Risk Prediction of Multidrug-Resistant Infections in Pediatric Burn Patients with Hospital-Acquired Infections: A Retrospective Cohort Study

Abstract

Background:

Multidrug-resistant organism (MDRO) infections pose a growing threat to pediatric burn patients, contributing to prolonged hospitalization, increased health care burden, and limited therapeutic options. Despite their clinical significance, predictive tools specific to MDRO infections in this population remain scarce. The authors aimed to identify independent clinical predictors of MDRO infection among pediatric burn patients with health care–associated infections (HAIs) and to construct a validated nomogram for individualized risk estimation.

Methods:

A retrospective cohort study was conducted at a tertiary-care pediatric burn center in China, including patients 1 month to 18 years of age who developed HAIs between January of 2020 and December of 2023. Multivariable logistic regression was used to identify independent risk factors, and a nomogram was constructed accordingly. Model performance was evaluated using the area under the receiver operating characteristic curve, calibration plot, decision curve analysis, and clinical impact curve.

Results:

Among 763 pediatric burn patients with HAIs, 93 (12.2%) developed MDRO infections. Six independent predictors were identified: total burn surface area, hospital length of stay, intensive care unit admission, mechanical ventilation, central venous catheter use, and acute inhalation injury. The nomogram demonstrated excellent discrimination (area under the receiver operating characteristic curve 0.962), good calibration, and favorable clinical utility across a range of threshold probabilities.

Conclusions:

This study presents a clinically applicable nomogram for early prediction of MDRO infection in pediatric burn patients with HAIs. The tool may support individualized risk stratification, guide targeted preventive interventions, and inform empirical antibiotic decision-making in burn care settings. Prospective validation is warranted.

Severe burns in children represent a significant global health burden, contributing to substantial morbidity and mortality, particularly due to health care–associated infections (HAIs) during hospitalization.¹ Multidrug-resistant organism (MDRO) infections, in particular, have become increasingly prevalent in pediatric burn units, complicating clinical management and therapeutic decision-making.² Surveillance data from burn centers worldwide have reported MDRO infection rates as high as 60% to 70% among hospitalized burn patients, with Pseudomonas aeruginosa, Acinetobacter baumannii, and Staphylococcus aureus being the most frequently isolated resistant pathogens.^3,4 In pediatric populations, especially among children younger than 2 years, immunologic immaturity, thin dermal barriers, and frequent need for invasive support further elevate the risk of acquiring MDRO infections.⁵ These infections are often associated with increased antimicrobial resistance, prolonged hospital length of stay (LOS), delayed wound healing, higher treatment costs, and elevated mortality risk.^6,7

In recent years, growing attention has been given to elucidating the clinical and procedural factors that predispose pediatric burn patients to MDRO infections. Studies have consistently demonstrated that large total burn surface area (TBSA), prolonged intensive care unit (ICU) admission, multiple surgical procedures, mechanical ventilation, central venous catheter (CVC) use, and urinary catheterization are major risk factors for the development of HAI infections in this population.^8,9 In addition, cumulative exposure to broad-spectrum antibiotics—particularly carbapenems and cephalosporins—has been linked to increased selective pressure and colonization by MDROs.^10,11

Emerging research has begun exploring predictive modeling approaches, including nomogram-based tools, to identify patients at high risk for infection. For example, a recent pediatric burn cohort study in China constructed a model incorporating TBSA, surgery, ICU admission, and catheter use to predict HAI occurrence, achieving an area under the receiver operating characteristic curve (AUC) of 0.93.⁸ However, few studies have focused specifically on predicting MDRO infections among patients with confirmed HAIs, and pediatric-specific risk models remain scarce. Furthermore, advances in data-driven modeling—such as machine learning–based prediction of antimicrobial resistance—are gaining traction in adult ICUs and trauma populations,^12,13 yet these approaches are underutilized in pediatric burn care.

In light of these knowledge gaps, we aimed to identify independent clinical risk factors associated with the development of MDRO infections in pediatric burn patients who had developed HAIs. Based on these variables, we constructed and validated a nomogram to facilitate early risk stratification. By focusing on this high-risk and underresearched population, we created a practical tool to guide clinical decision-making, inform antimicrobial stewardship, and reduce the burden of MDRO infections in pediatric burn care.

PATIENTS AND METHODS

Study Design and Population

This single-center retrospective cohort study was conducted at the Department of Burns and Reconstructive Surgery of Central Hospital Affiliated to Shandong First Medical University, a tertiary-care pediatric burn center in Jinan, China. Pediatric patients 1 month to 18 years of age who were hospitalized for thermal burn injuries between January of 2020 and December of 2023 were retrospectively screened.

The study population was restricted to patients who developed HAIs during hospitalization. HAIs were defined as infections that occurred 48 hours or later after admission and were not incubating at the time of hospital entry, in accordance with the diagnostic criteria issued by the Ministry of Health of the People’s Republic of China. Diagnosis was based on clinical signs and symptoms, laboratory tests, imaging results, and culture-confirmed microbiological evidence. All HAI cases were independently reviewed and confirmed by the hospital infection control team to ensure diagnostic accuracy.

Patients were excluded if they were admitted for nonthermal burn injuries (eg, chemical or electrical burns), were discharged or transferred within 48 hours of admission, had incomplete clinical or microbiological data, or had underlying malignancies or end-stage systemic diseases.

Among the patients with confirmed HAIs, those whose infections were caused by pathogens resistant to at least 1 agent in 3 or more antimicrobial categories were categorized into the multidrug-resistant (MDR) infection group, based on the internationally accepted definition of the Centers for Disease Control and Prevention and the European Centre for Disease Prevention and Control. Patients with HAIs caused by susceptible (non-MDR) organisms were classified into the non-MDR infection group. Patients without any documented infection were not included in the analysis.

Data Collection

Two independent researchers extracted clinical and microbiological data from the hospital infection surveillance system and electronic medical record system using a standardized data collection protocol. Variables collected included demographic data (age and sex); burn-related characteristics, including burn mechanism (scalds, flames, or others) and TBSA estimated using the Lund-Browder chart; hospitalization data (LOS and ICU admission); clinical interventions (surgical treatment, prophylactic antibiotic use [administered within 48 hours of admission], urinary catheterization, CVC placement, and mechanical ventilation); and injury complications (presence of acute inhalation injury, diagnosed clinically and confirmed with bronchoscopy when applicable). Prophylactic antibiotic use was defined as the empirical administration of systemic antibiotics in the absence of confirmed infection, based on clinical judgment and the perceived high risk of infection (eg, extensive burn area, ICU admission, clinical instability). This does not include standard preoperative surgical prophylaxis. Surgery was defined as a binary indicator reflecting whether the patient underwent at least 1 surgical procedure during hospitalization, regardless of duration or frequency. Urine catheterization was recorded as a binary variable (yes/no), defined as any instance of urinary catheter insertion during hospitalization.

All predictor variables were recorded before the onset of infection to ensure temporal consistency. Microbiological identification and antimicrobial susceptibility testing were performed using the VITEK 2 Compact System (bioMérieux) or the Kirby-Bauer disk diffusion method following Clinical and Laboratory Standards Institute guidelines.

Statistical Analysis

Continuous variables were summarized as mean ± SD and compared using the t test, depending on the distribution assessed using the Shapiro-Wilk test. Categorical variables were reported as frequencies and percentages and were compared using the chi-square test or Fisher exact test where appropriate.

Univariate logistic regression analysis was conducted to identify potential risk factors for MDR infection. Variables with a P value less than 0.10 in the univariate analysis were entered into a multivariable logistic regression model using backward stepwise selection based on the Akaike Information Criterion. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were reported for each independent predictor.

A predictive nomogram was developed based on the multivariable model using the “rms” and “regplot” packages in R version 4.2.2 (R Foundation for Statistical Computing). Model discrimination was assessed using the AUC, and calibration was evaluated with a calibration plot based on 500 bootstrap resamples. The clinical utility of the nomogram was further evaluated through decision curve analysis (DCA) and clinical impact curve analysis using the “DecisionCurve” package.

All statistical analyses were performed using SPSS version 22.0 (IBM) and R version 4.2.2. A 2-sided P value less than 0.05 was considered statistically significant.

RESULTS

As shown in Figure 1, a total of 1633 patients initially met the inclusion criteria. After applying the exclusion criteria—admitted for nonburn or nonthermal injuries (n = 632), missing clinical data (n = 45), transferred or discharged within 48 hours of admission (n = 29), or severe underlying diseases such as malignancies or end-stage organ failure (n = 164)—a total of 763 patients were ultimately included in the study.

Fig. 1.

Flowchart of patient selection. Boxes represent patient subgroups; arrows indicate selection flow.

These participants were subsequently categorized into 2 groups based on infection status: the MDR infection group (n = 93) and the non-MDR infection group (n = 670). A total of 102 MDRO isolates were identified from 93 pediatric burn patients with confirmed MDRO infections, indicating that several patients experienced polymicrobial infections. The most commonly isolated pathogens were S aureus (24.5%) and P aeruginosa (21.6%), followed by Staphylococcus epidermidis (13.7%) and Acinetobacter baumannii (11.8%). Additional MDRO species, including Enterobacter cloacae, Klebsiella pneumoniae, Escherichia coli, Proteus mirabilis, Candida albicans, and Enterococcus faecalis, were detected at lower frequencies. Rarely isolated organisms were grouped as “other species.” (See Table, Supplemental Digital Content 1, which shows the distribution of MDROs isolated from the MDR infection group, https://links.lww.com/PRS/I660.) These findings highlight the microbial diversity of MDRO infections in this high-risk population.

Table 1 compares the sociodemographic and clinical characteristics between the MDR infection group and the non-MDR infection group. No significant differences were observed in sex (P = 0.788) or age (P = 0.650) between the 2 groups. However, patients with MDR infections had a significantly higher incidence of flame burns (40.9% versus 31.5%) and acute inhalation injury (35.5% versus 2.8%; P < 0.001). Clinical interventions, including surgery (71.0% versus 27.9%), ICU admission (34.4% versus 2.2%), urinary catheter placement (25.8% versus 10.6%), mechanical ventilation (15.1% versus 0.7%), and antibiotic prophylaxis (65.6% versus 30.0%), were significantly more common in the MDR infection group (all P < 0.001). In addition, the MDR group had larger burn areas (TBSA 22.4% ± 15.8% versus 7.6% ± 4.9%; P < 0.001), longer hospital LOS (27.9 ± 8.3 days versus 8.7 ± 5.6 days; P < 0.001), and longer surgery duration (2.9 ± 0.5 hours versus 1.3 ± 0.6 hours; P < 0.001).

Table 1.

Sociodemographic Characteristics of Children with MDR or Non-MDR Infections^a

Characteristics	Total (n = 763)	MDR Infection Group (n = 93)	Non-MDR Infection Group (n = 670)	P
Male sex	461 (60.4)	55 (59.1)	406 (60.6)	0.788
Burn reasons				0.344
Scalds	491 (64.3)	52 (55.9)	439 (65.5)	0.177
Flames	249 (32.6)	38 (40.9)	211 (31.5)
Others	23 (3.1)	3 (3.2)	20 (3.0)
Acute inhalation injury	52 (6.8)	33 (35.5)	19 (2.8)	<0.001
Surgery	253 (33.2)	66 (71.0)	187 (27.9)	<0.001
ICU admission	47 (6.16)	32 (34.4)	15 (2.2)	<0.001
Urinary catheter	95 (12.5)	24 (25.8)	71 (10.6)	<0.001
Mechanical ventilation	19 (2.5)	14 (15.1)	5 (0.7)	<0.001
Antibiotic prophylaxis	262 (34.3)	61 (65.6)	201 (30.0)	<0.001
CVC	25 (3.3)	18 (19.4)	7 (1.0)	<0.001
Age, yrs	2.8 ± 2.9	2.7 ± 2.6	2.9 ± 3.0	0.650
TBSA, %	9.1 ± 7.6	22.4 ± 15.8	7.6 ± 4.9	<0.001
LOS, d	11.4 ± 4.5	27.9 ± 8.3	8.7 ± 5.6	<0.001
Length of surgery, h	2.4 ± 1.1	2.9 ± 0.5	1.3 ± 0.6	<0.001

^aData are presented as n (%) or mean ± SD.

To investigate the determinants of MDR infection among pediatric burn patients, both univariable and multivariable logistic regression analyses were performed (Table 2). The univariable analysis identified a range of clinical variables significantly associated with MDR infection. Greater TBSA, prolonged hospital LOS, and the presence of clinical interventions—including central venous catheterization, ICU admission, mechanical ventilation, urinary catheter use, and prophylactic antibiotic administration—were each significantly correlated with increased odds of MDR infection (all P < 0.001). In addition, acute inhalation injury and the need for surgical intervention were found to be significant risk factors. Variables such as patient sex, age, and burn etiology (scalds, flames, others) were not significantly associated with MDR infection in the univariable model.

Table 2.

Univariable and Multivariable Logistic Regression Analysis of Risk Factors for MDR Infection in Pediatric Patients^a

Variables	Univariable Logistic Regression Analysis		Multivariable Logistic Regression Analysis
	OR (95% CI)	P	OR (95% CI)	P
TBSA	1.25 (1.15, 1.35)	<0.001b	1.18 (1.08, 1.28)	0.001b
LOS	1.23 (1.15, 1.32)	<0.001b	1.21 (1.14, 1.30)	<0.001b
CVC	22.73 (9.19, 56.20)	<0.001b	7.40 (2.95, 18.12)	0.001b
ICU admission	22.91 (11.76, 44.64)	<0.001b	6.12 (1.95, 15.30)	0.009b
Mechanical ventilation	23.57 (8.27, 67.18)	<0.001b	8.75 (1.92, 20.48)	0.005b
Acute inhalation injury	18.84 (10.10, 35.15)	<0.001b	6.55 (1.89, 17.20)	0.003b
Antibiotic prophylaxis	4.45 (2.81, 7.04)	<0.001b	1.25 (0.72, 2.85)	0.315
Surgery	6.31 (3.91, 10.19)	<0.001b	1.07 (0.85, 2.12)	0.431
Urinary catheter	2.93 (1.73, 4.96)	<0.001b	1.22 (0.85, 1.49)	0.632
Scalds (burn reasons)	0.94 (0.61, 1.46)	0.821	0.78 (0.59, 1.22)	0.479
Flames (burn reasons)	0.67 (0.43, 1.04)	0.083	0.55 (0.28, 1.91)	0.261
Others (burn reasons)	1.50 (0.96, 2.34)	0.077	1.27 (0.81, 2.43)	0.187
Age	1.12 (0.98, 1.29)	0.056	1.04 (0.90, 1.18)	0.145
Sex	0.94 (0.61, 1.46)	0.821	0.97 (0.55, 2.30)	0.512

^aReference groups were female for sex and no for ICU admission, mechanical ventilation, CVC, urinary catheter, acute inhalation injury, surgery, and antibiotic prophylaxis. For burn reasons, each category was compared against the combination of the other 2 categories (ie, scalds versus flames + others, flames versus scalds + others, and others versus scalds + flames).

^bSignificant.

After adjustment for potential confounding factors in the multivariable logistic regression analysis, 6 independent predictors of MDR infection were identified. These included larger TBSA (OR, 1.18 [95% CI, 1.08 to 1.28]; P = 0.001), longer LOS (OR, 1.21 [95% CI, 1.14 to 1.30]; P < 0.001), use of a CVC (OR, 7.40 [95% CI, 2.95 to 18.12]; P = 0.001), ICU admission (OR, 6.12 [95% CI, 1.95 to 15.30]; P = 0.009), mechanical ventilation (OR, 8.75 [95% CI, 1.92 to 20.48]; P = 0.005), and acute inhalation injury (OR, 6.55 [95% CI, 1.89 to 17.20]; P = 0.003). Notably, some variables that were significant in the univariable model—such as prophylactic antibiotic use, urinary catheterization, and surgery—did not remain statistically significant in the multivariable analysis, suggesting that their effects may be mediated by other factors reflecting illness severity or the extent of critical care. Variables such as age, sex, and the underlying cause of burn injury were not associated with MDR infection in either model.

Based on the multivariable logistic regression model, a nomogram was developed to provide individualized risk prediction for MDR infection in pediatric burn patients (Fig. 2). The nomogram incorporated 6 independent predictors identified in the multivariable analysis: TBSA, LOS, use of CVC, ICU admission, mechanical ventilation, and presence of acute inhalation injury. Each predictor was assigned a corresponding score on the point scale at the top of the nomogram. TBSA and LOS were continuous variables contributing proportionally to the total score, whereas the remaining 4 predictors were binary variables (yes/no), with the presence of each condition increasing the total score accordingly. The individual scores were then summed to obtain a total point value, which was mapped to the bottom scale to estimate the probability of MDR infection, ranging from approximately 0.1 to 0.9. This nomogram provides a visual and practical tool for clinicians to assess the risk of MDR infection at an early stage and may support individualized decision-making regarding infection surveillance and preventive strategies in children with severe burns.

Fig. 2.

Nomogram for predicting MDR infection risk. Each variable is assigned a point value on the top scale. The total score is obtained by summing individual points and is mapped to the bottom scale to estimate the predicted probability of MDR infection.

The performance of the nomogram was evaluated using both discrimination and calibration metrics. As shown in Figure 3, the receiver operating characteristic curve demonstrated excellent discriminative ability of the model in identifying patients at risk of MDR infection. The AUC was 0.962, indicating high sensitivity and specificity across various thresholds. The calibration of the nomogram was assessed using a calibration plot with 500 bootstrap resamples. The plot showed a close agreement between the predicted probabilities and the actual observed outcomes, with both the apparent and bias-corrected curves closely following the ideal 45-degree reference line. The mean absolute error between the predicted and observed probabilities was 0.009, suggesting that the model predictions were well-calibrated and reliable across the full range of risk scores.

Fig. 3.

Discrimination and calibration of the prediction model. (Left) Receiver operating characteristic curve. The x axis shows 1 − specificity and the y axis shows sensitivity. (Right) Calibration plot. The x axis represents predicted probability and the y axis represents observed probability. The diagonal dashed line indicates perfect prediction; the solid line shows the actual model performance; the dotted line is the bias-corrected performance based on bootstrap resampling.

To evaluate the clinical applicability of the predictive nomogram, DCA was conducted (Fig. 4). The left panel of Figure 4 illustrates the net clinical benefit of using the nomogram across a range of threshold probabilities for predicting MDR infections. The decision curve revealed that, within a wide range of threshold probabilities (approximately 0.05 to 0.8), the nomogram consistently yielded a greater net benefit compared with either the “treat-all” or “treat-none” strategies. This indicates that using the nomogram to guide clinical decision-making would lead to more appropriate interventions with fewer unnecessary treatments or missed infections. The clinical impact curve further demonstrated the nomogram’s utility by displaying the estimated number of patients classified as high risk across different thresholds. The curve showed that, among those identified as high risk by the model, a substantial proportion were true-positives with MDR infection, particularly at lower thresholds. This supports the use of the model in screening and risk stratification efforts, especially when aiming to balance the benefits of early intervention against the risks of overtreatment.

Fig. 4.

DCA and clinical impact curve. (Left) Net benefit plotted against threshold probability. The black line represents the nomogram; the gray line assumes all patients are treated; the horizontal line assumes none are treated. (Right) Clinical impact curve showing the number of patients classified as high risk (blue line) and the number of true positives (red line) across different threshold probabilities.

DISCUSSION

In this retrospective study of pediatric burn patients with HAIs, we identified 6 independent clinical predictors associated with the development of MDRO infections: larger TBSA, prolonged LOS, ICU admission, mechanical ventilation, CVC use, and acute inhalation injury. Based on these factors, we developed a nomogram with excellent predictive performance (AUC 0.962), robust calibration, and demonstrable clinical utility. Our findings not only align with established literature but also contribute novel insights by focusing specifically on MDRO infection among pediatric patients with confirmed HAIs, a population often overlooked in predictive modeling.

Extensive burn size is a well-established risk factor for infection in burn patients, as it compromises the skin barrier and promotes systemic immunosuppression. Recent studies confirm that TBSA greater than 20% significantly increases the likelihood of MDRO colonization and infection, especially in pediatric patients.^14,15 In our study, TBSA was treated as a continuous variable, and higher TBSA percentages were independently associated with increased MDRO infection risk. Larger burn areas not only increase wound exposure to nosocomial organisms but also necessitate longer inpatient care, repeated surgical interventions, and higher antibiotic consumption, all of which are conducive to colonization and resistance selection.^16,17

Acute inhalation injury, another indicator of injury severity, similarly conferred higher risk. Inhalation injuries are often accompanied by a systemic inflammatory response, hypoxia, and the need for mechanical ventilation—factors that impair immune defense and increase vulnerability to respiratory pathogens.^18,19 Previous work has shown that inhalation injury often necessitates intubation and ICU-level care, increasing exposure to nosocomial pathogens such as P aeruginosa and A baumannii.^20,21 The combination of airway epithelial damage, impaired mucociliary clearance, and prolonged ventilation time substantially raises the risk of ventilator-associated pneumonia due to MDRO.²²

Prolonged hospital LOS is both a marker of disease severity and a contributor to nosocomial colonization.^11,23 Our findings are consistent with literature indicating that longer hospitalization correlates with increased MDRO acquisition, likely due to extended exposure to hospital flora and repeated antibiotic use.^11,24 In pediatric burn ICUs, extended LOS often reflects clinical instability, delayed wound closure, and complications such as sepsis or organ dysfunction—all of which further increase MDRO risk.^25,26 ICU admission was another strong predictor of MDRO infection in our cohort, which reflects the higher prevalence of invasive procedures, broad-spectrum antibiotic use, and critical illness among this population.²⁷ This is supported by recent epidemiologic data showing elevated MDRO colonization rates among critically ill burn patients, particularly those requiring repeated interventions and prolonged duration of ICU stay.²⁸

Invasive devices such as mechanical ventilators and CVCs have long been implicated in MDRO pathogenesis in burn patients.²⁵ Our results showed a significant association between both modalities and MDR infection, in line with reports that these devices serve as entry points for resistant bacteria and are often colonized in the ICU setting.²⁹ Mechanical ventilation has specifically been linked to ventilator-associated pneumonia caused by methicillin-resistant S aureus and MDR P aeruginosa, whereas CVCs are major contributors to bloodstream infections due to MDROs.^30,31 The use of these devices also correlates with increased exposure to antibiotics and cross-transmission risk within the ICU, reinforcing the need for stringent infection control practices.³² Given the necessity of mechanical ventilators and CVCs in critically ill burn patients, early removal and stringent aseptic protocols are essential.

Although variables such as surgical intervention, prophylactic antibiotic use, and urinary catheterization were significantly associated with MDR infection in the univariable analysis, they did not remain statistically significant in the multivariable logistic regression model. This attenuation may be due to confounding or multicollinearity with stronger predictors such as TBSA, ICU admission, and mechanical ventilation. For example, patients who underwent surgery or received empirical prophylactic antibiotics were typically those with more extensive burns or greater clinical instability—factors already accounted for in the multivariable model. As a result, the apparent associations observed in unadjusted analyses may have reflected underlying disease severity rather than independent effects. These findings emphasize the need for cautious interpretation of univariable associations and the importance of adjusting for confounding variables when developing predictive models.

The predictive model constructed in this study offers a practical and evidence-based tool for early identification of high-risk pediatric burn patients. By integrating routinely available variables, this nomogram may support clinicians in implementing targeted surveillance, isolation precautions, or adjusted empirical antibiotic strategies. This model also may aid in optimizing antimicrobial stewardship, reducing unnecessary broad-spectrum antibiotic exposure and limiting resistance pressure in pediatric burn ICUs. Compared with previous models that examined general infection risk, our nomogram provides a more focused framework for assessing MDRO-specific vulnerability among already infected patients, a clinically distinct and highly relevant subgroup.

This study has several limitations. This was a single-center retrospective analysis. Local epidemiologic patterns of antimicrobial resistance, institutional infection control policies, clinical workflows, and burn care protocols may differ significantly across centers and could influence the observed risk factors and model performance. Therefore, caution should be exercised in extrapolating our findings to other settings, and external validation in diverse institutions is essential to confirm the model’s generalizability. In addition, although younger age—particularly under 2 years—has been associated with increased MDRO susceptibility, age stratification was not performed in this study due to the absence of a significant association in univariate analysis (P = 0.650), the relatively narrow age distribution (mean age less than 3 years), and limited subgroup sample size. Future investigations with larger and more heterogeneous pediatric cohorts are warranted to elucidate potential age-related risk differentials. Dynamic variables such as previous colonization status, pathogen-specific resistance mechanisms, and cumulative antibiotic exposure were not included. Other relevant factors, such as the frequency and duration of surgical interventions, baseline nutritional status, time from burn to admission, previous hospitalization history, and adherence to infection control protocols, were also not accounted for. Future studies should consider incorporating these factors and evaluating the utility of machine learning–based predictive models. Moreover, integration of microbiome or genomic data may further enhance risk prediction accuracy in the era of precision medicine. Nevertheless, our findings provide a solid foundation for practical, risk-based MDRO prevention strategies in pediatric burn populations.

CONCLUSIONS

MDR infections remain a significant threat in pediatric burn care, leading to prolonged hospitalization, increased morbidity, and limited therapeutic options. In this study, we identified key clinical predictors of MDRO infection in pediatric patients with HAIs and constructed a robust, clinically applicable nomogram for individualized risk prediction. By incorporating variables such as TBSA, hospital LOS, ICU admission, invasive device use, and inhalation injury, our model provides a practical framework for early identification of high-risk patients. The use of such tools can improve infection control, guide empirical antibiotic decisions, and support targeted interventions in pediatric burn ICUs. Future research should focus on external validation, incorporation of dynamic and microbiological data, and adaptation of predictive analytics to real-time clinical workflows. Our findings represent a meaningful step toward precision-based infection prevention in a vulnerable patient population.

DISCLOSURE

The authors have no financial relationships or conflicts of interest to disclose.

ACKNOWLEDGMENTS

This work was supported by SingHealth Duke–NUS Academic Medicine Research Grant 2023 and the Youth Science Fund Cultivation Funding Program of Shandong First Medical University (grant 202201-130).

ETHICS STATEMENT

This study was approved by the ethics committee of the Central Hospital Affiliated to Shandong First Medical University (approval 2023-185-01). Given the retrospective nature of the study and the use of anonymized clinical data, the requirement for informed consent was waived by the committee. All procedures were conducted in accordance with the ethical standards of the institutional research committee and the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

DATA AVAILABILITY

The datasets generated or analyzed during the current study are not publicly available due to privacy regulations, but are available from the corresponding author on reasonable request.