Development and External Validation of a Machine Learning Model for Predicting Liver Injury in Children With Mycoplasma Pneumoniae Pneumonia: A Multicenter Retrospective Study

Development and External Validation of a Machine Learning Model for Predicting Liver Injury in Children With Mycoplasma Pneumoniae Pneumonia: A Multicenter Retrospective Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Development and External Validation of a Machine Learning Model for Predicting Liver Injury in Children With Mycoplasma Pneumoniae Pneumonia: A Multicenter Retrospective Study Yang Yu, Yang Jing, Yinyan Tang, Feng Liu, Hongjuan Wei This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9460650/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract Background: Mycoplasma pneumoniae pneumonia (MPP) is associated with a high risk of liver injury, which adversely affects clinical outcomes and healthcare costs. Early identification of at-risk children remains challenging, and no validated predictive tool is currently available. Methods: This multicenter retrospective study included 1,321 children with MPP from two centers in Nanjing, China. The development cohort was randomly split into training (67%) and test (33%) sets, with an independent external validation cohort (n = 640) from another hospital. Liver injury was defined as alanine aminotransferase (ALT) >80 U/L. Feature selection was performed using the boruta algorithm, and ten machine learning algorithms were developed and evaluated. Model performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy (ACC), sensitivity, specificity, calibration metrics, and decision curve analysis (DCA). The SHAP method was used for model interpretation. Results: Among 1,321 patients, 55 (4.2%) developed liver injury. The boruta algorithm identified six predictors: LDH, D-dimer, CK, age, IL-6, and CK-MB. The neural network (nnet) model demonstrated optimal performance, with AUC values of 0.914 in the training set, 0.829 in the test set, and 0.755 in the external validation set. The model showed acceptable calibration and modest clinical utility on decision curve analysis. SHAP analysis revealed LDH as the most important predictor. An interactive web-based application was developed to facilitate clinical implementation. Conclusions: We developed and validated a machine learning-based model using six readily available predictors that identifies children with MPP at risk for liver injury. The accompanying web-based tool may assist clinicians in early risk stratification and timely intervention. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Health care Health sciences/Medical research Health sciences/Risk factors Mycoplasma Pneumoniae Pneumonia Liver Injury Machine Learning Prediction Model Neural Network Children Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 1. Introduction MPP is a leading cause of community-acquired pneumonia in children, accounting for 10–40% of pediatric cases worldwide[1, 2], characterized by rapid progression and a high propensity for extrapulmonary complications[3, 4]. Among these, liver injury occurs in 10-30% of hospitalized children with MPP and is associated with prolonged hospitalization, increased healthcare costs, and potential adverse outcomes[5]. Early identification of children at risk for liver injury is critical for implementing preventive strategies such as avoiding hepatotoxic medications and intensifying monitoring. While serum ALT elevation provides the definitive diagnosis of liver injury, it is inherently a lagging indicator. In clinical practice, ALT levels measured at hospital admission may be normal or only mildly elevated, with significant elevations often occurring later during hospitalization. This diagnostic delay underscores a critical gap: the absence of a reliable tool for early identification of children with MPP who are at risk for subsequent liver injury. Existing predictive models, predominantly based on traditional logistic regression, are often limited by small sample sizes and lack external validation [6]. More fundamentally, they typically utilize parameters measured at or after the diagnosis of liver injury, rather than focusing on baseline admission data that could enable true preemptive risk stratification. Consequently, there is an urgent need for a robust, early-warning model using readily available admission data to facilitate timely monitoring and preventive intervention. Machine learning offers distinct advantages for clinical prediction by automatically identifying complex patterns, handling nonlinear relationships, and capturing variable interactions that conventional methods may miss[7, 8]. These techniques have been successfully applied in pediatric settings, including prediction of refractory MPP and myocardial injury[9]. Additionally, explainable artificial intelligence methods such as SHAP have enhanced model interpretability, addressing the "black box" concern and facilitating clinical adoption[10, 11]. To address these gaps, we conducted a multicenter retrospective study to develop and validate a machine learning-based model for predicting liver injury in children with MPP. We evaluated ten algorithms, selected the optimal model based on discriminative ability and generalizability, and developed an interactive web-based application to facilitate clinical implementation. 2. Materials and Methods 2.1 Dataset source This retrospective cohort study extracted data from electronic medical records (EMR) of two medical centers in Nanjing, China. The development cohort was obtained from Nanjing Children's Hospital, a tertiary pediatric center. Data were collected from patients admitted to respiratory wards between January 1, 2023, and December 31, 2023. The external validation cohort was collected from Nanjing Lishui People's Hospital during the same period. The study protocol was approved by the Institutional Review Board (IRB) of Children’s Hospital of Nanjing Medical University [Approval No. 201812257-1] and Nanjing Lishui People's Hospital [Approval No. 2024KY0802-02], with waiver of informed consent due to retrospective design and de-identified data. All procedures were performed in accordance with the Declaration of Helsinki. 2.2 Participants A total of 1,323 patients diagnosed with MPP were initially included. After excluding 2 patients with discharge against medical advice, 1,321 patients were enrolled and randomly divided into a training set (67%, n = 885) and a testing set (33%, n = 436). An independent external validation cohort consisting of 640 patients from Nanjing Lishui People's Hospital was also included (Figure 1). 2.3 Variable selection Based on previous literature and clinical relevance, the following three categories of variables were included: (a) Demographic and clinical characteristics: sex, age (years), cough (yes/no), fever (yes/no), fever duration before admission (days), and pulmonary consolidation on imaging (yes/no); (b) Laboratory parameters: CRP, WBC, LYMPH, NEUT, HGB, PLT, LDH, CK, CK-MB, D-dimer, IL-6. All laboratory variables were collected within 24 hours of admission. 2.4 Outcome definition Liver injury was defined as ALT >80 U/L based on pediatric reference ranges [5, 12]. 2.5 Statistical analysis All statistical analyses were performed using R Statistical Software (Version 4.2.2, http://www.R-project.org) and the Free Statistics analysis platform (Version 2.4, Beijing, China). A two-sided P value < 0.05 was considered statistically significant. Continuous variables were assessed for normality using the Shapiro–Wilk test. Normally distributed variables were presented as mean ± standard deviation (SD) and compared using Student’s t-test; non-normally distributed variables were presented as median (interquartile range) and compared using the Mann–Whitney U test. Categorical variables were expressed as numbers (percentages) and compared using the χ² test or Fisher’s exact test, as appropriate. Prior to feature selection, pairwise correlations among all candidate predictor variables were assessed. Feature selection was performed using the boruta algorithm via the boruta package in R. Variables with importance significantly higher than the maximum importance of shadow features were retained for model development. Multivariable logistic regression with mutual adjustment for the six selected variables was performed to assess their independent associations with liver injury. Ten supervised machine learning algorithms (catboost, gbm, kknn, lightgbm, naive bayes, nnet, ranger, rpart, svm, and xgboost) were developed to predict liver injury. All models were trained on the training set (n = 885), with hyperparameter tuning performed using 10-fold cross-validation. Model discrimination was assessed using AUC, ACC, Brier score, Fβ score, calibration error (CE), precision-recall AUC (PR-AUC), sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). Receiver operating characteristic (ROC) curves and calibration curves were plotted to visualize model performance. DCA was performed to evaluate clinical utility across a range of threshold probabilities. To enhance model interpretability, SHapley Additive exPlanations (SHAP) was applied to the final nnet model using the shapviz and fastshap packages. Variable importance was assessed by mean absolute SHAP values, and beeswarm, dependence, and individual force plots were generated to visualize feature effects. 2. 6 Model d evelopment, e valuation, and v isualization 2. 6 .1 Data cleaning After applying the inclusion and exclusion criteria detailed in Section 2.2, data completeness was assessed for all variables in the remaining cohort. All retained patients (n = 1,321) had complete data for the outcome variable (ALT >80 U/L) and all candidate predictor variables; therefore, no imputation was performed. 2. 6 .2 Feature selection Correlation analysis among all candidate variables is presented in Figure 2. Subsequently, feature selection was performed using the boruta algorithm, which iteratively compares the importance of original variables with randomized shadow features. Variables with importance significantly higher than the maximum importance of shadow features were classified as "confirmed" (depicted in green in Figure 3) and retained for model development. After 100 iterations, the algorithm confirmed six variables as important predictors: LDH, D-dimer, CK, age, IL-6, and CK-MB. These selected variables were subsequently used for model training and validation. 2. 6 .3 Modeling and evaluation Ten supervised machine learning algorithms were employed for model development: catboost, gbm, kknn, lightgbm, naive bayes, nnet, ranger, rpart, svm, and xgboost. All models were trained on the training set (n=885). Hyperparameter tuning was performed via 10-fold cross-validation on the training set. Model discrimination was assessed using AUC, ACC, Brier score, Fβ score, CE, PR-AUC, sensitivity, specificity, NPV, and PPV. ROC curves and calibration curves were plotted to visualize model discrimination and calibration, respectively. DCA was performed to evaluate the clinical utility of the models across a range of threshold probabilities. The performance of all ten models in the training set, internal test set, and external validation set was comprehensively evaluated. Based on comprehensive evaluation of discriminative ability and generalizability across three datasets, the nnet model was selected as the optimal model. The nnet model demonstrated robust performance with good discrimination and acceptable external validity. 2. 6 .4 Evaluation of the importance of variables SHAP is a post-hoc explanation framework for machine learning models based on game theory[13]. It quantifies the importance of each feature in the model by calculating the contribution value, known as the Shapley value, for each feature towards the predicted outcome. This study utilized the SHAP method to enhance the interpretability and transparency of the final nnet model. SHAP analysis was performed on the training set, and variable importance was assessed by mean absolute SHAP values. Beeswarm plots were generated to visualize the distribution and direction of SHAP values for each feature, illustrating the relationship between feature values and their impact on model predictions. Dependence plots were constructed to further explore the marginal effects of key predictors. Individual force plots were also generated to demonstrate the decision-making process for representative cases, facilitating local interpretability. 2. 6 . 5 Model visualization To facilitate clinical application, an interactive web-based application was developed using the shiny package in R, allowing clinicians to input the six selected predictors and obtain real-time predicted probabilities of liver injury. The application is publicly accessible at http://127.0.0.1:7728/. 3. Results 3.1 Baseline characteristics of the study population A total of 1,321 children diagnosed with MPP were included, of whom 55 (4.2%) developed liver injury (defined as ALT > 80 U/L). Baseline characteristics are presented in Table 1. Patients with liver injury had significantly longer hospital stays (7.5 ± 3.8 vs. 5.7 ± 2.3 days, P < 0.001) and higher costs (14,498.2 ± 7,622.6 vs. 9,822.9 ± 3,680.7 yuan, P < 0.001). They also exhibited higher rates of pulmonary consolidation (21.8% vs. 11.9%, P = 0.029), atelectasis (18.2% vs. 8.2%, P = 0.022), pleural effusion (23.6% vs. 5.2%, P < 0.001), and pulmonary embolism (12.7% vs. 2.4%, P < 0.001), and were more likely to receive oxygen therapy (23.6% vs. 8.1%, P < 0.001) and be classified as RMPP (Refractory Mycoplasma Pneumoniae Pneumonia) (90.9% vs. 76.5%, P = 0.013). Laboratory findings differed markedly between groups. The liver injury group had elevated WBC (12.8 ± 5.2 vs. 10.9 ± 4.7 ×10⁹/L, P = 0.004) and NEUT (8.8 ± 4.3 vs. 7.2 ± 4.0 ×10⁹/L, P = 0.005) counts, higher HGB (129.5 ± 13.6 vs. 126.1 ± 10.1 g/L, P = 0.016), ALT [131.0 (101.0, 197.5) vs. 15.0 (11.0, 22.0) U/L, P < 0.001], AST [50.0 (38.5, 92.5) vs. 24.0 (20.0, 30.0) U/L, P < 0.001], and LDH (574.4 ± 383.7 vs. 356.3 ± 136.5 U/L, P < 0.001), and substantially higher D-dimer [718.0 (391.0, 1270.0) vs. 231.0 (154.0, 382.8) ng/mL, P 0.05). 3.2 Feature s election To mitigate overfitting and enhance model parsimony, feature selection was performed using the boruta algorithm. A total of 17 candidate variables were assessed, encompassing demographic characteristics (age, sex), clinical features (cough, fever, pre-admission fever duration, pulmonary consolidation), and laboratory parameters (CRP, WBC, LYMPH, NEUT, HGB, PLT, LDH, CK, CK-MB, D-dimer, IL-6). Of these, six were confirmed as significant predictors (green-zone features in Figure 3): LDH, D-dimer, CK, age, IL-6, and CK-MB. These variables demonstrated importance scores significantly higher than their shadow counterparts, indicating robust predictive value for liver injury in children with MPP. The remaining variables, such as CRP, WBC, PLT, and cough, were classified as unimportant (red-zone features) and excluded. This strategy ensured retention of the most relevant predictors, enhancing model generalizability and interpretability. As a complementary analysis, multivariable logistic regression identified LDH (adj OR = 1.08, 95% CI : 1.05–1.10, P < 0.001) and age (adj OR = 1.15, 95% CI : 1.03–1.29, P = 0.016) as independent risk factors (Supplementary Table 1). 3.3 Evaluation of m odel p erformance Following feature selection, the six predictors were used to develop ten machine learning algorithms. The cohort was randomly divided into training (67%, n = 885) and internal test (33%, n = 436) sets, with an independent external validation cohort (n = 640). Model performance was assessed using AUC, ACC, sensitivity, specificity, PPV, NPV, Fβ score, Brier score, and calibration error (Tables 2–4) (Figure 4). 3.3.1 Training s et p erformance Several models demonstrated excellent discrimination in the training cohort. Lightgbm achieved perfect metrics (AUC = 1.000, ACC = 1.000), while catboost, kknn, and ranger showed near-perfect performance (AUC = 1.000, ACC > 0.940). Naive bayes performed relatively poorly (AUC = 0.804). Nnet showed moderate performance (AUC = 0.914) without overfitting (Tables 2). 3.3.2 Internal t est s et p erformance Performance attenuated in the test set as expected. Lightgbm maintained the highest ACC (0.938) and specificity (0.962). Naive bayes achieved the highest AUC (0.911) but moderate ACC (0.805). Notably, nnet demonstrated stable performance (AUC = 0.829, ACC = 0.814, sensitivity = 0.722) (Tables 3). Confusion matrices and performance metrics on the test set are shown in Figure 5 (A–J) . 3.3.3 External v alidation Most models declined substantially in the external cohort (lightgbm AUC = 0.635; catboost = 0.664; ranger = 0.726). Nnet maintained the most robust performance (AUC = 0.755, ACC = 0.902, Brier score = 0.054), indicating superior generalizability and acceptable calibration (Tables 4). Confusion matrices and performance metrics on the external validation set are shown in Figure 5 (K–T) . 3.3.4 Clinical u tility and f inal m odel s election Comprehensive performance metrics of the ten models across the three datasets are summarized in Tables 2–4 and visually compared in Figure 6. Calibration curves confirmed that nnet-predicted probabilities matched observed rates across all datasets (Figure 7). DCA showed that the nnet model provided limited net benefit within a narrow range of threshold probabilities (Figure 8). Considering discrimination, calibration, clinical utility, and generalizability, nnet was selected as the final model (Table 5). 3.4 Evaluation of v ariable i mportance To enhance model interpretability, SHAP analysis was performed on the final nnet model. 3.4.1 Feature i mportance and d irectional e ffects The SHAP summary plot (Figure 9A) revealed the directional effects of each predictor on model output. Variable importance ranked by mean absolute SHAP values (Figure 9B) identified LDH as the dominant predictor (0.122), followed by D-dimer (0.072), CK (0.067), age (0.059), IL-6 (0.047), and CK-MB (0.047). SHAP dependence plots further elucidated the functional forms of these relationships. LDH exhibited a monotonic positive association (Figure 10A), with SHAP values rising steeply as LDH increased. Age displayed a positive linear association (Figure 10D). 3.4.2 Individual i nterpretation SHAP force plots illustrate how these features drive individual predictions (Figure 11A). A high-risk case (predicted probability 0.742) was driven by elevated LDH (1081 U/L), D-dimer (422 ng/mL), and CK-MB (92 U/L) pushing the prediction upward. In contrast, a low-risk case (predicted probability 0.156) (Figure 11B) was characterized by young age (4.84 years) as the dominant protective factor pulling the prediction downward, overcoming the risk contributions from elevated D-dimer (616 ng/mL) and mildly elevated LDH (417 U/L). 3.5 Model v isualization To facilitate clinical application of the final nnet model, an interactive web-based predictive tool was developed. As illustrated in Figure 12, the web application allows clinicians to input the six selected predictors (LDH, D-dimer, CK, age, IL-6, and CK-MB) and generates the predicted probability of liver injury in real time. The application is publicly accessible at http://127.0.0.1:7728/. 4. Discussion 4. 1 Summary of Principal Findings In this study, we developed and validated a machine learning-based predictive model for early identification of liver injury in children with MPP. Using the boruta algorithm for feature selection, six key predictors were identified: LDH, D-dimer, CK, age, IL-6, and CK-MB. Among ten machine learning algorithms evaluated, the nnet model demonstrated optimal and stable performance across the training, internal test, and external validation cohorts, with AUC values of 0.914, 0.829, and 0.755, respectively. To facilitate clinical implementation, an interactive web-based application was developed to provide real-time risk predictions. 4. 2 Comparison with e xisting s tudies In this study, we developed and validated a novel machine learning model to predict liver injury in children with MPP, a complication currently lacking early warning tools. Previous studies have primarily focused on identifying risk factors for MPP-associated liver injury using traditional logistic regression approaches. For instance, several studies have reported that elevated inflammatory markers such as WBC, NEUT, and CRP are associated with disease severity and extrapulmonary complications in Mycoplasma pneumoniae infection[14, 15]. However, these studies often relied on univariate or multivariate regression analyses, which may not fully capture the complex nonlinear relationships among clinical and laboratory variables. In contrast, machine learning approaches, particularly ensemble methods and nnet, are better equipped to handle such complexity and have demonstrated superior predictive performance in various clinical contexts[16]. The performance of our nnet model (AUC = 0.829 in the test set) is comparable to or better than previously reported predictive models for other complications of MPP. For example, previous studies on predicting MPP have reported AUC values ranging from 0.75 to 0.85[17, 18]. The moderate decrease in performance observed in the external validation set (AUC = 0.755) is expected and reflects the inherent heterogeneity between different clinical centers, highlighting the importance of external validation for assessing model generalizability. 4. 3 Interpretation of k ey p redictors The SHAP analysis identified LDH as the dominant predictor of liver injury (mean SHAP value = 0.122). LDH is a nonspecific marker of cellular damage abundant in hepatocytes. Elevated LDH levels in MPP patients have been associated with disease severity and extrapulmonary complications[19]. Our study further demonstrates its emergence as a critical biomarker specifically for predicting liver injury in this context. D-dimer ranked second (0.072) in importance. As a well-established marker of fibrin degradation, elevated D-dimer reflects coagulation activation and endothelial injury—common pathophysiological features in severe Mycoplasma pneumoniae infection [20]. The strong association with liver injury may reflect systemic microvascular thrombosis and subsequent hepatic ischemia-reperfusion injury, a mechanism implicated in sepsis-associated liver dysfunction[21, 22]. Furthermore, Mycoplasma pneumoniae induces procoagulant activity through direct endothelial damage and cytokine-mediated coagulation cascade activation[23]. Notably, CK ranked third (0.067) but exhibited a protective inverse relationship (negative SHAP values), consistent with lower baseline CK levels observed in the liver injury group. This paradoxical finding may reflect the complex metabolic state in severe infection: higher CK could indicate preserved muscle mass and nutritional status, whereas low CK may signal systemic catabolism and multi-organ dysfunction[24]. Alternatively, CK elevation might parallel general tissue turnover without specific hepatic toxicity[25]. Older age (0.059) contributed to higher predicted risk, possibly reflecting more robust immune responses and greater inflammatory burden in older children[26]. Elevated IL‑6 levels (0.047) are strongly associated with severe Mycoplasma pneumoniae infection and reflect the intensity of the host inflammatory response, supporting its role as a biomarker of disease severity [27, 28] . CK‑MB (0.047), traditionally a cardiac marker, emerged as a risk factor for liver injury with importance comparable to IL‑6. This finding aligns with the observation that M. pneumoniae infection can induce concurrent cardiac and hepatic damage through shared immunopathological mechanisms[29, 30]. 4. 4 Clinical i mplications The developed model provides a practical tool for early identification of children with MPP who are at elevated risk of liver injury, utilizing only routine admission data. First, all six predictors (LDH, D-dimer, CK, age, IL-6, and CK-MB) are obtainable from admission workup, enabling preemptive risk stratification before ALT elevation occurs. Second, the web-based application facilitates real-time risk assessment for avoidance of hepatotoxic medications and intensified monitoring, potentially reducing prolonged hospitalization. Third, SHAP-based interpretability enhances clinician trust and supports shared decision-making with families, addressing the "black box" concern limiting clinical adoption of machine learning models. 4. 5 Strengths and l imitations This study has several strengths. The boruta algorithm ensured rigorous feature selection, reducing overfitting risk. The model was evaluated on both internal and external validation cohorts, demonstrating generalizability across clinical settings. SHAP analysis enhanced interpretability, addressing the “black box” concern. A publicly accessible web-based application was developed to facilitate clinical translation. Several limitations should be acknowledged. First, this retrospective study may have selection bias, and although external validation was performed using data from another hospital, the sample size was modest, particularly for the liver injury group (n = 55). Second, the retrospective design limited standardized follow-up protocols; prospective validation is needed to confirm the optimal timing and frequency of risk assessment in clinical practice. Third, although all six predictors are routinely available in tertiary hospitals, some indicators (e.g., IL-6) may not be measured in all primary or secondary care settings, which could limit the model's applicability in resource-limited settings. 5. Conclusion We developed and validated a machine learning-based model using six readily available predictors (LDH, D-dimer, CK, age, IL-6, and CK-MB) to identify liver injury risk in children with MPP. The nnet model demonstrated robust discrimination (AUC = 0.829 in internal test, 0.755 in external validation) and was deployed as an interactive web-based application. This tool may facilitate early risk stratification and timely intervention, potentially reducing the incidence of liver injury in MPP. Declarations Authors' contributions Hongjuan Wei conceptualized and supervised the study, obtained ethical approval, and critically revised the manuscript.Feng Liu contributed to the study design, data interpretation, and manuscript revision. Yang Yu and Yang Jing contributed equally as co-first authors, performing data analysis, visualization, and initial drafting. Yinyan Tang assisted with data collection and validation. All authors reviewed and approved the final manuscript, had full access to the data, and accept responsibility for its integrity and accuracy. Hongjuan Wei is the guarantor. Ethics approval and consent to participate The study protocol was approved by the Institutional Review Board (IRB) of Children’s Hospital of Nanjing Medical University [Approval No. 201812257-1] and Nanjing Lishui People's Hospital [Approval No. 2024KY0802-02]. All procedures performed were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments. Due to the retrospective nature of the study, the requirement for informed consent was waived by the committee. Declaration of competing interest The authors declare that they have no competing interests. Data availability The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Funding information No funding support in the data collection, analysis or preparation of the manuscript. References Ding G, Zhang X, Vinturache A, van Rossum A, Yin Y, Zhang Y. Challenges in the treatment of pediatric Mycoplasma pneumoniae pneumonia. Eur J Pediatr. 2024;183(7):3001-11. Liu N, Wang Y, Bai TM, Ma FF, Han TT, Zhu HL, et al. Epidemiological characteristics of Mycoplasma pneumoniae in hospitalized children before during and after the COVID-19 pandemic in xi'an China. Sci Rep. 2026;16(1):7577. Liu Y, Li P. Correlation between 25-hydroxyvitamin D, N-terminal pro-brain natriuretic peptide, inflammatory factors and myocardial injury in children with mycoplasma pneumoniae pneumonia. BMC Pediatr. 2025;25(1):914. Yang M, Xuan A, Zhu G. Diagnostic Efficacy of Combined N-Terminal Pro-Brain Natriuretic Peptide (NT-proBNP) and Cardiac Troponin I (cTnI) Testing in Myocardial Injury of Children with Mycoplasma Pneumoniae Pneumonia. J Multidiscip Healthc. 2025;18:3709-16. Yu Y, Ji R, Xia Y, Liu F. Multicenter Analysis of Clinical Characteristics and Risk Factors for Liver Injury in Severe Mycoplasma pneumoniae Pneumonia. Pediatr Infect Dis J. 2026;45(3):236-43. Chen J, Hou D, Song Y. Development and multi-database validation of interpretable machine learning models for predicting In-Hospital mortality in pneumonia patients: A comprehensive analysis across four healthcare systems. Respir Res. 2025;26(1):279. Hu Y, Zhang X, Slavin V, Belsti Y, Tiruneh SA, Callander E, et al. Beyond Comparing Machine Learning and Logistic Regression in Clinical Prediction Modelling: Shifting from Model Debate to Data Quality. J Med Internet Res. 2025;27:e77721. Wang J, Wu H, Cai H, Wang Y, Gu J. Machine Learning and Shapley Additive Explanations Value Integration for Predicting the Prognostic of Anti-N-Methyl-D-Aspartate Receptor Encephalitis: Model Development and Evaluation Study. JMIR Med Inform. 2025;13:e75020. He B, Li X, Dong R, Yao H, Zhou Q, Xu C, et al. Development of machine learning-based differential diagnosis model and risk prediction model of organ damage for severe Mycoplasma pneumoniae pneumonia in children. Sci Rep. 2025;15(1):9431. Watson GL, Staples G, Carver R, Bhargava A, López-Espina C, Schmalz L, et al. Interpretability of an FDA-authorized AI/ML sepsis diagnostic tool improved by SHAP values. JAMIA Open. 2026;9(1):ooag020. Hur S, Lee Y, Park J, Jeon YJ, Cho JH, Cho D, et al. Comparison of SHAP and clinician friendly explanations reveals effects on clinical decision behaviour. NPJ Digit Med. 2025;8(1):578. Meng Q, Li N, Yuan L, Gao X. Analysis of common causes of liver damage among children 12 years and younger in Weifang. J Int Med Res. 2021;49(4):3000605211006661. Pal M, Saha HN, Chakrabarti A. The Trust-Aware XAI (TAXAI) framework: a quantitative model for interpretable and reliable clinical AI systems. Sci Rep. 2026. Zhang X, Sun R, Jia W, Li P, Song C. A new dynamic nomogram for predicting the risk of severe Mycoplasma pneumoniae pneumonia in children. Sci Rep. 2024;14(1):8260. Liu F, Chen L, Wang MY, Shi WJ, Wang XP. Exploring high-risk factors for the prediction of severe mycoplasma pneumonia in children. Transl Pediatr. 2024;13(11):2003-11. Ye Y, Gao Z, Zhang Z, Chen J, Chu C, Zhou W. A machine learning model for predicting severe mycoplasma pneumoniae pneumonia in school-aged children. BMC Infect Dis. 2025;25(1):570. Qian Y, Tao Y, Wu L, Zhou C, Liu F, Xu S, et al. Model based on the automated AI-driven CT quantification is effective for the diagnosis of refractory Mycoplasma pneumoniae pneumonia. Sci Rep. 2024;14(1):16172. Lu W, Wu X, Xu Y, Wang T, Xiao A, Guo X, et al. Predictive value of bronchoscopy combined with CT score for refractory mycoplasma pneumoniae pneumonia in children. BMC Pulm Med. 2024;24(1):251. Wang LP, Hu ZH, Jiang JS, Jin J. Serum inflammatory markers in children with Mycoplasma pneumoniae pneumonia and their predictive value for mycoplasma severity. World J Clin Cases. 2024;12(22):4940-6. Chen L, Yin J, Liu X, Liu J, Xu B, Shen K. Thromboembolic complications of Mycoplasma pneumoniae pneumonia in children. Clin Respir J. 2023;17(3):187-96. Xu X, Yang T, An J, Li B, Dou Z. Liver injury in sepsis: manifestations, mechanisms and emerging therapeutic strategies. Front Immunol. 2025;16:1575554. Chen JW, Liu CY, Li S, Wu SW, Cai C, Lu MQ. Sepsis-associated liver injury: Mechanisms and potential therapeutic targets. World J Gastroenterol. 2024;30(42):4518-22. Yu Y, Su Y, Jin X, Zhang X, Yang Y, Shen Y. Serum expression of ESM-1 and Syndecan-1 and its relationship with disease severity in children with Mycoplasma pneumoniae pneumonia. Ital J Pediatr. 2025;51(1):247. Van de Moortel L, Speeckaert MM, Fiers T, Oeyen S, Decruyenaere J, Delanghe J. Low serum creatine kinase activity is associated with worse outcome in critically ill patients. J Crit Care. 2014;29(5):786-90. Kc O, Dahal PH, Koirala M, NtemMensah AD. Rhabdomyolysis and Neurological Manifestation With Progressive Weakness in a Young Adult: A Rare Extrapulmonary Presentation of Mycoplasma Pneumoniae. Cureus. 2021;13(12):e20552. Zhang X, Sun R, Jia W, Li P, Song C. Clinical Characteristics of Lung Consolidation with Mycoplasma pneumoniae Pneumonia and Risk Factors for Mycoplasma pneumoniae Necrotizing Pneumonia in Children. Infect Dis Ther. 2024;13(2):329-43. Fang W, Huang J, Wang J, Huang T, Lin D, Yin J. Blockade of interleukin-6 receptor attenuates apoptosis and modulates the inflammatory response in Mycoplasma pneumoniae infected A549 cells. Am J Transl Res. 2022;14(9):6187-95. Zhang M, Liu D, Song Y, Zhao J. Correlation between Interleukin-6, C-reactive protein, and lactate dehydrogenase levels and macrolide-resistant severe mycoplasma pneumoniae infection in children. BMC Infect Dis. 2025;26(1):62. Ni T, Zhao F. Predicting myocardial damage in children with mycoplasma pneumoniae pneumonia: a retrospective case-control study. BMC Infect Dis. 2025;25(1):733. Bongiovanni M. From Respiratory Pathogen to Systemic Threat: Rethinking Mycoplasma pneumoniae Infections. Microorganisms. 2026;14(2):419. Tables Table 1. Baseline characteristics of the study population. Variables Total (n = 1321) Non-Liver Injury (n = 1266) Liver Injury (n = 55) P Hospitalization Duration, day 5.8 ± 2.4 5.7 ± 2.3 7.5 ± 3.8 < 0.001 Hospitalization Cost, yuan 10017.6 ± 4029.0 9822.9 ± 3680.7 14498.2 ± 7622.6 < 0.001 Sex, n (%) 0.338 Male 637 (48.2) 607 (47.9) 30 (54.5) Female 684 (51.8) 659 (52.1) 25 (45.5) Age,year 7.0 ± 2.7 7.0 ± 2.7 7.8 ± 2.6 0.034 Cough, n (%) 1.000 No 16 ( 1.2) 16 (1.3) 0 (0) Yes 1305 (98.8) 1250 (98.7) 55 (100) Fever, n (%) 1.000 No 25 ( 1.9) 24 (1.9) 1 (1.8) Yes 1296 (98.1) 1242 (98.1) 54 (98.2) Pre-admission fever duration, day 8.2 ± 5.1 8.2 ± 5.2 8.6 ± 4.3 0.564 Pleural reaction/Pleurisy, n (%) 1.000 No 1319 (99.8) 1264 (99.8) 55 (100) Yes 2 ( 0.2) 2 (0.2) 0 (0) Pleural effusion, n (%) < 0.001 No 1242 (94.0) 1200 (94.8) 42 (76.4) Yes 79 ( 6.0) 66 (5.2) 13 (23.6) Pulmonary consolidation, n (%) 0.029 No 1158 (87.7) 1115 (88.1) 43 (78.2) Yes 163 (12.3) 151 (11.9) 12 (21.8) Pulmonary atelectasis, n (%) 0.022 No 1207 (91.4) 1162 (91.8) 45 (81.8) Yes 114 ( 8.6) 104 (8.2) 10 (18.2) CRP, mg/L 4.0 (0.7, 11.2) 4.0 (0.7, 11.0) 4.7 (0.7, 14.7) 0.583 WBC, 10 9 /L 11.0 ± 4.7 10.9 ± 4.7 12.8 ± 5.2 0.004 LYMPH, 10 9 /L 2.8 ± 1.6 2.8 ± 1.6 2.8 ± 1.8 0.996 NEUT, 10 9 /L 7.3 ± 4.0 7.2 ± 4.0 8.8 ± 4.3 0.005 HGB, g/L 126.3 ± 10.3 126.1 ± 10.1 129.5 ± 13.6 0.016 PLT, 10 9 /L 363.1 ± 122.7 363.0 ± 122.9 365.4 ± 119.8 0.884 ALT, U/L 15.0 (11.0, 23.0) 15.0 (11.0, 22.0) 131.0 (101.0, 197.5) < 0.001 AST, U/L 25.0 (20.0, 31.0) 24.0 (20.0, 30.0) 50.0 (38.5, 92.5) < 0.001 LDH, U/L 365.4 ± 160.5 356.3 ± 136.5 574.4 ± 383.7 < 0.001 CK, U/L 57.0 (37.0, 90.0) 57.5 (37.0, 90.0) 40.0 (23.0, 76.0) 0.002 CKMB, U/L 21.0 (16.0, 26.0) 20.0 (16.0, 26.0) 24.0 (17.5, 34.5) 0.006 D-dimer, ng/mL 236.0 (155.0, 414.0) 231.0 (154.0, 382.8) 718.0 (391.0, 1270.0) < 0.001 IL-6, pg/mL 3.5 (2.2, 7.7) 3.5 (2.2, 7.3) 4.3 (2.2, 21.5) 0.076 Oxygen, n (%) < 0.001 No 1205 (91.2) 1163 (91.9) 42 (76.4) Yes 116 ( 8.8) 103 (8.1) 13 (23.6) RMPP, n (%) 0.013 No 302 (22.9) 297 (23.5) 5 (9.1) Yes 1019 (77.1) 969 (76.5) 50 (90.9) PE, n (%) < 0.001 No 1283 (97.1) 1235 (97.6) 48 (87.3) Yes 38 ( 2.9) 31 (2.4) 7 (12.7) Note: Continuous variables are presented as mean ± SD (normally distributed) or median (IQR) (skewed); categorical variables as n (%). Liver injury was defined as ALT >80 U/L. Group comparisons were performed using Student's t-test, Mann-Whitney U test, or χ²/Fisher's exact test, as appropriate. PE: Pulmonary Embolism. Table 2. Comparison of the predictive ability of several models in the training set. Model AUC ACC Brier Fβ CE PR-AUC Sensitivity Specificity NPV PPV catboost 1.000 0.993 0.009 0.925 0.007 1.000 1.000 0.993 1.000 0.860 gbm 0.882 0.827 0.118 0.282 0.173 0.312 0.811 0.828 0.990 0.170 kknn 1.000 0.941 0.037 0.587 0.059 1.000 1.000 0.939 1.000 0.416 lightgbm 1.000 1.000 0.000 1.000 0.000 1.000 1.000 1.000 1.000 1.000 Naive bayes 0.804 0.771 0.168 0.204 0.229 0.221 0.703 0.774 0.984 0.119 nnet 0.914 0.816 0.116 0.294 0.184 0.343 0.919 0.811 0.996 0.175 ranger 1.000 0.998 0.013 0.974 0.002 1.000 1.000 0.998 1.000 0.949 rpart 0.872 0.852 0.119 0.328 0.148 0.234 0.865 0.851 0.993 0.203 svm 0.980 0.923 0.053 0.514 0.077 0.545 0.973 0.921 0.999 0.350 xgboost 0.937 0.896 0.166 0.425 0.104 0.474 0.919 0.895 0.996 0.276 Table 3. Comparison of the predictive power of several models in the test set. Model AUC ACC Brier Fβ CE PR-AUC Sensitivity Specificity NPV PPV catboost 0.840 0.911 0.060 0.235 0.089 0.342 0.333 0.935 0.970 0.182 gbm 0.895 0.856 0.104 0.308 0.144 0.292 0.778 0.859 0.989 0.192 kknn 0.785 0.869 0.106 0.260 0.131 0.155 0.556 0.883 0.979 0.169 lightgbm 0.869 0.938 0.049 0.341 0.062 0.350 0.389 0.962 0.973 0.304 Naive bayes 0.911 0.805 0.136 0.261 0.195 0.273 0.833 0.804 0.991 0.155 nnet 0.829 0.814 0.122 0.243 0.186 0.162 0.722 0.818 0.986 0.146 ranger 0.878 0.929 0.053 0.279 0.071 0.267 0.333 0.955 0.971 0.240 rpart 0.720 0.842 0.128 0.242 0.158 0.165 0.611 0.852 0.981 0.151 svm 0.835 0.897 0.074 0.286 0.103 0.223 0.500 0.914 0.977 0.200 xgboost 0.776 0.849 0.175 0.214 0.151 0.185 0.500 0.864 0.976 0.136 Table 4. Comparison of the predictive power of several models in the external validation set. Model AUC ACC Brier Fβ CE PR-AUC Sensitivity Specificity NPV PPV catboost 0.664 0.955 0.034 0.121 0.045 0.150 0.133 0.974 0.979 0.111 gbm 0.755 0.919 0.070 0.071 0.081 0.180 0.133 0.938 0.978 0.049 kknn 0.771 0.931 0.054 0.083 0.069 0.065 0.133 0.950 0.979 0.061 lightgbm 0.635 0.953 0.038 0.118 0.047 0.166 0.133 0.973 0.979 0.105 Naive bayes 0.728 0.909 0.068 0.171 0.091 0.212 0.400 0.922 0.985 0.109 nnet 0.755 0.902 0.054 0.160 0.098 0.168 0.400 0.914 0.984 0.100 ranger 0.726 0.970 0.038 0.174 0.030 0.180 0.133 0.990 0.979 0.250 rpart 0.492 0.870 0.096 0.088 0.130 0.030 0.267 0.885 0.980 0.053 svm 0.751 0.958 0.037 0.069 0.042 0.068 0.067 0.979 0.978 0.071 xgboost 0.584 0.864 0.172 0.065 0.136 0.073 0.200 0.880 0.979 0.038 Table 5. Performance metrics of the selected nnet model in the training, test, and external validation sets. AUC ACC Brier Fβ PR- AUC Precision Recall Sensitivity Specificity CE Log loss NPV PPV Train.cv.folds 0.734 0.78 0.132 0.161 0.126 0.100 0.564 0.564 0.792 0.220 0.496 0.976 0.100 Train 0.914 0.816 0.116 0.294 0.343 0.175 0.919 0.919 0.811 0.184 0.336 0.996 0.175 Test 0.829 0.814 0.122 0.243 0.162 0.146 0.722 0.722 0.818 0.186 0.395 0.986 0.146 External.validaion 0.755 0.902 0.054 0.160 0.168 0.100 0.400 0.400 0.914 0.098 0.178 0.984 0.100 Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterials.doc Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 06 May, 2026 Reviewers invited by journal 24 Apr, 2026 Editor invited by journal 23 Apr, 2026 Editor assigned by journal 20 Apr, 2026 Submission checks completed at journal 20 Apr, 2026 First submitted to journal 19 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9460650","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":633896804,"identity":"d26cb868-8ec8-4bc6-b17b-bd052822409e","order_by":0,"name":"Yang Yu","email":"","orcid":"","institution":"Nanjing Lishui People’s Hospital, Zhongda Hospital Lishui Branch, Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Yu","suffix":""},{"id":633896805,"identity":"61aa1d65-8e8a-47ef-8cf6-74e9ddfa5370","order_by":1,"name":"Yang Jing","email":"","orcid":"","institution":"Nanjing Lishui People’s Hospital, Zhongda Hospital Lishui Branch, Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Jing","suffix":""},{"id":633896806,"identity":"30f6c64c-66ee-41f5-8734-0e9027c96258","order_by":2,"name":"Yinyan Tang","email":"","orcid":"","institution":"Nanjing Lishui People’s Hospital, Zhongda Hospital Lishui Branch, Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Yinyan","middleName":"","lastName":"Tang","suffix":""},{"id":633896807,"identity":"87a4bf09-e3cb-4cf3-9ebd-5aa7b90e7497","order_by":3,"name":"Feng Liu","email":"","orcid":"","institution":"Children's Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Liu","suffix":""},{"id":633896808,"identity":"204e6f92-25f9-4c9d-a40c-a36f271133fe","order_by":4,"name":"Hongjuan Wei","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYDCCA1BagoGHweCDgY0dMVoYG2BaDGcUpCWTpoWZ58MhCA8f4DueY/7wZ9thOckZuQeKbQwOMDOwHz66AZ8WyTNvDBsk2w4bS0vkJRjnGNzhY+BJS7uBT4vBjRzDBsO2w4nzpHMMgFqeMTNI8JgR1pLYdrgerMXC4DBjA1FaDrYdTpAGaWEgRovkmWeFMxvOpRvOnP/GwLDHIC2ZjZBf+I4nb/j4o8xaXuLMGTODH39s7PjZDx/Dq4WBIcOAgZGtGcRiMwCT+JWDQPoDBoY/dSAW8wPCqkfBKBgFo2AkAgDDmVD8IwxmzAAAAABJRU5ErkJggg==","orcid":"","institution":"Nanjing Lishui People’s Hospital, Zhongda Hospital Lishui Branch, Southeast University","correspondingAuthor":true,"prefix":"","firstName":"Hongjuan","middleName":"","lastName":"Wei","suffix":""}],"badges":[],"createdAt":"2026-04-19 08:54:52","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9460650/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9460650/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108977819,"identity":"7d6792f9-650a-46b6-bbc5-c5271707a8b1","added_by":"auto","created_at":"2026-05-11 11:33:03","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":140560,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of this study.\u003c/p\u003e","description":"","filename":"Binder21.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/f711bafabe84fd2a617c8b4d.png"},{"id":108939173,"identity":"7dfc1f1b-71a7-4c3c-af49-9d2940c46659","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":246278,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation between variables.\u003c/p\u003e","description":"","filename":"Binder22.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/0a4353388832d58fc2d5d0ec.png"},{"id":108978228,"identity":"e1a18e2c-11ba-4cff-8918-71feaaebd9c9","added_by":"auto","created_at":"2026-05-11 11:35:14","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":65908,"visible":true,"origin":"","legend":"\u003cp\u003eFeature selection based on the boruta algorithm.\u003c/p\u003e\n\u003cp\u003eNote: The horizontal axis is the name of each variable, and the vertical axis is the Z value of each variable. The box plot shows the Z value of each variable during model calculation. The green boxes represent important variables, and the red boxes represent unimportant variables.\u003c/p\u003e","description":"","filename":"Binder23.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/3e8b999675e4711c3a9b3530.png"},{"id":108977749,"identity":"24064227-dab5-4b3f-8d1f-a533e93cd4db","added_by":"auto","created_at":"2026-05-11 11:32:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":154822,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for the ten models in the training set (A) , test set (B) and external validation set(C).\u003c/p\u003e","description":"","filename":"Binder24.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/2b5d7b4baad96e8b4f196665.png"},{"id":108977903,"identity":"9bfb854c-69c8-4849-9951-219619b13ef3","added_by":"auto","created_at":"2026-05-11 11:33:24","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":679525,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion matrices and performance metrics of ten models on the test set and external validation set.\u003c/p\u003e","description":"","filename":"Binder25.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/729c49dc06c598a6b760d992.png"},{"id":108939176,"identity":"7c8958e6-bcc2-47e7-a631-a56a1b7732d7","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":432825,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion matrix and radar plot for the ten models in the training set, test set and external validation set.\u003c/p\u003e","description":"","filename":"Binder26.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/a28a620d8b610ef7cef8a748.png"},{"id":108939179,"identity":"a399ba72-a052-45d1-bcc4-1ac206c6c066","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":124180,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration curve for the ten models in the training set (A) , test set (B) and external validation set(C).\u003c/p\u003e","description":"","filename":"Binder27.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/ae2cad8b614c04bdb0d3881f.png"},{"id":108939177,"identity":"95304c60-c2b1-4c77-801b-6c8bc0afda16","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":140055,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curves for the ten models in the training set (A) , test set (B) and external validation set(C).\u003c/p\u003e","description":"","filename":"Binder28.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/a4ea2b079fc3e1046174f6f3.png"},{"id":108977690,"identity":"38232e90-0c70-4123-a444-adac9c29c1bb","added_by":"auto","created_at":"2026-05-11 11:32:33","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":5713,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis for the nnet model in predicting liver injury in MPP patients.\u003c/p\u003e\n\u003cp\u003eNote: (A) Beeswarm plot of feature contributions. (B) Variable importance ranked by mean SHAP value.\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/a942dd4e668875b0f9f37c64.png"},{"id":108939175,"identity":"4bee0f01-e135-400c-9923-b1a8de895531","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":5713,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP dependence plots for the six selected features in predicting liver injury in MPP patients.\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/db83d4c3e7ee966864ce4cdc.png"},{"id":108939181,"identity":"7981dd0c-876a-4fa9-b0af-18b4ef6ac3ca","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":79113,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP force plots for representative individual cases.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNote: \u003cstrong\u003eCase \u003c/strong\u003e(A \u003cstrong\u003eand \u003c/strong\u003eC)\u003cstrong\u003e was predicted as positive, while case (B and D) was predicted as negative.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure11.SHAPforceplotsforrepresentativeindividualcases.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/00833ebd2cab005213e233e1.png"},{"id":108978012,"identity":"4e35acc7-787b-4f25-a632-4de047ce241e","added_by":"auto","created_at":"2026-05-11 11:33:42","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":101830,"visible":true,"origin":"","legend":"\u003cp\u003eA web-based calculator for predicting liver injury of in patients with MPP.\u003c/p\u003e","description":"","filename":"Figure12.AwebbasedcalculatorforpredictingliverinjuryofinpatientswithMPP.png","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/101e85d38ba04b1ab0920bdf.png"},{"id":109203828,"identity":"2b28ffd5-1d32-41a3-ac37-d863e3152df5","added_by":"auto","created_at":"2026-05-13 14:47:33","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2682390,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/72e2ff27-40ad-494e-a685-bf93f573e7ba.pdf"},{"id":108939171,"identity":"5628c236-bd77-4362-b6e5-e69a954e57c2","added_by":"auto","created_at":"2026-05-11 05:05:59","extension":"doc","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":38400,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.doc","url":"https://assets-eu.researchsquare.com/files/rs-9460650/v1/89b5540548e20a66e880ac17.doc"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eDevelopment and External Validation of a Machine Learning Model for Predicting Liver Injury in Children With Mycoplasma Pneumoniae Pneumonia: A Multicenter Retrospective Study \u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eMPP is a leading cause of community-acquired pneumonia in children, accounting for 10–40% of pediatric cases worldwide[1, 2], characterized by rapid progression and a high propensity for extrapulmonary complications[3, 4]. Among these, liver injury occurs in 10-30% of hospitalized children with MPP and is associated with prolonged hospitalization, increased healthcare costs, and potential adverse outcomes[5]. Early identification of children at risk for liver injury is critical for implementing preventive strategies such as avoiding hepatotoxic medications and intensifying monitoring. While serum ALT elevation provides the definitive diagnosis of liver injury, it is inherently a lagging indicator. In clinical practice, ALT levels measured at hospital admission may be normal or only mildly elevated, with significant elevations often occurring later during hospitalization. This diagnostic delay underscores a critical gap: the absence of a reliable tool for early identification of children with MPP who are at risk for subsequent liver injury. Existing predictive models, predominantly based on traditional logistic regression, are often limited by small sample sizes and lack external validation [6]. More fundamentally, they typically utilize parameters measured at or after the diagnosis of liver injury, rather than focusing on baseline admission data that could enable true preemptive risk stratification. Consequently, there is an urgent need for a robust, early-warning model using readily available admission data to facilitate timely monitoring and preventive intervention. Machine learning offers distinct advantages for clinical prediction by automatically identifying complex patterns, handling nonlinear relationships, and capturing variable interactions that conventional methods may miss[7, 8]. These techniques have been successfully applied in pediatric settings, including prediction of refractory MPP and myocardial injury[9]. Additionally, explainable artificial intelligence methods such as SHAP have enhanced model interpretability, addressing the \"black box\" concern and facilitating clinical adoption[10, 11]. To address these gaps, we conducted a multicenter retrospective study to develop and validate a machine learning-based model for predicting liver injury in children with MPP. We evaluated ten algorithms, selected the optimal model based on discriminative ability and generalizability, and developed an interactive web-based application to facilitate clinical implementation.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003e2.1 Dataset source\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective cohort study extracted data from electronic medical records (EMR) of two medical centers in Nanjing, China. The development cohort was obtained from Nanjing Children\u0026apos;s Hospital, a tertiary pediatric center. Data were collected from patients admitted to respiratory wards between January 1, 2023, and December 31, 2023. The external validation cohort was collected from Nanjing Lishui People\u0026apos;s Hospital during the same period. The study protocol was approved by the Institutional Review Board (IRB) of Children\u0026rsquo;s Hospital of Nanjing Medical University [Approval No. 201812257-1] and Nanjing Lishui People\u0026apos;s Hospital [Approval No. 2024KY0802-02], with waiver of informed consent due to retrospective design and de-identified data. All procedures were performed in accordance with the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.2 Participants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 1,323 patients diagnosed with MPP were initially included. After excluding 2 patients with discharge against medical advice, 1,321 patients were enrolled and randomly divided into a training set (67%, n = 885) and a testing set (33%, n = 436). An independent external validation cohort consisting of 640 patients from Nanjing Lishui People\u0026apos;s Hospital was also included (Figure 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.3 Variable selection\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBased on previous literature and clinical relevance, the following three categories of variables were included: (a) Demographic and clinical characteristics: sex, age (years), cough (yes/no), fever (yes/no), fever duration before admission (days), and pulmonary consolidation on imaging (yes/no); (b) Laboratory parameters: CRP, WBC, LYMPH, NEUT, HGB, PLT, LDH, CK, CK-MB, D-dimer, IL-6. All laboratory variables were collected within 24 hours of admission.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.4\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eOutcome definition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLiver injury was defined as ALT \u0026gt;80 U/L based on pediatric reference ranges\u0026nbsp;[5, 12].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.5 Statistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll statistical analyses were performed using R Statistical Software (Version 4.2.2, http://www.R-project.org) and the Free Statistics analysis platform (Version 2.4, Beijing, China). A two-sided \u003cem\u003eP\u003c/em\u003e value \u0026lt; 0.05 was considered statistically significant. Continuous variables were assessed for normality using the Shapiro\u0026ndash;Wilk test. Normally distributed variables were presented as mean \u0026plusmn; standard deviation (SD) and compared using Student\u0026rsquo;s t-test; non-normally distributed variables were presented as median (interquartile range) and compared using the Mann\u0026ndash;Whitney U test. Categorical variables were expressed as numbers (percentages) and compared using the \u0026chi;\u0026sup2; test or Fisher\u0026rsquo;s exact test, as appropriate. Prior to feature selection, pairwise correlations among all candidate predictor variables were assessed. Feature selection was performed using the boruta algorithm via the boruta package in R. Variables with importance significantly higher than the maximum importance of shadow features were retained for model development.\u0026nbsp;\u003cem\u003eMultivariable logistic regression with mutual adjustment for the six selected variables was performed to assess their independent associations with liver injury.\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003eTen supervised machine learning algorithms (catboost,\u0026nbsp;gbm,\u0026nbsp;kknn,\u0026nbsp;lightgbm, naive bayes,\u0026nbsp;nnet,\u0026nbsp;ranger,\u0026nbsp;rpart,\u0026nbsp;svm, and\u0026nbsp;xgboost) were developed to predict liver injury. All models were trained on the training set (n = 885), with hyperparameter tuning performed using 10-fold cross-validation. Model discrimination was assessed using AUC,\u0026nbsp;ACC,\u0026nbsp;Brier score, F\u0026beta; score, calibration error (CE), precision-recall AUC (PR-AUC), sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). Receiver operating characteristic (ROC) curves and calibration curves were plotted to visualize model performance. DCA was performed to evaluate clinical utility across a range of threshold probabilities.\u0026nbsp;To enhance model interpretability, SHapley Additive exPlanations (SHAP) was applied to the final\u0026nbsp;nnet model using the shapviz and fastshap packages. Variable importance was assessed by mean absolute SHAP values, and beeswarm, dependence, and individual force plots were generated to visualize feature effects.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eModel\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ed\u003c/strong\u003e\u003cstrong\u003eevelopment,\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ee\u003c/strong\u003e\u003cstrong\u003evaluation, and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ev\u003c/strong\u003e\u003cstrong\u003eisualization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003cstrong\u003e.1 Data cleaning\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter applying the inclusion and exclusion criteria detailed in Section 2.2, data completeness was assessed for all variables in the remaining cohort. All retained patients (n = 1,321) had complete data for the outcome variable (ALT \u0026gt;80 U/L) and all candidate predictor variables; therefore, no imputation was performed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003cstrong\u003e.2 Feature selection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCorrelation analysis among all candidate variables is presented in Figure 2. Subsequently, feature selection was performed using the boruta algorithm, which iteratively compares the importance of original variables with randomized shadow features. Variables with importance significantly higher than the maximum importance of shadow features were classified as \u0026quot;confirmed\u0026quot; (depicted in green in Figure 3) and retained for model development. After 100 iterations, the algorithm confirmed six variables as important predictors: LDH, D-dimer, CK, age, IL-6, and CK-MB. These selected variables were subsequently used for model training and validation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003cstrong\u003e.3 Modeling and evaluation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTen supervised machine learning algorithms were employed for model development: catboost, gbm, kknn, lightgbm, naive bayes, nnet, ranger, rpart, svm, and xgboost. All models were trained on the training set (n=885). Hyperparameter tuning was performed via 10-fold cross-validation on the training set. Model discrimination was assessed using AUC, ACC, Brier score, F\u0026beta; score, CE, PR-AUC, sensitivity, specificity, NPV, and PPV. ROC curves and calibration curves were plotted to visualize model discrimination and calibration, respectively. DCA was performed to evaluate the clinical utility of the models across a range of threshold probabilities. The performance of all ten models in the training set, internal test set, and external validation set was comprehensively evaluated. Based on comprehensive evaluation of discriminative ability and generalizability across three datasets, the nnet model was selected as the optimal model. The nnet model demonstrated robust performance with good discrimination and acceptable external validity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003cstrong\u003e.4 Evaluation of the importance of variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSHAP is a post-hoc explanation framework for machine learning models based on game theory[13]. It quantifies the importance of each feature in the model by calculating the contribution value, known as the Shapley value, for each feature towards the predicted outcome. This study utilized the SHAP method to enhance the interpretability and transparency of the final nnet model. SHAP analysis was performed on the training set, and variable importance was assessed by mean absolute SHAP values. Beeswarm plots were generated to visualize the distribution and direction of SHAP values for each feature, illustrating the relationship between feature values and their impact on model predictions. Dependence plots were constructed to further explore the marginal effects of key predictors. Individual force plots were also generated to demonstrate the decision-making process for representative cases, facilitating local interpretability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e\u003cstrong\u003e5\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Model visualization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo facilitate clinical application, an interactive web-based application was developed using the shiny package in R, allowing clinicians to input the six selected predictors and obtain real-time predicted probabilities of liver injury. The application is publicly accessible at http://127.0.0.1:7728/.\u003c/p\u003e"},{"header":"3. Results","content":"\u003cp\u003e\u003cstrong\u003e3.1 Baseline characteristics of the study population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 1,321 children diagnosed with MPP were included, of whom 55 (4.2%) developed liver injury (defined as ALT \u0026gt; 80 U/L). Baseline characteristics are presented in Table 1.\u003c/p\u003e\n\u003cp\u003ePatients with liver injury had significantly longer hospital stays (7.5 \u0026plusmn; 3.8 vs. 5.7 \u0026plusmn; 2.3 days, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) and higher costs (14,498.2 \u0026plusmn; 7,622.6 vs. 9,822.9 \u0026plusmn; 3,680.7 yuan, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001). They also exhibited higher rates of pulmonary consolidation (21.8% vs. 11.9%, \u003cem\u003eP\u003c/em\u003e = 0.029), atelectasis (18.2% vs. 8.2%, \u003cem\u003eP\u003c/em\u003e = 0.022), pleural effusion (23.6% vs. 5.2%, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001), and pulmonary embolism (12.7% vs. 2.4%, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001), and were more likely to receive oxygen therapy (23.6% vs. 8.1%, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) and be classified as RMPP (Refractory Mycoplasma Pneumoniae Pneumonia)\u0026nbsp;(90.9% vs. 76.5%, \u003cem\u003eP\u003c/em\u003e = 0.013).\u003c/p\u003e\n\u003cp\u003eLaboratory findings differed markedly between groups. The liver injury group had elevated WBC (12.8 \u0026plusmn; 5.2 vs. 10.9 \u0026plusmn; 4.7 \u0026times;10⁹/L, \u003cem\u003eP\u003c/em\u003e = 0.004) and NEUT (8.8 \u0026plusmn; 4.3 vs. 7.2 \u0026plusmn; 4.0 \u0026times;10⁹/L, \u003cem\u003eP\u003c/em\u003e = 0.005) counts, higher HGB (129.5 \u0026plusmn; 13.6 vs. 126.1 \u0026plusmn; 10.1 g/L, \u003cem\u003eP\u003c/em\u003e = 0.016), ALT [131.0 (101.0, 197.5) vs. 15.0 (11.0, 22.0) U/L, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001], AST [50.0 (38.5, 92.5) vs. 24.0 (20.0, 30.0) U/L, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001], and LDH (574.4 \u0026plusmn; 383.7 vs. 356.3 \u0026plusmn; 136.5 U/L, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001), and substantially higher D-dimer [718.0 (391.0, 1270.0) vs. 231.0 (154.0, 382.8) ng/mL, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001], but lower CK [40.0 (23.0, 76.0) vs. 57.5 (37.0, 90.0) U/L, \u003cem\u003eP\u003c/em\u003e = 0.002]. No significant differences were observed in sex, cough, fever, pre-admission fever duration, CRP, LYMPH, or PLT (all \u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.2 Feature\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003cstrong\u003eelection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo mitigate overfitting and enhance model parsimony, feature selection was performed using the boruta algorithm. A total of 17 candidate variables were assessed, encompassing demographic characteristics (age, sex), clinical features (cough, fever, pre-admission fever duration, pulmonary consolidation), and laboratory parameters (CRP, WBC, LYMPH, NEUT, HGB, PLT, LDH, CK, CK-MB, D-dimer, IL-6). Of these, six were confirmed as significant predictors (green-zone features in Figure 3): LDH, D-dimer, CK, age, IL-6, and CK-MB. These variables demonstrated importance scores significantly higher than their shadow counterparts, indicating robust predictive value for liver injury in children with MPP.\u003c/p\u003e\n\u003cp\u003eThe remaining variables, such as CRP, WBC, PLT, and cough, were classified as unimportant (red-zone features) and excluded. This strategy ensured retention of the most relevant predictors, enhancing model generalizability and interpretability.\u003c/p\u003e\n\u003cp\u003eAs a complementary analysis, multivariable logistic regression identified LDH (adj \u003cem\u003eOR\u003c/em\u003e = 1.08, 95% \u003cem\u003eCI\u003c/em\u003e: 1.05\u0026ndash;1.10, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) and age (adj \u003cem\u003eOR\u003c/em\u003e = 1.15, 95% \u003cem\u003eCI\u003c/em\u003e: 1.03\u0026ndash;1.29, \u003cem\u003eP\u003c/em\u003e = 0.016) as independent risk factors (Supplementary Table 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3 Evaluation of\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003em\u003c/strong\u003e\u003cstrong\u003eodel\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ep\u003c/strong\u003e\u003cstrong\u003eerformance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing feature selection, the six predictors were used to develop ten machine learning algorithms. The cohort was randomly divided into training (67%, n = 885) and internal test (33%, n = 436) sets, with an independent external validation cohort (n = 640). Model performance was assessed using AUC, ACC, sensitivity, specificity, PPV, NPV, F\u0026beta; score, Brier score, and calibration error (Tables 2\u0026ndash;4) (Figure 4).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3.1 Training\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003cstrong\u003eet\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ep\u003c/strong\u003e\u003cstrong\u003eerformance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSeveral models demonstrated excellent discrimination in the training cohort. Lightgbm achieved perfect metrics (AUC = 1.000, ACC = 1.000), while catboost, kknn, and ranger showed near-perfect performance (AUC = 1.000, ACC \u0026gt; 0.940). Naive bayes performed relatively poorly (AUC = 0.804). Nnet showed moderate performance (AUC = 0.914) without overfitting (Tables 2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3.2 Internal\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003et\u003c/strong\u003e\u003cstrong\u003eest\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003cstrong\u003eet\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ep\u003c/strong\u003e\u003cstrong\u003eerformance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePerformance attenuated in the test set as expected. Lightgbm maintained the highest ACC (0.938) and specificity (0.962). Naive bayes achieved the highest AUC (0.911) but moderate ACC (0.805). Notably, nnet demonstrated stable performance (AUC = 0.829, ACC = 0.814, sensitivity = 0.722) (Tables 3). Confusion matrices and performance metrics on the test set are shown in \u003cstrong\u003eFigure\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e5\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;(A\u0026ndash;J)\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3.3 External\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ev\u003c/strong\u003e\u003cstrong\u003ealidation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMost models declined substantially in the external cohort (lightgbm AUC = 0.635; catboost = 0.664; ranger = 0.726). Nnet maintained the most robust performance (AUC = 0.755, ACC = 0.902, Brier score = 0.054), indicating superior generalizability and acceptable calibration (Tables 4). Confusion matrices and performance metrics on the external validation set are shown in\u003cstrong\u003eFigure\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e5\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;(K\u0026ndash;T)\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3.4 Clinical\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eu\u003c/strong\u003e\u003cstrong\u003etility and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ef\u003c/strong\u003e\u003cstrong\u003einal\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003em\u003c/strong\u003e\u003cstrong\u003eodel\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003cstrong\u003eelection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eComprehensive performance metrics of the ten models across the three datasets are summarized in Tables 2\u0026ndash;4 and visually compared in Figure 6. Calibration curves confirmed that nnet-predicted probabilities matched observed rates across all datasets (Figure 7). DCA showed that the nnet model provided limited net benefit within a narrow range of threshold probabilities (Figure 8). Considering discrimination, calibration, clinical utility, and generalizability, nnet was selected as the final model (Table 5).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.4 Evaluation of\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ev\u003c/strong\u003e\u003cstrong\u003eariable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ei\u003c/strong\u003e\u003cstrong\u003emportance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo enhance model interpretability, SHAP analysis was performed on the final nnet model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.4.1 Feature\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ei\u003c/strong\u003e\u003cstrong\u003emportance and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ed\u003c/strong\u003e\u003cstrong\u003eirectional\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ee\u003c/strong\u003e\u003cstrong\u003effects\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe SHAP summary plot (Figure 9A) revealed the directional effects of each predictor on model output. Variable importance ranked by mean absolute SHAP values (Figure 9B) identified LDH as the dominant predictor (0.122), followed by D-dimer (0.072), CK (0.067), age (0.059), IL-6 (0.047), and CK-MB (0.047). SHAP dependence plots further elucidated the functional forms of these relationships. LDH exhibited a monotonic positive association (Figure 10A), with SHAP values rising steeply as LDH increased. Age displayed a positive linear association (Figure 10D).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.4.2 Individual\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ei\u003c/strong\u003e\u003cstrong\u003enterpretation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSHAP force plots illustrate how these features drive individual predictions (Figure 11A). A high-risk case (predicted probability 0.742) was driven by elevated LDH (1081 U/L), D-dimer (422 ng/mL), and CK-MB (92 U/L) pushing the prediction upward. In contrast, a low-risk case (predicted probability 0.156) (Figure 11B) was characterized by young age (4.84 years) as the dominant protective factor pulling the prediction downward, overcoming the risk contributions from elevated D-dimer (616 ng/mL) and mildly elevated LDH (417 U/L).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.5 Model\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ev\u003c/strong\u003e\u003cstrong\u003eisualization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo facilitate clinical application of the final nnet model, an interactive web-based predictive tool was developed. As illustrated in Figure 12, the web application allows clinicians to input the six selected predictors (LDH, D-dimer, CK, age, IL-6, and CK-MB) and generates the predicted probability of liver injury in real time. The application is publicly accessible at http://127.0.0.1:7728/.\u003c/p\u003e"},{"header":"4. Discussion","content":"\u003cp\u003e\u003cstrong\u003e4.\u003c/strong\u003e\u003cstrong\u003e1\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eSummary of Principal Findings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, we developed and validated a machine learning-based predictive model for early identification of liver injury in children with MPP. Using the boruta algorithm for feature selection, six key predictors were identified: LDH, D-dimer, CK, age, IL-6, and CK-MB. Among ten machine learning algorithms evaluated, the nnet model demonstrated optimal and stable performance across the training, internal test, and external validation cohorts, with AUC values of 0.914, 0.829, and 0.755, respectively. To facilitate clinical implementation, an interactive web-based application was developed to provide real-time risk predictions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.\u003c/strong\u003e\u003cstrong\u003e2\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eComparison with\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ee\u003c/strong\u003e\u003cstrong\u003existing\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003cstrong\u003etudies\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, we developed and validated a novel machine learning model to predict liver injury in children with MPP, a complication currently lacking early warning tools. Previous studies have primarily focused on identifying risk factors for MPP-associated liver injury using traditional logistic regression approaches. For instance, several studies have reported that elevated inflammatory markers such as WBC, NEUT, and CRP are associated with disease severity and extrapulmonary complications in Mycoplasma pneumoniae infection[14, 15]. However, these studies often relied on univariate or multivariate regression analyses, which may not fully capture the complex nonlinear relationships among clinical and laboratory variables. In contrast, machine learning approaches, particularly ensemble methods and nnet, are better equipped to handle such complexity and have demonstrated superior predictive performance in various clinical contexts[16].\u0026nbsp;The performance of our nnet model (AUC = 0.829 in the test set) is comparable to or better than previously reported predictive models for other complications of MPP. For example, previous studies on predicting MPP\u0026nbsp;have reported AUC values ranging from 0.75 to 0.85[17, 18]. The moderate decrease in performance observed in the external validation set (AUC = 0.755) is expected and reflects the inherent heterogeneity between different clinical centers, highlighting the importance of external validation for assessing model generalizability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.\u003c/strong\u003e\u003cstrong\u003e3\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eInterpretation of\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ek\u003c/strong\u003e\u003cstrong\u003eey\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ep\u003c/strong\u003e\u003cstrong\u003eredictors\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe SHAP analysis identified LDH as the dominant predictor of liver injury (mean SHAP value = 0.122). LDH is a nonspecific marker of cellular damage abundant in hepatocytes. Elevated LDH levels in MPP patients have been associated with disease severity and extrapulmonary complications[19]. Our study further demonstrates its emergence as a critical biomarker specifically for predicting liver injury in this context.\u003c/p\u003e\n\u003cp\u003eD-dimer ranked second (0.072) in importance. As a well-established marker of fibrin degradation, elevated D-dimer reflects coagulation activation and endothelial injury—common pathophysiological features in severe Mycoplasma pneumoniae infection\u0026nbsp;[20]. The strong association with liver injury may reflect systemic microvascular thrombosis and subsequent hepatic ischemia-reperfusion injury, a mechanism implicated in sepsis-associated liver dysfunction[21, 22]. Furthermore, \u003cem\u003eMycoplasma\u0026nbsp;\u003c/em\u003epneumoniae induces procoagulant activity through direct endothelial damage and cytokine-mediated coagulation cascade activation[23].\u003c/p\u003e\n\u003cp\u003eNotably, CK ranked third (0.067) but exhibited a protective inverse relationship (negative SHAP values), consistent with lower baseline CK levels observed in the liver injury group. This paradoxical finding may reflect the complex metabolic state in severe infection: higher CK could indicate preserved muscle mass and nutritional status, whereas low CK may signal systemic catabolism and multi-organ dysfunction[24]. Alternatively, CK elevation might parallel general tissue turnover without specific hepatic toxicity[25].\u003c/p\u003e\n\u003cp\u003eOlder age (0.059) contributed to higher predicted risk, possibly reflecting more robust immune responses and greater inflammatory burden in older children[26].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eElevated IL‑6 levels\u003c/strong\u003e(0.047)\u003cstrong\u003e\u0026nbsp;are strongly associated with severe\u0026nbsp;\u003c/strong\u003e\u003cem\u003eMycoplasma pneumoniae\u003c/em\u003e\u003cstrong\u003e\u0026nbsp;infection and reflect the\u003c/strong\u003e\u003cstrong\u003eintensity of the host inflammatory response, supporting its role as a biomarker of disease severity\u003c/strong\u003e\u003cstrong\u003e[27, 28]\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCK‑MB (0.047), traditionally a cardiac marker, emerged as a risk factor for liver injury with importance comparable to IL‑6. This finding aligns with the observation that M. pneumoniae infection can induce concurrent cardiac and hepatic damage through shared immunopathological mechanisms[29, 30].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.\u003c/strong\u003e\u003cstrong\u003e4\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eClinical\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ei\u003c/strong\u003e\u003cstrong\u003emplications\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe developed model provides a practical tool for early identification of children with MPP who are at elevated risk of liver injury, utilizing only routine admission data. First, all six predictors (LDH, D-dimer, CK, age, IL-6, and CK-MB) are obtainable from admission workup, enabling preemptive risk stratification before ALT elevation occurs. Second, the web-based application facilitates real-time risk assessment for avoidance of hepatotoxic medications and intensified monitoring, potentially reducing prolonged hospitalization. Third, SHAP-based interpretability enhances clinician trust and supports shared decision-making with families, addressing the \"black box\" concern limiting clinical adoption of machine learning models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.\u003c/strong\u003e\u003cstrong\u003e5\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eStrengths and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003el\u003c/strong\u003e\u003cstrong\u003eimitations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study has several strengths. The boruta algorithm ensured rigorous feature selection, reducing overfitting risk. The model was evaluated on both internal and external validation cohorts, demonstrating generalizability across clinical settings. SHAP analysis enhanced interpretability, addressing the “black box” concern. A publicly accessible web-based application was developed to facilitate clinical translation.\u003c/p\u003e\n\u003cp\u003eSeveral limitations should be acknowledged. First, this retrospective study may have selection bias, and although external validation was performed using data from another hospital, the sample size was modest, particularly for the liver injury group (n = 55). Second, the retrospective design limited standardized follow-up protocols; prospective validation is needed to confirm the optimal timing and frequency of risk assessment in clinical practice. Third, although all six predictors are routinely available in tertiary hospitals, some indicators (e.g., IL-6) may not be measured in all primary or secondary care settings, which could limit the model's applicability in resource-limited settings.\u0026nbsp;\u003cstrong\u003e\u003cbr clear=\"all\"\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eWe developed and validated a machine learning-based model using six readily available predictors (LDH, D-dimer, CK, age, IL-6, and CK-MB) to identify liver injury risk in children with MPP. The nnet model demonstrated robust discrimination (AUC = 0.829 in internal test, 0.755 in external validation) and was deployed as an interactive web-based application. This tool may facilitate early risk stratification and timely intervention, potentially reducing the incidence of liver injury in MPP.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHongjuan Wei conceptualized and supervised the study, obtained ethical approval, and critically revised the manuscript.Feng Liu contributed to the study design, data interpretation, and manuscript revision. Yang Yu and Yang Jing contributed equally as co-first authors, performing data analysis, visualization, and initial drafting. Yinyan Tang assisted with data collection and validation. All authors reviewed and approved the final manuscript, had full access to the data, and accept responsibility for its integrity and accuracy. Hongjuan Wei is the guarantor.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study protocol was approved by the Institutional Review Board (IRB) of Children’s Hospital of Nanjing Medical University [Approval No. 201812257-1] and Nanjing Lishui People's Hospital [Approval No. 2024KY0802-02]. All procedures performed were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments. Due to the retrospective nature of the study, the requirement for informed consent was waived by the committee.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;No funding support in the data collection, analysis or preparation of the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eDing G, Zhang X, Vinturache A, van Rossum A, Yin Y, Zhang Y. Challenges in the treatment of pediatric Mycoplasma pneumoniae pneumonia. Eur J Pediatr. 2024;183(7):3001-11.\u003c/li\u003e\n \u003cli\u003eLiu N, Wang Y, Bai TM, Ma FF, Han TT, Zhu HL, et al. Epidemiological characteristics of Mycoplasma pneumoniae in hospitalized children before during and after the COVID-19 pandemic in xi\u0026apos;an China. Sci Rep. 2026;16(1):7577.\u003c/li\u003e\n \u003cli\u003eLiu Y, Li P. Correlation between 25-hydroxyvitamin D, N-terminal pro-brain natriuretic peptide, inflammatory factors and myocardial injury in children with mycoplasma pneumoniae pneumonia. BMC Pediatr. 2025;25(1):914.\u003c/li\u003e\n \u003cli\u003eYang M, Xuan A, Zhu G. Diagnostic Efficacy of Combined N-Terminal Pro-Brain Natriuretic Peptide (NT-proBNP) and Cardiac Troponin I (cTnI) Testing in Myocardial Injury of Children with Mycoplasma Pneumoniae Pneumonia. J Multidiscip Healthc. 2025;18:3709-16.\u003c/li\u003e\n \u003cli\u003eYu Y, Ji R, Xia Y, Liu F. Multicenter Analysis of Clinical Characteristics and Risk Factors for Liver Injury in Severe Mycoplasma pneumoniae Pneumonia. Pediatr Infect Dis J. 2026;45(3):236-43.\u003c/li\u003e\n \u003cli\u003eChen J, Hou D, Song Y. Development and multi-database validation of interpretable machine learning models for predicting In-Hospital mortality in pneumonia patients: A comprehensive analysis across four healthcare systems. Respir Res. 2025;26(1):279.\u003c/li\u003e\n \u003cli\u003eHu Y, Zhang X, Slavin V, Belsti Y, Tiruneh SA, Callander E, et al. Beyond Comparing Machine Learning and Logistic Regression in Clinical Prediction Modelling: Shifting from Model Debate to Data Quality. J Med Internet Res. 2025;27:e77721.\u003c/li\u003e\n \u003cli\u003eWang J, Wu H, Cai H, Wang Y, Gu J. Machine Learning and Shapley Additive Explanations Value Integration for Predicting the Prognostic of Anti-N-Methyl-D-Aspartate Receptor Encephalitis: Model Development and Evaluation Study. JMIR Med Inform. 2025;13:e75020.\u003c/li\u003e\n \u003cli\u003eHe B, Li X, Dong R, Yao H, Zhou Q, Xu C, et al. Development of machine learning-based differential diagnosis model and risk prediction model of organ damage for severe Mycoplasma pneumoniae pneumonia in children. Sci Rep. 2025;15(1):9431.\u003c/li\u003e\n \u003cli\u003eWatson GL, Staples G, Carver R, Bhargava A, L\u0026oacute;pez-Espina C, Schmalz L, et al. Interpretability of an FDA-authorized AI/ML sepsis diagnostic tool improved by SHAP values. JAMIA Open. 2026;9(1):ooag020.\u003c/li\u003e\n \u003cli\u003eHur S, Lee Y, Park J, Jeon YJ, Cho JH, Cho D, et al. Comparison of SHAP and clinician friendly explanations reveals effects on clinical decision behaviour. NPJ Digit Med. 2025;8(1):578.\u003c/li\u003e\n \u003cli\u003eMeng Q, Li N, Yuan L, Gao X. Analysis of common causes of liver damage among children 12 years and younger in Weifang. J Int Med Res. 2021;49(4):3000605211006661.\u003c/li\u003e\n \u003cli\u003ePal M, Saha HN, Chakrabarti A. The Trust-Aware XAI (TAXAI) framework: a quantitative model for interpretable and reliable clinical AI systems. Sci Rep. 2026.\u003c/li\u003e\n \u003cli\u003eZhang X, Sun R, Jia W, Li P, Song C. A new dynamic nomogram for predicting the risk of severe Mycoplasma pneumoniae pneumonia in children. Sci Rep. 2024;14(1):8260.\u003c/li\u003e\n \u003cli\u003eLiu F, Chen L, Wang MY, Shi WJ, Wang XP. Exploring high-risk factors for the prediction of severe mycoplasma pneumonia in children. Transl Pediatr. 2024;13(11):2003-11.\u003c/li\u003e\n \u003cli\u003eYe Y, Gao Z, Zhang Z, Chen J, Chu C, Zhou W. A machine learning model for predicting severe mycoplasma pneumoniae pneumonia in school-aged children. BMC Infect Dis. 2025;25(1):570.\u003c/li\u003e\n \u003cli\u003eQian Y, Tao Y, Wu L, Zhou C, Liu F, Xu S, et al. Model based on the automated AI-driven CT quantification is effective for the diagnosis of refractory Mycoplasma pneumoniae pneumonia. Sci Rep. 2024;14(1):16172.\u003c/li\u003e\n \u003cli\u003eLu W, Wu X, Xu Y, Wang T, Xiao A, Guo X, et al. Predictive value of bronchoscopy combined with CT score for refractory mycoplasma pneumoniae pneumonia in children. BMC Pulm Med. 2024;24(1):251.\u003c/li\u003e\n \u003cli\u003eWang LP, Hu ZH, Jiang JS, Jin J. Serum inflammatory markers in children with Mycoplasma pneumoniae pneumonia and their predictive value for mycoplasma severity. World J Clin Cases. 2024;12(22):4940-6.\u003c/li\u003e\n \u003cli\u003eChen L, Yin J, Liu X, Liu J, Xu B, Shen K. Thromboembolic complications of Mycoplasma pneumoniae pneumonia in children. Clin Respir J. 2023;17(3):187-96.\u003c/li\u003e\n \u003cli\u003eXu X, Yang T, An J, Li B, Dou Z. Liver injury in sepsis: manifestations, mechanisms and emerging therapeutic strategies. Front Immunol. 2025;16:1575554.\u003c/li\u003e\n \u003cli\u003eChen JW, Liu CY, Li S, Wu SW, Cai C, Lu MQ. Sepsis-associated liver injury: Mechanisms and potential therapeutic targets. World J Gastroenterol. 2024;30(42):4518-22.\u003c/li\u003e\n \u003cli\u003eYu Y, Su Y, Jin X, Zhang X, Yang Y, Shen Y. Serum expression of ESM-1 and Syndecan-1 and its relationship with disease severity in children with Mycoplasma pneumoniae pneumonia. Ital J Pediatr. 2025;51(1):247.\u003c/li\u003e\n \u003cli\u003eVan de Moortel L, Speeckaert MM, Fiers T, Oeyen S, Decruyenaere J, Delanghe J. Low serum creatine kinase activity is associated with worse outcome in critically ill patients. J Crit Care. 2014;29(5):786-90.\u003c/li\u003e\n \u003cli\u003eKc O, Dahal PH, Koirala M, NtemMensah AD. Rhabdomyolysis and Neurological Manifestation With Progressive Weakness in a Young Adult: A Rare Extrapulmonary Presentation of Mycoplasma Pneumoniae. Cureus. 2021;13(12):e20552.\u003c/li\u003e\n \u003cli\u003eZhang X, Sun R, Jia W, Li P, Song C. Clinical Characteristics of Lung Consolidation with Mycoplasma pneumoniae Pneumonia and Risk Factors for Mycoplasma pneumoniae Necrotizing Pneumonia in Children. Infect Dis Ther. 2024;13(2):329-43.\u003c/li\u003e\n \u003cli\u003eFang W, Huang J, Wang J, Huang T, Lin D, Yin J. Blockade of interleukin-6 receptor attenuates apoptosis and modulates the inflammatory response in Mycoplasma pneumoniae infected A549 cells. Am J Transl Res. 2022;14(9):6187-95.\u003c/li\u003e\n \u003cli\u003eZhang M, Liu D, Song Y, Zhao J. Correlation between Interleukin-6, C-reactive protein, and lactate dehydrogenase levels and macrolide-resistant severe mycoplasma pneumoniae infection in children. BMC Infect Dis. 2025;26(1):62.\u003c/li\u003e\n \u003cli\u003eNi T, Zhao F. Predicting myocardial damage in children with mycoplasma pneumoniae pneumonia: a retrospective case-control study. BMC Infect Dis. 2025;25(1):733.\u003c/li\u003e\n \u003cli\u003eBongiovanni M. From Respiratory Pathogen to Systemic Threat: Rethinking Mycoplasma pneumoniae Infections. Microorganisms. 2026;14(2):419.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1. Baseline characteristics of the study population.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"99%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003eTotal (n = 1321)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003eNon-Liver Injury\u003c/p\u003e\n \u003cp\u003e(n = 1266)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003eLiver Injury\u003c/p\u003e\n \u003cp\u003e(n = 55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eHospitalization Duration, day\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e5.8 \u0026plusmn; 2.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e5.7 \u0026plusmn; 2.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e7.5 \u0026plusmn; 3.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eHospitalization Cost, yuan\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e10017.6 \u0026plusmn; 4029.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e9822.9 \u0026plusmn; 3680.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e14498.2 \u0026plusmn; 7622.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eSex, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.338\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Male\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e637 (48.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e607 (47.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e30 (54.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e684 (51.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e659 (52.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e25 (45.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eAge,year\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e7.0 \u0026plusmn; 2.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e7.0 \u0026plusmn; 2.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e7.8 \u0026plusmn; 2.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eCough, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e16 ( 1.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e16 (1.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e0 (0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1305 (98.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1250 (98.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e55 (100)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eFever, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e25 ( 1.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e24 (1.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e1 (1.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1296 (98.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1242 (98.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e54 (98.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePre-admission fever duration, day\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e8.2 \u0026plusmn; 5.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e8.2 \u0026plusmn; 5.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e8.6 \u0026plusmn; 4.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.564\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePleural reaction/Pleurisy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1319 (99.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1264 (99.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e55 (100)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e2 ( 0.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e2 (0.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e0 (0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePleural effusion, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1242 (94.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1200 (94.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e42 (76.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e79 ( 6.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e66 (5.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e13 (23.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePulmonary consolidation, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.029\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1158 (87.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1115 (88.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e43 (78.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e163 (12.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e151 (11.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e12 (21.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePulmonary atelectasis, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.022\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1207 (91.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1162 (91.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e45 (81.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e114 ( 8.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e104 (8.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e10 (18.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eCRP, mg/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e4.0 (0.7, 11.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e4.0 (0.7, 11.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e4.7 (0.7, 14.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.583\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eWBC, 10\u003csup\u003e9\u003c/sup\u003e/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e11.0 \u0026plusmn; 4.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e10.9 \u0026plusmn; 4.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e12.8 \u0026plusmn; 5.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eLYMPH, 10\u003csup\u003e9\u003c/sup\u003e/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e2.8 \u0026plusmn; 1.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e2.8 \u0026plusmn; 1.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e2.8 \u0026plusmn; 1.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.996\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNEUT, 10\u003csup\u003e9\u003c/sup\u003e/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e7.3 \u0026plusmn; 4.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e7.2 \u0026plusmn; 4.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e8.8 \u0026plusmn; 4.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eHGB, g/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e126.3 \u0026plusmn; 10.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e126.1 \u0026plusmn; 10.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e129.5 \u0026plusmn; 13.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.016\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePLT, 10\u003csup\u003e9\u003c/sup\u003e/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e363.1 \u0026plusmn; 122.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e363.0 \u0026plusmn; 122.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e365.4 \u0026plusmn; 119.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.884\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eALT, U/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e15.0 (11.0, 23.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e15.0 (11.0, 22.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e131.0 (101.0, 197.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eAST, U/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e25.0 (20.0, 31.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e24.0 (20.0, 30.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e50.0 (38.5, 92.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eLDH, U/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e365.4 \u0026plusmn; 160.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e356.3 \u0026plusmn; 136.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e574.4 \u0026plusmn; 383.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eCK,\u0026nbsp;U/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e57.0 (37.0, 90.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e57.5 (37.0, 90.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e40.0 (23.0, 76.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eCKMB, U/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e21.0 (16.0, 26.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e20.0 (16.0, 26.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e24.0 (17.5, 34.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eD-dimer, ng/mL\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e236.0 (155.0, 414.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e231.0 (154.0, 382.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e718.0 (391.0, 1270.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eIL-6,\u0026nbsp;pg/mL\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e3.5 (2.2, 7.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e3.5 (2.2, 7.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e4.3 (2.2, 21.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.076\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eOxygen, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1205 (91.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1163 (91.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e42 (76.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e116 ( 8.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e103 (8.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e13 (23.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eRMPP, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.013\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e302 (22.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e297 (23.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e5 (9.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1019 (77.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e969 (76.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e50 (90.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePE, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e1283 (97.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e1235 (97.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e48 (87.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 18px;\"\u003e\n \u003cp\u003e38 ( 2.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 24px;\"\u003e\n \u003cp\u003e31 (2.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 19px;\"\u003e\n \u003cp\u003e7 (12.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 7px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNote: Continuous variables are presented as mean \u0026plusmn; SD (normally distributed) or median (IQR) (skewed); categorical variables as n (%). Liver injury was defined as ALT \u0026gt;80 U/L. Group comparisons were performed using Student\u0026apos;s t-test, Mann-Whitney U test, or \u0026chi;\u0026sup2;/Fisher\u0026apos;s exact test, as appropriate. PE: Pulmonary Embolism.\u003c/p\u003e\n\u003cp\u003eTable 2. Comparison of the predictive ability of several models in the training set.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"98%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eModel\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eACC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBrier\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF\u0026beta;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCE\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR-AUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003ecatboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.993\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.009\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.925\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.993\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.860\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003egbm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.882\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.827\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.118\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.282\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.173\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.312\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e0.811\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.828\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.990\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.170\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003ekknn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.941\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.037\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.587\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.059\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.939\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.416\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003elightgbm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eNaive bayes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.804\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.771\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.168\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.204\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.229\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.221\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e0.703\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.774\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.984\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.119\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003ennet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.816\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.116\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.294\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.184\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.343\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e0.919\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.811\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.996\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.175\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eranger\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.998\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.013\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.974\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.998\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.949\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003erpart\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.872\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.852\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.119\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.328\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.234\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e0.865\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.851\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.993\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.203\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003esvm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.980\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.923\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.053\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.514\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.077\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.545\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e0.973\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.921\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.999\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.350\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003exgboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.937\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.896\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.166\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.425\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.104\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.474\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e0.919\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.895\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.996\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.276\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 3. Comparison of the predictive power of several models in the test set.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"98%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003eModel\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eACC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBrier\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF\u0026beta;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCE\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR-AUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 7px;\"\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003ecatboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.840\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.911\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.060\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.235\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.089\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.342\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.333\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.935\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.970\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.182\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003egbm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.895\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.856\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.104\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.308\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.144\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.292\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.778\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.859\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.989\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.192\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003ekknn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.785\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.869\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.260\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.131\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.155\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.883\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.169\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003elightgbm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.869\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.938\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.049\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.341\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.062\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.350\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.389\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.962\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.973\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.304\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003eNaive bayes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.911\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.805\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.136\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.261\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.195\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.273\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.833\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.804\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.991\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.155\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003ennet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.829\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.814\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.122\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.243\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.186\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.162\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.986\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.146\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003eranger\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.878\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.929\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.053\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.279\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.071\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.267\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.333\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.955\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.971\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.240\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003erpart\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.720\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.842\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.128\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.242\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.158\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.165\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.611\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.852\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.981\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.151\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003esvm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.835\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.897\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.074\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.286\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.103\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.223\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.977\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.200\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003exgboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.776\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.849\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.175\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.214\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.151\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.185\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.864\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.976\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.136\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 4. Comparison of the predictive power of several models in the external validation set.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"98%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eModel\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eACC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBrier\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF\u0026beta;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCE\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR-AUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003ecatboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.664\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.955\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.121\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.045\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.150\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.133\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.974\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.111\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003egbm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.755\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.919\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.070\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.071\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.081\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.180\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.133\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.938\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.978\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.049\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003ekknn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.771\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.931\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.083\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.069\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.065\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.133\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.950\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.061\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003elightgbm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.635\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.953\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.038\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.118\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.047\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.166\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.133\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.973\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.105\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eNaive bayes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.728\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.909\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.068\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.171\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.091\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.212\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.400\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.922\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.985\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.109\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003ennet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.755\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.902\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.160\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.168\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.400\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.984\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003eranger\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.726\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.970\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.038\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.174\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.030\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.180\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.133\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.990\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.250\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003erpart\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.492\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.870\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.096\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.088\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.130\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.030\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.267\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.885\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.980\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.053\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003esvm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.751\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.958\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.037\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.069\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.042\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.068\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.067\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.978\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.071\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003exgboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.584\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.864\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.172\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.065\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.136\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.073\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.200\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.880\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\n \u003cp\u003e0.038\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 5. Performance metrics of the selected nnet model in the training, test, and external validation sets.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"687\" class=\"fr-table-selection-hover\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eACC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBrier\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF\u0026beta;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR-\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003eRecall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCE\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003eLog loss\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 46px;\"\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 94px;\"\u003e\n \u003cp\u003eTrain.cv.folds\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.734\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.132\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.161\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.126\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.564\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.564\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.792\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.220\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.496\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.976\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 94px;\"\u003e\n \u003cp\u003eTrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.816\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.116\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.294\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.343\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.175\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.919\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.919\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.811\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.184\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.336\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.996\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.175\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 94px;\"\u003e\n \u003cp\u003eTest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.829\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.814\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.122\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.243\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.162\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.146\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.186\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.395\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.986\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.146\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 94px;\"\u003e\n \u003cp\u003eExternal.validaion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.755\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.902\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.160\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.168\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.400\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.400\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.178\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.984\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"bottom\" style=\"width: 46px;\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Mycoplasma Pneumoniae Pneumonia, Liver Injury, Machine Learning, Prediction Model, Neural Network, Children","lastPublishedDoi":"10.21203/rs.3.rs-9460650/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9460650/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground: \u003c/strong\u003eMycoplasma pneumoniae pneumonia (MPP) is associated with a high risk of liver injury, which adversely affects clinical outcomes and healthcare costs. Early identification of at-risk children remains challenging, and no validated predictive tool is currently available.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eThis multicenter retrospective study included 1,321 children with MPP from two centers in Nanjing, China. The development cohort was randomly split into training (67%) and test (33%) sets, with an independent external validation cohort (n = 640) from another hospital. Liver injury was defined as alanine aminotransferase (ALT) \u0026gt;80 U/L. Feature selection was performed using the boruta algorithm, and ten machine learning algorithms were developed and evaluated. Model performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy (ACC), sensitivity, specificity, calibration metrics, and decision curve analysis (DCA). The SHAP method was used for model interpretation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eAmong 1,321 patients, 55 (4.2%) developed liver injury. The boruta algorithm identified six predictors: LDH, D-dimer, CK, age, IL-6, and CK-MB. The neural network (nnet) model demonstrated optimal performance, with AUC values of 0.914 in the training set, 0.829 in the test set, and 0.755 in the external validation set. The model showed acceptable calibration and modest clinical utility on decision curve analysis. SHAP analysis revealed LDH as the most important predictor. An interactive web-based application was developed to facilitate clinical implementation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions: \u003c/strong\u003eWe developed and validated a machine learning-based model using six readily available predictors that identifies children with MPP at risk for liver injury. The accompanying web-based tool may assist clinicians in early risk stratification and timely intervention.\u003c/p\u003e","manuscriptTitle":"Development and External Validation of a Machine Learning Model for Predicting Liver Injury in Children With Mycoplasma Pneumoniae Pneumonia: A Multicenter Retrospective Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-11 05:05:54","doi":"10.21203/rs.3.rs-9460650/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"206115630244269816844120596373630051647","date":"2026-05-07T02:52:01+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-24T06:43:03+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-23T15:17:51+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-20T12:39:42+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-20T12:39:12+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-04-19T08:45:43+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"310766c7-dd27-4a1e-a2e3-ea625a9770a4","owner":[],"postedDate":"May 11th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"206115630244269816844120596373630051647","date":"2026-05-07T02:52:01+00:00","index":27,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":67812116,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":67812117,"name":"Health sciences/Diseases"},{"id":67812118,"name":"Health sciences/Health care"},{"id":67812119,"name":"Health sciences/Medical research"},{"id":67812120,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2026-05-11T05:05:54+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-11 05:05:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9460650","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9460650","identity":"rs-9460650","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}