e-ISSN: 2722-3558 Sriwijaya Journal of Surgery [SJS] https://sriwijayasurgery.
Diagnostic Accuracy of the Modified ABSI Versus the Revised Baux Score for InHospital Mortality Prediction in Adult Burn Patients at a Tertiary Indonesian Burn Centre Ilham Taufan Ikramiawan1*.
Abda Arif2.
Theodorus3 1Department of Surgery.
Faculty of Medicine.
Universitas Sriwijaya.
Palembang.
Indonesia 2Department of Plastic Surgery.
Faculty of Medicine.
Universitas Sriwijaya.
Palembang.
Indonesia 3Department of Pharmacology.
Faculty of Medicine.
Universitas Sriwijaya.
Palembang.
Indonesia
ARTICLE
ABSTRACT
INFO
Keywords:
Burn surgery Diagnostic accuracy Modified ABSI Mortality prediction Revised Baux score *Corresponding author:
Ilham Taufan Ikramiawan E-mail address:
ilhamtaufan0@gmail.
All authors have reviewed and approved the Anal version of the manuscript.
https://doi.
org/10.
37275/sjs.
Introduction: Rapid bedside risk stratification of burn-injured patients informs intensive-care admission, fluid-resuscitation intensity, and timing of early surgical excision and grafting.
The Modified Abbreviated Burn Severity Index .
ABSI) and the Revised Baux (R-Bau.
score are widely cited but have not previously been compared head-to-head in a Sumatran tertiary burn unit.
Methods: STARD-aligned retrospective diagnostic-accuracy study of consecutive thermal-burn patients admitted to the Burn Unit.
Dr.
Mohammad Hoesin General Hospital Palembang (JanuaryAeDecember 2.
Patients with concurrent major trauma or incomplete records were excluded.
Both scores were calculated from medical records by an investigator blinded to the outcome.
In-hospital mortality was the reference standard.
Receiveroperating-characteristic.
Wilson-CI sensitivity/specificity.
DeLong AUC comparison.
McNemar paired test, and Cohen's analyses were performed in SPSS v25 ( = 0.
Results: Of 45 included patients .
ean age 40.
8% male.
4% fire burn.
, 20 .
4%) died.
Optimal cutoffs were mABSI Ou 5 (AUC 0.
811, 95% CI 0.
688Ae0.
0%, specificity 0%.
Youden 0.
and R-Baux > 60 (AUC 0.
808, 95% CI 0.
678Ae0.
0%, specificity 82.
Youden 0.
Inter-test agreement was substantial ( = 0.
69, 95% CI 0.
48Ae0.
p < 0.
, with overall accuracy of 4% when R-Baux was treated as the comparator.
Conclusion: Both scores discriminated burn mortality well.
the Modified ABSI offered higher sensitivity and Youden index, supporting its evaluation as a primary triage tool in similar Indonesian tertiary burn units, while the Revised Baux retained complementary specificity for confirmatory risk classification.
Introduction post-traumatic The preventable death and long-term disability, with the World Health Organization estimating more than operative interventions, intensive nutritional and 180,000 deaths each year and a heavier per-capita rehabilitative support, and lifelong reconstructive needs combine to make severe burn one of the most (LMIC.
1,2 Beyond the headline mortality figures, burn resource-intensive forms of trauma encountered by injury accounts for an estimated eleven million surgical services worldwide.
Burn injury remains a leading global cause of middle-income presentations to medical attention annually and Mortality from major burn in South-East Asia and contributes substantially to global disability-adjusted South Asia exceeds that of high-income regions by an life-years through scarring, contracture, amputation, order of magnitude, driven by late presentation, transport peripheral-hospital together determine the trajectory of care.
management, restricted intensive-care capacity, and Many bedside risk-stratification scores have been constrained access to early excision, autografting and the Abbreviated Burn Severity Index .
Pakistani, and the Revised Baux score (Osler 2.
remain the Vietnamese, and Indian series consistently report most widely cited.
14,15 The Modified ABSI, introduced by Bartels and colleagues in 2020, simplifies the hospitalised burn patients, in stark contrast to the 4Ae original by retaining four parameters .
ge, %TBSA, 8% reported by contemporary North American and full-thickness depth, inhalation injur.
and removing European registries.
Within Indonesia, hospital-based it has shown an area under the receiver- series from Cipto Mangunkusumo Jakarta and Dr.
operating-characteristic Soetomo Surabaya report inpatient mortality of 25Ae 14 The Revised Baux sums age and %TBSA 35% for severe burns, broadly congruent with the 44% and adds 17 points for inhalation injury, with AUC in-hospital mortality observed at Dr.
Mohammad values of 0.
94Ae0.
96 reported in the canonical Dutch Hoesin General Hospital Palembang (RSMH) in the and London validations and confirmed in a recent present cohort.
5-7 This high baseline mortality is not North American multicentre study.
15-17 Comparative solely an indicator of disease severity.
it also reflects methodologies have evolved in parallel Ai including the constrained surgical infrastructure typical of an Karimi's pediatric Baux derivation.
Halgas's multi- Indonesian critical-care (AUC) Brusselaers's intensive-care beds, mechanical ventilators, blood- Belgian-Hungarian external-validation work Ai adding product availability, and operating-theatre time must to a robust methodological lineage.
18-20 Head-to-head be allocated dynamically across competing acute data are nonetheless scarce, particularly in Indonesia, surgical demands.
single-centre Acute burn injury triggers a hyperinflammatory examined R-Baux at the national referral burn unit and hypermetabolic cascade Ai capillary leak driving and none has compared it directly with the Modified burn shock, immune dysregulation predisposing to ABSI.
sepsis, and a persistent open wound surface that Three conceptual considerations make a direct sustains catabolism and acts as a portal of entry for head-to-head comparison particularly informative in nosocomial pathogens.
These mechanisms drive the late causes of burn mortality, principally sepsis, multiple-organ dysfunction syndrome, and acute substantively from the cohorts in which both indices respiratory distress syndrome.
1,8 Inhalation injury were derived: patients are younger, present later, are independently doubles or triples mortality across more frequently transferred from peripheral facilities cohorts and is uniformly the strongest non-anatomical with partial early management, and have a higher 9,10 Modern burn-surgery practice therefore proportion of full-thickness involvement at admission.
emphasises early tangential excision and autografting.
Second, intensive-care-bed availability is rate-limiting structured Surviving Sepsis After Burn Campaign in most Indonesian provincial centres, which means source-control bundles, aggressive nutrition support, that triage decisions rest more heavily on bedside- and graded mobilisation Ai interventions whose computable scores than on subsequent laboratory or benefit is most apparent in patients accurately radiological information that may not be available in stratified for risk at admission.
The clinical leverage Third, afforded by accurate early stratification is therefore substantial: it informs not merely a probability of variable absent from R-Baux Ai making its sensitivity death but a sequence of operational decisions that advantage particularly relevant when full-thickness Indonesian Indonesian First.
Modified Ai
case-mix
ABSI
surgical-decision involvement is common.
support, and inpatient rehabilitation.
The study was RSMH Palembang serves as the tertiary referral designed and reported in accordance with the STARD burn centre for South Sumatra, admitting 92 burn 2015 statement for diagnostic-accuracy research and patients during calendar year 2025 .
with full- the STROBE checklist for observational studies, and thickness involvemen.
and reporting an in-hospital the QUADAS-2 risk-of-bias domains .
atient selection, mortality of 34% in its 2019Ae2021 series.
The index test, reference standard, flow and timin.
were institution provides a representative example of a considered during interpretation.
Indonesian burn-care dedicated burn unit, an integrated plastic-surgery and burn-subspecialty service, and supporting intensivecare.
Patients and eligibility Consecutive admitted with acute thermal burn injury between 1 rehabilitation services, but with the constrained January 2025 and 31st December 2025 were screened.
throughput characteristic of a public-sector tertiary Inclusion criteria were: .
admission diagnosis of acute referral centre.
The primary aim of this study was burn injury irrespective of aetiology.
availability of therefore to compare the diagnostic accuracy of the complete admission data for age, %TBSA, inhalation- Modified ABSI and Revised Baux scores for predicting injury status, and burn depth.
documented in- in-hospital mortality at RSMH Palembang.
Secondary hospital outcome .
live at discharge or in-hospital aims were to quantify inter-test agreement using Exclusion criteria were concomitant major Cohen's and to characterise the additive value of full- thickness depth as a discriminating variable.
preventing score calculation.
The decision to exclude hypothesised, a priori, that the Modified ABSI would poly-trauma cases reflected the methodological aim of demonstrate higher sensitivity than R-Baux owing to its inclusion of burn-depth information, while the two composite trauma scoring.
both indices were derived indices would show overall comparable discrimination for isolated thermal burn and have not been validated as measured by AUC.
in patients with substantive concurrent injury.
Of 92 .
oly-traum.
burn-specific admissions, 47 were excluded .
oncurrent major Methods trauma n = 18.
incomplete records n = .
and 45 Study design and setting patients formed the analytic sample, as illustrated in This was a single-centre, retrospective, blinded the patient-flow diagram (Figure .
diagnostic-accuracy study conducted at the Burn Unit.
Department of Surgery.
Dr.
Mohammad Hoesin Sample size calculation General Hospital.
Palembang, the tertiary referral Sample size was estimated using n = ZA y P y .
Oe burn centre for South Sumatra Province.
Indonesia.
P) / iA for a single proportion with anticipated RSMH is a 1,011-bed government-operated tertiary sensitivity P = 0.
ased on Shah et al.
), precision i teaching hospital affiliated with the Faculty of = 0.
05, and Z = 1.
The required minimum was 45 Medicine, patients, which was satisfied.
21 Post hoc, with the catchment population of approximately 8.
5 million observed sensitivities .
% for mABSI, 52% for R- people across Southern Sumatra.
The Burn Unit Bau.
, the study had 70% power to detect a 20- operates as part of the Plastic and Reconstructive percentage-point difference in paired sensitivities at Surgery service and provides integrated burn-shock = 0.
05 (McNemar exac.
secondary estimates therefore carry wider confidence intervals than the planning microbiology-guided calculation suggested.
We deliberately anchored the Universitas Sriwijaya, intensive-care planning sensitivity to a published external-validation estimate rather than to an internal pilot estimate, in occupational therapy, and pressure-garment therapy order to provide a reproducible and externally on healing wounds.
1,8,11,12,13,22
defensible justification.
Index tests: Modified ABSI and Revised Baux Burn-care pathway and operative management The Modified ABSI was calculated as the sum of age All patients received structured care according to category .
Ae5 point.
, inhalation-injury indicator institutional burn protocol.
On arrival, patients .
, full-thickness burn indicator .
and %TBSA category .
Ae10 point.
, giving a total range of 2Ae1714.
structured ABCDE approach.
Suspected inhalation The Revised Baux score was calculated as age .
injury Ai based on closed-space exposure, facial or %TBSA 17 if inhalation injury was present (R = .
or 0 otherwise.
15,17 For both indices, age was taken hoarseness, or progressive stridor Ai prompted early from the admission record.
%TBSA was estimated endotracheal intubation, with fibreoptic bronchoscopy using the Lund and Browder chart for patients aged performed when available to confirm and grade airway less than 15 years and the Wallace rule of nines for Modified Parkland fluid resuscitation .
mL y older patients.
burn depth was assessed by the kg y %TBSA Ringer's lactate over the first 24 hours, attending burn surgeon at admission, refined intra- half delivered in the first 8 hours from injur.
was operatively at first excision, and the more accurate of titrated to target urine output of 0.
5 mL y kg y hour in the two assessments .
ypically the intra-operative finding, where availabl.
was used.
and inhalation- Multimodal analgesia, tetanus prophylaxis where .
arbonaceous sputum, hoarseness, facial indicated, and prophylactic antibiotics in selected closed-space exposur.
supplemented by fibreoptic high-risk patients were administered according to bronchoscopy when performed.
Both indices were institutional protocol.
derived independently by the principal investigator Operative management followed a structured earlyexcision Escharotomy (ITI) a 10% random sample was double- urgently for circumferential full-thickness extremity or abstracted by the senior author (AA) with 100% torso burns at risk of compartment syndrome or restricted ventilation.
Tangential excision with split- abstraction process.
thickness autografting was the default approach, performed at the discretion of the attending Reference standard and outcome burn surgeon and typically within the first week of The reference standard was in-hospital mortality, admission for full-thickness or deep-dermal burns.
defined as death from any cause prior to hospital Donor sites were dressed with hydrocolloid or paraffin Patients transferred to other facilities for gauze, depending on availability.
For very large burns ongoing care were classified according to their final where autograft donor sites were limited, staged RSMH disposition.
Early deaths .
ithin 24 hours of excision with biological or synthetic temporary cover was used.
Postoperative care included high-protein enteral nutrition support targeting 25Ae30 kcal/kg/day exclusion would systematically bias the discriminative 5Ae2.
0 g protein/kg/day, structured silver-based performance of the scoring systems by removing the most severely injured patients on whom such systems surveillance per the 2023 Surviving Sepsis After Burn are arguably most informative.
Date and time of death Campaign, were recorded for descriptive purposes.
cause of death 24Ae48 burn-registry was abstracted where documented in the medical depth assessment reduced the risk of operator- index-test structured exclusion criteria reduced the risk of Statistical analysis confounding by concurrent injury.
The risk of selection Data were analysed with IBM SPSS Statistics for bias from the 51% record-exclusion was acknowledged Windows, version 25 (IBM Corp.
Armonk.
NY).
a priori and is discussed transparently in the Continuous variables were summarised as mean A SD or median .
after Kolmogorov-Smirnov/ShapiroWilk normality testing.
Categorical variables were Ethical considerations summarised as n (%).
Between-group comparisons The study protocol received written approval from used the independent-samples t-test .
ith Levene test for equality of variance.
or Mann-Whitney U test for Mohammad Hoesin General Hospital.
Palembang and continuous data and Pearson NA or Fisher exact test the Faculty of Medicine.
Universitas Sriwijaya, under Receiver- Ethical 087/kepkrsmhfkunsri/2025, issued on 15th January optimal cutoffs by the Youden index.
Sensitivity.
The institutional Ethics Committee waived the specificity, positive predictive value (PPV) and negative requirement for individual informed consent because predictive value (NPV) were reported with Wilson 95% the analysis used de-identified retrospective medical- confidence intervals.
Areas under the curve were record data collected during routine clinical care, with compared by the DeLong test.
Paired sensitivities and no additional risk to participants.
The study was specificities were compared by McNemar's exact test, conducted in accordance with the principles of the exploiting the within-patient nature of the comparison.
Declaration of Helsinki .
3 revisio.
and the Inter-test agreement was quantified by Cohen's with Indonesian Ministry of Health regulations on health 95% CI.
the prevalence- and bias-adjusted (PABAK) was reported as a sensitivity analysis to address the maintained throughout the study.
all identifying marginal-distribution skew that can deflate raw information was removed prior to analysis, and a estimates in 2 y 2 tables.
Effect sizes .
ank-biserial unique study identifier was used for each record.
operating-characteristic (ROC) Health Research Ethics Clearance Patient Committee Certificate Dr.
No.
correlation for non-parametric comparisons.
Cohen's d for parametric comparison.
were reported alongside Results p-values throughout.
All tests were two-sided.
p < 0.
Patient flow and demographics was considered statistically significant.
Multiple- Of 92 burn admissions during the study period, 47 comparison correction was not applied to the primary records were excluded .
oncurrent major trauma n = analyses, but the multiple-comparison concern is incomplete records n = .
, and 45 patients formed acknowledged in the limitations.
the analytic sample, as summarised in the patient-flow diagram (Figure .
In-hospital mortality was 20/45 Quality control and bias mitigation .
4%).
The full demographic and clinical profile.
Several procedural steps were taken to mitigate the stratified by survival status, is detailed in Table 1:
risks of bias inherent in single-centre retrospective patients were predominantly male .
/45, 77.
8%) with diagnostic-accuracy a mean age of 40.
8 years .
ange 1Ae.
1% were during index-test computation reduced the risk of aged 0Ae40 years and 46.
7% were 41Ae70 years.
The incorporation bias.
double-abstraction of a random dominant aetiology was fire/flame burn .
/45, subsample reduced the risk of measurement error in 4%), reflecting the regional pattern of domestic and score calculation.
intra-operative refinement of burn- occupational thermal injury, and inhalation injury Outcome was documented in 6/45 .
3%) of patients.
Although 0%), this difference the inhalation-injury rate was numerically higher significance in this small sample .
= 0.
among decedents .
/20, 20.
0%) than survivors .
/25.
Figure 1.
STROBE-aligned patient-flow diagram for the diagnostic-accuracy comparison of Modified ABSI and Revised Baux scores at Dr.
Mohammad Hoesin General Hospital.
Palembang (JanuaryAeDecembe.
As shown in Table 1, gender .
= 0.
and age burns predominated in the survivor cohort.
This group .
= 0.
did not differ between deceased and pattern is consistent with the international burn surviving patients, whereas burn aetiology .
= 0.
literature in which fire/flame injury Ai particularly and burn depth .
= 0.
showed statistically when sustained in closed-space contexts Ai is The %TBSA, aetiological pattern is clinically informative: fire/flame inhalation injury, and greater overall severity than burn was dramatically over-represented among non- other burn mechanisms.
/20, 85.
0%) compared with survivors Total body surface area (%TBSA) distribution also .
/25, 48.
0%), while electrical, scald, and chemical differed significantly between groups, as detailed in Table 2 .
= 0.
, with within-row mortality rising well-established monotonically beyond 30% TBSA and reaching 100% surface-area involvement and uniformly poor outcome in the 71Ae100% bands.
Fully half of the deceased in this resource-limited setting.
The relationship patients .
/20, 50.
0%) had %TBSA in the 31Ae50% range, where mortality climbed steeply from 25% in monotonicity above 20%, supporting the inclusion of the 11Ae20% band to 62.
5% in the 31Ae40% band and %TBSA as a major component of both scoring systems 75% in the 41Ae50% band.
Above 70% TBSA, mortality under evaluation.
%TBSA
was uniformly 100% .
/5 patient.
, confirming the Table 1.
Demographic and clinical characteristics of burn patients stratified by in-hospital mortality .
= .
Characteristic Deceased Survivors Total p value .
= .
= .
= .
Male 15 .
Female 5 .
0Ae40 years 11 .
41Ae70 years 8 .
71Ae80 years 1 .
Ai Ai 8 A 19.
Ai Fire/flame 17 .
Electrical 3 .
Scald .
ot wate.
Chemical 0 .
Inhalation injury, n (%) 4 .
Full-thickness burn, n (%) 8 .
Gender, n (%) Age group, n (%) Mean age .
, mean A SD Burn etiology, n (%) *Fisher's exact test .
referred for sparse cell.
APearson NA test.
p < 0.
05 considered statistically significant.
Table 2.
Distribution of total body surface area (%TBSA) burned by survival status, with within-row mortality.
%TBSA category Deceased Survivors Total Within-row .
= .
= .
n (%) mortality (%) 1Ae10 4 .
11Ae20 12 .
21Ae30 6 .
31Ae40 8 .
41Ae50 4 .
51Ae60 2 .
61Ae70 4 .
71Ae80 2 .
81Ae90 1 .
91Ae100 2 .
Overall (Pearson NA, 45 .
p = 0.
Modified ABSI and Revised Baux scores.
Score-level results are summarised in Table 3 and overlap between deceased and surviving patients Ai visualised in Figure 2.
The mean Modified ABSI was particularly visible in the lower boxplot whisker for 02 A 2.
ange 2Ae.
deceased patients deceased patients and the upper whisker for survivors 50 A 2.
edian 6, range 4Ae.
Ai illustrates the practical limitation of any single A 1.
edian 5, range 2Ae.
in survivors (Mann- dichotomised cutoff: there will always be a transition Whitney U, p < 0.
rank-biserial correlation 0.
zone in which patients with intermediate scores The mean Revised Baux was 79.
3 A 29.
ange 20Ae require clinical judgment beyond the score itself.
deceased patients scored 92.
6 A 27.
edian 90, range 51Ae.
4 A 24.
edian Diagnostic accuracy and ROC analysis 60, range 1Ae.
in survivors .
ndependent-samples t- ROC analysis identified optimal cutoffs of mABSI Ou test, p < 0.
Cohen's d = 1.
Both scores were 5 (AUC 0.
811, 95% CI 0.
688Ae0.
p < 0.
and R- therefore significantly higher in non-survivors with Baux > 60 (AUC 0.
808, 95% CI 0.
678Ae0.
large effect sizes, as illustrated by the boxplot The full diagnostic-accuracy profile, including distributions in Figure 2.
The within-distribution Wilson 95% confidence intervals around sensitivity, specificity.
PPV, and NPV, is presented in Table 4.
The test confirmed that the Modified ABSI demonstrated visual comparison of these indices is shown in Figure significantly higher paired sensitivity than R-Baux .
the corresponding forest plot of point estimates with = 0.
, at the cost of slightly lower specificity.
This confidence intervals is displayed in Figure 4.
and the finding directly supports our a priori hypothesis that the inclusion of full-thickness depth as a Modified are plotted in Figure 5.
Discriminative performance of ABSI component would lift its sensitivity above that of the two scores was statistically indistinguishable on R-Baux the AUC scale (DeLong difference 0.
003, 95% CI
receiver-operating-characteristic Oe0.
150 to 0.
p = 0.
, but the McNemar exact Table 3.
Modified ABSI and Revised Baux scores by survival status with effect sizes .
= .
Score Deceased .
= .
mean Survivors .
= .
mean p-value .
ffect siz.
A SD .
edian, rang.
A SD .
edian, rang.
Modified ABSI 50 A 2.
, 4Ae.
84 A 1.
, 2Ae.
< 0.
001* .
b = 0.
Revised Baux 6 A 27.
, 51Ae.
4 A 24.
, 1Ae.
< 0.
001A .
= 1.
*Mann-Whitney U test, rank-biserial correlation rb.
AIndependent samples t-test (Levene p = 0.
Cohen's d.
Figure 2.
Distribution of Modified ABSI and Revised Baux scores by survival status .
oxplot with overlaid individual *** p < 0.
Table 4.
Diagnostic accuracy of the Modified ABSI and Revised Baux scores at their optimal Youden cutoffs for inhospital mortality, with paired comparisons.
Diagnostic index .
% CI) Modified ABSI .
utoff Ou .
Revised Baux .
utoff > .
688Ae0.
678Ae0.
Sensitivity, % (Wilson 95% CI) 0 .
6Ae87.
3Ae72.
Specificity, % (Wilson 95% CI) 0 .
3Ae90.
5Ae93.
Positive predictive value, % (Wilson 95% CI) 3 .
3Ae92.
4Ae95.
Negative predictive value, % (Wilson 95% CI) 2 .
1Ae86.
9Ae76.
Youden index (J) Cutoff p-value < 0.
< 0.
Difference in AUC (DeLong tes.
% CI Oe0.
150 to p = 0.
Ai McNemar test .
aired sensitivity, mABSI vs R-Bau.
p = 0.
Ai AUC (DeLong 95% CI) Figure 3.
Diagnostic accuracy of Modified ABSI versus Revised Baux for in-hospital mortality .
= 45 burn patient.
, comparing sensitivity, specificity, predictive values, and Youden index for the two scores.
Figure 4.
Forest plot of diagnostic-accuracy estimates with 95% confidence intervals .
lue squares = significant.
squares = not significan.
Figure 5.
Receiver-operating-characteristic curves for the Modified ABSI and Revised Baux in predicting in-hospital mortality .
= .
Operating points correspond to the Youden-derived cutoffs reported in Table 4.
Agreement between the two scores ABSI showed sensitivity 75.
8%, specificity 100%.
PPV Cohen's for the agreement between Modified ABSI 100%.
NPV 69.
5% and overall accuracy 84.
The Ou 5 and Revised Baux > 60 was 0.
% CI 0.
48Ae p < 0.
, corresponding to substantial discordances in one direction and zero in the other Ai the prevalence- and bias-adjusted is clinically informative: it suggests that the Modified (PABAK) was 0.
69, confirming that the estimate was ABSI does not produce false-positive flags relative to not deflated by marginal-distribution skew.
The 2 y 2 R-Baux in this cohort, while it captures additional cross-tabulation, presented in Table 5, showed 22 patients positive on both scores and 16 negative on nonetheless went on to die.
This asymmetry aligns both, with seven patients positive on R-Baux but with the McNemar paired sensitivity result and negative on mABSI and none in the reverse direction.
reinforces the case for the Modified ABSI as a primary When R-Baux was treated as the comparator.
Modified triage screen.
R-Baux Ai low-risk Table 5.
Cross-tabulation of dichotomised Modified ABSI .
utoff Ou .
and Revised Baux .
utoff > .
= .
Modified ABSI R-Baux > 60 .
redicted deat.
R-Baux O 60 .
redicted surviva.
Total Ou 5 .
redicted deat.
O 4 .
redicted surviva.
Total Cohen's = 0.
% CI 0.
48Ae0.
PABAK 0.
p < 0.
001 Ai substantial agreement.
Discussion single-centre Baux.
This central finding directly supports our a Indonesian priori hypothesis and offers practical guidance for accuracy study, both the Modified ABSI and Revised burn-care services in similar resource-limited settings.
Baux scores demonstrated good and statistically in-hospital Our R-Baux published international benchmarks but somewhat mortality, with overlapping AUCs of 0.
811 and 0.
lower than the AUC of 0.
94Ae0.
96 reported by Dokter et (DeLong p = 0.
, as quantified in Table 4 and in 4,389 Dutch patients and by Heng et al.
visualised in Figures 3Ae5.
The Modified ABSI achieved London tertiary unit.
15,16 This attenuation likely a higher Youden index .
47 vs 0.
and significantly reflects the smaller sample .
= .
, the higher higher paired sensitivity .
0% vs 52.
McNemar p baseline mortality .
% vs 7Ae13% in those cohort.
= 0.
, while the Revised Baux retained slightly and the higher proportion of advanced-presentation higher specificity .
0% vs 75.
0%).
Substantial inter- injuries typical of Indonesian referral practice.
Our score agreement ( = 0.
PABAK 0.
Table .
and AUC of 0.
808 falls within the range reported in other a 100% specificity when Modified ABSI was tested LMIC and Asian validations: Lam et al.
(Vietna.
against Revised Baux as a comparator support the reported an AUC of 0.
Lip et al.
(Malaysi.
reported conclusion that the two indices identify largely 94.
Henderson et al.
's 2024 multicentre North overlapping high-risk groups but that the Modified American validation reported 0.
93, and Wardhana et ABSI captures additional non-survivors missed by R- (Jakart.
broadly comparable to ours.
6,17,23,24 The convergence antibiotic stewardship, and aggressive supportive of these AUC estimates across geographically diverse Accurate early stratification Ai such as that cohorts supports the conceptual transportability of R- provided by the Modified ABSI Ai supports compliance Baux as a discriminator, while the variation in with these bundles by triggering early intensive-care absolute sensitivity and specificity across cohorts admission, central-line placement, and proactive confirms that institution-specific optimal cutoffs are sepsis surveillance.
From the perspective of a burn needed before clinical adoption.
surgeon working in a resource-limited Indonesian The Modified ABSI's provincial centre, the most actionable implication of Unlike R-Baux, an early high score is therefore not the prognostic Modified ABSI explicitly weights full-thickness burn statement itself but the operational sequence it depth Ai a variable directly linked to the surgical work triggers: early ICU admission, early surgical excision, of excision, grafting and re-debridement, to systemic structured antimicrobial stewardship, and early family catabolism through the persistent open wound, to counselling about realistic outcome expectations.
nosocomial-pathogen The clinical implications for surgical practice in Indonesia are concrete and operationally specific.
Full-thickness burn injury represents First, both scores can and should be calculated at the complete loss of dermal vascular supply, eccrine admission within minutes of presentation.
bedside computability is one of their main advantages re-epithelialisation is no longer possible over machine-learning prediction tools that require without grafting, the wound surface remains a portal of entry for pathogens until covered, and the systemic asymmetry of regret in burn surgery Ai under-triage of inflammatory response is more sustained than in a high-risk patient is much worse than over-triage of partial-thickness injury.
In our cohort, as detailed in a low-risk patient Ai the Modified ABSI's higher Table 1, 8/20 .
%) of decedents had full-thickness sensitivity, demonstrated in Table 4 and Figure 3, burn versus 8/25 .
%) of survivors.
although this supports its evaluation as a primary triage screen in difference did not reach significance in isolation .
= resource-limited settings where intensive-care beds .
, its inclusion as a Modified ABSI component are scarce.
Third.
R-Baux retains complementary appears to lift several full-thickness patients into the value as a confirmatory specificity-anchor and as the high-risk band who would have remained sub- score most familiar to anaesthetists and intensivists threshold on R-Baux.
Shah et al.
reported a similar Combining them Ai for instance, pattern in a north-Indian developing-country cohort, treating any patient positive on either index as high- finding the Modified ABSI marginally superior to R- risk Ai would yield the highest aggregate sensitivity for Baux for sensitivity while R-Baux retained calibration triage decisions, although this carries a corresponding for specificity.
1,8,14
Sepsis remains the leading proximate cause of late over-triage Second, samples7,21 burn death and is multifactorial Ai driven by institutional protocol might be: calculate both scores translocation of gut flora across an injured mucosal on the standardised admission template.
admit any barrier, immunoparalysis from the hyperinflammatory patient positive on either index directly to the burn response, and the persistent open wound surface that intensive-care area pending further assessment.
serves as a portal of entry for nosocomial pathogens.
use the dual-score concordance to inform family The 2023 Surviving Sepsis After Burn Campaign counselling and operative planning.
emphasises early surgical source-control through Our Asian Indonesian data.
Wang et al.
reported in-hospital mortality of 39% among severe burns in Beijing with R-Baux and ABSI both predictive at AUC > 0.
Indonesian context.
It is also worth situating the present comparison Salsabilla et al.
at Dr.
Soetomo Surabaya reported R- Baux inhalation-injury mortality-prediction research.
Several research groups mortality with AUC 0.
88, broadly congruent with our have explored machine-learning approaches to burn- Tomita et al.
and Akkad et al.
emphasise that mortality prediction using random-forest, gradient- age-stratified mortality patterns drive much of the boosting, and deep-learning architectures.
These discriminative power of both indices, supporting the approaches sometimes achieve marginally higher AUC age-adjustment approach evident in Table 1.
26,27
than the traditional regression-based scores, but they Wardhana et al.
's Cipto Mangunkusumo series require complete laboratory and physiological data, remains the closest available Indonesian comparator are computationally opaque to bedside clinicians, and their performance has not been validated in resource- 6,7 The convergence between our findings limited settings where the necessary input data are and the Surabaya.
Jakarta, and broader Asian data frequently incomplete.
The Modified ABSI and R-Baux strengthens the case for institutional protocols that retain a substantial advantage in such settings incorporate both indices as part of the standard burn- precisely because they require only four to five admission workup across Indonesian tertiary referral admission variables Ai all routinely collected at admission Ai and produce a single integer score that Beyond the immediate discriminative comparison, can be calculated mentally at the bedside in under a several broader methodological considerations are This bedside computability is a feature, not a worth noting.
First.
AUC measures discrimination Ai the ability of the score to separate cases from non- Indonesian cases Ai but does not measure calibration, which is laboratory turnaround times can be measured in the ability of the score to predict the absolute hours rather than minutes.
The role of machine probability of mortality at each score band.
Calibration learning in Indonesian burn care should therefore be matters as much as discrimination for surgical seen as complementary rather than substitutive:
decision-making, since clinicians use the score to machine-learning models may add incremental value inform absolute-risk discussions with patients and when complete data are available, while traditional Future prospective Indonesian validation scoring systems remain the default at the moment of should report both calibration slope and calibration- admission when triage decisions must be made on the in-the-large, ideally with calibration plots overlaid on information at hand.
the diagonal of perfect calibration.
Second, a more Integrated Reclassification Discrimination Improvement Improvement provincial-referral From a health-systems perspective, the work has contemporary diagnostic-accuracy analysis would also Net Indonesia substantially in trauma-care infrastructure over the switching from R-Baux to Modified ABSI.
continuous metrics quantify the practical clinical concentrated at a small number of tertiary referral benefit of switching scores in a way that dichotomous sensitivity and specificity cannot.
Third, decision- throughput, and outcomes.
The development of curve analysis Ai which weighs sensitivity and standardised, bedside-computable risk-stratification specificity by the clinical asymmetry of false positives tools Ai calibrated to local case-mix and validated and false negatives Ai would provide a complementary perspective on the net benefit of each score in the improvement in the consistency and quality of burn-care Ai case-mix.
Indonesian burn care.
The Modified ABSI and R-Baux bronchoscopy, which was not uniformly performed.
are good candidates for such standardisation precisely Reference standard: only in-hospital mortality is because they require only data routinely collected at longer-term outcomes .
-day, 90-day, admission and avoid dependence on laboratory or functiona.
and surgical-process measures .
perative radiological investigations that may not be available in time, blood loss, length of stay, ventilator days, real time across all institutions.
A coordinated transfusion rat.
were not extracted from the records Indonesian and warrant prospective study, in line with the modelled on the Australian and New Zealand Burns multivariable predictors emphasised by Choi et al.
Association registry Ai would substantially advance Stoica et al.
, and Reiff et al.
28-30 Flow and timing:
both the calibration of these scores to local case-mix and the broader benchmarking of Indonesian burn Charlson comorbidity index, time from injury to outcomes against international peers.
burn-registry Ai
ASA
physical-status, surgeon-experience The strengths of this study are fourfold.
First, it is.
Statistical considerations: with n = 45 to our knowledge, the first head-to-head Modified ABSI the 95% confidence intervals around individual versus Revised Baux comparison performed at an sensitivity and specificity estimates are wide, and the Indonesian tertiary burn centre outside Jakarta, post hoc power for detecting the observed difference in addressing a clear geographical and methodological paired sensitivities is limited.
The institution-specific gap in the regional literature.
Second, blinded outcome Youden cutoffs derived here .
ABSI Ou 5.
R-Baux > .
ascertainment with double-abstraction quality-control are lower than the published international cutoffs reduces risk of differential misclassification.
Third, .
ABSI Ou 7.
R-Baux Ou .
and should not be reporting of agreement (.
PABAK) in addition to uncritically adopted at other centres without local discrimination (AUC.
DeLon.
and paired sensitivity (McNema.
Ai all summarised in Tables 4 and 5 Ai comparison correction across the multiple secondary provides a comprehensive diagnostic-accuracy profile tests should temper the strength of the inferences that exceeds the methodological norm in the regional drawn from any individual subgroup comparison.
Finally, burn-surgery literature.
Fourth, effect sizes .
ankbiserial correlation.
Cohen's .
accompany p values Conclusion throughout, providing inferential transparency that Among Dr.
supports independent re-interpretation of the results.
Mohammad Hoesin General Hospital.
Palembang Limitations are substantial and warrant explicit during 2025, both the Modified ABSI .
utoff Ou 5.
AUC
acknowledgement, organised by the QUADAS-2 risk- .
and the Revised Baux score .
utoff > 60.
AUC
of-bias domains.
Patient selection: the single-centre, .
discriminated in-hospital mortality well, with record-exclusion substantial inter-test agreement ( = 0.
and (Figure .
limits external validity to other Indonesian complementary performance profiles, as summarised burn units that may differ in case-mix and operative in Tables 4 and 5 and Figures 3Ae5.
The Modified ABSI offered higher paired sensitivity and Youden index, supporting its evaluation as a primary triage tool in plausibly correlated with both severity and outcome.
similar Indonesian tertiary burn units.
the Revised Index test: burn-depth assessment is operator- Baux retained slightly higher specificity and remains dependent, and although intra-operative refinement a reasonable confirmatory score.
The mechanistic was used where available, residual measurement error basis for the Modified ABSI's sensitivity advantage Ai cannot be excluded.
inhalation-injury determination explicit incorporation of full-thickness burn depth, a similarly depends on the availability of fibreoptic surgically meaningful variable absent from R-Baux Ai supports its preferential use in Indonesian referral Wardhana A.
Valeria M.
Apriza RP, et al.
populations enriched for full-thickness involvement.
Comparison between ABSI and BOBI score for Burn surgeons in similar resource-limited settings burns mortality prediction in Indonesia's should consider routinely calculating both indices at national referral burn center: a 5-year study.
admission to inform fluid-resuscitation intensity.
Burns Open.
: 92Ae96.
intensive-care surgical-excision Wardhana A.
Mulyantara IG.
Kekalih A.
Larger Implementation of revised Baux score to multicentre prospective studies Ai incorporating ASA predict mortality of burn-injured patients in physical-status.
Charlson comorbidity index, time-to- burn unit of Dr.
Cipto Mangunkusumo Hospital.
Jakarta.
New Ropanasuri J Surg.
calibration-in-the-large, decision-curve analysis, and longer-term functional outcomes Ai are now needed to derive transportable cutoffs and to test 1.
: 23Ae26.
Greenhalgh DG.
Hill DM.
Burmeister DM, et whether a composite Modified ABSI / R-Baux rule Surviving Sepsis After Burn Campaign.
outperforms either score alone in Indonesian and Burns.
: 1487Ae97.
ASEAN burn-care contexts.
A coordinated Indonesian Korkmaz HI.
Niessen FB.
Tuk B, et al.
burn-registry initiative would substantially accelerate Inhalation this validation work and improve the consistency and outcomes in burn patients.
Burns.
quality of burn care across the country's tertiary 49.
: 547Ae55.
referral network.
El-Hebawy RH.
Ghareeb FM.
Inhalation injury as a prognostic factor for mortality in burn
References