Application of Explainable Machine Learning in Early Diagnosis Models for Risk Prediction of Severe Mycoplasma pneumoniae Pneumonia in Children

Application of Explainable Machine Learning in Early Diagnosis Models for Risk Prediction of Severe Mycoplasma pneumoniae Pneumonia in Children | 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 Research Article Application of Explainable Machine Learning in Early Diagnosis Models for Risk Prediction of Severe Mycoplasma pneumoniae Pneumonia in Children Shuai Yu, Yaya Ren, Jiangang Song, Yuxin Zhu, Hua Jiang, Yuanxia Li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7886677/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 07 Apr, 2026 Read the published version in European Journal of Medical Research → Version 1 posted 9 You are reading this latest preprint version Abstract Background Mycoplasma pneumoniae is a leading cause of community-acquired pneumonia in children, with severe cases (SMPP) posing a significant threat to pediatric health. Current diagnostic approaches rely primarily on imaging and clinical signs, lacking objective and quantitative tools for early risk prediction. Traditional statistical models face limitations in capturing the complexity of this condition, while machine learning (ML) methods offer the potential to uncover nonlinear relationships. However, the "black box" nature of many ML models hinders their clinical application. This study aimed to develop an interpretable ML model for the early prediction of SMPP risk in children and to enhance model transparency using Shapley Additive Explanations (SHAP) methods, thereby facilitating informed clinical decision-making. Methods: The study retrospectively analyzed data from 286 inpatients with MPP admitted to the Affiliated Hospital of Yan'an University between August 2023 and August 2024. Patients were divided into MPP (n = 163) and SMPP (n = 123) groups based on their clinical condition. Forty-four clinical variables, including symptoms, laboratory parameters, and imaging features, were collected. Pearson correlation analysis, Mann-Whitney U test, chi-square test, and LASSO regression were employed to identify key predictors. Seven machine learning models (CatBoost, XGBoost, LightGBM, SVM, KNN, LR, GNB) were constructed using Python. Hyperparameters were optimized through 5-fold cross-validation and grid search, and model performance was evaluated by accuracy, AUC, and other metrics. Model interpretability was analyzed using the SHAP method. Results Twenty-one key features, such as thermal peak, thermal path, D-dimer, CRP, pleural effusion, and bronchoscopic manifestations, were evaluated. Among the seven machine learning models, the CatBoost model demonstrated superior performance, achieving an AUC of .961, an accuracy of .907, a recall of .923, and an F1 score of .900. The performance discrepancy between the training and test sets was minimal, indicating robust generalization. SHAP visual analysis identified thermal peak, thermal range, D-dimer, and CRP as the most significant positive predictors. Decision curve analysis further validated the CatBoost model's higher clinical net benefit across a broad threshold range. Conclusions This study developed and internally validated an interpretable machine learning model using the CatBoost algorithm to effectively predict early risk of SMPP in children, outperforming traditional methods. The model incorporates multidimensional clinical features, emphasizing the significance of bronchoscopic findings, and offers a quantitative tool for early identification of high-risk children. Future multi-center, prospective studies are necessary to further assess the model's generalizability and clinical applicability. Severe Mycoplasma pneumoniae pneumonia Machine learning Predictive modeling Children CatBoost Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 1 Introduction Mycoplasma pneumoniae (MP) is a cell wall-deficient prokaryotic organism and a prevalent infectious pathogen [ 1 ] . Epidemiological studies have demonstrated that MP accounts for 40% of community-acquired pneumonia cases in children, primarily affecting school-age individuals and adolescents. In addition to pulmonary complications, MP infection can also precipitate a range of extrapulmonary manifestations, such as liver injury, encephalitis, hemolytic anemia, and thrombosis [ 2 ] . Alarmingly, the rise in antibiotic misuse has led to increased MP resistance, resulting in a growing incidence of severe Mycoplasma pneumoniae pneumonia (SMPP) in children [ 3 ] . SMPP can lead to severe intra- and extrapulmonary complications such as plastic bronchitis, pleural effusion, pulmonary consolidation necrosis, pulmonary embolism, myocardial injury, abnormal liver and renal function, anemia, and encephalitis [ 4 ] . These complications can diminish children's quality of life, impose a greater family burden, and even pose life-threatening risks [ 5 ] . The optimal treatment window for MPP is 5 to 10 days post-fever onset, with early recognition of SMPP being crucial for effective treatment [ 6 ] . The diagnosis of SMPP currently relies primarily on imaging techniques and clinical signs [ 7 ] . However, the clinical and imaging features of SMPP exhibit substantial heterogeneity and are highly similar to those of viral infections [ 8 ] . Consequently, there are significant differences in the clinical presentation of pneumonia among children [ 9 ] , posing a significant challenge to the timely and accurate diagnosis of SMPP and the implementation of appropriate treatment strategies [ 10 ] . Comprehensive guidelines, such as CURB-65 and PSI scores, have been established in the adult field to assess pneumonia severity in response to the diverse manifestations of the disease [ 11 ] . However, a lack of objective, quantitative, economical, and convenient diagnostic systems persists for evaluating childhood pneumonia. Previous studies have collected clinical data and examined indicators to predict the progression and possible etiology of pneumonia in children. Increases in respiratory rate, heart rate, and oxygen saturation have been found to indicate the diagnosis and progression of pediatric pneumonia [ 12 ] , and elevations in C-reactive protein and procalcitonin may be associated with bacterial infection and mortality [ 13 ] . The sensitivity and specificity of these indicators are not high, and some studies have reported conflicting results [ 14 ] . Consequently, clinical practice often varies with individual doctors' experiences, leading to either overdiagnosis and overtreatment or delays in diagnosis and treatment. Thus, predicting the occurrence of SMPP in children using existing clinical data is crucial. With the rising incidence of SMPP, research increasingly focuses on its early prediction. Current predictive models primarily utilize traditional logistic regression, which frequently faces challenges with data imbalance and may not achieve the accuracy required for modern diagnostic and therapeutic standards. No comprehensive model yet systematically integrates clinical features to effectively quantify the early predictive power of SMPP in children [ 15 ] . In recent years, machine learning has increasingly supplanted traditional statistical models in developing clinical diagnostic and predictive models, enhancing diagnostic and prognostic accuracy [ 16 ]−[ 17 ] . Machine learning (ML) employs sophisticated algorithms and statistical methods to predict disease progression, adverse outcomes, and treatment efficacy with precision [ 18 ] . It excels in data mining and classification, finding extensive applications across various medical domains [ 19 ]−[ 20 ] . However, ML algorithms often prioritize statistical features over clinically significant variables [ 21 ] . Shapley Additive Explanations (SHAP) is a prominent post-hoc interpretability algorithm that has gained significant attention in the field of ML [ 22 ] . Grounded in game theory's Shapley values, SHAP quantifies the impact of each feature on the model's output by calculating the average marginal contribution of features across all possible subsets. This approach not only ensures fairness and consistency of interpretation but also reveals the intricate interactions between features and their underlying mechanisms influencing the model's predictions, thereby providing reliable theoretical explanations for the model's behavior. The growing importance of ML interpretability can be attributed to the inherent "black box" nature of traditional ML models, underscoring the need for transparent and explainable AI systems [ 23 ] . Therefore, this study aims to construct a ML model for predicting the risk of SMPP, and adopt the SHAP method to interpret and visualize the model, thereby assisting clinicians in accurately identifying children at high risk of SMPP. 2 Subjects and Methods 2.1 Study Participants A total of 350 pediatric inpatients with Mycoplasma pneumoniae pneumonia (MPP) were initially enrolled in this study at the Affiliated Hospital of Yan'an University between August 2023 and August 2024. All participants underwent bronchoscopic alveolar lavage (BAL). Following the application of the predetermined inclusion and exclusion criteria and the removal of cases with extreme or aberrant values, 286 MPP patients were ultimately included in the final analysis. 2.2 SMPP diagnoses and inclusion and exclusion criteria 2.2.1 Diagnostic criteria MPP diagnostic criteria : According to the 2023 edition of the Guidelines for Diagnosis and Treatment of Mycoplasma Pneumonia in Children, the diagnostic criteria include [ 24 ] : (1) respiratory symptoms such as fever and cough, along with lung auscultation findings of dry and moist rales; (2) imaging results indicative of pneumonia. To confirm the diagnosis, these clinical and imaging criteria must be met, along with at least one of the following conditions: ①a single serum MP antibody titer of 1:160 or higher, or a fourfold or greater increase in the double serum MP antibody titer during the disease course; ②a positive result for MP DNA or RNA testing. Diagnostic criteria for SMPP : Based on confirmed MPP diagnosis, the criteria for severe pneumonia were established according to the "Guidelines for the Diagnosis and Management of Community-Acquired Pneumonia in Children (2019 Edition)" [ 25 ] , including any one of the following manifestations: ①poor general condition; ②impaired consciousness, cyanosis, tachypnea (respiratory rate ≥ 70 breaths/min in infants, or ≥ 50 breaths/min in children over 1 year old); ③signs of respiratory distress (grunting, nasal flaring/subcostal retractions), intermittent apnea, or oxygen saturation < 92%; ④ extreme fever or persistent high fever beyond 5 days; ⑤signs of dehydration or refusal to feed; ⑥chest CT revealing ≥ 2/3 lung infiltration in one lung, multi-lobar infiltration, pleural effusion, pneumothorax, atelectasis, pulmonary necrosis, or lung abscess; ⑦presence of extrapulmonary complications. 2.2.2 Inclusion criteria ①Age ≤ 14 years; ②Meeting the diagnostic criteria for MPP [ 24 ] ; ③Fulfilling the indications for bronchoscopy [ 26 ] ; ④Availability of complete clinical data. 2.2.3 Exclusion criteria ①absence of essential clinical data; ②presence of congenital diseases (e.g., congenital heart disease, Down syndrome); ③hematological diseases; ④concurrent infections or trauma at other sites; ⑤primary or acquired immunodeficiency; and ⑥severe pneumonia induced by bronchial foreign body aspiration. 2.3 Ethics approval This study was approved by the Ethics Committee of the Affiliated Hospital of Yan'an University (Approval No.IIT-R-20250178). 2.4 Study Group Allocation and Data Collection 2.4.1 Participant Assignment Based on disease severity, pediatric patients with pneumonia were divided into the MPP group (n = 163) and the SMPP group (n = 123). 2.4.2 Data Collection Clinical information was collected from paediatric medical records in medical records browsing system of Affiliated Hospital of Yan 'an University. Specific records included: ① Basic information: age, sex, BMI, length of stay, cost, single infection, mixed infection.② Symptoms + signs: respiratory symptoms (fever, heat spike, cough, expectoration, wheezing, shortness of breath); physical signs (wheezing, phlegm, moist rales, dyspnea, respiratory failure, etc.);③ extrapulmonary complications (Digestive system, cardiovascular system, nervous system);④ Laboratory examination: White blood cell count [WBC (10e9/L)], neutrophil percentage (NEU %), lymphocyte percentage (LYM %), hemoglobin (HB), platelet count [PLT (10e9/L)], C-reactive protein (CRP), procalcitonin (PCT), erythrocyte sedimentation rate (ESR), albumin (ALB), lactate dehydrogenase (LDH), alpha hydroxybutyrate dehydrogenase (HBDH), fibrinogen (FIB), fibrinogen degradation product (FDP), Plasma D-dimer, etc.; ⑤Imaging findings: Chest CT findings. Fasting venous blood samples were obtained within 24 hours post-admission for analysis. Chest CT scans were conducted either three days before admission or two days following admission, with results duly documented. 2.5 Development and Internal Validation of the Prediction Model 2.5.1 Evaluating the Stability of the Prediction Model The dataset was randomly partitioned into training and test subsets using the `train_test_split` function from the `sklearn.model_selection` module in Python 3.9.6, with a 70:30 ratio. To ensure reproducibility, the random seed was set to 42. 2.5.2 Feature Selection The study initially assessed 44 clinical characteristics, utilizing a sequential screening approach with Pearson correlation, Mann-Whitney U test, chi-square test, and Least Absolute Shrinkage and Selection Operator (LASSO) regression to identify the most pertinent variables. 2.5.3 Prediction Model Selection and Evaluation Based on SMPP risk characteristic variables, seven machine learning models were developed using Python 3.9.6, including Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Logistic Regression (LR), and Light Gradient Boosting Machine (LightGBM). Hyperparameters were optimized through grid search and random search, with the optimal parameter combinations selected via 5-fold cross-validation, using accuracy as the primary evaluation metric. After training, the best-performing model was applied to the test set for prediction. Model performance was comprehensively evaluated using accuracy, classification reports (including precision, recall, and F1-score), and confusion matrices. To enable intuitive comparison of model performance, receiver operating characteristic (ROC) curves, precision-recall curves, and decision curve analysis plots were generated for all models. The area under the curve (AUC) was used to quantify model discriminative ability, ultimately establishing a comprehensive model performance evaluation framework. 2.5.4 Model Explainability Machine learning-based predictive models can be interpreted using Shapley Additive Explanations (SHAP). SHAP is constructed based on insights from cooperative game theory, where all input features are regarded as "contributors." For each prediction instance, the model generates an output, and the SHAP value represents the numerical value assigned to each feature within that instance. Essentially, SHAP attributes the output value to the Shapley value of each feature—that is, it computes the SHAP value for every feature to quantify its influence on the final prediction. This approach enables precise measurement of the contribution and impact of each feature on the model’s output, reflecting not only the magnitude of each feature’s influence in every sample but also the direction of that influence (positive or negative). Furthermore, SHAP provides highly powerful data visualization capabilities to facilitate the interpretation of both the overall model and individual predictions. 2.6 Statistical Analysis Statistical analyses in this study were performed using SPSS 26.0. Normality of quantitative data was assessed using skewness and kurtosis tests, and homogeneity of variance was evaluated with Levene's test. Normally distributed continuous data are presented as mean ± standard deviation (x̅ ± s). For data with homogeneous variance, inter-group comparisons were conducted using the t-test; otherwise, the t'-test was applied. Non-normally distributed data are expressed as median (interquartile range) [M (Q1, Q3)], and the Wilcoxon rank-sum test was used for inter-group comparisons. Categorical data are summarized as frequency and proportion n (%). Fisher's exact test was employed for inter-group comparisons when at least 20% of expected frequencies were less than 5; otherwise, the Pearson chi-square test was applied. Machine learning models were developed using Python 3.9.6 in a Jupyter Notebook environment, leveraging key libraries including NumPy, pandas, matplotlib.pyplot, seaborn, and openpyxl. The SHAP algorithm was implemented to visualize model features and assess their contribution to the model predictions. 3 Results 3.1 Comparison of Clinical Characteristics Between the MPP and SMPP Groups Among 286 children infected with Mycoplasma pneumoniae, 163 had MPP and 123 had SMPP. A comparison between the MPP and SMPP groups revealed statistically significant differences (P < .05) in hospital days, costs, and infections (single/mixed), as well as in fever duration, peak temperature, wheezing, tachypnea, respiratory failure, sputum characteristics, and impacts on the digestive and cardiovascular systems. Significant differences were also noted in WBC, NEU, LYM, HB, CRP, PCT, ESR, ALB, LDH, HBDH, FDP, D-dimer, pulmonary consolidation, atelectasis, pleural effusion, lesion scope (patchy or floccular opacities/large shadow), pebble mucosa, plastic sputum plug, mucosal erosion, and inflammatory stricture. See Table 1 for details. Table 1 Comparison of Clinical Characteristics Between Children in the MPP and SMPP Groups [n (%), \(\:\stackrel{-}{\text{x}}\pm\:\text{s}\) ,p50 (p25,p75)] Variable MPP(n = 163) SMPP(n = 123) Statistic(t/χ2/Z) P value Age (years) 6.74 ± 2.43 6.49 ± 2.53 0.862 0.389 Gender 0.048 0.826 Boys 79(48.47) 58(47.15) Girls 84(51.53) 65(52.85) BMI 15.45(14.37, 17.56) 15.97(14.79, 17.36) -1.323 0.186 hospital days 9.00(8.00, 10.00) 10.00(9.00, 12.00) -4.577 <0.001 cost 8181.55(7179.06, 9441.93) 10097.64(8668.75, 12283.76) -7.473 <0.001 Infection 16.286 <0.001 Single 42(25.77) 9(7.32) Mixed 121(74.23) 114(92.68) fever days(d) 4.00(2.00, 6.00) 7.00(6.00, 9.00) -9.759 <0.001 peak temperature 38.90(38.30, 39.30) 39.80(39.38, 40.00) -10.757 <0.001 Cough 162(99.39) 122(99.19) / 1.000 Expectoration 162(99.39) 123(100) / 1.000 Wheezing 2(1.23) 7(5.70) / 0.042 Tachypnea 1(0.61) 9(7.32) / 0.003 Dyspnea 1(0.61) 5(4.07) / 0.088 Respiratory Failure 0(0.00) 4(3.25) / 0.033 Wheeze Sounds 2(1.23) 5(4.07) / 0.144 sputum sound 114(69.94) 103(83.74) 7.294 0.007 Moist Rales 160(98.16) 123(100) / 0.262 digestive system 21(12.88) 57(46.34) 39.564 <0.001 Cardiovascular System 0(0.00) 6(4.88) / 0.006 Nervous System 0(0.00) 1(0.81) / 1 WBC (10e9/L) 7.26(5.93, 9.19) 8.86(6.52, 12.33) -4.004 <0.001 NEU (%) 60.20(52.30, 68.50) 66.10(57.00, 77.90) -3.904 <0.001 LYM (%) 28.20(20.50, 37.60) 23.60(15.69, 32.40) -3.601 <0.001 HB (g/L) 127.00(121.00, 134.00) 126.00(119.00, 130.00) -2.348 0.019 PLT (10e9/L) 273.00(236.00, 332.00) 288.00(235.00, 362.00) -0.995 0.32 CRP (mg/L) 10.00(7.11, 15.91) 23.33(12.12, 50.56) -6.93 <0.001 PCT (ng/ml) 0.11(0.05, 0.29) 0.26(0.12, 0.93) -4.669 <0.001 ESR (mm/h) 19.00(12.00, 37.00) 39.00(23.00, 57.00) -5.234 <0.001 ALB (g/L) 41.60(39.60, 43.90) 39.90(36.50, 42.60) -4.429 <0.001 LDH (U/L) 279.00(249.00, 330.00) 328.00(267.00, 426.00) -4.187 <0.001 HBDH (U/L) 243.00(210.00, 288.00) 284.00(226.00, 358.00) -3.447 0.001 FIB (g/L) 4.29(3.77, 4.81) 4.40(3.83, 4.88) -1.111 0.266 FDP 2.50(2.50, 2.50) 2.60(2.50, 4.50) -5.925 <0.001 D-dimer (mg/L) 0.38(0.22, 0.64) 1.37(0.51, 3.44) -7.826 <0.001 Pulmonary Consolidation 44(26.99) 59(47.97) 13.382 <0.001 Atelectasis 0(0.00) 9(7.32) / <0.001 pleural effusion 0(0.00) 35(28.46) 52.85 <0.001 Lesion Scope 40.896 <0.001 Patchy or Flocculent Opacities 130(79.75) 53(43.09) large shadow 33(20.25) 70(56.91) cobblestone mucosa 21(12.88) 37(30.08) 12.825 <0.001 plastic sputum plug 24(14.72) 48(39.02) 21.976 <0.001 mucosal erosion 7(4.29) 26(21.14) 19.485 <0.001 inflammatory stricture 8(4.91) 21(17.07) 11.386 0.001 3.2 Results of Feature Selection The present study examined 44 initial features, with categorical variables processed using one-hot encoding. One-hot encoding is a widely-adopted technique for converting class features into a format suitable for machine learning and deep learning algorithms. This approach enables computers to better comprehend class features, thereby enhancing learning and predictive capabilities, and ultimately improving the performance of machine learning and deep learning models. The aforementioned 44 clinical features were sequentially subjected to the following three distinct methods for further screening: 3.2.1 To eliminate features exhibiting high inter-feature correlations, the Pearson correlation coefficient was employed. This statistic, also known as the Pearson product-moment correlation coefficient, quantifies the linear relationship between two variables, providing values from − 1 to 1. A value of 1 indicates perfect positive correlation, -1 signifies perfect negative correlation, and denotes no correlation. Consequently, features with correlation coefficients |r| ≥ .85 were discarded, retaining only those with |r| < .85. Following this criterion, 40 features were preserved after the removal of 4 features: NEU, patchy or cloudy shadows, LDH, and single infection. 3.2.2 To ensure the significance of the features, the Mann-Whitney U test and chi-square test were employed to identify variables associated with the outcome. Features with P-values greater than .05 were excluded, resulting in the elimination of 13 features. Consequently, 27 features were retained for further screening. These included: hospital stay, cost, fever days, heat peak, WBC, LYM, HB, CRP, PCT, ESR, ALB, HBDH, FDP, D-dimer, shortness of breath, dyspnea, sputum sound, digestive system, mucosal cobblestone protrusion, molded sputum thrombus, mucosal erosion, inflammatory stenosis, mixed infection, lung consolidation, atelectasis, pleural effusion, and large patchy shadow. 3.2.3 The LASSO regression technique was employed to screen features and ensure the stability and efficacy of the model. This method generates a penalty function λ to compress the variable coefficients in the regression model, thereby preventing overfitting and addressing the issue of severe collinearity, which is a widely used approach in predictive modeling. The LASSO regression was applied to select valuable feature sets from the 10-fold cross-validation results. As the penalty parameter λ was gradually increased from 10^-6 to 10^2, the number of variables included in the model decreased. When λ was set to 0.007565, the LASSO regression model exhibited the best predictive performance, eliminating 6 features and retaining 21 features with non-zero coefficients in the final model. The final set of retained features included days in hospital, days in fever, thermal peak, WBC, LYM, HB, CRP, PCT, ESR, ALB, FDP, D-dimer, sputum sound, digestive system, mucosal cobblestone ridge, molded sputum plug, mucosal erosion, inflammatory stenosis, mixed infection, pleural effusion, and large patchy shadow (Fig. 1). Figure 1 Feature Selection Process Using the Lasso Model Figure 1 depicts the changes in various metrics during LASSO regression. Panel a shows the mean squared error (MSE) as a function of the regularization parameter λ, where λ is plotted on the x-axis and MSE on the y-axis. Panel b shares the same x-axis as panel a, but the y-axis represents the coefficients of the model variables. This panel illustrates the relative importance of the variables. As λ increases, the coefficients of less important variables are driven to zero, while the coefficients of more important variables become less sensitive to the penalty imposed by the λ parameter, allowing them to be retained in the final model. The correlation matrix presented in Fig. 2 illustrates the relationships between the selected features of the predictive model. Negative correlations are depicted in blue, while positive correlations are shown in red. The intensity of the color corresponds to the strength of the correlation, with darker shades indicating stronger associations. 3.3 Results of the Machine Learning-Based Prediction Model Table 2 outlines the performance metrics—accuracy, precision, recall, F1 score, and AUC—of seven models (KNN, LightGBM, XGBoost, Catboost, SVM, LR, and GNB) on both training and test datasets. Table 2 Performance of the Machine Learning Prediction Model Model Accuracy Precision Recall F1 Score AUC ROC Training Set SVM 1.000 1.000 1.000 1.000 1.000 LightGBM 1.000 1.000 1.000 1.000 1.000 XGBoost 0.985 0.988 0.976 0.982 0.999 CatBoost 0.950 0.974 0.905 0.938 0.993 KNN 0.905 0.971 0.798 0.876 0.978 LR 0.905 0.901 0.869 0.885 0.962 GNB 0.820 0.914 0.631 0.746 0.934 Test Set CatBoost 0.907 0.878 0.923 0.900 0.961 SVM 0.884 0.854 0.897 0.875 0.940 XGBoost 0.884 0.854 0.897 0.875 0.945 LightGBM 0.884 0.837 0.923 0.878 0.961 LR 0.872 0.833 0.897 0.864 0.963 KNN 0.849 0.882 0.769 0.822 0.899 GNB 0.814 0.829 0.744 0.784 0.907 Figure 3 ROC Curves of the Different Machine Learning Models with Confidence Intervals As shown in Table 2 , Fig. 3 , and Fig. 4 , among the seven machine learning models evaluated, the CatBoost model demonstrated optimal overall performance, exhibiting particularly stable and outstanding results in the test set. Its test set AUC reached .961, performing at a similarly high level as LightGBM (.961) and LR (.963), and was slightly superior to XGBoost (.945), SVM (.940), GNB (.907), and KNN (.899). Furthermore, CatBoost achieved a test set Accuracy of .907, exceeding all other models (the second highest being .884). It also attained a Recall of .923, tying with LightGBM for the highest score in the test set and significantly outperforming models such as KNN (.769) and GNB (.744). Its Precision (.878) and F1 Score (.900) also ranked among the top performers in the test set, only slightly lower than its own best metric values, indicating well-balanced and excellent performance across all evaluation metrics. In the training set, SVM and LightGBM achieve the highest metrics (Accuracy, Precision, Recall, F1 Score, AUC), all at 1.000, indicating potential overfitting. In contrast, their test set AUCs (SVM .940, LightGBM .961) and Accuracies (both .884) are notably lower, highlighting insufficient generalization stability. XGBoost ranks just below SVM and LightGBM in the training set (AUC .999, Accuracy .985), but its test set AUC (.945) and Accuracy (.884) underperform compared to CatBoost, reflecting slightly inferior overall performance.The performance of the KNN, LR, and GNB models was relatively consistent across the training and test sets. While the LR model exhibited a slightly higher AUC (0.963) on the test set compared to CatBoost, its Accuracy (0.872) and Precision (0.833) were lower. Additionally, the AUC (0.962) of the LR model's training set was similar to its test set performance, indicating a balanced but not optimal overall performance. The KNN and GNB models were among the lower-performing models in the set of seven. The KNN model's Recall (0.769) on the test set and the GNB model's Accuracy (0.814) on the test set were relatively weak compared to the other models. The CatBoost model demonstrates superior stability compared to other models, with minimal differences in key evaluation metrics: an AUC difference of .032 (.993 for the training set and .961 for the test set) and an accuracy difference of .043 (.950 for the training set and .907 for the test set). These differences are notably smaller than those observed in SVM (AUC difference of .060, accuracy difference of .116) and LightGBM (AUC difference of .039, accuracy difference of .116). This stability indicates that CatBoost maintains reliable predictive capabilities despite changes in data distribution. Consequently, CatBoost is selected as the final prediction model due to its stable and accurate performance, effectively balancing training set performance, test set generalization, and metric consistency. Figure 5 Decision Curve Analysis (DCA) of the Various Machine Learning Models Decision curve analysis (DCA) was conducted on seven machine learning models in the training and test sets to evaluate the net clinical benefit of each model in informing clinical decision-making. DCA quantified the net benefit (i.e., the minimum acceptable probability of requiring further intervention) by contrasting the intervention strategy for each model with the default strategies of "Treat All" and "Treat None" [ 27 ] . The decision curves depicted in Fig. 5 illustrate the net benefit of the model across a range of decision thresholds, for both the training and test datasets. The dotted line representing "Intervention for all patients" and the solid black line denoting "Intervention for no patients" serve as reference strategies for comparison. The leftward positioning of the training set decision curve relative to the test set curve suggests the model may be overfit to the training data. In the training set, CatBoost, LightGBM, and similar models consistently outperform K-nearest neighbor and Gaussian naive Bayes, as their curves remain within the broad decision threshold and yield net returns surpassing the reference lines. This suggests their greater utility in aiding clinical decision-making during training. On the test set, the CatBoost model maintains a higher net benefit across a broader threshold range, significantly outperforming the two default strategies, thus demonstrating its effectiveness in clinical decision support post-generalization. The CatBoost model demonstrated robust decision-benefit performance across the training and test datasets, suggesting its potential to provide effective support for clinical decision-making. By judiciously setting the decision threshold, this model can facilitate the achievement of improved net benefit. 3.4 Model Explainability The feature importance analysis of the optimal CatBoost model, depicted in Fig. 6, reveals the relative significance of each predictor variable. The feature importance is quantified as the mean absolute value of a feature's influence on the target variable. The results indicate that the thermal peak has the highest predictive power across all prediction ranges, followed by thermal range, D-dimer, and CRP. To further elucidate the nature of these relationships, SHAP values are employed to identify both positive and negative associations between the predictors and the target outcome. Figure 6 SHAP Feature Importance Bar Plot Fig. 7 SHAP Beeswarm Plot Figure 8 presents a SHAP waterfall plot that clearly deconstructs the prediction logic for a high-risk severe case (actual severity: severe, model-predicted probability: .932). Starting from the baseline value, E[f(x)] = − .474, which represents the model's average prediction over the dataset, features such as peak body temperature (+ .95) and duration of fever (+ .81), represented by red bars, significantly drove the prediction probability toward the severe outcome. In contrast, features like WBC and FDP, indicated by blue bars, exerted weaker negative effects on the prediction of severity. An additional 12 features provided minor positive contributions. The cumulative effect of these multi-feature interactions elevated the model's output value, ultimately resulting in a high-probability severe prediction. This visualization intuitively demonstrates the contribution of each feature to the final prediction outcome. Figure 9 displays a SHAP decision plot, which reveals the model's decision-making process in distinguishing SMPP from non-SMPP cases by stacking the prediction paths of individual samples. Each line represents a patient case, originating from the baseline value at the bottom and extending to the model's output value on the x-axis, thereby clearly demonstrating the cumulative contribution of different features to the prediction outcome for each case. As illustrated, while "peak body temperature" and "D-dimer" serve as universally important positive drivers, the specific combination and sequence of features influencing the final prediction vary substantially across different samples. This visually demonstrates that the CatBoost model employs a highly individualized and flexible decision logic when synthesizing multiple indicators for judgment, adapting to the distinct clinical manifestations presented by different cases. 4 Discussion This study developed and internally validated an interpretable machine learning model for early prediction of SMPP. Employing the SHAP method, we elucidated the relative contribution of each feature to the predictive model, thereby enhancing its transparency and clinical reliability. Traditional logistic regression models, commonly used for SMPP risk prediction [ 28 ]−[ 29 ]−[ 30 ]−[ 31 ] , struggle with complex relationships and collinearity issues, potentially compromising prediction accuracy and stability. In contrast, machine learning models excel at capturing nonlinear relationships and interactions within data, thereby enhancing prediction accuracy and robustness [ 22 ] . Catia et al. [ 32 ] employed machine learning to predict mortality in patients with community-acquired pneumonia, facilitating early detection and intervention for high-risk individuals. Similarly, Zhang et al. [ 33 ] developed an XGBoost model to predict volume responsiveness in oliguric ICU patients with acute kidney injury, identifying urinary creatinine, serum urea nitrogen, and age as significant factors. In our study, we integrated clinical features, laboratory results, and imaging data, demonstrating that CatBoost achieved outstanding predictive performance, with an AUC of .961 and an accuracy of .907 in the test set. CatBoost [ 34 ]−[ 35 ] is a gradient boosting framework utilizing decision trees, renowned for its adept handling of categorical variables compared to conventional decision tree models. It computes target statistics for classification features to mitigate distribution bias between training and test data, enhancing model accuracy and generalization. CatBoost delivers impressive predictive outcomes without extensive parameter tuning, minimizing overtuning risk and fostering robust model development. Recent evidence highlights CatBoost's superior accuracy in various medical prediction domains, including acute pancreatitis [ 36 ] and cancer prognosis [ 37 ] . Our study corroborates CatBoost's exceptional discrimination capabilities over traditional linear methods in children's SMPP, a complex task marked by heterogeneous and nonlinear clinical manifestations, thus supporting early clinical precision interventions. The ability of traditional ML models to handle large-scale, high-dimensional data is well-established. However, these models often lack the capacity to elucidate the dose-response relationship between individual feature variables and the prediction outcome. Additionally, interpreting the inner workings of ML models and effectively communicating their predictions to clinicians have posed significant challenges [ 38 ] . The visualization of SHAP, through techniques such as bar charts and summary graphs, offers a means to quantify and graphically depict the relationship between clinical variables and disease risk [ 21 ] . The SHAP method is employed to interpret and visualize the CatBoost model, using bar and summary charts to break down a single sample's prediction into feature contribution values, thereby elucidating the model's decision process [ 39 ] . Several clinical factors potentially linked to SMPP development in children were identified: peak body temperature, temperature range, D-dimer levels, CRP levels, and pleural effusion. Elevated temperature and prolonged fever duration are risk factors for SMPP, possibly due to severe infection leading to intense lung inflammation, resulting in sustained high fever. Without timely intervention, this condition may progress to SMPP [ 40 ]−[ 41 ] . CRP, an acute-phase protein, serves as a prevalent inflammatory marker. Its levels rise with inflammation [ 42 ] . A retrospective cohort study involving 2,377 COVID-19 patients at New York University Langone Health Center indicated that those with elevated baseline D-dimer were more prone to critical illness compared to those with normal levels (43.9% vs. 18.5%) [ 43 ] . Research by Megan Carolina Cerda-Mancillas et al. [ 44 ] demonstrated a correlation between D-dimer plasma levels and SMPP severity. Pulmonary inflammation can lead to fluid exudation and pleural effusion, exacerbating dyspnea and potentially spreading inflammation, thereby contributing to SMPP. Radiologically, lower lobe infections are more frequently associated with SMPP than infections in other lung regions [ 45 ] . The model notably attributes predictive significance to bronchoscopic features such as "mucosal cobblestones" and "shaped sputum plugs," representing a pivotal advancement in this research. These features serve as "end-organ evidence" of severe airway inflammation and epithelial damage induced by Mycoplasma pneumonia. As noted by C. Tran et al. [ 46 ] , airway mucus embolism primarily results from increased bronchial secretions, which are difficult to expel, leading to bronchial obstruction and compromised respiration. The blockage caused by mucus plugs can result in symptoms such as dyspnea, cough, wheezing, and chest discomfort, potentially escalating to oxygen deficiency, cyanosis, respiratory failure, and other critical conditions. This study objectively and quantitatively establishes the independent predictive value of these microscopic features in assessing the early risk of SMPP using interpretable AI techniques. This approach not only offers clinicians a more robust decision-making framework beyond conventional laboratory parameters but also underscores the critical importance of early bronchoscopy in risk assessment and the exploration of underlying mechanisms in children who exhibit poor responses to initial treatment. This study offers several advantages: Firstly, it incorporates a multidimensional set of clinical characteristics, including manifestations, laboratory results, and imaging tests, which are commonly accessible in medical settings, thereby enhancing the predictive model's generalizability. Secondly, the model employs a machine learning algorithm, which is better suited than traditional logistic regression for handling complex, nonlinear data. Given that diseases often arise from multiple factors, machine learning algorithms are more effective for early prediction of SMPP, delivering precise results. Additionally, while transparency and interpretability are crucial in clinical practice, machine learning models often struggle with these aspects. This study addresses interpretability by utilizing SHAP, an interpretability tool in artificial intelligence. In summary, the model effectively captures complex data and provides interpretable outputs, offering valuable support for the early clinical identification of SMPP. This study is subject to several limitations. The model was developed retrospectively using single-center data, introducing inherent biases in the data collection process. Additionally, model validation was conducted solely through internal methods, limiting the assessment of its broader applicability. To enhance the model's universality, future research should incorporate external datasets for validation. Importantly, the proposed model lacks prospective validation, which will be a focus of forthcoming investigations. 5 Conclusions The present study utilized machine learning techniques to develop seven predictive models based on key feature variables. Of these, the CatBoost model demonstrated robust predictive performance through internal validation. To corroborate these findings, the authors recommend conducting multicenter validation and large-scale prospective investigations. Abbreviations MP Mycoplasma pneumoniae MPP Mycoplasma pneumoniae pneumonia SMPP severe Mycoplasma pneumoniae pneumonia ML Machine learning SHAP Shapley Additive Explanations BAL bronchoscopic alveolar lavage LASSO Least Absolute Shrinkage and Selection Operator WBC White blood cell count NEU % Laboratory tests included neutrophil percentage LYM % lymphocyte percentage HB hemoglobin PLT platelet count CRP C-reactive protein PCT procalcitonin ESR erythrocyte sedimentation rate ALB albumin LDH lactate dehydrogenase HBDH alpha hydroxybutyrate dehydrogenase FIB fibrinogen FDP fibrin (pro) degradation product CatBoost Categorical Boosting XGBoost Extreme Gradient Boosting GNB Gaussian Naive Bayes SVM Support Vector Machine KNN K-Nearest Neighbors LR Logistic Regression LightGBM Light Gradient Boosting Machine PR Precision-Recall DCA Decision Curve Analysis Declarations Ethics approval and consent to participate This retrospective study involved the analysis of medical records from 286 hospitalized children diagnosed with pneumonia at the Affiliated Hospital of Yan'an University. Data collection was conducted non-invasively through the retrieval of anonymized routine medical records from the hospital archives. Due to the retrospective and non-invasive nature of the study, along with the de-identification of the data utilized, the research team sought and obtained a waiver of informed consent from the Ethics Committee of Yan'an University Hospital, China. The waiver application included a detailed rationale for the exemption. The Ethics Committee rigorously evaluated the study design, data handling procedures, and grounds for exemption in line with the principles of the Declaration of Helsinki. Subsequently, the committee formally granted approval for the study (No.IIT-R-20250178) to ensure adherence to ethical standards and regulatory protocols. Availability of data and materials The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation. Funding No funding required Author information Authors and Affiliations Affiliated Hospital of Yan’an University, Yan’an, Shaanxi, China Shuai Yu, Yaya Ren, Jiangang Song, Yuxin Zhu, Hua Jiang, Yuanxia Li Authors' contributions YS：Writing – original draft, Writing – review & editing. RYY: Writing – original draft, Writing – review & editing. SJG: Writing – review & editing. ZYX: Writing – review & editing. JH：Writing – review & editing. LYX: Writing – review & editing. All authors reviewed the manuscript. Corresponding author Correspondence to Yuanxia Li. Consent for publication Not applicable. Competing Interests The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Acknowledgements Not applicable.CRediT roles Clinical trial number Not applicable. References Ding G, Zhang X, Vinturache A, van Rossum AMC, Yin Y, Zhang Y. Challenges in the treatment of pediatric Mycoplasma pneumoniae pneumonia. Eur J Pediatr. 2024;183(7):3001–3011. doi: 10.1007/s00431-024-05519-1 . Epub 2024 Apr 18. PMID: 38634891. Li YT, Zhang J, Wang MZ, Ma YM, Zhi K, Dai FL,et al. Changes in coagulation markers in children with Mycoplasma pneumoniae pneumonia and their predictive value for Mycoplasma severity. Ital J Pediatr. 2023;49(1):143. doi: 10.1186/s13052-023-01545-1 . PMID: 37858230; PMCID: PMC10588045. Fang C, Mao Y, Jiang M, Yin W. Pediatric Critical Illness Score, Clinical Characteristics and Comprehensive Treatment of Children with Severe Mycoplasma Pneumoniae Pneumonia. Front Surg. 2022;9:897550. doi: 10.3389/fsurg.2022.897550 . PMID: 35693303; PMCID: PMC9174934. 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. doi: 10.1038/s41598-024-58784-3 . PMID: 38589453; PMCID: PMC11002011. Ding X, Li S, Liu X, Zhan X, Wang Z, Wang M,et al.Study on the Correlation Between the Expression of NF-Ƙb in the Alveolar Lavage Fluid of Children with Severe Mycoplasma Pneumoniae Pneumonia, Its Clinical Characteristics, and Cellular Immunity. Infect Drug Resist. 2024;17:2469–2484. doi: 10.2147/IDR.S411361. PMID: 38915319; PMCID: PMC11194830. Yan C, Xue GH, Zhao HQ, Feng YL, Cui JH, Yuan J. Current status of Mycoplasma pneumoniae infection in China. World J Pediatr. 2024;20(1):1–4. doi: 10.1007/s12519-023-00783-x . Epub 2024 Jan 8. PMID: 38185707; PMCID: PMC10827902. Kant R, Kumar N, Malik YS, Everett D, Saluja D, Launey T, et al. Critical insights from recent outbreaks of Mycoplasma pneumoniae: decoding the challenges and effective interventions strategies. Int J Infect Dis. 2024;147:107200. doi: 10.1016/j.ijid.2024.107200 . Epub 2024 Aug 6. PMID: 39117175. Gao L, Sun Y. Laboratory diagnosis and treatment of Mycoplasma pneumoniae infection in children: a review. Ann Med. 2024;56(1):2386636. doi: 10.1080/07853890.2024.2386636. Epub 2024 Aug 3. PMID: 39097794; PMCID: PMC11299444. Zhao MC, Wang L, Qiu FZ, Zhao L, Guo WW, Yang S,et al.Impact and clinical profiles of Mycoplasma pneumoniae co-detection in childhood community-acquired pneumonia. BMC Infect Dis. 2019;19(1):835. doi: 10.1186/s12879-019-4426-0 . PMID: 31601192; PMCID: PMC6788033. Dean P, Schumacher D, Florin TA. Defining Pneumonia Severity in Children: A Delphi Study. Pediatr Emerg Care. 2021;37(12):e1482-e1490. doi: 10.1097/PEC.0000000000002088 . PMID: 32205802; PMCID: PMC8223262. Metlay JP, Waterer GW, Long AC, Anzueto A, Brozek J, Crothers K, et al.Diagnosis and Treatment of Adults with Community-acquired Pneumonia. An Official Clinical Practice Guideline of the American Thoracic Society and Infectious Diseases Society of America. Am J Respir Crit Care Med. 2019;200(7):e45-e67. doi: 10.1164/rccm.201908-1581ST . PMID: 31573350; PMCID: PMC6812437. Shah SN, Bachur RG, Simel DL, Neuman MI. Does This Child Have Pneumonia?: The Rational Clinical Examination Systematic Review. JAMA. 2017;318(5):462–471. doi: 10.1001/jama.2017.9039 . Erratum in: JAMA. 2017;318(13):1284. doi: 10.1001/jama.2017.13624. PMID: 28763554. Zhang Y, Shi Q, Zhong G, Lei X, Lin J, Fu Z,et al.Biomarker-based score for predicting in-hospital mortality of children admitted to the intensive care unit. J Investig Med. 2021;69(8):1458–1463. doi: 10.1136/jim-2021-001855 . Epub 2021 Jul 5. PMID: 34226249. Doerflinger M, Haeusler GM, Li-Wai-Suen CSN, Clark JE, Slavin M, Babl FE, et al. Procalcitonin and Interleukin-10 May Assist in Early Prediction of Bacteraemia in Children With Cancer and Febrile Neutropenia. Front Immunol. 2021;12:641879. doi: 10.3389/fimmu.2021.641879. PMID: 34093531; PMCID: PMC8173204. 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. doi: 10.1186/s12879-025-10958-8 . PMID: 40259232; PMCID: PMC12013137. Araújo ALD, Moraes MC, Pérez-de-Oliveira ME, Silva VMD, Saldivia-Siracusa C, Pedroso CM,et al. Machine learning for the prediction of toxicities from head and neck cancer treatment: A systematic review with meta-analysis. Oral Oncol. 2023;140:106386. doi: 10.1016/j.oraloncology.2023.106386. Epub 2023 Apr 4. PMID: 37023561. Black JE, Kueper JK, Williamson TS. An introduction to machine learning for classification and prediction. Fam Pract. 2023;40(1):200–204. doi: 10.1093/fampra/cmac104 . PMID: 36181463. An Q, Rahman S, Zhou J, Kang JJ. A Comprehensive Review on Machine Learning in Healthcare Industry: Classification, Restrictions, Opportunities and Challenges. Sensors (Basel). 2023;23(9):4178. doi: 10.3390/s23094178 . PMID: 37177382; PMCID: PMC10180678. Yi F, Yang H, Chen D, Qin Y, Han H, Cui J, et al.XGBoost-SHAP-based interpretable diagnostic framework for alzheimer's disease. BMC Med Inform Decis Mak. 2023;23(1):137. doi: 10.1186/s12911-023-02238-9 . PMID: 37491248; PMCID: PMC10369804. Anh, D.T., Takakura, H., Asai, M. et al. Application of machine learning in the diagnosis of vestibular disease. Sci Rep 12, 20805 (2022). https://doi.org/10.1038/s41598-022-24979-9 Xue B, Li D, Lu C, King CR, Wildes T, Avidan MS, et al.Use of Machine Learning to Develop and Evaluate Models Using Preoperative and Intraoperative Data to Identify Risks of Postoperative Complications. JAMA Netw Open. 2021;4(3):e212240. doi: 10.1001/jamanetworkopen.2021.2240 . PMID: 33783520; PMCID: PMC8010590. Ahlquist KD, Sugden LA, Ramachandran S. Enabling interpretable machine learning for biological data with reliability scores. PLoS Comput Biol. 2023;19(5):e1011175. doi: 10.1371/journal.pcbi.1011175 . PMID: 37235578; PMCID: PMC10249903. AlMansoori ME, Jemimah S, Abuhantash F, AlShehhi A. Predicting early Alzheimer's with blood biomarkers and clinical features. Sci Rep. 2024;14(1):6039. doi: 10.1038/s41598-024-56489-1. Erratum in: Sci Rep. 2024;14(1):20210. doi: 10.1038/s41598-024-71400-8 . PMID: 38472245; PMCID: PMC10933308. Zhao S Y, Qian S Y, Chen Z M, et al. Guidelines for the Diagnosis and Treatment of Mycoplasma Pneumoniae Pneumonia in Children (2023 Edition) [J]. (in Chinese).Infectious Disease Information, 2023, 36(04): 291–297. DOI: CNKI:SUN:CRBX.0.2023-04-019. National Health Commission of the People's Republic of China, National Administration of Traditional Chinese Medicine. Guidelines for the Diagnosis and Treatment of Community-Acquired Pneumonia in Children (2019 Edition) [J]. (in Chinese).Chinese Journal of Clinical Infectious Diseases, 2019, 12(1): 6–13. DOI: 10.3760/cma.j.issn.1674-2397.2019.01.002 . Expert Group on Pediatric Respiratory Endoscopy Technology, Talent Exchange Service Center of the National Health Commission; Endoscopy Professional Committee, Pediatrician Branch of the Chinese Medical Doctor Association; Pediatric Respiratory Endoscopy Professional Committee, Endoscopist Branch of the Chinese Medical Doctor Association; et al. Guidelines for Flexible Bronchoscopy in Chinese Pediatrics (2018 Edition) [J]. (in Chinese).Chinese Journal of Practical Pediatrics and Child Health, 2018, 33(13):983–989. DOI: 10.3760/cma.j.issn.2095-428X.2018.13.006 . Vickers AJ, van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res. 2019;3:18. doi: 10.1186/s41512-019-0064-7 . PMID: 31592444; PMCID: PMC6777022. 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. doi: 10.1038/s41598-024-58784-3 . PMID: 38589453; PMCID: PMC11002011. Wang S, Wan Y, Zhang W. The Clinical Value of Systemic Immune Inflammation Index (SII) in Predicting the Severity of Hospitalized Children with Mycoplasma Pneumoniae Pneumonia: A Retrospective Study. Int J Gen Med. 2024;17:935–942. doi: 10.2147/IJGM.S451466. PMID: 38495920; PMCID: PMC10944171. Qiu J, Ge J, Cao L. D-dimer: The Risk Factor of Children's Severe Mycoplasma Pneumoniae Pneumonia. Front Pediatr. 2022;10:828437. doi: 10.3389/fped.2022.828437 . PMID: 35498793; PMCID: PMC9039299. Jiang C, Bao S, Shen W, Wang C. Predictive value of immune-related parameters in severe Mycoplasma pneumoniae pneumonia in children. Transl Pediatr. 2024;13(9):1521–1528. doi: 10.21037/tp-24-172 . Epub 2024 Sep 13. PMID: 39399713; PMCID: PMC11467233. Cilloniz C, Ward L, Mogensen ML, Pericàs JM, Méndez R, Gabarrús A,et al. Machine-Learning Model for Mortality Prediction in Patients With Community-Acquired Pneumonia: Development and Validation Study. Chest. 2023;163(1):77–88. doi: 10.1016/j.chest.2022.07.005. Epub 2022 Jul 16. PMID: 35850287. Zhang Z, Ho KM, Hong Y. Machine learning for the prediction of volume responsiveness in patients with oliguric acute kidney injury in critical care. Crit Care. 2019;23(1):112. doi: 10.1186/s13054-019-2411-z . PMID: 30961662; PMCID: PMC6454725. Hancock JT, Khoshgoftaar TM. CatBoost for big data: an interdisciplinary review. J Big Data. 2020;7(1):94. doi: 10.1186/s40537-020-00369-8 . Epub 2020 Nov 4. PMID: 33169094; PMCID: PMC7610170. @inproceedings{Ostroumova2017CatBoostUB, title={CatBoost: unbiased boosting with categorical features}, author={Liudmila Ostroumova and Gleb Gusev and Aleksandr Vorobev and Anna Veronika Dorogush and Andrey Gulin}, booktitle={Neural Information Processing Systems}, year={2017}, url={https://api.semanticscholar.org/CorpusID:5044218} } Kui B, Pintér J, Molontay R, Nagy M, Farkas N, Gede N, et al.Hungarian Pancreatic Study Group. EASY-APP: An artificial intelligence model and application for early and easy prediction of severity in acute pancreatitis. Clin Transl Med. 2022;12(6):e842. doi: 10.1002/ctm2.842 . PMID: 35653504; PMCID: PMC9162438. Geeitha S, Ravishankar K, Cho J, Easwaramoorthy SV. Integrating cat boost algorithm with triangulating feature importance to predict survival outcome in recurrent cervical cancer. Sci Rep. 2024;14(1):19828. doi: 10.1038/s41598-024-67562-0 . PMID: 39191808; PMCID: PMC11349876. Lei T, Guo J, Wang P, Zhang Z, Niu S, Zhang Q,et al. Establishment and Validation of Predictive Model of Tophus in Gout Patients. J Clin Med. 2023;12(5):1755. doi: 10.3390/jcm12051755 . PMID: 36902542; PMCID: PMC10002994. Fan Z, Jiang J, Xiao C, Chen Y, Xia Q, Wang J,et al.Construction and validation of prognostic models in critically Ill patients with sepsis-associated acute kidney injury: interpretable machine learning approach. J Transl Med. 2023;21(1):406. doi: 10.1186/s12967-023-04205-4 . PMID: 37349774; PMCID: PMC10286378. 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):329–343. doi: 10.1007/s40121-023-00914-x . Epub 2024 Jan 24. PMID: 38265626; PMCID: PMC10904708. Tsai TA, Tsai CK, Kuo KC, Yu HR. Rational stepwise approach for Mycoplasma pneumoniae pneumonia in children. J Microbiol Immunol Infect. 2021;54(4):557–565. doi: 10.1016/j.jmii.2020.10.002 . Epub 2020 Oct 17. PMID: 33268306. 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. doi: 10.1038/s41598-024-58784-3 . PMID: 38589453; PMCID: PMC11002011. Yadav RK, Kumar D, Gupta A, Sharma P. C-reactive protein and procalcitonin: As predictor biomarkers of severity and outcome in children with community-acquired pneumonia. Trop Doct. 2024;54(3):262–267. doi: 10.1177/00494755241250371. Epub 2024 May 2. PMID: 38693837. Kolotylo TR, Moskaliuk VD, Syrota BV, Balaniuk IV, Melenko SR, Chernetska NV, et al. EVALUATION OF D-DIMER LEVEL AS A BIOMARKER OF DISEASE SEVERITY AND MORTALITY IN PATIENTS WITH COVID-19. Wiad Lek. 2023;76(7):1636–1641. doi: 10.36740/WLek202307118 . PMID: 37622508. Luo XQ, Luo J, Wang CJ, Luo ZX, Tian DY, Xie XH. Clinical features of severe Mycoplasma pneumoniae pneumonia with pulmonary complications in childhood: A retrospective study. Pediatr Pulmonol. 2023;58(10):2815–2822. doi: 10.1002/ppul.26593 . Epub 2023 Jul 11. PMID: 37431970. Tran C, Singh GV, Haider E, Boylan C, Venegas C, Riaz S, et al. Luminal mucus plugs are spatially associated with airway wall thickening in severe COPD and asthma: A single-centered, retrospective, observational study. Respir Med. 2022;202:106982. doi: 10.1016/j.rmed.2022.106982. Epub 2022 Sep 9. PMID: 36116144. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 07 Apr, 2026 Read the published version in European Journal of Medical Research → Version 1 posted Editorial decision: Revision requested 28 Dec, 2025 Reviews received at journal 27 Dec, 2025 Reviewers agreed at journal 17 Dec, 2025 Reviews received at journal 19 Nov, 2025 Reviewers agreed at journal 19 Nov, 2025 Reviewers invited by journal 19 Nov, 2025 Editor assigned by journal 23 Oct, 2025 Submission checks completed at journal 23 Oct, 2025 First submitted to journal 17 Oct, 2025 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-7886677","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":547549951,"identity":"181560e9-05ae-4429-bd87-b220267f51a2","order_by":0,"name":"Shuai Yu","email":"","orcid":"","institution":"Affiliated Hospital of Yan’an University","correspondingAuthor":false,"prefix":"","firstName":"Shuai","middleName":"","lastName":"Yu","suffix":""},{"id":547549953,"identity":"443372fa-7452-49e3-87c1-14500d2c3b5d","order_by":1,"name":"Yaya Ren","email":"","orcid":"","institution":"Affiliated Hospital of Yan’an University","correspondingAuthor":false,"prefix":"","firstName":"Yaya","middleName":"","lastName":"Ren","suffix":""},{"id":547549958,"identity":"42e307ad-5f70-4d28-9fb1-915c165d2f9e","order_by":2,"name":"Jiangang Song","email":"","orcid":"","institution":"Affiliated Hospital of Yan’an University","correspondingAuthor":false,"prefix":"","firstName":"Jiangang","middleName":"","lastName":"Song","suffix":""},{"id":547549961,"identity":"4fc4b6e9-fc98-4d1f-9abe-4a491e828809","order_by":3,"name":"Yuxin Zhu","email":"","orcid":"","institution":"Affiliated Hospital of Yan’an University","correspondingAuthor":false,"prefix":"","firstName":"Yuxin","middleName":"","lastName":"Zhu","suffix":""},{"id":547549963,"identity":"c783fcd6-8496-4803-97be-e7213fa82694","order_by":4,"name":"Hua Jiang","email":"","orcid":"","institution":"Affiliated Hospital of Yan’an University","correspondingAuthor":false,"prefix":"","firstName":"Hua","middleName":"","lastName":"Jiang","suffix":""},{"id":547549964,"identity":"639e66b1-89b5-4fcf-b444-98ef2ac4821c","order_by":5,"name":"Yuanxia Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyElEQVRIiWNgGAWjYBACefnHxz9IVNjIsbE3EKnFsCEtjcHiTJoxH88BYq05kGPGUNlyKHGeRAKROhgbjqU9uNlwwJhN8vHGGww1NtEEtbAzNh83nLnjjhybdFqxBcOxtNwGgrY0syVIS555ZswmnWMmwdhwmLAWhmM8BtJ/2w4ntkmeIVbLGR4zCUmQFgkeIrUYzmBLNpAABjIbD9AvCcT4RV6C+eADUFTKtx/eeONDjQ0RDkMCBkRHDZIWUnWMglEwCkbByAAAn8Y/8kBew44AAAAASUVORK5CYII=","orcid":"","institution":"Affiliated Hospital of Yan’an University","correspondingAuthor":true,"prefix":"","firstName":"Yuanxia","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2025-10-17 12:23:26","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7886677/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7886677/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s40001-026-04319-7","type":"published","date":"2026-04-07T15:57:02+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":96970683,"identity":"c42103f7-4fc6-4191-a5d5-54cb7249a4c9","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1146572,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript2025.10.11.docx","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/e5ce54b4ecb16da7ec1ebaad.docx"},{"id":96970678,"identity":"f08ca0c4-78c9-46d0-8386-15f0f7ab2f2d","added_by":"auto","created_at":"2025-11-28 07:14:42","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9957,"visible":true,"origin":"","legend":"","description":"","filename":"d2ca250496914d6e8229cc792ddeaa65.json","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/95982059e5fae0e98527f739.json"},{"id":97137585,"identity":"ed1f2e8c-b4e0-467b-b2b9-35df7cd412ae","added_by":"auto","created_at":"2025-12-01 09:57:57","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170544,"visible":true,"origin":"","legend":"","description":"","filename":"d2ca250496914d6e8229cc792ddeaa651enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/1f99582e6c9f178594b6fed1.xml"},{"id":97137713,"identity":"ca76a74d-199a-4f87-afbd-808419d5111d","added_by":"auto","created_at":"2025-12-01 09:58:05","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":114556,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/97e69c32eabcd577f05b48b2.png"},{"id":96970686,"identity":"83671c2d-b281-48f1-b6d6-d98b2888f6f9","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1116143,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/6dccefca0139022092ceecb6.jpeg"},{"id":97137885,"identity":"19c59122-c554-4870-9faa-dd5a7899154b","added_by":"auto","created_at":"2025-12-01 09:58:17","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":177154,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/ffa3ea9c750e51024ad71c9e.jpeg"},{"id":97138318,"identity":"ee2a3d0a-4a36-4d40-acca-1341dd7b56fc","added_by":"auto","created_at":"2025-12-01 09:58:45","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":184762,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/3330fb2b892844ba477563a7.png"},{"id":96970704,"identity":"521b36a7-1299-4098-9d34-5f7349d0d6f2","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":332221,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/222d9af92e840d24029479d9.jpeg"},{"id":96970696,"identity":"ca08eac5-2c3b-4264-b88a-7ba16efc22e6","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90777,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/6af7f11716379a000e2b81fb.png"},{"id":97137740,"identity":"04aebf80-b248-45cf-a003-3b8c4a320327","added_by":"auto","created_at":"2025-12-01 09:58:06","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":95729,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/77c2ce828eee8d1a66756546.png"},{"id":97137632,"identity":"bf1d4eab-3392-47c3-b5f0-56239db93d17","added_by":"auto","created_at":"2025-12-01 09:58:00","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47810,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/ec318572cd9b11a48fd587ca.png"},{"id":97138026,"identity":"8ed09f93-38ef-4e0e-91b8-aae8ebacf749","added_by":"auto","created_at":"2025-12-01 09:58:25","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":73271,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/217a46bfe74faf1467f95951.png"},{"id":96970701,"identity":"31ab78a6-368e-49ad-85ab-69e6bc618a58","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82186,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/b0c526f009374d5aad883def.png"},{"id":96970708,"identity":"78007565-6181-46bc-8313-b832f4c81729","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":119662,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/fbb0a70aa2c137fc4f810db0.png"},{"id":96970694,"identity":"051e0abd-0575-40b2-9fa8-849278303a83","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":25809,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/20711046f11fc0cedf8561a5.png"},{"id":96970706,"identity":"d3d340a4-bad7-4fc0-bb2d-1e2236d05f4b","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":147574,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/65f708928493261a9484524d.png"},{"id":96970698,"identity":"6463d5b0-e482-4d5e-b932-e7484ae7a579","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36330,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/f957d0ebe86889f4bb731ecf.png"},{"id":96970700,"identity":"fa1e5eec-6995-4a35-992c-15507276eae4","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44364,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/6743fc23d737862ac9164057.png"},{"id":96970705,"identity":"be9552e3-67c4-46fa-b377-dc56b35b55f1","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":66182,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/94dad43659e122cc4d44506d.png"},{"id":96970699,"identity":"0b60a88f-ef78-4d21-aac2-903dd2df293a","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27569,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/039b09a9a65bd4b86ae02655.png"},{"id":97138302,"identity":"d0c3abd9-b57f-4c73-b37b-b06e7cf369f8","added_by":"auto","created_at":"2025-12-01 09:58:45","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20239,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/44bda981c854fe25c21cf629.png"},{"id":97136954,"identity":"ed9cfd49-42b4-4f29-8f17-42ff213a538d","added_by":"auto","created_at":"2025-12-01 09:57:12","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12323,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/5e7109bcd5290025f3ca795d.png"},{"id":96970703,"identity":"529e8f57-6fe6-4d55-886a-ee6c53b2f56e","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16761,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/320e1bdce3fa30fe69b18acd.png"},{"id":96970710,"identity":"1ce920fd-dcda-41b3-833f-d5052431b586","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22470,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/f498182fc54b9de0829b87ec.png"},{"id":96970707,"identity":"43e31779-aca3-4e42-a7c5-6fd91d5bdd2e","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29033,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/c2b3ab625a6d50486b6b2f80.png"},{"id":96970712,"identity":"7e76a1c2-6bfe-40ec-9b8c-fd7e5aafcd38","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"xml","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":164889,"visible":true,"origin":"","legend":"","description":"","filename":"d2ca250496914d6e8229cc792ddeaa651structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/c93df705c38dac78aac84828.xml"},{"id":96970714,"identity":"664e82c0-32aa-4f32-bcd7-6fab0979654c","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"html","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":180932,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/568d214b36572a26cbc26e95.html"},{"id":96970681,"identity":"bb287f28-653d-43ad-96ef-f18462fed1a2","added_by":"auto","created_at":"2025-11-28 07:14:42","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":152340,"visible":true,"origin":"","legend":"\u003cp\u003eFeature Selection Process Using the Lasso Model\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/fe0955ba0760b531c18b6fe7.png"},{"id":97137002,"identity":"4ab63ad4-f47e-4465-a9bf-c4065229fed7","added_by":"auto","created_at":"2025-12-01 09:57:16","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":129683,"visible":true,"origin":"","legend":"\u003cp\u003eFeature Correlation Heatmap of the Final Model\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/685ce64915b113eb79bf0a8a.png"},{"id":97138738,"identity":"9b84529d-2fd8-42e1-9dfa-a7c34da61f3f","added_by":"auto","created_at":"2025-12-01 09:59:16","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":170205,"visible":true,"origin":"","legend":"\u003cp\u003eROC Curves of the Different Machine Learning Models with Confidence Intervals\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/4511be51aef1157bbc7efd2b.png"},{"id":96970680,"identity":"935e8856-6fde-443f-9d3b-97a5e43f5b5f","added_by":"auto","created_at":"2025-11-28 07:14:42","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":87930,"visible":true,"origin":"","legend":"\u003cp\u003ePrecision-Recall (PR) Curves of the Various Machine Learning Models\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/ea3ae752021b283f41c97cc1.png"},{"id":97137084,"identity":"60857eee-48a2-415d-b66a-070e23caff09","added_by":"auto","created_at":"2025-12-01 09:57:22","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":93000,"visible":true,"origin":"","legend":"\u003cp\u003eDecision Curve Analysis (DCA) of the Various Machine Learning Models\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/17034a1771c33ffda66ddb38.png"},{"id":96970687,"identity":"7d8d9433-b02d-4db5-84a3-29233b4ad38c","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":81832,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP Feature Importance Bar Plot\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/462250a16238cc4120a6e9e6.png"},{"id":96970688,"identity":"43544880-5a87-412a-b4b3-01b66b450969","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":31829,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP Beeswarm Plot\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/1927329176f4fc939cc90c0c.png"},{"id":96970691,"identity":"d6e7536e-c4bf-4af2-8c05-d3498d1f9e19","added_by":"auto","created_at":"2025-11-28 07:14:43","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":77569,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eForce Plot for Individual Prediction Interpretation\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/d22e672ae27869de973d502f.png"},{"id":97137612,"identity":"09266d17-9a2b-4a2c-961c-d00f8520b45a","added_by":"auto","created_at":"2025-12-01 09:57:59","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":114344,"visible":true,"origin":"","legend":"\u003cp\u003eDecision Plot of SHAP Values\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/f6a6cf889bc713ae07a28c54.png"},{"id":97137873,"identity":"fcf8b138-da71-4505-8f5f-e8e4f46d0ac2","added_by":"auto","created_at":"2025-12-01 09:58:17","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":138776,"visible":true,"origin":"","legend":"\u003cp\u003eMultidimensional Analysis of Peak Temperature Characteristics and Impact Across Categories\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/0ecbdf0cb9cee54f232d36d4.png"},{"id":106808739,"identity":"7e707f59-494d-41e3-8056-1aeffc95eee9","added_by":"auto","created_at":"2026-04-13 15:59:51","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2307485,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7886677/v1/abd1abea-c8fa-451c-b3d9-54dec8ae016a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Application of Explainable Machine Learning in Early Diagnosis Models for Risk Prediction of Severe Mycoplasma pneumoniae Pneumonia in Children","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eMycoplasma pneumoniae (MP) is a cell wall-deficient prokaryotic organism and a prevalent infectious pathogen\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Epidemiological studies have demonstrated that MP accounts for 40% of community-acquired pneumonia cases in children, primarily affecting school-age individuals and adolescents. In addition to pulmonary complications, MP infection can also precipitate a range of extrapulmonary manifestations, such as liver injury, encephalitis, hemolytic anemia, and thrombosis \u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. Alarmingly, the rise in antibiotic misuse has led to increased MP resistance, resulting in a growing incidence of severe Mycoplasma pneumoniae pneumonia (SMPP) in children\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. SMPP can lead to severe intra- and extrapulmonary complications such as plastic bronchitis, pleural effusion, pulmonary consolidation necrosis, pulmonary embolism, myocardial injury, abnormal liver and renal function, anemia, and encephalitis \u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. These complications can diminish children's quality of life, impose a greater family burden, and even pose life-threatening risks\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. The optimal treatment window for MPP is 5 to 10 days post-fever onset, with early recognition of SMPP being crucial for effective treatment\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e .\u003c/p\u003e\u003cp\u003eThe diagnosis of SMPP currently relies primarily on imaging techniques and clinical signs \u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. However, the clinical and imaging features of SMPP exhibit substantial heterogeneity and are highly similar to those of viral infections\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. Consequently, there are significant differences in the clinical presentation of pneumonia among children \u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e, posing a significant challenge to the timely and accurate diagnosis of SMPP and the implementation of appropriate treatment strategies\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. Comprehensive guidelines, such as CURB-65 and PSI scores, have been established in the adult field to assess pneumonia severity in response to the diverse manifestations of the disease \u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. However, a lack of objective, quantitative, economical, and convenient diagnostic systems persists for evaluating childhood pneumonia. Previous studies have collected clinical data and examined indicators to predict the progression and possible etiology of pneumonia in children. Increases in respiratory rate, heart rate, and oxygen saturation have been found to indicate the diagnosis and progression of pediatric pneumonia \u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e, and elevations in C-reactive protein and procalcitonin may be associated with bacterial infection and mortality\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. The sensitivity and specificity of these indicators are not high, and some studies have reported conflicting results \u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. Consequently, clinical practice often varies with individual doctors' experiences, leading to either overdiagnosis and overtreatment or delays in diagnosis and treatment. Thus, predicting the occurrence of SMPP in children using existing clinical data is crucial. With the rising incidence of SMPP, research increasingly focuses on its early prediction. Current predictive models primarily utilize traditional logistic regression, which frequently faces challenges with data imbalance and may not achieve the accuracy required for modern diagnostic and therapeutic standards. No comprehensive model yet systematically integrates clinical features to effectively quantify the early predictive power of SMPP in children \u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn recent years, machine learning has increasingly supplanted traditional statistical models in developing clinical diagnostic and predictive models, enhancing diagnostic and prognostic accuracy \u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. Machine learning (ML) employs sophisticated algorithms and statistical methods to predict disease progression, adverse outcomes, and treatment efficacy with precision \u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. It excels in data mining and classification, finding extensive applications across various medical domains\u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. However, ML algorithms often prioritize statistical features over clinically significant variables \u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. Shapley Additive Explanations (SHAP) is a prominent post-hoc interpretability algorithm that has gained significant attention in the field of ML \u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. Grounded in game theory's Shapley values, SHAP quantifies the impact of each feature on the model's output by calculating the average marginal contribution of features across all possible subsets. This approach not only ensures fairness and consistency of interpretation but also reveals the intricate interactions between features and their underlying mechanisms influencing the model's predictions, thereby providing reliable theoretical explanations for the model's behavior. The growing importance of ML interpretability can be attributed to the inherent \"black box\" nature of traditional ML models, underscoring the need for transparent and explainable AI systems \u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. Therefore, this study aims to construct a ML model for predicting the risk of SMPP, and adopt the SHAP method to interpret and visualize the model, thereby assisting clinicians in accurately identifying children at high risk of SMPP.\u003c/p\u003e"},{"header":"2 Subjects and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Study Participants\u003c/h2\u003e\u003cp\u003eA total of 350 pediatric inpatients with Mycoplasma pneumoniae pneumonia (MPP) were initially enrolled in this study at the Affiliated Hospital of Yan'an University between August 2023 and August 2024. All participants underwent bronchoscopic alveolar lavage (BAL). Following the application of the predetermined inclusion and exclusion criteria and the removal of cases with extreme or aberrant values, 286 MPP patients were ultimately included in the final analysis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 SMPP diagnoses and inclusion and exclusion criteria\u003c/h2\u003e\u003cdiv id=\"Sec5\" class=\"Section3\"\u003e\u003ch2\u003e2.2.1 Diagnostic criteria\u003c/h2\u003e\u003cp\u003e\u003cb\u003eMPP diagnostic criteria\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eAccording to the 2023 edition of the Guidelines for Diagnosis and Treatment of Mycoplasma Pneumonia in Children, the diagnostic criteria include\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e :\u003c/p\u003e\u003cp\u003e(1) respiratory symptoms such as fever and cough, along with lung auscultation findings of dry and moist rales; (2) imaging results indicative of pneumonia.\u003c/p\u003e\u003cp\u003eTo confirm the diagnosis, these clinical and imaging criteria must be met, along with at least one of the following conditions: ①a single serum MP antibody titer of 1:160 or higher, or a fourfold or greater increase in the double serum MP antibody titer during the disease course; ②a positive result for MP DNA or RNA testing.\u003c/p\u003e\u003cp\u003e\u003cb\u003eDiagnostic criteria for SMPP\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eBased on confirmed MPP diagnosis, the criteria for severe pneumonia were established according to the \"Guidelines for the Diagnosis and Management of Community-Acquired Pneumonia in Children (2019 Edition)\"\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e, including any one of the following manifestations: ①poor general condition; ②impaired consciousness, cyanosis, tachypnea (respiratory rate\u0026thinsp;\u0026ge;\u0026thinsp;70 breaths/min in infants, or \u0026ge;\u0026thinsp;50 breaths/min in children over 1 year old); ③signs of respiratory distress (grunting, nasal flaring/subcostal retractions), intermittent apnea, or oxygen saturation\u0026thinsp;\u0026lt;\u0026thinsp;92%; ④ extreme fever or persistent high fever beyond 5 days; ⑤signs of dehydration or refusal to feed; ⑥chest CT revealing\u0026thinsp;\u0026ge;\u0026thinsp;2/3 lung infiltration in one lung, multi-lobar infiltration, pleural effusion, pneumothorax, atelectasis, pulmonary necrosis, or lung abscess; ⑦presence of extrapulmonary complications.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.2.2 Inclusion criteria\u003c/h2\u003e\u003cp\u003e①Age\u0026thinsp;\u0026le;\u0026thinsp;14 years; ②Meeting the diagnostic criteria for MPP \u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e; ③Fulfilling the indications for bronchoscopy \u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e; ④Availability of complete clinical data.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e2.2.3 Exclusion criteria\u003c/h2\u003e\u003cp\u003e①absence of essential clinical data; ②presence of congenital diseases (e.g., congenital heart disease, Down syndrome); ③hematological diseases; ④concurrent infections or trauma at other sites; ⑤primary or acquired immunodeficiency; and ⑥severe pneumonia induced by bronchial foreign body aspiration.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Ethics approval\u003c/h2\u003e\u003cp\u003e This study was approved by the Ethics Committee of the Affiliated Hospital of Yan'an University (Approval No.IIT-R-20250178).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Study Group Allocation and Data Collection\u003c/h2\u003e\u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\u003ch2\u003e2.4.1 Participant Assignment\u003c/h2\u003e\u003cp\u003eBased on disease severity, pediatric patients with pneumonia were divided into the MPP group (n\u0026thinsp;=\u0026thinsp;163) and the SMPP group (n\u0026thinsp;=\u0026thinsp;123).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\u003ch2\u003e2.4.2 Data Collection\u003c/h2\u003e\u003cp\u003eClinical information was collected from paediatric medical records in medical records browsing system of Affiliated Hospital of Yan 'an University. Specific records included: ① Basic information: age, sex, BMI, length of stay, cost, single infection, mixed infection.② Symptoms\u0026thinsp;+\u0026thinsp;signs: respiratory symptoms (fever, heat spike, cough, expectoration, wheezing, shortness of breath); physical signs (wheezing, phlegm, moist rales, dyspnea, respiratory failure, etc.);③ extrapulmonary complications (Digestive system, cardiovascular system, nervous system);④ Laboratory examination: White blood cell count [WBC (10e9/L)], neutrophil percentage (NEU %), lymphocyte percentage (LYM %), hemoglobin (HB), platelet count [PLT (10e9/L)], C-reactive protein (CRP), procalcitonin (PCT), erythrocyte sedimentation rate (ESR), albumin (ALB), lactate dehydrogenase (LDH), alpha hydroxybutyrate dehydrogenase (HBDH), fibrinogen (FIB), fibrinogen degradation product (FDP), Plasma D-dimer, etc.; ⑤Imaging findings: Chest CT findings.\u003c/p\u003e\u003cp\u003eFasting venous blood samples were obtained within 24 hours post-admission for analysis. Chest CT scans were conducted either three days before admission or two days following admission, with results duly documented.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Development and Internal Validation of the Prediction Model\u003c/h2\u003e\u003cdiv id=\"Sec13\" class=\"Section3\"\u003e\u003ch2\u003e2.5.1 Evaluating the Stability of the Prediction Model\u003c/h2\u003e\u003cp\u003eThe dataset was randomly partitioned into training and test subsets using the `train_test_split` function from the `sklearn.model_selection` module in Python 3.9.6, with a 70:30 ratio. To ensure reproducibility, the random seed was set to 42.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\u003ch2\u003e2.5.2 Feature Selection\u003c/h2\u003e\u003cp\u003eThe study initially assessed 44 clinical characteristics, utilizing a sequential screening approach with Pearson correlation, Mann-Whitney U test, chi-square test, and Least Absolute Shrinkage and Selection Operator (LASSO) regression to identify the most pertinent variables.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section3\"\u003e\u003ch2\u003e2.5.3 Prediction Model Selection and Evaluation\u003c/h2\u003e\u003cp\u003eBased on SMPP risk characteristic variables, seven machine learning models were developed using Python 3.9.6, including Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Logistic Regression (LR), and Light Gradient Boosting Machine (LightGBM). Hyperparameters were optimized through grid search and random search, with the optimal parameter combinations selected via 5-fold cross-validation, using accuracy as the primary evaluation metric. After training, the best-performing model was applied to the test set for prediction. Model performance was comprehensively evaluated using accuracy, classification reports (including precision, recall, and F1-score), and confusion matrices. To enable intuitive comparison of model performance, receiver operating characteristic (ROC) curves, precision-recall curves, and decision curve analysis plots were generated for all models. The area under the curve (AUC) was used to quantify model discriminative ability, ultimately establishing a comprehensive model performance evaluation framework.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e2.5.4 Model Explainability\u003c/h2\u003e\u003cp\u003eMachine learning-based predictive models can be interpreted using Shapley Additive Explanations (SHAP). SHAP is constructed based on insights from cooperative game theory, where all input features are regarded as \"contributors.\" For each prediction instance, the model generates an output, and the SHAP value represents the numerical value assigned to each feature within that instance. Essentially, SHAP attributes the output value to the Shapley value of each feature\u0026mdash;that is, it computes the SHAP value for every feature to quantify its influence on the final prediction. This approach enables precise measurement of the contribution and impact of each feature on the model\u0026rsquo;s output, reflecting not only the magnitude of each feature\u0026rsquo;s influence in every sample but also the direction of that influence (positive or negative). Furthermore, SHAP provides highly powerful data visualization capabilities to facilitate the interpretation of both the overall model and individual predictions.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Statistical Analysis\u003c/h2\u003e\u003cp\u003eStatistical analyses in this study were performed using SPSS 26.0. Normality of quantitative data was assessed using skewness and kurtosis tests, and homogeneity of variance was evaluated with Levene's test. Normally distributed continuous data are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (x̅ \u0026plusmn; s). For data with homogeneous variance, inter-group comparisons were conducted using the t-test; otherwise, the t'-test was applied. Non-normally distributed data are expressed as median (interquartile range) [M (Q1, Q3)], and the Wilcoxon rank-sum test was used for inter-group comparisons. Categorical data are summarized as frequency and proportion n (%). Fisher's exact test was employed for inter-group comparisons when at least 20% of expected frequencies were less than 5; otherwise, the Pearson chi-square test was applied.\u003c/p\u003e\u003cp\u003eMachine learning models were developed using Python 3.9.6 in a Jupyter Notebook environment, leveraging key libraries including NumPy, pandas, matplotlib.pyplot, seaborn, and openpyxl. The SHAP algorithm was implemented to visualize model features and assess their contribution to the model predictions.\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Results","content":"\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Comparison of Clinical Characteristics Between the MPP and SMPP Groups\u003c/h2\u003e\u003cp\u003eAmong 286 children infected with Mycoplasma pneumoniae, 163 had MPP and 123 had SMPP. A comparison between the MPP and SMPP groups revealed statistically significant differences (P\u0026thinsp;\u0026lt;\u0026thinsp;.05) in hospital days, costs, and infections (single/mixed), as well as in fever duration, peak temperature, wheezing, tachypnea, respiratory failure, sputum characteristics, and impacts on the digestive and cardiovascular systems. Significant differences were also noted in WBC, NEU, LYM, HB, CRP, PCT, ESR, ALB, LDH, HBDH, FDP, D-dimer, pulmonary consolidation, atelectasis, pleural effusion, lesion scope (patchy or floccular opacities/large shadow), pebble mucosa, plastic sputum plug, mucosal erosion, and inflammatory stricture. See Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e for details.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparison of Clinical Characteristics Between Children in the MPP and SMPP Groups [n (%),\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\text{x}}\\pm\\:\\text{s}\\)\u003c/span\u003e\u003c/span\u003e,p50 (p25,p75)]\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMPP(n\u0026thinsp;=\u0026thinsp;163)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSMPP(n\u0026thinsp;=\u0026thinsp;123)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eStatistic(t/χ2/Z)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.74\u0026thinsp;\u0026plusmn;\u0026thinsp;2.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.49\u0026thinsp;\u0026plusmn;\u0026thinsp;2.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.862\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.389\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.048\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.826\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBoys\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e79(48.47)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e58(47.15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGirls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e84(51.53)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e65(52.85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15.45(14.37, 17.56)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15.97(14.79, 17.36)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-1.323\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.186\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ehospital days\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e9.00(8.00, 10.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10.00(9.00, 12.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-4.577\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ecost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8181.55(7179.06, 9441.93)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10097.64(8668.75, 12283.76)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-7.473\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInfection\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e16.286\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSingle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42(25.77)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9(7.32)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMixed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e121(74.23)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e114(92.68)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003efever days(d)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.00(2.00, 6.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.00(6.00, 9.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-9.759\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003epeak temperature\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e38.90(38.30, 39.30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e39.80(39.38, 40.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-10.757\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCough\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e162(99.39)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e122(99.19)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eExpectoration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e162(99.39)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e123(100)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWheezing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2(1.23)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7(5.70)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.042\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTachypnea\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1(0.61)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9(7.32)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDyspnea\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1(0.61)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5(4.07)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.088\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRespiratory Failure\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0(0.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4(3.25)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.033\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWheeze Sounds\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2(1.23)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5(4.07)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.144\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003esputum sound\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e114(69.94)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e103(83.74)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e7.294\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.007\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMoist Rales\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e160(98.16)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e123(100)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.262\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003edigestive system\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21(12.88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e57(46.34)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e39.564\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCardiovascular System\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0(0.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6(4.88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.006\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNervous System\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0(0.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1(0.81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWBC (10e9/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7.26(5.93, 9.19)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.86(6.52, 12.33)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-4.004\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNEU (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e60.20(52.30, 68.50)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66.10(57.00, 77.90)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-3.904\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLYM (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e28.20(20.50, 37.60)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23.60(15.69, 32.40)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-3.601\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHB (g/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e127.00(121.00, 134.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e126.00(119.00, 130.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-2.348\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.019\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePLT (10e9/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e273.00(236.00, 332.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e288.00(235.00, 362.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.995\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.32\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRP (mg/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10.00(7.11, 15.91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23.33(12.12, 50.56)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-6.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePCT (ng/ml)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.11(0.05, 0.29)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.26(0.12, 0.93)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-4.669\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eESR (mm/h)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e19.00(12.00, 37.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e39.00(23.00, 57.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-5.234\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eALB (g/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e41.60(39.60, 43.90)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e39.90(36.50, 42.60)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-4.429\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLDH (U/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e279.00(249.00, 330.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e328.00(267.00, 426.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-4.187\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHBDH (U/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e243.00(210.00, 288.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e284.00(226.00, 358.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-3.447\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFIB (g/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.29(3.77, 4.81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.40(3.83, 4.88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-1.111\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.266\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFDP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.50(2.50, 2.50)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.60(2.50, 4.50)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-5.925\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eD-dimer (mg/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.38(0.22, 0.64)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.37(0.51, 3.44)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-7.826\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePulmonary Consolidation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44(26.99)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e59(47.97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e13.382\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAtelectasis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0(0.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9(7.32)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e/\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003epleural effusion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0(0.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35(28.46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e52.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLesion Scope\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e40.896\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePatchy or Flocculent Opacities\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e130(79.75)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e53(43.09)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003elarge shadow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e33(20.25)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e70(56.91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ecobblestone mucosa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21(12.88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37(30.08)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e12.825\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eplastic sputum plug\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e24(14.72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e48(39.02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e21.976\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emucosal erosion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7(4.29)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26(21.14)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e19.485\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003einflammatory stricture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8(4.91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21(17.07)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e11.386\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Results of Feature Selection\u003c/h2\u003e\u003cp\u003eThe present study examined 44 initial features, with categorical variables processed using one-hot encoding. One-hot encoding is a widely-adopted technique for converting class features into a format suitable for machine learning and deep learning algorithms. This approach enables computers to better comprehend class features, thereby enhancing learning and predictive capabilities, and ultimately improving the performance of machine learning and deep learning models.\u003c/p\u003e\u003cp\u003eThe aforementioned 44 clinical features were sequentially subjected to the following three distinct methods for further screening:\u003c/p\u003e\u003cp\u003e3.2.1 To eliminate features exhibiting high inter-feature correlations, the Pearson correlation coefficient was employed. This statistic, also known as the Pearson product-moment correlation coefficient, quantifies the linear relationship between two variables, providing values from \u0026minus;\u0026thinsp;1 to 1. A value of 1 indicates perfect positive correlation, -1 signifies perfect negative correlation, and denotes no correlation. Consequently, features with correlation coefficients |r| \u0026ge; .85 were discarded, retaining only those with |r| \u0026lt; .85. Following this criterion, 40 features were preserved after the removal of 4 features: NEU, patchy or cloudy shadows, LDH, and single infection.\u003c/p\u003e\u003cp\u003e3.2.2 To ensure the significance of the features, the Mann-Whitney U test and chi-square test were employed to identify variables associated with the outcome. Features with P-values greater than .05 were excluded, resulting in the elimination of 13 features. Consequently, 27 features were retained for further screening. These included: hospital stay, cost, fever days, heat peak, WBC, LYM, HB, CRP, PCT, ESR, ALB, HBDH, FDP, D-dimer, shortness of breath, dyspnea, sputum sound, digestive system, mucosal cobblestone protrusion, molded sputum thrombus, mucosal erosion, inflammatory stenosis, mixed infection, lung consolidation, atelectasis, pleural effusion, and large patchy shadow.\u003c/p\u003e\u003cp\u003e3.2.3 The LASSO regression technique was employed to screen features and ensure the stability and efficacy of the model. This method generates a penalty function λ to compress the variable coefficients in the regression model, thereby preventing overfitting and addressing the issue of severe collinearity, which is a widely used approach in predictive modeling. The LASSO regression was applied to select valuable feature sets from the 10-fold cross-validation results. As the penalty parameter λ was gradually increased from 10^-6 to 10^2, the number of variables included in the model decreased. When λ was set to 0.007565, the LASSO regression model exhibited the best predictive performance, eliminating 6 features and retaining 21 features with non-zero coefficients in the final model. The final set of retained features included days in hospital, days in fever, thermal peak, WBC, LYM, HB, CRP, PCT, ESR, ALB, FDP, D-dimer, sputum sound, digestive system, mucosal cobblestone ridge, molded sputum plug, mucosal erosion, inflammatory stenosis, mixed infection, pleural effusion, and large patchy shadow (Fig.\u0026nbsp;1).\u003c/p\u003e\u003cp\u003eFigure 1 Feature Selection Process Using the Lasso Model\u003c/p\u003e\u003cp\u003eFigure 1 depicts the changes in various metrics during LASSO regression. Panel a shows the mean squared error (MSE) as a function of the regularization parameter λ, where λ is plotted on the x-axis and MSE on the y-axis. Panel b shares the same x-axis as panel a, but the y-axis represents the coefficients of the model variables. This panel illustrates the relative importance of the variables. As λ increases, the coefficients of less important variables are driven to zero, while the coefficients of more important variables become less sensitive to the penalty imposed by the λ parameter, allowing them to be retained in the final model.\u003c/p\u003e\u003cp\u003eThe correlation matrix presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the relationships between the selected features of the predictive model. Negative correlations are depicted in blue, while positive correlations are shown in red. The intensity of the color corresponds to the strength of the correlation, with darker shades indicating stronger associations.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Results of the Machine Learning-Based Prediction Model\u003c/h2\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e outlines the performance metrics\u0026mdash;accuracy, precision, recall, F1 score, and AUC\u0026mdash;of seven models (KNN, LightGBM, XGBoost, Catboost, SVM, LR, and GNB) on both training and test datasets.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerformance of the Machine Learning Prediction Model\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"7\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eF1 Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eAUC ROC\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTraining Set\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.985\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.988\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.976\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.982\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.999\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCatBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.950\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.974\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.905\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.938\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.993\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.905\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.971\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.798\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.876\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.978\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.905\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.901\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.869\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.885\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.962\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGNB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.820\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.914\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.631\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.746\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.934\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTest Set\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCatBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.878\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.900\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.961\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.884\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.854\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.897\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.875\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.940\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.884\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.854\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.897\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.875\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.945\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.884\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.837\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.878\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.961\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.872\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.897\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.864\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.963\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.849\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.882\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.769\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.822\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.899\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGNB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.814\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.829\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.744\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.784\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.907\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003e ROC Curves of the Different Machine Learning Models with Confidence Intervals\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003e, and Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e, among the seven machine learning models evaluated, the CatBoost model demonstrated optimal overall performance, exhibiting particularly stable and outstanding results in the test set. Its test set AUC reached .961, performing at a similarly high level as LightGBM (.961) and LR (.963), and was slightly superior to XGBoost (.945), SVM (.940), GNB (.907), and KNN (.899). Furthermore, CatBoost achieved a test set Accuracy of .907, exceeding all other models (the second highest being .884). It also attained a Recall of .923, tying with LightGBM for the highest score in the test set and significantly outperforming models such as KNN (.769) and GNB (.744). Its Precision (.878) and F1 Score (.900) also ranked among the top performers in the test set, only slightly lower than its own best metric values, indicating well-balanced and excellent performance across all evaluation metrics.\u003c/p\u003e\u003cp\u003eIn the training set, SVM and LightGBM achieve the highest metrics (Accuracy, Precision, Recall, F1 Score, AUC), all at 1.000, indicating potential overfitting. In contrast, their test set AUCs (SVM .940, LightGBM .961) and Accuracies (both .884) are notably lower, highlighting insufficient generalization stability. XGBoost ranks just below SVM and LightGBM in the training set (AUC .999, Accuracy .985), but its test set AUC (.945) and Accuracy (.884) underperform compared to CatBoost, reflecting slightly inferior overall performance.The performance of the KNN, LR, and GNB models was relatively consistent across the training and test sets. While the LR model exhibited a slightly higher AUC (0.963) on the test set compared to CatBoost, its Accuracy (0.872) and Precision (0.833) were lower. Additionally, the AUC (0.962) of the LR model's training set was similar to its test set performance, indicating a balanced but not optimal overall performance. The KNN and GNB models were among the lower-performing models in the set of seven. The KNN model's Recall (0.769) on the test set and the GNB model's Accuracy (0.814) on the test set were relatively weak compared to the other models.\u003c/p\u003e\u003cp\u003eThe CatBoost model demonstrates superior stability compared to other models, with minimal differences in key evaluation metrics: an AUC difference of .032 (.993 for the training set and .961 for the test set) and an accuracy difference of .043 (.950 for the training set and .907 for the test set). These differences are notably smaller than those observed in SVM (AUC difference of .060, accuracy difference of .116) and LightGBM (AUC difference of .039, accuracy difference of .116). This stability indicates that CatBoost maintains reliable predictive capabilities despite changes in data distribution. Consequently, CatBoost is selected as the final prediction model due to its stable and accurate performance, effectively balancing training set performance, test set generalization, and metric consistency.\u003c/p\u003e\u003cp\u003eFigure 5 Decision Curve Analysis (DCA) of the Various Machine Learning Models\u003c/p\u003e\u003cp\u003eDecision curve analysis (DCA) was conducted on seven machine learning models in the training and test sets to evaluate the net clinical benefit of each model in informing clinical decision-making. DCA quantified the net benefit (i.e., the minimum acceptable probability of requiring further intervention) by contrasting the intervention strategy for each model with the default strategies of \"Treat All\" and \"Treat None\"\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e .\u003c/p\u003e\u003cp\u003eThe decision curves depicted in Fig.\u0026nbsp;5 illustrate the net benefit of the model across a range of decision thresholds, for both the training and test datasets. The dotted line representing \"Intervention for all patients\" and the solid black line denoting \"Intervention for no patients\" serve as reference strategies for comparison. The leftward positioning of the training set decision curve relative to the test set curve suggests the model may be overfit to the training data.\u003c/p\u003e\u003cp\u003eIn the training set, CatBoost, LightGBM, and similar models consistently outperform K-nearest neighbor and Gaussian naive Bayes, as their curves remain within the broad decision threshold and yield net returns surpassing the reference lines. This suggests their greater utility in aiding clinical decision-making during training. On the test set, the CatBoost model maintains a higher net benefit across a broader threshold range, significantly outperforming the two default strategies, thus demonstrating its effectiveness in clinical decision support post-generalization.\u003c/p\u003e\u003cp\u003eThe CatBoost model demonstrated robust decision-benefit performance across the training and test datasets, suggesting its potential to provide effective support for clinical decision-making. By judiciously setting the decision threshold, this model can facilitate the achievement of improved net benefit.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Model Explainability\u003c/h2\u003e\u003cp\u003eThe feature importance analysis of the optimal CatBoost model, depicted in Fig.\u0026nbsp;6, reveals the relative significance of each predictor variable. The feature importance is quantified as the mean absolute value of a feature's influence on the target variable. The results indicate that the thermal peak has the highest predictive power across all prediction ranges, followed by thermal range, D-dimer, and CRP. To further elucidate the nature of these relationships, SHAP values are employed to identify both positive and negative associations between the predictors and the target outcome.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure 6 SHAP Feature Importance Bar Plot Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e7\u003c/span\u003e SHAP Beeswarm Plot\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e8\u003c/span\u003e presents a SHAP waterfall plot that clearly deconstructs the prediction logic for a high-risk severe case (actual severity: severe, model-predicted probability: .932). Starting from the baseline value, E[f(x)]\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;.474, which represents the model's average prediction over the dataset, features such as peak body temperature (+\u0026thinsp;.95) and duration of fever (+\u0026thinsp;.81), represented by red bars, significantly drove the prediction probability toward the severe outcome. In contrast, features like WBC and FDP, indicated by blue bars, exerted weaker negative effects on the prediction of severity. An additional 12 features provided minor positive contributions. The cumulative effect of these multi-feature interactions elevated the model's output value, ultimately resulting in a high-probability severe prediction. This visualization intuitively demonstrates the contribution of each feature to the final prediction outcome.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e9\u003c/span\u003e displays a SHAP decision plot, which reveals the model's decision-making process in distinguishing SMPP from non-SMPP cases by stacking the prediction paths of individual samples. Each line represents a patient case, originating from the baseline value at the bottom and extending to the model's output value on the x-axis, thereby clearly demonstrating the cumulative contribution of different features to the prediction outcome for each case. As illustrated, while \"peak body temperature\" and \"D-dimer\" serve as universally important positive drivers, the specific combination and sequence of features influencing the final prediction vary substantially across different samples. This visually demonstrates that the CatBoost model employs a highly individualized and flexible decision logic when synthesizing multiple indicators for judgment, adapting to the distinct clinical manifestations presented by different cases.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eThis study developed and internally validated an interpretable machine learning model for early prediction of SMPP. Employing the SHAP method, we elucidated the relative contribution of each feature to the predictive model, thereby enhancing its transparency and clinical reliability.\u003c/p\u003e\u003cp\u003eTraditional logistic regression models, commonly used for SMPP risk prediction\u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e, struggle with complex relationships and collinearity issues, potentially compromising prediction accuracy and stability. In contrast, machine learning models excel at capturing nonlinear relationships and interactions within data, thereby enhancing prediction accuracy and robustness\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. Catia et al. \u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e employed machine learning to predict mortality in patients with community-acquired pneumonia, facilitating early detection and intervention for high-risk individuals. Similarly, Zhang et al.\u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e developed an XGBoost model to predict volume responsiveness in oliguric ICU patients with acute kidney injury, identifying urinary creatinine, serum urea nitrogen, and age as significant factors. In our study, we integrated clinical features, laboratory results, and imaging data, demonstrating that CatBoost achieved outstanding predictive performance, with an AUC of .961 and an accuracy of .907 in the test set. CatBoost\u003csup\u003e[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e is a gradient boosting framework utilizing decision trees, renowned for its adept handling of categorical variables compared to conventional decision tree models. It computes target statistics for classification features to mitigate distribution bias between training and test data, enhancing model accuracy and generalization. CatBoost delivers impressive predictive outcomes without extensive parameter tuning, minimizing overtuning risk and fostering robust model development. Recent evidence highlights CatBoost's superior accuracy in various medical prediction domains, including acute pancreatitis \u003csup\u003e[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e and cancer prognosis \u003csup\u003e[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]\u003c/sup\u003e. Our study corroborates CatBoost's exceptional discrimination capabilities over traditional linear methods in children's SMPP, a complex task marked by heterogeneous and nonlinear clinical manifestations, thus supporting early clinical precision interventions.\u003c/p\u003e\u003cp\u003eThe ability of traditional ML models to handle large-scale, high-dimensional data is well-established. However, these models often lack the capacity to elucidate the dose-response relationship between individual feature variables and the prediction outcome. Additionally, interpreting the inner workings of ML models and effectively communicating their predictions to clinicians have posed significant challenges \u003csup\u003e[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]\u003c/sup\u003e. The visualization of SHAP, through techniques such as bar charts and summary graphs, offers a means to quantify and graphically depict the relationship between clinical variables and disease risk \u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. The SHAP method is employed to interpret and visualize the CatBoost model, using bar and summary charts to break down a single sample's prediction into feature contribution values, thereby elucidating the model's decision process \u003csup\u003e[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]\u003c/sup\u003e. Several clinical factors potentially linked to SMPP development in children were identified: peak body temperature, temperature range, D-dimer levels, CRP levels, and pleural effusion. Elevated temperature and prolonged fever duration are risk factors for SMPP, possibly due to severe infection leading to intense lung inflammation, resulting in sustained high fever. Without timely intervention, this condition may progress to SMPP \u003csup\u003e[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]\u0026minus;[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]\u003c/sup\u003e. CRP, an acute-phase protein, serves as a prevalent inflammatory marker. Its levels rise with inflammation \u003csup\u003e[\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]\u003c/sup\u003e. A retrospective cohort study involving 2,377 COVID-19 patients at New York University Langone Health Center indicated that those with elevated baseline D-dimer were more prone to critical illness compared to those with normal levels (43.9% vs. 18.5%) \u003csup\u003e[\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]\u003c/sup\u003e. Research by Megan Carolina Cerda-Mancillas et al. \u003csup\u003e[\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]\u003c/sup\u003e demonstrated a correlation between D-dimer plasma levels and SMPP severity. Pulmonary inflammation can lead to fluid exudation and pleural effusion, exacerbating dyspnea and potentially spreading inflammation, thereby contributing to SMPP. Radiologically, lower lobe infections are more frequently associated with SMPP than infections in other lung regions \u003csup\u003e[\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]\u003c/sup\u003e. The model notably attributes predictive significance to bronchoscopic features such as \"mucosal cobblestones\" and \"shaped sputum plugs,\" representing a pivotal advancement in this research. These features serve as \"end-organ evidence\" of severe airway inflammation and epithelial damage induced by Mycoplasma pneumonia. As noted by C. Tran et al. \u003csup\u003e[\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]\u003c/sup\u003e, airway mucus embolism primarily results from increased bronchial secretions, which are difficult to expel, leading to bronchial obstruction and compromised respiration. The blockage caused by mucus plugs can result in symptoms such as dyspnea, cough, wheezing, and chest discomfort, potentially escalating to oxygen deficiency, cyanosis, respiratory failure, and other critical conditions. This study objectively and quantitatively establishes the independent predictive value of these microscopic features in assessing the early risk of SMPP using interpretable AI techniques. This approach not only offers clinicians a more robust decision-making framework beyond conventional laboratory parameters but also underscores the critical importance of early bronchoscopy in risk assessment and the exploration of underlying mechanisms in children who exhibit poor responses to initial treatment.\u003c/p\u003e\u003cp\u003eThis study offers several advantages: Firstly, it incorporates a multidimensional set of clinical characteristics, including manifestations, laboratory results, and imaging tests, which are commonly accessible in medical settings, thereby enhancing the predictive model's generalizability. Secondly, the model employs a machine learning algorithm, which is better suited than traditional logistic regression for handling complex, nonlinear data. Given that diseases often arise from multiple factors, machine learning algorithms are more effective for early prediction of SMPP, delivering precise results. Additionally, while transparency and interpretability are crucial in clinical practice, machine learning models often struggle with these aspects. This study addresses interpretability by utilizing SHAP, an interpretability tool in artificial intelligence. In summary, the model effectively captures complex data and provides interpretable outputs, offering valuable support for the early clinical identification of SMPP.\u003c/p\u003e\u003cp\u003eThis study is subject to several limitations. The model was developed retrospectively using single-center data, introducing inherent biases in the data collection process. Additionally, model validation was conducted solely through internal methods, limiting the assessment of its broader applicability. To enhance the model's universality, future research should incorporate external datasets for validation. Importantly, the proposed model lacks prospective validation, which will be a focus of forthcoming investigations.\u003c/p\u003e"},{"header":"5 Conclusions","content":"\u003cp\u003eThe present study utilized machine learning techniques to develop seven predictive models based on key feature variables. Of these, the CatBoost model demonstrated robust predictive performance through internal validation. To corroborate these findings, the authors recommend conducting multicenter validation and large-scale prospective investigations.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eMP\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMycoplasma pneumoniae\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eMPP\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMycoplasma pneumoniae pneumonia\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eSMPP\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003esevere Mycoplasma pneumoniae pneumonia\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eML\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMachine learning\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eSHAP\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eShapley Additive Explanations\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eBAL\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ebronchoscopic alveolar lavage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eLASSO\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLeast Absolute Shrinkage and Selection Operator\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eWBC\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eWhite blood cell count\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eNEU %\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLaboratory tests included neutrophil percentage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eLYM %\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003elymphocyte percentage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eHB\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ehemoglobin\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003ePLT\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eplatelet count\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eCRP\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eC-reactive protein\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003ePCT\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eprocalcitonin\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eESR\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eerythrocyte sedimentation rate\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eALB\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ealbumin\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eLDH\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003elactate dehydrogenase\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eHBDH\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ealpha hydroxybutyrate dehydrogenase\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eFIB\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003efibrinogen\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eFDP\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003efibrin (pro) degradation product\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eCatBoost\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eCategorical Boosting\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eXGBoost\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eExtreme Gradient Boosting\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eGNB\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eGaussian Naive Bayes\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eSVM\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSupport Vector Machine\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eKNN\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eK-Nearest Neighbors\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eLR\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLogistic Regression\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eLightGBM\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLight Gradient Boosting Machine\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003ePR\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ePrecision-Recall\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u003cb\u003eDCA\u003c/b\u003e\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDecision Curve Analysis\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study involved the analysis of medical records from 286 hospitalized children diagnosed with pneumonia at the Affiliated Hospital of Yan'an University. Data collection was conducted non-invasively through the retrieval of anonymized routine medical records from the hospital archives. Due to the retrospective and non-invasive nature of the study, along with the de-identification of the data utilized, the research team sought and obtained a waiver of informed consent from the Ethics Committee of Yan'an University Hospital, China. The waiver application included a detailed rationale for the exemption. The Ethics Committee rigorously evaluated the study design, data handling procedures, and grounds for exemption in line with the principles of the Declaration of Helsinki. Subsequently, the committee formally granted approval for the study (No.IIT-R-20250178) to ensure adherence to ethical standards and regulatory protocols.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo funding required\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthors and Affiliations\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAffiliated Hospital of Yan’an University, Yan’an, Shaanxi, China\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eShuai Yu, Yaya Ren, Jiangang Song, Yuxin Zhu, Hua Jiang, Yuanxia Li\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYS：Writing – original draft, Writing – review \u0026amp; editing. RYY: Writing – original draft, Writing – review \u0026amp; editing. SJG: Writing – review \u0026amp; editing. ZYX: Writing – review \u0026amp; editing. JH：Writing – review \u0026amp; editing. LYX: Writing – review \u0026amp; editing. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCorresponding author\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCorrespondence to Yuanxia Li.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.CRediT roles\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eDing G, Zhang X, Vinturache A, van Rossum AMC, Yin Y, Zhang Y. Challenges in the treatment of pediatric Mycoplasma pneumoniae pneumonia. Eur J Pediatr. 2024;183(7):3001\u0026ndash;3011. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00431-024-05519-1\u003c/span\u003e\u003cspan address=\"10.1007/s00431-024-05519-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2024 Apr 18. PMID: 38634891.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi YT, Zhang J, Wang MZ, Ma YM, Zhi K, Dai FL,et al. Changes in coagulation markers in children with Mycoplasma pneumoniae pneumonia and their predictive value for Mycoplasma severity. Ital J Pediatr. 2023;49(1):143. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13052-023-01545-1\u003c/span\u003e\u003cspan address=\"10.1186/s13052-023-01545-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 37858230; PMCID: PMC10588045.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFang C, Mao Y, Jiang M, Yin W. Pediatric Critical Illness Score, Clinical Characteristics and Comprehensive Treatment of Children with Severe Mycoplasma Pneumoniae Pneumonia. Front Surg. 2022;9:897550. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fsurg.2022.897550\u003c/span\u003e\u003cspan address=\"10.3389/fsurg.2022.897550\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 35693303; PMCID: PMC9174934.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\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. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-58784-3\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-58784-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 38589453; PMCID: PMC11002011.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDing X, Li S, Liu X, Zhan X, Wang Z, Wang M,et al.Study on the Correlation Between the Expression of NF-Ƙb in the Alveolar Lavage Fluid of Children with Severe Mycoplasma Pneumoniae Pneumonia, Its Clinical Characteristics, and Cellular Immunity. Infect Drug Resist. 2024;17:2469\u0026ndash;2484. doi: 10.2147/IDR.S411361. PMID: 38915319; PMCID: PMC11194830.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan C, Xue GH, Zhao HQ, Feng YL, Cui JH, Yuan J. Current status of Mycoplasma pneumoniae infection in China. World J Pediatr. 2024;20(1):1\u0026ndash;4. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12519-023-00783-x\u003c/span\u003e\u003cspan address=\"10.1007/s12519-023-00783-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2024 Jan 8. PMID: 38185707; PMCID: PMC10827902.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKant R, Kumar N, Malik YS, Everett D, Saluja D, Launey T, et al. Critical insights from recent outbreaks of Mycoplasma pneumoniae: decoding the challenges and effective interventions strategies. Int J Infect Dis. 2024;147:107200. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijid.2024.107200\u003c/span\u003e\u003cspan address=\"10.1016/j.ijid.2024.107200\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2024 Aug 6. PMID: 39117175.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGao L, Sun Y. Laboratory diagnosis and treatment of Mycoplasma pneumoniae infection in children: a review. Ann Med. 2024;56(1):2386636. doi: 10.1080/07853890.2024.2386636. Epub 2024 Aug 3. PMID: 39097794; PMCID: PMC11299444.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao MC, Wang L, Qiu FZ, Zhao L, Guo WW, Yang S,et al.Impact and clinical profiles of Mycoplasma pneumoniae co-detection in childhood community-acquired pneumonia. BMC Infect Dis. 2019;19(1):835. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12879-019-4426-0\u003c/span\u003e\u003cspan address=\"10.1186/s12879-019-4426-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 31601192; PMCID: PMC6788033.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDean P, Schumacher D, Florin TA. Defining Pneumonia Severity in Children: A Delphi Study. Pediatr Emerg Care. 2021;37(12):e1482-e1490. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/PEC.0000000000002088\u003c/span\u003e\u003cspan address=\"10.1097/PEC.0000000000002088\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 32205802; PMCID: PMC8223262.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMetlay JP, Waterer GW, Long AC, Anzueto A, Brozek J, Crothers K, et al.Diagnosis and Treatment of Adults with Community-acquired Pneumonia. An Official Clinical Practice Guideline of the American Thoracic Society and Infectious Diseases Society of America. Am J Respir Crit Care Med. 2019;200(7):e45-e67. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1164/rccm.201908-1581ST\u003c/span\u003e\u003cspan address=\"10.1164/rccm.201908-1581ST\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 31573350; PMCID: PMC6812437.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShah SN, Bachur RG, Simel DL, Neuman MI. Does This Child Have Pneumonia?: The Rational Clinical Examination Systematic Review. JAMA. 2017;318(5):462\u0026ndash;471. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.2017.9039\u003c/span\u003e\u003cspan address=\"10.1001/jama.2017.9039\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Erratum in: JAMA. 2017;318(13):1284. doi: 10.1001/jama.2017.13624. PMID: 28763554.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Shi Q, Zhong G, Lei X, Lin J, Fu Z,et al.Biomarker-based score for predicting in-hospital mortality of children admitted to the intensive care unit. J Investig Med. 2021;69(8):1458\u0026ndash;1463. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jim-2021-001855\u003c/span\u003e\u003cspan address=\"10.1136/jim-2021-001855\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2021 Jul 5. PMID: 34226249.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDoerflinger M, Haeusler GM, Li-Wai-Suen CSN, Clark JE, Slavin M, Babl FE, et al. Procalcitonin and Interleukin-10 May Assist in Early Prediction of Bacteraemia in Children With Cancer and Febrile Neutropenia. Front Immunol. 2021;12:641879. doi: 10.3389/fimmu.2021.641879. PMID: 34093531; PMCID: PMC8173204.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\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. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12879-025-10958-8\u003c/span\u003e\u003cspan address=\"10.1186/s12879-025-10958-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 40259232; PMCID: PMC12013137.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAra\u0026uacute;jo ALD, Moraes MC, P\u0026eacute;rez-de-Oliveira ME, Silva VMD, Saldivia-Siracusa C, Pedroso CM,et al. Machine learning for the prediction of toxicities from head and neck cancer treatment: A systematic review with meta-analysis. Oral Oncol. 2023;140:106386. doi: 10.1016/j.oraloncology.2023.106386. Epub 2023 Apr 4. PMID: 37023561.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBlack JE, Kueper JK, Williamson TS. An introduction to machine learning for classification and prediction. Fam Pract. 2023;40(1):200\u0026ndash;204. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/fampra/cmac104\u003c/span\u003e\u003cspan address=\"10.1093/fampra/cmac104\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 36181463.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAn Q, Rahman S, Zhou J, Kang JJ. A Comprehensive Review on Machine Learning in Healthcare Industry: Classification, Restrictions, Opportunities and Challenges. Sensors (Basel). 2023;23(9):4178. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/s23094178\u003c/span\u003e\u003cspan address=\"10.3390/s23094178\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 37177382; PMCID: PMC10180678.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYi F, Yang H, Chen D, Qin Y, Han H, Cui J, et al.XGBoost-SHAP-based interpretable diagnostic framework for alzheimer's disease. BMC Med Inform Decis Mak. 2023;23(1):137. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12911-023-02238-9\u003c/span\u003e\u003cspan address=\"10.1186/s12911-023-02238-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 37491248; PMCID: PMC10369804.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAnh, D.T., Takakura, H., Asai, M. et al. Application of machine learning in the diagnosis of vestibular disease. Sci Rep 12, 20805 (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-022-24979-9\u003c/span\u003e\u003cspan address=\"10.1038/s41598-022-24979-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXue B, Li D, Lu C, King CR, Wildes T, Avidan MS, et al.Use of Machine Learning to Develop and Evaluate Models Using Preoperative and Intraoperative Data to Identify Risks of Postoperative Complications. JAMA Netw Open. 2021;4(3):e212240. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jamanetworkopen.2021.2240\u003c/span\u003e\u003cspan address=\"10.1001/jamanetworkopen.2021.2240\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 33783520; PMCID: PMC8010590.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAhlquist KD, Sugden LA, Ramachandran S. Enabling interpretable machine learning for biological data with reliability scores. PLoS Comput Biol. 2023;19(5):e1011175. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pcbi.1011175\u003c/span\u003e\u003cspan address=\"10.1371/journal.pcbi.1011175\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 37235578; PMCID: PMC10249903.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlMansoori ME, Jemimah S, Abuhantash F, AlShehhi A. Predicting early Alzheimer's with blood biomarkers and clinical features. Sci Rep. 2024;14(1):6039. doi: 10.1038/s41598-024-56489-1. Erratum in: Sci Rep. 2024;14(1):20210. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-71400-8\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-71400-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 38472245; PMCID: PMC10933308.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao S Y, Qian S Y, Chen Z M, et al. Guidelines for the Diagnosis and Treatment of Mycoplasma Pneumoniae Pneumonia in Children (2023 Edition) [J]. (in Chinese).Infectious Disease Information, 2023, 36(04): 291\u0026ndash;297. DOI: CNKI:SUN:CRBX.0.2023-04-019.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNational Health Commission of the People's Republic of China, National Administration of Traditional Chinese Medicine. Guidelines for the Diagnosis and Treatment of Community-Acquired Pneumonia in Children (2019 Edition) [J]. (in Chinese).Chinese Journal of Clinical Infectious Diseases, 2019, 12(1): 6\u0026ndash;13. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3760/cma.j.issn.1674-2397.2019.01.002\u003c/span\u003e\u003cspan address=\"10.3760/cma.j.issn.1674-2397.2019.01.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eExpert Group on Pediatric Respiratory Endoscopy Technology, Talent Exchange Service Center of the National Health Commission; Endoscopy Professional Committee, Pediatrician Branch of the Chinese Medical Doctor Association; Pediatric Respiratory Endoscopy Professional Committee, Endoscopist Branch of the Chinese Medical Doctor Association; et al. Guidelines for Flexible Bronchoscopy in Chinese Pediatrics (2018 Edition) [J]. (in Chinese).Chinese Journal of Practical Pediatrics and Child Health, 2018, 33(13):983\u0026ndash;989. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3760/cma.j.issn.2095-428X.2018.13.006\u003c/span\u003e\u003cspan address=\"10.3760/cma.j.issn.2095-428X.2018.13.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVickers AJ, van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res. 2019;3:18. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s41512-019-0064-7\u003c/span\u003e\u003cspan address=\"10.1186/s41512-019-0064-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 31592444; PMCID: PMC6777022.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\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. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-58784-3\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-58784-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 38589453; PMCID: PMC11002011.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang S, Wan Y, Zhang W. The Clinical Value of Systemic Immune Inflammation Index (SII) in Predicting the Severity of Hospitalized Children with Mycoplasma Pneumoniae Pneumonia: A Retrospective Study. Int J Gen Med. 2024;17:935\u0026ndash;942. doi: 10.2147/IJGM.S451466. PMID: 38495920; PMCID: PMC10944171.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQiu J, Ge J, Cao L. D-dimer: The Risk Factor of Children's Severe Mycoplasma Pneumoniae Pneumonia. Front Pediatr. 2022;10:828437. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fped.2022.828437\u003c/span\u003e\u003cspan address=\"10.3389/fped.2022.828437\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 35498793; PMCID: PMC9039299.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJiang C, Bao S, Shen W, Wang C. Predictive value of immune-related parameters in severe Mycoplasma pneumoniae pneumonia in children. Transl Pediatr. 2024;13(9):1521\u0026ndash;1528. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/tp-24-172\u003c/span\u003e\u003cspan address=\"10.21037/tp-24-172\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2024 Sep 13. PMID: 39399713; PMCID: PMC11467233.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCilloniz C, Ward L, Mogensen ML, Peric\u0026agrave;s JM, M\u0026eacute;ndez R, Gabarr\u0026uacute;s A,et al. Machine-Learning Model for Mortality Prediction in Patients With Community-Acquired Pneumonia: Development and Validation Study. Chest. 2023;163(1):77\u0026ndash;88. doi: 10.1016/j.chest.2022.07.005. Epub 2022 Jul 16. PMID: 35850287.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Z, Ho KM, Hong Y. Machine learning for the prediction of volume responsiveness in patients with oliguric acute kidney injury in critical care. Crit Care. 2019;23(1):112. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13054-019-2411-z\u003c/span\u003e\u003cspan address=\"10.1186/s13054-019-2411-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 30961662; PMCID: PMC6454725.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHancock JT, Khoshgoftaar TM. CatBoost for big data: an interdisciplinary review. J Big Data. 2020;7(1):94. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s40537-020-00369-8\u003c/span\u003e\u003cspan address=\"10.1186/s40537-020-00369-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2020 Nov 4. PMID: 33169094; PMCID: PMC7610170.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e@inproceedings{Ostroumova2017CatBoostUB, title={CatBoost: unbiased boosting with categorical features}, author={Liudmila Ostroumova and Gleb Gusev and Aleksandr Vorobev and Anna Veronika Dorogush and Andrey Gulin}, booktitle={Neural Information Processing Systems}, year={2017}, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eurl={https://api.semanticscholar.org/CorpusID:5044218}\u003c/span\u003e\u003cspan address=\"http://url={https://api.semanticscholar.org/CorpusID:5044218}\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e }\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKui B, Pint\u0026eacute;r J, Molontay R, Nagy M, Farkas N, Gede N, et al.Hungarian Pancreatic Study Group. EASY-APP: An artificial intelligence model and application for early and easy prediction of severity in acute pancreatitis. Clin Transl Med. 2022;12(6):e842. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ctm2.842\u003c/span\u003e\u003cspan address=\"10.1002/ctm2.842\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 35653504; PMCID: PMC9162438.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGeeitha S, Ravishankar K, Cho J, Easwaramoorthy SV. Integrating cat boost algorithm with triangulating feature importance to predict survival outcome in recurrent cervical cancer. Sci Rep. 2024;14(1):19828. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-67562-0\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-67562-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 39191808; PMCID: PMC11349876.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLei T, Guo J, Wang P, Zhang Z, Niu S, Zhang Q,et al. Establishment and Validation of Predictive Model of Tophus in Gout Patients. J Clin Med. 2023;12(5):1755. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/jcm12051755\u003c/span\u003e\u003cspan address=\"10.3390/jcm12051755\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 36902542; PMCID: PMC10002994.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFan Z, Jiang J, Xiao C, Chen Y, Xia Q, Wang J,et al.Construction and validation of prognostic models in critically Ill patients with sepsis-associated acute kidney injury: interpretable machine learning approach. J Transl Med. 2023;21(1):406. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12967-023-04205-4\u003c/span\u003e\u003cspan address=\"10.1186/s12967-023-04205-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 37349774; PMCID: PMC10286378.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\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\u0026ndash;343. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s40121-023-00914-x\u003c/span\u003e\u003cspan address=\"10.1007/s40121-023-00914-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2024 Jan 24. PMID: 38265626; PMCID: PMC10904708.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTsai TA, Tsai CK, Kuo KC, Yu HR. Rational stepwise approach for Mycoplasma pneumoniae pneumonia in children. J Microbiol Immunol Infect. 2021;54(4):557\u0026ndash;565. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jmii.2020.10.002\u003c/span\u003e\u003cspan address=\"10.1016/j.jmii.2020.10.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2020 Oct 17. PMID: 33268306.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\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. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-58784-3\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-58784-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 38589453; PMCID: PMC11002011.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYadav RK, Kumar D, Gupta A, Sharma P. C-reactive protein and procalcitonin: As predictor biomarkers of severity and outcome in children with community-acquired pneumonia. Trop Doct. 2024;54(3):262\u0026ndash;267. doi: 10.1177/00494755241250371. Epub 2024 May 2. PMID: 38693837.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKolotylo TR, Moskaliuk VD, Syrota BV, Balaniuk IV, Melenko SR, Chernetska NV, et al. EVALUATION OF D-DIMER LEVEL AS A BIOMARKER OF DISEASE SEVERITY AND MORTALITY IN PATIENTS WITH COVID-19. Wiad Lek. 2023;76(7):1636\u0026ndash;1641. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.36740/WLek202307118\u003c/span\u003e\u003cspan address=\"10.36740/WLek202307118\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 37622508.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLuo XQ, Luo J, Wang CJ, Luo ZX, Tian DY, Xie XH. Clinical features of severe Mycoplasma pneumoniae pneumonia with pulmonary complications in childhood: A retrospective study. Pediatr Pulmonol. 2023;58(10):2815\u0026ndash;2822. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ppul.26593\u003c/span\u003e\u003cspan address=\"10.1002/ppul.26593\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2023 Jul 11. PMID: 37431970.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTran C, Singh GV, Haider E, Boylan C, Venegas C, Riaz S, et al. Luminal mucus plugs are spatially associated with airway wall thickening in severe COPD and asthma: A single-centered, retrospective, observational study. Respir Med. 2022;202:106982. doi: 10.1016/j.rmed.2022.106982. Epub 2022 Sep 9. PMID: 36116144.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"european-journal-of-medical-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ejmr","sideBox":"Learn more about [European Journal of Medical Research](http://eurjmedres.biomedcentral.com)","snPcode":"40001","submissionUrl":"https://submission.nature.com/new-submission/40001/3","title":"European Journal of Medical Research","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Severe Mycoplasma pneumoniae pneumonia, Machine learning, Predictive modeling, Children, CatBoost","lastPublishedDoi":"10.21203/rs.3.rs-7886677/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7886677/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eMycoplasma pneumoniae is a leading cause of community-acquired pneumonia in children, with severe cases (SMPP) posing a significant threat to pediatric health. Current diagnostic approaches rely primarily on imaging and clinical signs, lacking objective and quantitative tools for early risk prediction. Traditional statistical models face limitations in capturing the complexity of this condition, while machine learning (ML) methods offer the potential to uncover nonlinear relationships. However, the \"black box\" nature of many ML models hinders their clinical application. This study aimed to develop an interpretable ML model for the early prediction of SMPP risk in children and to enhance model transparency using Shapley Additive Explanations (SHAP) methods, thereby facilitating informed clinical decision-making.\u003c/p\u003e\u003ch2\u003eMethods:\u003c/h2\u003e\u003cp\u003eThe study retrospectively analyzed data from 286 inpatients with MPP admitted to the Affiliated Hospital of Yan'an University between August 2023 and August 2024. Patients were divided into MPP (n\u0026thinsp;=\u0026thinsp;163) and SMPP (n\u0026thinsp;=\u0026thinsp;123) groups based on their clinical condition. Forty-four clinical variables, including symptoms, laboratory parameters, and imaging features, were collected. Pearson correlation analysis, Mann-Whitney U test, chi-square test, and LASSO regression were employed to identify key predictors. Seven machine learning models (CatBoost, XGBoost, LightGBM, SVM, KNN, LR, GNB) were constructed using Python. Hyperparameters were optimized through 5-fold cross-validation and grid search, and model performance was evaluated by accuracy, AUC, and other metrics. Model interpretability was analyzed using the SHAP method.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eTwenty-one key features, such as thermal peak, thermal path, D-dimer, CRP, pleural effusion, and bronchoscopic manifestations, were evaluated. Among the seven machine learning models, the CatBoost model demonstrated superior performance, achieving an AUC of .961, an accuracy of .907, a recall of .923, and an F1 score of .900. The performance discrepancy between the training and test sets was minimal, indicating robust generalization. SHAP visual analysis identified thermal peak, thermal range, D-dimer, and CRP as the most significant positive predictors. Decision curve analysis further validated the CatBoost model's higher clinical net benefit across a broad threshold range.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eThis study developed and internally validated an interpretable machine learning model using the CatBoost algorithm to effectively predict early risk of SMPP in children, outperforming traditional methods. The model incorporates multidimensional clinical features, emphasizing the significance of bronchoscopic findings, and offers a quantitative tool for early identification of high-risk children. Future multi-center, prospective studies are necessary to further assess the model's generalizability and clinical applicability.\u003c/p\u003e","manuscriptTitle":"Application of Explainable Machine Learning in Early Diagnosis Models for Risk Prediction of Severe Mycoplasma pneumoniae Pneumonia in Children","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-28 07:14:38","doi":"10.21203/rs.3.rs-7886677/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-12-29T04:15:29+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-27T07:40:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"25646527702531475471790805206588787556","date":"2025-12-18T03:04:21+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-19T12:57:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"339634971077439008411661427601908155488","date":"2025-11-19T11:38:40+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-19T11:30:10+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-23T14:51:07+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-23T14:49:17+00:00","index":"","fulltext":""},{"type":"submitted","content":"European Journal of Medical Research","date":"2025-10-17T12:18:17+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"european-journal-of-medical-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ejmr","sideBox":"Learn more about [European Journal of Medical Research](http://eurjmedres.biomedcentral.com)","snPcode":"40001","submissionUrl":"https://submission.nature.com/new-submission/40001/3","title":"European Journal of Medical Research","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c3fbdfd1-a879-4193-9f72-f7f7eb46dda5","owner":[],"postedDate":"November 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-04-13T15:59:39+00:00","versionOfRecord":{"articleIdentity":"rs-7886677","link":"https://doi.org/10.1186/s40001-026-04319-7","journal":{"identity":"european-journal-of-medical-research","isVorOnly":false,"title":"European Journal of Medical Research"},"publishedOn":"2026-04-07 15:57:02","publishedOnDateReadable":"April 7th, 2026"},"versionCreatedAt":"2025-11-28 07:14:38","video":"","vorDoi":"10.1186/s40001-026-04319-7","vorDoiUrl":"https://doi.org/10.1186/s40001-026-04319-7","workflowStages":[]},"version":"v1","identity":"rs-7886677","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7886677","identity":"rs-7886677","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}