burns 39 (2013) 1331–1340

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/burns

Review

Predicting length of stay in thermal burns: A systematic

review of prognostic factors

Amer Hussain, Ken W. Dunn *

University Hospital South Manchester, Southmoor Road, Manchester M23 9LT, United Kingdom

article info abstract

Article history: Background: Continued improvement in all aspects of the management of thermal injury
Accepted 23 April 2013 has resulted in marked improvements in the traditionally reported outcome of mortality.
This has resulted in the search for alternative parameters that can be monitored to indicate
Keywords: the performance of burn services. Length of stay (LOS) in hospitalised burn patients has long
Burns been considered reflective of injury-associated morbidity, cost and the quality of care, which
Length of stay can be monitored consistently across services.
Outcome Aim: We undertook a systematic review of published literature pertaining to LOS prognos-
Severity-adjustment tication in thermal burns to identify the relevant factors, quantify the risk associated with
Prediction these factors and identify predictive prognostic models.
Methods: Electronic searches were performed on MEDLINE, CINHAL, EMBASE, Web of
Science1, the Cochrane collection and a general web search was performed using Google1.
The searches were complemented by a manual search of the contents of leading burns
journals. Quality of the studies included in the review was evaluated against published
standards for prognostic studies.
Results: Fourteen studies were included in the review after meeting the inclusion/exclusion
criteria. Age and %TBSA were the strongest predictors of LOS in these studies. Other
significant predictors included % full thickness burn, female gender, inhalation injury,
surgery including escharotomy and the depth of burn. Nine studies reported multivariate
models for predicting LOS in patients sustaining thermal injury. None of these models were
validated and the goodness-of-fit statistic (R2) ranged from 0.15 to 0.75.
Conclusion: This review has demonstrated that %TBSA and age are the best predictors of LOS
in published literature. Current prognostic models do not explain a significant proportion of
variation in LOS.
# 2013 Elsevier Ltd and ISBI. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332
2. Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332
2.1. Literature search strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332
2.2. Inclusion and exclusion criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332

* Corresponding author at: Tel.: +44 161 291 6323; fax: +44 161 291 6323.
E-mail addresses: amh@doctors.net.uk (A. Hussain), ken.dunn@uhsm.nhs.uk (K.W. Dunn).
0305-4179/$36.00 # 2013 Elsevier Ltd and ISBI. All rights reserved.
http://dx.doi.org/10.1016/j.burns.2013.04.026
1332 burns 39 (2013) 1331–1340

2.3. Quality assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332

2.4. Data synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333
3.1. Quality assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333
3.2. Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336
4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336
4.1. Limitations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1338
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1338
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1339

1. Introduction EMBASE (1980-September 2011) databases via the NHS

Evidence interface. Currently there is no widely acknowl-
Most descriptions of measuring performance and outcomes in edged optimum search strategy for identifying prognostic
burn care have traditionally looked at mortality [1–5]. This is studies and using both natural language and MeSH terms
primarily because it is easy to define and measure with reliable according to published recommendations for identifying
data available over many decades allowing useful comparison prognostic studies optimised the database search [31]. Key
between time periods. However with continued improve- terms explored were: burns, outcome, length of stay,
ments in all aspects of the management of burn mortality prognosis, prediction, index, scoring and model. The search
rates have shown marked improvement in developed coun- was restricted to English language publications. The titles
tries from 54 to 100% at the beginning of the twentieth century and abstracts were reviewed and the list of studies for
to currently reported rates of around 4–6% [6–11]. As a result, inclusion in the review was narrowed by evaluation against
the usefulness of mortality rates as a measure of quality of the inclusion/exclusion criteria. Searches were also con-
care and burn service performance is increasingly being ducted on the Web of Science1 database, the Cochrane
questioned [12–14]. Outcomes such as quality of life and collection and a general web search was performed using
functional status are instead being examined as long-term Google1. The bibliographies of the retrieved full text articles
consequence of burn care [15–21]. At present these outcomes were also reviewed to identify any missed publications. The
are difficult to measure in a consistent, efficient and electronic and bibliography searches were complemented
meaningful manner across different services, age groups by a manual search of the contents of leading burns
and ethnic groups. Alternatively LOS data is easy to collect and journals: Burns (1974-September 2011), the Journal of Burn
measure across different services. The initial descriptions of Care and Rehabilitation (1980–2005) and Journal of Burn Care
LOS in hospitalised burn patients focused on the analysis of and Research (2006-September 2011).
this outcome to predict burn service bed utilisation [22,23].
More recently it has been shown to provide an indirect 2.2. Inclusion and exclusion criteria
indication of morbidity and the incidence of clinical complica-
tions, and a direct correlation with functional and aesthetic To be included in the review, studies must have produced a
outcomes as well as cost of care associated with thermal prognostic estimation of LOS by identifying possible risk
injury [24–28]. Several cohort studies have identified risk factors. In studies that devised or employed prognostic
factors that have an impact on LOS [29,30]. Risk stratification model(s), this model must have predicted LOS. Only multivar-
has many uses, including the identification of high-risk iate studies were included in the review and both retrospective
patient groups, guiding clinical decision-making, patient/ and prospective studies were included. Studies involving
family counselling, and the evaluation of new therapies and patients of all ages and burn sizes were considered but those
techniques. Risk-adjusted prediction models that allow the focusing on extremes of age (children and/or elderly) were
comparison of observed and expected LOS while taking excluded to allow evaluation of comparable study popula-
patient and injury characteristics into account have consider- tions. Studies focusing primarily on non-thermal burns (e.g.
able value in monitoring the performance of services. chemical, electrical) and those investigating the effect of
The aims of this study were: individual factors on LOS were excluded. Studies from
multiple etiologies were included if thermal burns comprised
 to undertake a systematic review of published literature the majority of the patient population (>50%) but those
pertaining to LOS prognostication in thermal burns, focusing on specific causation patterns (e.g. self inflicted
 to identify prognostic factors that impact on LOS, burns) were excluded. Those studies reporting composite
 to quantify the risk associated with these factors, prediction models were excluded unless the model predicted
 to identify predictive prognostic models. LOS as the outcome measure.

2. Materials and methods 2.3. Quality assessment

2.1. Literature search strategy The methodological quality, internal and external validity and
applicability of the studies in the review were assessed using
An electronic search was performed on MEDLINE (1950- guidelines published by the National Institute for Health and
September 2011), CINHAL (1981-September 2011) and Clinical Excellence for evaluating cohort studies [32].
 burns 39 (2013) 1331–1340 1333

2.4. Data synthesis
Data extraction and analysis were performed on specifically
designed forms and the results summarised in tabular form.
Meta-analysis of the included studies was not possible
because of the wide variation in design, data analysis and
reporting of results between the studies.
3. Results
Of the 24 full text articles reviewed, 14 fulfilled the inclusion
and exclusion criteria (Fig. 1) [23,24,26,29,30,33–41]. The
excluded studies comprised 4 studies that focused on patients
at extremes of age [42–45]. Four studies were excluded, as they
did not undertake multivariate analysis [22,46–48]. A further 2
studies were excluded for limiting the patient populations to
specific causation patterns [49,50].
3.1. Quality assessment
The main outcome of each of the studies was clearly defined
i.e. multivariate analysis of the correlation of independent
variables with LOS. Patient numbers and characteristics (age,
%TBSA and gender) were described for the populations in each
of the included studies. These are representative of the
spectrum of burn patients receiving acute care in burn centres.
Twelve studies (86%) analysed LOS in a cohort of thermal burn
patients admitted to specialised burn centres while 2 studies
undertook a review of pooled data from multiple centres
contributing to a national database [26,39]. All study popula-
tions comprised acute burns and in 3 studies clinical and
demographic data was collected and analysed prospectively
[34–36]. Predictive variables investigated were clearly de-
scribed in all studies. Blinding of the outcome, i.e. LOS, from
the individuals undertaking the analysis was not addressed in
any of the studies. The ‘‘follow-up’’ was complete for all
patients with no losses.
The acute treatment of patients sustaining burns was
described in 21% (3/14) studies [34,37,38]. None of the studies
undertook specific interventions. Discharge for the purposes
of LOS calculation was defined in only 28% (4/14) studies; it was
defined as hospital discharge in 3 studies [35,38,39]. Only two
studies defined discharge in terms of wound healing, with one
defining this criterion as 98% of burn wound closed to allow
discharge, and the other defining it as sufficient healing to
allow further wound management in the community (by
general practitioner and in outpatient ward) (Table 1) [30,38].
Study population exclusions were described in 71% (10/14)
studies with all deaths excluded from the analysis in 6 studies
[23,30,34–36,39] and deaths prior to surgery excluded from 1
study [29] (Table 1). Separate multivariate analysis was
performed for survivors and non-survivors in one study [38].
Power calculations were performed in 2 studies [30,35].
Multivariate regression analysis was performed in all
studies. None of the studies reported odds ratios for the
identified prognostic factors as a part of the regression
analysis. All studies reported results using ‘‘p’’ values and
all studies describing prognostic models for predicting LOS
reported goodness-fit-statistics (R2) for the models.

Electronic database
Bibliography search searches
n=1 n= 3981

Potential studies
n= 134 Hand search of
General web search relevant journal
n= 5,890,000 contents
First thousand reviewed n= 1

Full text retrieved

n= 24
10 studies excluded
• studies focusing on
patients at extremes of age
(4)
• studies not undertaking
multivariate analysis (4)
• studies limited to specific
injury causation (2)

14 studies included in
review

Fig. 1 – Search strategy and results.

 1334
Table 1 – Summary of prognostic studies of LOS in hospitalised thermal injury patients.

Author(s) Year Type of study Study population Age (years) TBSA (%) Mean LOS (days) Mortality (%)
Moores et al. [33] 1975 Retrospective cohort 495 72.2% patients <40 years 52% patients <20% NS 14.3
study
Sanderson et al. [23] 1981 Prospective cohort 92 54.3% patients <30 years 57.6% patients <20% TBSA NS NS
study
Bowser et al. [24] 1983 Retrospective cohort 806 7  6 (paediatric); 43  21 14  16 (paediatric); 28  29 (paediatric); 1.4 (paediatric);
study (adult, elderly) 29  26 (adult, elderly) 34  36 (adult, elderly) 25 (adult)
Krob et al. [40] 1991 Retrospective cohort 511 31 8.6% (range 0–100%) 10 4.1
study
Wong and Nigam [34] 1995 Prospective cohort 336 29.5 3.3 NS 4.5
study
Saffle et al. [26] 1995 Retrospective database 6417 80% patients >50 years 14.12  16.6 11 days non-survivor; 5.1
review 13.7 survivors
Peck et al. [35] 1996 Prospective cohort 196 45 9.2  11 13.9 7.7
study
Attia et al. [36] 2000 Prospective cohort 533 22.95  16.71 90% survivors <40% 16.2  19.5 survivors; 33
study TBSA; 90% non- 5.3  5.1 non-survivors

burns 39 (2013) 1331–1340

survivors >40%
TBSA
Ho et al. [37] 2002 Retrospective cohort 286 23 18 21 8.7
study
Meshulam-Derazon 2006 Retrospective cohort 249 36  23 14  15 22.9  17.1 5.2
et al. [29] study
Gravante et al. [30] 2007 Retrospective cohort 261 45.5  26.3 23.2  19.5 28  22.7 6
study
Andel et al. [38] 2007 Retrospective cohort 365 41 (IQR 28–56) survivors: 12 (IQR 6–22) 18 (IQR 12–23) survivors: 24
study 58 (IQR 43–76) non- survivors: 45 (IQR 8 (IQR 4–14) non-survivors
survivors 31–70) non-survivors
Chong et al. 2009 Retrospective cohort 482 35  19.6 13.5  18 8.61  7.02 patients 4.5
study undergoing surgery;
26.3  18.8 patients not
undergoing surgery
Johnson et al. 2010 Retrospective database 52,712 53% patients <30 years Mean not calculated NS NS
review

Author(s) Definition of Exclusions Indexes analysed Type of Prediction Statistically significant Comments
discharge analysis model predictors
produced
Moores et al. [33] N Non-lethal burns with LOS N Multivariate Y Age, %TBSA, % full R2 = 0.158
<2 days; burns <4% TBSA regression thickness burn, % partial
thickness burn,
Pseudomonas infection
Sanderson N Admissions >7 days N Multivariate Y %TBSA, % full thickness burn, R2 = 0.572
et al. [23] post-burn, deaths regression skin grafting, feet burn,
respiratory complications,
cardiovascular complications,
other complications
(metabolic, reconstructive), sepsis
Bowser et al. [24] N None N Multivariate Y Age, %TBSA, % full R2 paediatric
linear thickness burn, female LOS model
regression gender, escharotomy, = 0.592; R2 adult LOS
intravenous resuscitation, model = 0.561
colloid resuscitation
Krob et al. [40] N None Baux Score, Multiple N Baux Score, Zawacki Score, ABSI –
Zawacki Score, ABSI logistic
regression
Wong and N 16 deaths N Multivariate Y Age, %TBSA, %full thickness burn, R2 = 0.57
Nigam [34] linear interval between injury and
regression admission, type of burn,
inhalation injury, place
of injury
Saffle et al. [26] N Patients under N Multivariate linear Y Age, %TBSA, % full R2 = 0.695

burns 39 (2013) 1331–1340

15 years age regression thickness burn,
female gender, inhalation injury,
surgical procedures,
ventilator support days, death
Peck et al. [35] Hospital discharge Patients admitted for N Multivariate Y %TBSA, length of intensive R2 = 0.75
reconstructive surgery, linear care stay, length of antibiotics
deaths regression use, comorbidites,
concomitant injuries
Attia et al. [36] N 176 deaths N Multivariate Y Inhalation injury, R2 = 0.444
linear depth of burn, female
regression gender, %TBSA, degree
of burn, clothing ignition
Ho et al. [37] N None N Multivariate Y %TBSA, inhalation R2 = 0.2
linear injury, female gender
regression
Meshulam-Derazon N Conservatively treated N Multivariate Y Age, %TBSA, degree of R2 = 0.4
et al. [29] burns; deaths prior to surgery linear burn, inhalation injury,
regression time to surgery
Gravante et al. [30] 98% wound healed 38 (16 deaths, 15 electrical Baux Score, Unit Burn Multivariate N Age, %TBSA, %full thickness Roi’s index most
burns, 7 chemical burns) Standard (UBS) index, linear burn, caloric significant
Roi’s model, ABSI regression delivery, protein delivery, predictor R2 = 0.054
Baux Score, Unit Burn
Standard (UBS)
index, Roi’s model, ABSI
Andel et al. [38] Hospital discharge, None ABSI Multivariate N Age, %TBSA Total ABSI score not a
wound healed allowing linear significant predictor
community regression of LOS
management

1335
1336 burns 39 (2013) 1331–1340

3.2. Summary of results

0.03 days for burns >20%

0.01 days for all burns,
if surgery not required
LOS significantly less

LOS/%TBSA = 0.94 

LOS/%TBSA = 1.08 
The results of the included studies are shown in Table 1.
Comments

Multivariate analysis showed %TBSA (12 studies) and age (8

studies) to be the strongest predictors of LOS. Other significant
predictors include % full thickness burn (6 studies), female
gender (5 studies), inhalation injury (5 studies), surgery

TBSA
including escharotomy (3 studies) and depth of burn (2
studies). Less frequently cited statistically significant pre-
dictors of LOS are listed in Table 1. As none of the included
Statistically significant

studies reported odds ratios it is not possible to quantify the

surgery, bacterial growth,
Age <6 years, occupation

risk associated with the identified prognostic factors.

(international patients)
predictors

%TBSA, female gender

Nine studies produced risk-adjusted prediction models

(military personnel),

based on prognostic factors identified in the multivariate

analysis (Table 2) [23,24,26,29,33–37]. None of the studies
referral source

undertook validation of the produced models. R2 for these

models ranged from 0.158 to 0.75 [33,35] (Table 1).
Three studies analysed the ability of burn-specific mortali-
ty prognostic models to predict LOS. One study demonstrated
Prediction

that Zawacki Score [51], Abbreviated Burn Severity Index

produced
model

(ABSI) [52], Baux Score [53] are better predictors of LOS

compared to generic trauma scores (Injury Severity Score,
N

Trauma Score, Glasgow Coma Score) ( p < 0.01) [40]. One study
showed Baux Score, ABSI, Roi’s model [54] and Unit Burn
Standard (UBS) index [55] to be significant predictors of LOS
analysis
Type of

Multivariate

LOS (days) = 1  %TBSA Multivariate

with Roi’s model being the most significant predictor (R2 0.054)
regression
analysis

[30]. One study found variables of age and %TBSA to be

linear

significant predictors of LOS compared to total ABSI score

( p < 0.01) [38].
One study assessed the predictive accuracy of a ‘‘rule of
Indexes analysed

thumb’’ model [LOS (days) = 1  %TBSA] in data from 52,712

patients [39]. It showed LOS was 0.94  0.01 days for all burns,
and 1.08  0.03 days for burns >20%TBSA. Although neither
value was exactly 1 day, they were not sufficiently different
from 1 to statistically invalidate this rule of thumb.
N

4. Discussion
50 patients with overnight

outpatients; records with

Deaths; re-admissions;

missing variables (age,

Exclusions

TBSA, outcome, LOS)

Significant advances in all aspects of burn care over the second

half of last century have dramatically reduced the overall
admission only

mortality associated with burn [10,11]. These reductions have

rendered obsolete the role of traditional prediction and
monitoring of mortality as the sole means of benchmarking
the quality of care [9,10,52–54]. LOS for hospitalised burn
patients offers an alternative parameter that is easy to collect
in a consistent manner across different services provided
criteria for hospital discharge are defined and standardised. It
Definition of
discharge

also provides an indication of the short and long-term

consequences of burn care [24–28]. LOS has the additional
discharge

advantage that it is a continuous, quantitative variable. As

Hospital

such, a study attempting multivariate analysis of prognostic

Table 1 (Continued )

risk factors would require a smaller sample size to gain

N

sufficient statistical power compared to mortality [35,38]. The

identification of risk factors can aid in targeting clinical care to
Johnson et al.

high-risk patients, risk-stratification for the purposes of

Chong et al.
Author(s)

research and contribute to improving the quality of care

through comparisons between patient groups and services.
This systematic review has demonstrated that LOS is a
useful alternative outcome measure to mortality in thermally
 burns 39 (2013) 1331–1340 1337

Table 2 – Composite LOS prediction models.

Authors
Moores et al. [33]
Sanderson et al. [23]
Bowser et al. [24]
Wong and Nigam [34]
Saffle et al. [26]
Peck et al. [35]
Attia et al. [36]
Ho et al. [37]
Meshulam-Derazon et al. [29]
Models
LOS = 46.66 + 2.424 (%FTBSA) + 0.501
(%PTBSA) + 14.84 (pseudomonas infection) + 63.87
(survival) + 0.00298 (age2) 0.0247 (%FTBSA2)
LOS = 18.16 + 0.9 (%third degree burn) + 8.95 (skin
grafting) + 0.002 (%TBSA) + 7.56 (feet burn) + 11.14
(respiratory complications) + 6.61 (other
complications)
LOS (paediatric) = 11.47 + 0.9159 (%TBSA) + 1.082
(%FTBSA); LOS (adult) = 10.03 + 1.767
(%FTBSA) + 0.046 (%TBSA) 13.87 (escharot-
omy) + 7.11 (gender) 11.89 (IV resuscita-
tion) + 25.77 (colloid resuscitation) + 0.17 (age)
LOS = 1.90 + 0.93 (%TBSA) + 3.2 (%FTBSA) + 0.14
(age) + 6.97 (status of respiratory injury)
LOS = 2.2182 + 5.387 (surgical procedures) + 0.6991
(days ventilator support) + 0.3216 (%TBSA) 20.96
(disposition) + 0.0901 (age) 1.2392
(gender) 0.0327 (full thickness burn) 0.0025
(inhalation injury)
LOS = burns size + smoke inhaltion + intensive
care LOS + length of anitbiotic use + other medical
conditions + other trauma
LOS = 17.81 + 6.89 (inhalation injury) + 7.74
(depth) + 4.31 (gender) + 0.365 (%TBSA) + 1.95 (de-
gree) + 5.27 (clothing ignition)
LOS = 8.7 + 2.1 (%TBSA) 0.0018 (%TBSA2)) + 1.67
(inhalation injury) 9.4 (gender)
LOS = 20.3 + 0.58 (%TBSA) + 10.9 (time to opera-
tion) + 14.1 (smoke inhalation) + 5.5 (degree
of burn) + 4.7 (age group); bedside LOS
model = 18 + (%TBSA)/3
Comments
Pseudomonas infection = 1 if present other-
wise 0; survival = 1 if yes otherwise 0;
PTBSA = partial thickness burn surface area;
FTBSA = full thickness burn surface area
Skin grafting, feet burn, respiratory
complication and other complications
= 1 if present otherwise 0
Escharotomy = 1 if yes otherwise 0; IV (in-
travenous) resuscitation = 1 if yes otherwise
0; colloid resuscitation 1 if yes otherwise 0
Status of respiratory injury = 1 if present
otherwise 0
Disposition = 1 if died otherwise 0; gender = 1
if male otherwise 0; inhalation injury = 1 if
present otherwise 0
Discrete variables (smoke inhalation, other
medical conditions and other trauma) scored
as present (Y) or absent (N). Continuous
variables = burn size, intensive care LOS,
length of antibiotic use
Gender = 1 if female otherwise 0; inhalation
injury = 1 if present otherwise 0; depth = 1 if
deep, 0 if superficial; clothing ignition = 1 if
present otherwise 0; degree = not coded
Inhalation injury = 1 if present otherwise 0;
gender = 1 if female, 2 if male
Time to operation = 1 if <7 days post admis-
sion otherwise 2; smoke inhalation = 1 if
present otherwise 0; degree of burn = 2 if
second degree, 2.5 if mixed second and third
degree, 3 if third degree; age group = 1 if 17
years or less, otherwise 2

injured patients. There have been several good quality studies
investigating prognostic factors for LOS. The most commonly
identified significant prognostic factors being %TBSA and age.
Other significant factors include % full thickness burn, female
gender, inhalation injury, surgery and depth of burn. However
wide variations in study designs and reporting of results limit
the quantification of risk associated with these factors. In
addition, multivariate analysis of LOS in studies included both
survivors and non-survivors, which can skew prognostic
findings by virtue of non-survivors usually having sustained
severe injuries [24,26,29,33,37,40,41]. Also with significant
reductions in mortality it is inappropriate to undertake
multivariate analysis including both survivors and non-
survivors. It is preferable to exclude non-survivors from the
analysis to identify prognostic factors for LOS in survivors
[23,30,34–36,39] or to divide them in two separate analyses [38].
The review has also identified risk-adjusted prognostic
models to predict LOS in hospitalised burn patients with no
studies reporting the validation of these models
[23,24,26,29,33–37] (Table 2). The generalisation and applica-
bility of these models are limited by the lack of a standardised
criteria constituting discharge following burn care with only
one study defining it as hospital discharge rather than
discharge from the burn unit [35]. Local healthcare systems,
culture, pathways and clinical behaviour dictate discharge
practices and reported data. The LOS for less severe burns can
be paradoxically prolonged because of confounding factors
such as co-morbidities and social circumstances. It can be
shortened by discharging patients from an acute setting, only
to immediately admit them to a rehabilitation service, or by
discharging them for overnight leave at home, only to readmit
them the next day. In the UK and continental Europe the
availability of inpatient rehabilitation beds is low and most
patients admitted to burn services are discharged to either
their own residence or a permanent care facility once fit for
discharge. Hence in reports originating from these countries
burn centre LOS is nearly equal to total inpatient LOS. In the
USA however an estimated 37% of burn treatment facilities
have designated inpatient rehabilitation beds, 42% patients
transferred to these beds have a TBSA >40% and the LOS for
patients in these beds ranges from 3 to 100 days [56,57]. Hence
patients can be discharged to rehabilitation beds from the
burn centre still receiving a considerable period of inpatient
care but the burn centre LOS is paradoxically low.
The identified prediction models only explain 15–75%
variation in LOS for the study populations (R2 0.15–0.75). This
1338 burns 39 (2013) 1331–1340

would suggest that prognostic factors not analysed in these strategy for identifying prognostic studies and although a
studies might explain a significant proportion of variation in search strategy of high sensitivity was used it is possible that
LOS and these need to be identified to improve the accuracy of some studies may have been missed.
prediction models [27,58].
Constructing models to predict LOS is complicated in
conditions such as burns, which are associated with a 5. Conclusion
significant mortality risk because factors that are associated
with predicting mortality are also linked to morbidity and LOS. This review has demonstrated that prognostic factors associ-
A recent systematic review of mortality prediction models in ated with LOS have been investigated in several good quality
burns has interestingly identified all the prognostic factors for studies. These studies show that %TBSA and age are the best
LOS identified in this review as also being significant predictors of LOS in published literature. Other significant
independent predictors of mortality [59]. As a consequence, prognostic factors include % full thickness burn, female
the accuracy of risk-adjusted models predicting mortality has gender, inhalation injury, surgery and depth of burn. The
been assessed for their ability to predict LOS [30,38,40]. These literature however does not contain any reports of ‘‘risks’’ or
studies have demonstrated that mortality prediction models ‘‘odds’’ associated with these factors making it impossible to
(Baux Score, Zawacki Score, ABSI, Roi’s model, UBS) by quantify the risk. The review has also identified prognostic
incorporating variables such as age, %TBSA and inhalation models for predicting LOS but currently these models do not
are also significant predictors of LOS (Table 1). However explain a significant proportion of the variation in LOS. These
Gravante et al. showed in multivariate analysis of multiple results are useful for general counselling of patients/families
mortality prediction models (Baux Score, ABSI, Roi’s model, as well as guiding further research into outcome prognostica-
UBS) that Roi’s model was the only one that predicted LOS [30]. tion and benchmarking of burn services.
This correlation was direct with every half-point increase in If LOS is to be used as robust outcome measure further
the index corresponding to approximately 10 days LOS. research should focus on larger studies with standardised
Significantly this model did not predict the observed LOS in definitions of what constitutes discharge from burn care.
94.5% of cases (R2 0.054). Andel et al. also demonstrated that These should preferably involve all age and TBSA variations to
although age ( p = 0.0002) and %TBSA ( p < 0.000) significantly
increased hospital LOS, a similar relationship was not seen
between total ABSI score and hospital LOS ( p = 0.4725) [38].
These studies further highlight that prognostic factors other Table 3 – Recommendations for future studies.
than those incorporated in mortality prediction models
1. Study type
impact on LOS. Retrospective
In the absence of accurate prognostic models a crude Prospective
standardisation of LOS is to express it as a function of burn size
2. Source of data
i.e. LOS (days)/%TBSA allowing comparison between services.
Case notes
This ratio has previously been employed to indicate the Institutional database
efficiency of burn care with a ratio of <1 being described as the Regional/National database
goal of burn treatment [26,60] for survivors of burn injury. A
3. Patient population
recent analysis of the 52,712 records from the North American Age
National Burn Repository (NBR) has demonstrated the linear %TBSA/%FTBSA/%PTBSA
relationship of %TBSA with LOS [39]. This analysis, which did Inhalation injury
not provide a definition of what constituted ‘‘discharge’’ from Gender
hospital or burn service, demonstrated 1 day LOS/%TBSA to be Type of burn
Injury management protocols
an achievable goal for small to moderate size burns but the
Inclusion and exclusion criteria for patients
same does not hold true for burns >60% TBSA which received
additional days. This may be related to complications, % full 4. LOS
Clear definition of ‘‘discharge’’
thickness burn, or inhalation injury that are all likely to be
Discharge pathway
associated with large burns. Though the 1 day/%TBSA rule
Total inpatient LOS
represents a useful starting point for estimating LOS, a model ICU LOS
that accurately characterises LOS for surviving burn patients Burn centre LOS
has considerable applications for benchmarking quality of Rehabilitation LOS
care and evaluation of therapies. 5. Statistical analysis
Missing data declared
4.1. Limitations Type of analysis
Significant variables predicting LOS
This review was limited to English-language publications and Risk ratios attributable to identified factors
Selection protocol for variables in modela
may have excluded relevant publications in other languages.
Interactions between selected variablesa
Prognostic studies are more prone to publication bias and as
‘‘Goodness-of-fit’’ of modela
only published material was used, this review, in common Model cross-validationa
with all systematic reviews, is prone to publication bias. a
For studies reporting LOS prediction models.
Currently there is no widely acknowledged optimum search
 burns 39 (2013) 1331–1340
investigate additional predictive factors of the LOS that could
help develop more accurate and comprehensive prognostic
models. To achieve these objectives we recommend a
minimum reporting dataset for such studies based on
methodological and reporting standards for prognostic studies
[31,61–65] (Table 3).
Conflict of interest
The authors declare that we have no conflicts of interest such
as employment, consultancies, stock ownership, honoraria,
paid expert testimony, patent applications/registrations,
grants or other funding.
references
[1] Bull JP, Squire JR. A study of mortality in a burns unit;
standards for the evaluation of alternative methods of
treatment. Ann Surg 1949;130:160–73.
[2] Bull JP, Fisher AJ. A study of mortality in a burns unit: a
revised estimate. Ann Surg 1954;139:269–74.
[3] Bull JP. Revised analysis of mortality due to burns. Lancet
1971;2:1133.
[4] Feller I, Crane K, Flanders S. Baseline data on the mortality
of burn patients. QRB Qual Rev Bull 1979;5:4–8.
[5] Rashid A, Khanna A, Gowar JP, Bull JP. Revised estimates of
mortality from burns in the last 20 years at the Birmingham
Burns Centre. Burns 2001;27:723–30.
[6] Sneve H. The treatment of burns and skin grafting. JAMA
1905;45:1–8.
[7] Dunbar J. Review of the burn cases treated in the Glagow
Royal Infirmary during the past hundred years (1833–1933)
with some observations on present day treatment. Glasgow
Med J 1934;122:239–55.
[8] Steinvall I, Fredrikson M, Bak Z, Sjoberg F. Mortality after
thermal injury: no sex-related difference. J Trauma
2011;70:959–64.
[9] Osler T, Glance LG, Hosmer DW. Simplified estimates of the
probability of death after burn injuries: extending and
updating the Baux score. J Trauma 2010;68:
690–7.
[10] Belgian Outcome Burn Injury Study Group. Development
and validation of a model for prediction of mortality in
patients with acute burn injury. Br J Surg 2009;96:
111–7.
[11] Miller SF, Bessey P, Lentz CW, Jeng JC, Schurr M, Browning
S. National burn repository 2007 report: a synopsis of the
2007 call for data. J Burn Care Res 2008;29:862–70.
discussion 71.
[12] Pereira C, Murphy K, Herndon D. Outcome measures in
burn care. Is mortality dead? Burns 2004;30:761–71.
[13] Jaskille AD, Shupp JW, Pavlovich AR, Fidler P, Jordan MH,
Jeng JC. Outcomes from burn injury-should decreasing
mortality continue to be our compass? Clin Plast Surg
2009;36:701–8.
[14] Shakespeare PG. Prognostic indicators and burns. Burns
2003;29:105–6.
[15] Druery M, Brown TL, Muller M. Long term functional
outcomes and quality of life following severe burn injury.
Burns 2005;31:692–5.
[16] Pope SJ, Solomons WR, Done DJ, Cohn N, Possamai AM.
Body image, mood and quality of life in young burn
survivors. Burns 2007;33:747–55.
1339
[17] Cromes GF, Holavanahalli R, Kowalske K, Helm P.
Predictors of quality of life as measured by the Burn
Specific Health Scale in persons with major burn injury. J
Burn Care Rehabil 2002;23:229–34.
[18] Anzarut A, Chen M, Shankowsky H, Tredget EE. Quality-of-
life and outcome predictors following massive burn injury.
Plast Reconstr Surg 2005;116:791–7.
[19] Kvannli L, Finlay V, Edgar DW, Wu A, Wood FM. Using the
Burn Specific Health Scale – brief as a measure of quality of
life after a burn – what score should clinicians expect?
Burns 2011;37:54–60.
[20] Xiao J, Cai BR. Functional and occupational outcome in
patients surviving massive burns. Burns 1995;21:415–21.
[21] van Baar ME, Essink-Bot ML, Oen IM, Dokter J, Boxma H,
van Beeck EF. Functional outcome after burns: a review.
Burns 2006;32:1–9.
[22] Curreri PW, Luterman A, Braun Jr DW, Shires GT. Burn
injury. Analysis of survival and hospitalization time for 937
patients. Ann Surg 1980;192:472–8.
[23] Sanderson LM, Buffler PA, Perry RR, Blackwell SJ. A
multivariate evaluation of determinants of length of stay in
a hospital burn unit. J Burn Care Rehabil 1981;2:142–9.
[24] Bowser BH, Caldwell FT, Baker JA, Walls RC. Statistical
methods to predict morbidity and mortality: self assessment
techniques for burn units. Burns Incl Therm Inj 1983;9:318–26.
[25] Manson WL, Pernot PC, Fidler V, Sauer EW, Klasen HJ.
Colonization of burns and the duration of hospital stay of
severely burned patients. J Hosp Infect 1992;22:55–63.
[26] Saffle JR, Davis B, Williams P. Recent outcomes in the
treatment of burn injury in the United States: a report from
the American Burn Association Patient Registry. J Burn Care
Rehabil 1995;16:219–32. discussion 88-9.
[27] Gravante G, Montone A, Esposito G. Length of
hospitalization: an important parameter for burned
patients. J Burn Care Res 2007;28:537–8.
[28] Gillespie R, Carroll W, Dimick AR, Haith L, Heimbach D,
Kibbee E, et al. Diagnosis-related groupings (DRGs) and
wound closure: roundtable discussion. J Burn Care Rehabil
1987;8:199–209.
[29] Meshulam-Derazon S, Nachumovsky S, Ad-El D, Sulkes J,
Hauben DJ. Prediction of morbidity and mortality on
admission to a burn unit. Plast Reconstr Surg 2006;118:
116–20.
[30] Gravante G, Delogu D, Esposito G, Montone A. Analysis of
prognostic indexes and other parameters to predict the
length of hospitalization in thermally burned patients.
Burns 2007;33:312–5.
[31] Altman DG. Systematic reviews in health care: systematic
reviews of evaluations of prognostic variables. BMJ
2001;323:224–8.
[32] NICE National Institute for Health and Clinical Excellence
Guidelines Manual Appendix D-Methodology Checklist:
Cohort Studies; 2007.
[33] Moores B, Rahman MM, Settle JA, Browning FSC. On the
predictability of the length of patient stay in a burns unit.
Burns 1975;1:291–6.
[34] Wong MK, Ngim RC. Burns mortality and hospitalization
time – a prospective statistical study of 352 patients in an
Asian National Burn Centre. Burns 1995;21:39–46.
[35] Peck MD, Mantelle L, Ward CG. Comparison of length of
hospital stay to mortality rate in a regional burn center. J
Burn Care Rehabil 1996;17:39–44.
[36] Attia AF, Reda AA, Mandil AM, Arafa MA, Massoud N.
Predictive models for mortality and length of hospital stay
in an Egyptian burns centre. East Mediterr Health J
2000;6:1055–61.
[37] Ho WS, Ying SY, Burd A. Outcome analysis of 286 severely
burned patients: retrospective study. Hong Kong Med J
2002;8:235–9.
1340 burns 39 (2013) 1331–1340
[38] Andel D, Kamolz LP, Niedermayr M, Hoerauf K, Schramm
W, Andel H. Which of the abbreviated burn severity index
variables are having impact on the hospital length of stay? J
Burn Care Res 2007;28:163–6.
[39] Johnson LS, Shupp JW, Pavlovich AR, Pezzullo JC, Jeng JC,
Jordan MH. Hospital length of stay – does 1% TBSA really
equal 1 day? J Burn Care Res 2011;32:13–9.
[40] Krob MJ, D’Amico FJ, Ross DL. Do trauma scores accurately
predict outcomes for patients with burns? J Burn Care
Rehabil 1991;12:560–3.
[41] Chong SJ, Song C, Tan TW, Kusumawijaja G, Chew KY.
Multi-variate analysis of burns patients in the Singapore
General Hospital Burns Centre (2003–2005). Burns
2009;35:215–20.
[42] Sheridan R, Weber J, Prelack K, Petras L, Lydon M,
Tompkins R. Early burn center transfer shortens the length
of hospitalization and reduces complications in children
with serious burn injuries. J Burn Care Rehabil 1999;20:
347–50.
[43] Lumenta DB, Hautier A, Desouches C, Gouvernet J, Giorgi R,
Manelli JC, et al. Mortality and morbidity among elderly
people with burns – evaluation of data on admission. Burns
2008;34:965–74.
[44] Pomahac B, Matros E, Semel M, Chan RK, Rogers SO,
Demling R, et al. Predictors of survival and length of stay in
burn patients older than 80 years of age: does age really
matter? J Burn Care Res 2006;27:265–9.
[45] Thombs BD. Patient and injury characteristics, mortality
risk, and length of stay related to child abuse by burning:
evidence from a national sample of 15,802 pediatric
admissions. Ann Surg 2008;247:519–23.
[46] Pavlovich AR, Shupp JW, Jeng JC. Is length of stay linearly
related to burn size? A glimmer from the national burn
repository. J Burn Care Res 2009;30:229–30.
[47] Frye KE, Izenberg SD, Williams MD, Luterman A. Simulated
biologic intelligence used to predict length of stay and
survival of burns. J Burn Care Rehabil 1996;17:540–6.
[48] Yang CS, Wei CP, Yuan CC, Schoung JY. Predicting the
length of hospital stay of burn patients: comparisons of
prediction accuracy among different clinical stages. Dec
Support Syst 2010;50:325–35.
[49] Albayrak Y, Cakir C, Albayrak A, Aylu B. A comparison of
the morbidity and mortality of tandir burns and non-tandir
burns: experience in two centers. Ulus Travma Acil Cerrahi
Derg 2011;17:323–8.
[50] Thombs BD, Bresnick MG. Mortality risk and length of stay
associated with self-inflicted burn injury: evidence from a
national sample of 30,382 adult patients. Crit Care Med
2008;36:118–25.
[51] Zawacki BE, Azen SP, Imbus SH, Chang YT. Multifactorial
probit analysis of mortality in burned patients. Ann Surg
1979;189:1–5.
[52] Tobiasen J, Hiebert JM, Edlich RF. The abbreviated burn
severity index. Ann Emerg Med 1982;11:260–2.
[53] Baux S, Mimoun M, Saade H, Lioret N, Esteve M, Nolland XB,
et al. Burns in the elderly. Burns 1989;15:239–40.
[54] Roi LD, Flora Jr JD, Davis TM, Wolfe RA. Two new burn
severity indices. J Trauma 1983;23:1023–9.
[55] Sachs A, Watson J. Four years’ experience at a specialised
burns centre, The McIndoe Burns Centre 1965–68. Lancet
1969;1:718–21.
[56] Cromes GF, Helm PA. The status of burn rehabilitation
services in the United States: results of a national survey. J
Burn Care Rehabil 1992;13:656–62.
[57] Sliwa JA, Heinemann A, Semik P. Inpatient rehabilitation
following burn injury: patient demographics and functional
outcomes. Arch Phys Med Rehabil 2005;86:1920–3.
[58] Gravante G, Delogu D, Esposito G, Montone A. The length of
hospitalization and morbidity of burned patients. Plast
Reconstr Surg 2007;120:362. author reply-3.
[59] Hussain A, Choukairi F, Dunn K. Predicting survival in
thermal injury: a systematic review of methodology of
composite prediction models. Burns 2013. S0305-
4179(12)00402-0 [pii]10.1016/j.burns.2012.12.010.
[60] Herndon DN, Barrow RE, Kunkel KR, Broemeling L, Rutan
RL. Effects of recombinant human growth hormone on
donor-site healing in severely burned children. Ann Surg
1990;212:424–9. discussion 30-1.
[61] Altman DG, Vergouwe Y, Royston P, Moons KG. Prognosis
and prognostic research: validating a prognostic model.
BMJ 2009;338:b605.
[62] Royston P, Moons KG, Altman DG, Vergouwe Y. Prognosis
and prognostic research: developing a prognostic model.
BMJ 2009;338:b604.
[63] Coste J, Fermanian J, Venot A. Methodological and
statistical problems in the construction of composite
measurement scales: a survey of six medical and
epidemiological journals. Stat Med 1995;14:331–45.
[64] Wasson JH, Sox HC, Neff RK, Goldman L. Clinical prediction
rules. Applications and methodological standards. N Engl J
Med 1985;313:793–9.
[65] Laupacis A, Sekar N, Stiell IG. Clinical prediction rules. A
review and suggested modifications of methodological
standards. JAMA 1997;277:488–94.