Aug 2026· PLoS ONE· Vol 21· 0 citations· 42 references
Medicine
TL;DR
ML models can distinguish high-risk KC groups based on clinical risk factors, facilitating risk stratification and early lifestyle interventions and confirmed by univariable logistic regression.
Abstract
Purpose To evaluate keratoconus (KC) risk factors and to develop a machine-learning (ML) model for KC and myopia classification. Methods In this retrospective single-center cross-sectional study, demographic and lifestyle data from patients with KC and individuals from a preoperative refractive surgery clinic were collected from January 20, 2024, to December 1, 2024. Univariable and multivariable regression analyses were used to identify key risk factors. Additionally, random forest (RF)-recursive feature elimination (RFE), extreme gradient boosting (XGBoost)-RFE, and univariable logistic regression were applied to select factors for ML models. Seven ML models were developed for a lifestyle-based classification system, with the performance being validated through discrimination and calibration, and interpretability being improved using SHapley Additive exPlanations (SHAP). Results Analysis of 711 patients (mean [standard deviation] age, 26.6 [7.1] years; 439 males [61.7%]) revealed 275 with KC. Multivariable regression analysis identified seven risk factors for KC, including male sex, higher body-mass index (BMI), lower education level, more distant childhood residence, allergic conjunctivitis, and increased eye-rubbing intensity and frequency. After feature selection of 24 variables, the neural-network model demonstrated the highest performance (area under the receiver operating characteristic curve [AUROC] = 0.79), followed by RF (AUROC = 0.77) and XGBoost (AUROC = 0.76). SHAP analysis consistently highlighted eye-rubbing intensity, sex, BMI, and childhood residence among the top 10 factors across the top three models, which were also confirmed by univariable logistic regression. Conclusion ML models can distinguish high-risk KC groups based on clinical risk factors, facilitating risk stratification and early lifestyle interventions.
OBJECTIVE
This study aims to develop and evaluate a machine learning-based risk stratification tool for predicting hearing loss in high-risk neonates using clinical risk factors to support targeted early identification.
METHODS
A total of 270 infants, 105 with hearing loss, 165 with normal hearing who had passed initial screening but possessed clinical risk factors were retrospectively analyzed. Clinical variables included prematurity, low birth weight, hyperbilirubinemia, phototherapy, NICU stay duration, and family history. Five models (Random Forest, XGBoost, CatBoost, K-nearest neighbors, and Logistic Regression) were developed using an 80/20 train-test split and stratified 5-fold cross-validation. Additionally, a web-based clinical decision support tool was developed using the Streamlit framework to provide real-time risk assessment.
RESULTS
The XGBoost model achieved the highest performance with 85.2% accuracy and an AUC of 87.1%. SHAP analysis identified NICU stay duration and positive family history as the most influential predictors for neonatal hearing loss.
CONCLUSION
Machine learning models, particularly XGBoost, provide robust risk stratification for high-risk neonates. Rather than replacing universal screenings, these tools can complement existing programs by identifying high-risk infants who require prioritized diagnostic follow-up and closer clinical monitoring. The developed web application (available at https://newbornhearing.streamlit.app/) offers a practical interface for clinical use.
LEVEL OF EVIDENCE: 3
Sevgi Kutlu, Emre Söylemez, Betül Temir et al.· The Laryngoscope· 0 citations
Hyperuricemia (HUA) imposes a growing public health burden, calling for better risk stratification tools. In this cross-sectional study of 4906 Chinese adults undergoing routine health checks (overall HUA prevalence: 26.0%), we built machine learning-based predictive models using a stratified 80/20 data split. To avoid variable selection bias, we applied LASSO regression with tenfold cross-validation, which identified 12 core predictors from routine clinical and demographic data. Among four algorithms tested, the Gradient Boosting (GB) model showed the best discrimination (AUC = 0.770) and good calibration (slope = 0.982, intercept = - 0.007, Brier score = 0.157). SHAP analysis revealed serum creatinine (Scr), HDL cholesterol (HDL-C), and body mass index (BMI) as the top predictors. Notably, SHAP interaction plots uncovered a nonlinear rise in risk above a Scr threshold and a compounded risk when high Scr coincided with low HDL-C. In summary, the well-calibrated GB model offers a reliable, data-driven tool for HUA risk screening, with insights into marker interactions to guide targeted prevention and early intervention.
Chu-xia Tan, Xue He, Lijun Li et al.· Scientific Reports· 0 citations
An interpretable XGBoost-based ML model that facilitates early risk stratification and targeted interventions for CSVD is validated, readily transferable to resource-limited settings and on embedding the nomogram into electronic-health-record decision support.
Xi Zhu, Xu-Hui Liu, Xujie Wang et al.· Frontiers in Neurology· 0 citations
Objective This study aimed to develop and validate a distinct, stable, and interpretable predictive model using machine learning techniques to identify individuals at high risk of pneumonia early after admission. The goal was to provide a potential quantitative reference for implementing preventive interventions in clinical practice. Methods A retrospective nested case–control design was adopted. A total of 822 patients with hemorrhagic stroke admitted between January 2019 and October 2024 were enrolled. Feature selection was performed using LASSO regression to eliminate multicollinearity and identify key predictors. Five machine learning algorithms—logistic regression (LRC), gradient boosting classifier (GBC), random forest classifier (RFC), multilayer perceptron classifier (MLPC), and support vector machine classifier (SVC)—were employed to construct predictive models. Hyperparameters were optimized through 10-fold cross-validation and grid search. Model performance was comprehensively evaluated on an independent test set using metrics including area under the curve (AUC), accuracy, sensitivity, precision, and F1-score. Finally, SHAP (SHapley Additive exPlanations) values were applied to interpret the optimal model and elucidate the contribution of each feature to the prediction. Results LASSO regression selected 14 key predictors from 57 initial variables. Among the five models, the logistic regression model achieved the best performance on the test set. SHAP-based interpretability analysis revealed that the most influential factors for pneumonia risk prediction were, in descending order: left lower limb muscle strength, total cholesterol (TC), right lower limb muscle strength, low-density lipoprotein cholesterol (LDL-C), white blood cell count (WBC), consciousness status, D-dimer, age, systolic blood pressure (SBP), and bleeding location. Conclusion This study successfully developed a logistic regression-based predictive model for pneumonia risk in patients with hemorrhagic stroke. The model demonstrated favorable discrimination and stability. It provides an objective, quantitative basis for early identification of high-risk patients, stratified management, and precise prevention and control, supporting a shift from reactive to proactive complication management.
Darong Lu, Wanting Shi, Li Wu et al.· Frontiers in Medicine· 0 citations
Objective Hypertension is a common yet frequently underdiagnosed comorbidity in psoriasis patients. Early identification and blood pressure control are critical to improving outcomes. Although machine learning (ML) is widely used in disease prediction, a model for hypertension risk within the psoriasis population remains unavailable. This study aims to develop and validate such a model in patients with psoriasis. Methods In this retrospective study, 2,957 psoriasis patients from a single tertiary center were used for model development and internal validation, and 567 psoriasis participants from the National Health and Nutrition Examination Survey (NHANES) served as the external validation cohort. After missForest imputation and consensus feature selection, the Synthetic Minority Oversampling Technique (SMOTE) was applied to the training set only. Nine machine learning algorithms were trained and evaluated for discrimination, calibration, and clinical utility. Shapley Additive Explanations (SHAP) were used for model interpretation. Results Nine nonredundant predictors were retained. The SMOTE-enhanced logistic regression model showed the most balanced and generalizable performance, with area under the receiver operating characteristic curve values of 0.850, 0.816, and 0.789 in the training, internal validation, and external validation cohorts, respectively, together with acceptable calibration and favorable net clinical benefit. SHAP identified age, dyslipidemia, and type 2 diabetes mellitus as the leading contributors. The final model was deployed as a publicly accessible web application. Conclusions This interpretable and externally validated machine learning model provides a practical tool for hypertension risk stratification in psoriasis patients and may support earlier identification and individualized preventive management in clinical practice.
Guo-Hua Xue, Xiao-Yang Guo, Jia-Qi Chen et al.· Digital Health· 0 citations
The LASSO-based GBM model demonstrated stable predictive performance and acceptable clinical utility and was selected as the final model for predicting postoperative recurrence of anal fistula.
Yun-Hao Zhou, Da-Wei Wang, Min Tang et al.· Frontiers in Surgery· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
