"Weibing" is a fundamental concept in traditional Chinese medicine (TCM), representing a transitional state characterized by diminished self-regulatory abilities without overt physiological or social dysfunction. This perspective delves into the biological foundations and quantifiable markers of Weibing, aiming to establish a research framework for early disease intervention. Here, we propose the "Health Quadrant Classification" system, which divides the state of human body into health, sub-health, disease-susceptible state, and disease. We suggest the disease-susceptible stage emerges as a pivotal point for TCM interventions. To understand the intrinsic dynamics of this state, we propose laboratory and clinical studies utilizing time-series experiments and stress-induced disease susceptibility models. At the molecular level, bio-omics technologies and bioinformatics approaches are highlighted for uncovering intricate changes during disease progression. Furthermore, we discuss the application of mathematical models and artificial intelligence in developing early warning systems to anticipate and avert the transition from health to disease. This approach resonates with TCM's preventive philosophy, emphasizing proactive health maintenance and disease prevention. Ultimately, our perspective underscores the significance of integrating modern scientific methodologies with TCM principles to propel Weibing research and early intervention strategies forward.

Objective:To develop and validate a clinical warning model for severe pediatric community-acquired pneumonia (CAP) using electronic health records.Methods:A retrospective cohort study was conducted, analyzing clinical data of 15 750 children hospitalized for CAP at Wuhan Children′s Hospital between January 1, 2019, and December 31, 2023. Patient data were randomly split into training and testing sets at a 7∶3 ratio. Six supervised machine learning models were constructed in the training set, optimized using five-fold cross-validation, and evaluated in the testing set. Model performance was assessed using ROC-AUC, sensitivity, specificity, positive predictive value, negative predictive value, calibration curves, and clinical decision curve analysis at optimal thresholds. The best-performing model was selected, and SHapley Additive exPlanations (SHAP) were used to interpret feature importance. A program interface was developed based on the model results, enabling integration into clinical decision support systems for automated early warning.Results:A total of 15 750 participants, ranging in age from 28 days to 18 years, were included in the study. The median age was 2 years [interquartile range (IQR): 0-4 years], with 9 555 males (60.67%) and 6 195 females (39.33%). Among them, 2 211 (14.04%) developed severe pneumonia. In the prediction models, XGB outperformed other models with an ROC-AUC of 0.884 (95% CI: 0.870-0.898), sensitivity (0.803, 95% CI: 0.772-0.832), specificity (0.828, 95% CI: 0.816-0.839). Calibration analysis showed strong agreement between predicted and observed risks (Brier score: 0.081, 95% CI: 0.075-0.086). The analysis based on the SHAP method revealed that respiratory rate, heart rate, T-lymphocyte subsets, and red blood cell volume distribution width-SD are predictive factors for severe progression of community-acquired pneumonia (CAP) in children. Conclusion:An interpretable machine learning model was developed for the early detection and personalized treatment planning of severe CAP in children, providing valuable support to clinicians.

Objective:To develop and validate a clinical warning model for severe pediatric community-acquired pneumonia (CAP) using electronic health records.Methods:A retrospective cohort study was conducted, analyzing clinical data of 15 750 children hospitalized for CAP at Wuhan Children′s Hospital between January 1, 2019, and December 31, 2023. Patient data were randomly split into training and testing sets at a 7∶3 ratio. Six supervised machine learning models were constructed in the training set, optimized using five-fold cross-validation, and evaluated in the testing set. Model performance was assessed using ROC-AUC, sensitivity, specificity, positive predictive value, negative predictive value, calibration curves, and clinical decision curve analysis at optimal thresholds. The best-performing model was selected, and SHapley Additive exPlanations (SHAP) were used to interpret feature importance. A program interface was developed based on the model results, enabling integration into clinical decision support systems for automated early warning.Results:A total of 15 750 participants, ranging in age from 28 days to 18 years, were included in the study. The median age was 2 years [interquartile range (IQR): 0-4 years], with 9 555 males (60.67%) and 6 195 females (39.33%). Among them, 2 211 (14.04%) developed severe pneumonia. In the prediction models, XGB outperformed other models with an ROC-AUC of 0.884 (95% CI: 0.870-0.898), sensitivity (0.803, 95% CI: 0.772-0.832), specificity (0.828, 95% CI: 0.816-0.839). Calibration analysis showed strong agreement between predicted and observed risks (Brier score: 0.081, 95% CI: 0.075-0.086). The analysis based on the SHAP method revealed that respiratory rate, heart rate, T-lymphocyte subsets, and red blood cell volume distribution width-SD are predictive factors for severe progression of community-acquired pneumonia (CAP) in children. Conclusion:An interpretable machine learning model was developed for the early detection and personalized treatment planning of severe CAP in children, providing valuable support to clinicians.