can sometimes have substantial impacts on LOS.
Finally, it is important to note the data analysis utilized ABA SUPPLEMENTARY DATA
NBR Version 8.0, which includes burn patients from 2002 to
Supplementary data is available at Journal of Burn Care &
2011. Trends may have accelerated or changed since this time,
Research online.
which is not discernable due to the lag in data availability.
Version 8.0 was the most recent dataset available at the time
of analysis, and thus this study reflects the most up-to-date
ACKNOWLEDGEMENTS
analysis possible. Given the recent release of the 2019 update
to the NBR research dataset, future work could repeat this The authors acknowledge the helpful comments on the man-
regression analysis to provide updated predictive equations uscript from Katie Bush, Narayan Iyer, Andy Quick, Jeremiah
and to compare trends in influential independent variables Sparks, and Tom Walsh.
over time. Furthermore, it should be noted that while this
study uses one of the most robust and rich datasets for burns
available in the United States, the NBR is not exhaustive in REFERENCES
its inclusion of important factors known to impact patient 1. Pruitt B, Mason A. Epidemiological, demographic, and outcome charac-
outcomes, including LOS. For example, recent retrospective teristics of burn injury. In: Herndon D, editor. Total burn care. London:
Saunders Co; 1996. p. 13.
analysis of electronic medical records found that other im- 2. American Burn Association. Burn incidence and treatment in the United
portant variables, including but not limited to socioeconomic States: 2016, 2016, accessed 3 Dec. 2018; available from http://www.
status and key comorbid conditions beyond burns (clotting ameriburn.org/resources_factsheet.php.
3. Klein MB, Goverman J, Hayden DL et al. Benchmarking outcomes in the
disorders, anemia, and admission serum ethanol level) are critically injured burn population. Ann Surg 2014;259(5):833–41.
important variables associated with LOS not captured in the 4. Stone Ii R, Natesan S, Kowalczewski CJ et al. Advancements in regen-
erative strategies through the continuum of burn care. Front Pharmacol
NBR.32 Similar to the conclusions of the Smith et al study, we 2018;9:672.
also recommend that variables in the NBR be expanded to 5. Rowan MP, Cancio LC, Elster EA et al. Burn wound healing and treat-
capture more important concepts that help to better under- ment: review and advancements. Crit Care 2015;19:243.
6. Moores B, Rahman M, Settle J et al. On the predictability of the length
stand predictive factors for patient outcomes. of patient stay in a burns unit. Burns 1975;1(4):291–6.
7. Taylor S, Curri T, Lawless M, Sen S, Greenhalgh DG, Palmieri TL.
Predicting resource utilization in burn treatment. J Burn Care Res
CONCLUSIONS 2014;35 Suppl 2:S235–46.
8. Hussain A, Dunn KW. Predicting length of stay in thermal burns: a sys-
tematic review of prognostic factors. Burns 2013;39(7):1331–40.
This analysis uses regression of national aggregate data from 9. Meshulam-Derazon S, Nachumovsky S, Ad-El D, Sulkes J, Hauben DJ.
NBR to examine the predictive relationship between patient Prediction of morbidity and mortality on admission to a burn unit. Plast
and burn characteristics and number of procedures (debride- Reconstr Surg 2006;118(1):116–20.
10. Peck MD, Mantelle L, Ward CG. Comparison of length of hospital
ment, excision, and autograft) as well as LOS for surviving, stay to mortality rate in a regional burn center. J Burn Care Rehabil
severely burned patients. A key finding of this analysis is that 1996;17(1):39–44.
11. Saffle JR, Davis B, Williams P. Recent outcomes in the treatment of burn
the general rule of assuming one day of inpatient day per per- injury in the United States: a report from the American Burn Association
cent TBSA likely underestimates LOS for adults and pediatrics Patient Registry. J Burn Care Rehabil 1995;16(3 Pt 1):219–32; discus-
with large burns and overestimates for pediatrics with TBSA sion 288.
12. Wong MK, Ngim RC. Burns mortality and hospitalization time–a pro-
less than 20%, given a lack of consideration of other important spective statistical study of 352 patients in an Asian National Burn Centre.
factors such as HAI, exact patient age, depth of burn, sex, and Burns 1995;21(1):39–46.
inhalation injury. 13. Krob MJ, D’Amico FJ, Ross DL. Do trauma scores accurately
predict outcomes for patients with burns? J Burn Care Rehabil
These estimates also provide a benchmark against which 1991;12(6):560–563.
burn centers can compare the number of procedures and 14. Bowser BH, Caldwell FT, Baker JA, Walls RC. Statistical methods to pre-
dict morbidity and mortality: self-assessment techniques for burn units.
LOS for various patient profiles (eg, per age, TBSA). By Burns Incl Therm Inj 1983;9(5):318–26.
identifying patients with characteristics that lead to excess re- 15. American Burn Association. National Burn Repository. 2012 Report.
source use, burn centers can examine more closely why these Dataset Version 8.0.
16. American Burn Association. National burn repository (NBR), accessed
patients need more care, creating opportunities for more tai- 18 June 2018; available from https://ameriburn.org/quality-care/
lored care practices based on patient and burn characteristics. registry-programs/.
 Journal of Burn Care & Research
1044  Kruger et al September/October 2020

17. American Burn Association. National burn repository 2017 update: re- 25. Lim J, Liew S, Chan H et al. Is the length of time in acute burn surgery
port of the data from 2008–2017, 2018; accessed 8 Jan. 2019; available associated with poorer outcomes? Burns 2014;40(2):235–40.
from http://www.ameriburn.org/resources_factsheet.php. 26. Bilir P, Kruger E, Kowal S et al. Abstract: inpatient cost of acute care
18. Hussain A, Choukairi F, Dunn K. Predicting survival in thermal injury: a for severe burn patients: validation of economic model for adults and
systematic review of methodology of composite prediction models. Burns children. Baltimore (MD): ISPOR; 2018.
2013;39(5):835–50. 27. Kowal S, Kruger E, Bilir P et al. Cost-effectiveness of the use of autolo-
19. StataCorp. Stata statistical software: release 15. College Station (TX): gous cell harvesting device compared to standard of care for treatment of
StataCorp LLC; 2017; available from https://www.stata.com/support/ severe burns in the United States. Adv Ther 2019;7:1–5.
faqs/resources/citing-software-documentation-faqs/. 28. Bittner EA, Shank E, Woodson L, Martyn JA. Acute and perioperative
20. Faraway JJ. Extending the linear model with R: generalized linear, mixed care of the burn-injured patient. Anesthesiology 2015;122(2):448–64.
effects and nonparametric regression models. Vol 124. CRC Press; 2016. 29. Devgan L, Bhat S, Aylward S, Spence RJ. Modalities for the assessment of
21. Ng SK, Yau KKW, Lee AH. Modelling inpatient length of stay by a hier- burn wound depth. J Burns Wounds 2006;5:e2.
archical mixture regression via the EM algorithm. Math Comput Model 30. Hoeksema H, Van de Sijpe K, Tondu T et al. Accuracy of early burn depth
2003;37(3):365–75. assessment by laser Doppler imaging on different days post burn. Burns
22. Leung KM, Elashoff RM, Rees KS, Hasan MM, Legorreta AP. 2009;35(1):36–45.
Hospital- and patient-related characteristics determining maternity 31. Newman MC. Regression analysis of log‐transformed data: Statistical bias
length of stay: a hierarchical linear model approach. Am J Public Health and its correction. Environ Toxicol Chem 1993;12(6):1129–33.
1998;88(3):377–381. 32. Smith RR, Hill DM, Hickerson WL, Velamuri SR. Analysis of factors
23. Nau R. Statistical forecasting: notes on regression and time series analysis. impacting length of stay in thermal and inhalation injury. Burns

Downloaded from https://academic.oup.com/jbcr/article/41/5/1037/5813026 by guest on 09 April 2025

Durham: Fuqua School of Business, Duke University; 2017; accessed 23 Apr. 2019;45(7):1593–9.
2018; available from https://people.duke.edu/~rnau/411home.htm. 33. Foster K, Bilir P, Kruger E et al. Cost-effectiveness (CE) of an autologous
24. Park JH, Heggie KM, Edgar DW, Bulsara MK, Wood FM. Does the type regenerative epithelial suspension (RES) versus standard of care (SOC)
of skin replacement surgery influence the rate of infection in acute burn for treatment of severe burns in the United States: Maricopa burn hos-
injured patients? Burns 2013;39(7):1386–90. pital. New Delhi: ISBI; 2018.

Join us on Twitter to share your

insights on specific articles
in the most recent issue of
JBCR. Various topics for
discussion will be posted by
the moderators.

Follow @BurnJournalClub
and be sure to tag your
tweets with #BurnJC