Identifying cardiovascular disease risk in the U.S. population using environmental volatile organic compounds exposure: A machine learning predictive model based on the SHAP methodology

Development of advanced lung cancer inflammation index-based machine learning models for predicting stroke and mortality: A comparative and interpretable study.

Objectives:To derive and externally validate supervised machine learning (ML) models predictive of cardiac surgery-associated acute kidney injury (CS-AKI).Design:Retrospective cohort analysis.Setting:Multicenter (4), cardiac surgical centers from January 2019 to February 2022.Patients:Seven days to 18 years old who had undergone cardiac surgery.Interventions:None.Measurements and Main Results:CS-AKI was defined using Kidney Disease: Improving Global Outcomes criteria, with stages 2/3 classified as severe, during the first 7 postoperative days. Data analysis followed two approaches: 1) combining three centers for derivation and using a fourth for external validation and 2) randomly dividing the entire dataset into derivation and validation cohorts in a 4:1 ratio. Forty ML models were developed across five derivation-validation pairs using four ML algorithms (light gradient-boosting machine, extreme gradient boosting, categorical boosting, and histogram gradient boosting) to predict two outcomes (any and severe CS-AKI) utilizing preoperative, intraoperative, and immediate postoperative variables. SHapley Additive exPlanations was used for input variable importance analysis. A cohort of 1100 patients was analyzed. Any CS-AKI and severe CS-AKI occurred in 49.1% and 23.1% patients, respectively. Wide range of variations in external validation of model performance were observed among all 40 ML models. For any CS-AKI, the range in metrics were: area under the receiver operating characteristic curve (AUROC) 0.64–0.83, sensitivity 0.29–0.86, specificity 0.46–0.95, positive predictive value (PPV) 0.50–0.85, and negative predictive value (NPV) 0.60–0.86. For severe CS-AKI, we found the range in metrics with AUROC 0.65–0.77, sensitivity 0.04–0.58, specificity 0.77–0.99, PPV 0.32–0.75, and NPV 0.78–0.90. Preoperative serum creatinine, cardiopulmonary bypass, aortic cross-clamp duration, weight, and age at surgery were the most important predictors associated with CS-AKI.Conclusions:This analysis of a retrospective multicenter dataset shows that external performance of ML models vary, highlighting challenges in generalizability, which may be due to center-based differences in practice.

Exposure to volatile organic compounds (VOCs) is widespread and has been implicated in the pathogenesis of various chronic diseases. However, the specific relationship between VOC exposure and the risk of osteoarthritis (OA) remains poorly characterized. This study aimed to investigate the associations between a broad spectrum of VOC metabolites and OA risk, and to identify the most influential VOC metabolites. We analyzed data from the National Health and Nutrition Examination Survey (NHANES) 2011-2018, comprising 3683 US adults. OA status was self-reported. Exposure levels to 17 VOCs were assessed using their urinary metabolites. After data splitting (70% training, 30% testing), multiple machine learning models were trained and evaluated. The optimal model was interpreted using SHapley Additive exPlanations (SHAP) to identify key predictors and elucidate their dose-response relationships with OA risk. The Linear Discriminant Analysis (LDA) model demonstrated the best predictive performance (AUC = 0.755). SHAP interpretation revealed that besides age, specific VOC metabolites were among the top predictors of OA. N-Acetyl-S-(3,4-dihydroxybutyl)-l-cysteine (DHBMA, a metabolite of 1,3-butadiene) and N-Acetyl-S-(3-hydroxypropyl-2-methyl)-l-cysteine (HMPMA, a metabolite of crotonaldehyde) were identified as novel and significant risk factors. Further analysis delineated non-linear, dose-response relationships between these VOCs and OA risk. Subgroup analyses suggested that the associations were consistent across different demographics. In summary, this study developed a machine learning model based on VOC exposure that effectively predicts osteoarthritis risk. LDA model achieved robust performance, with SHAP interpretation identifying DHBMA and HMPMA as novel and significant risk factors, in addition to known demographic predictors. Subgroup analyses further confirmed the consistent and non-linear association of these VOC metabolites with OA across diverse populations. These findings underscore the value of integrating environmental exposure data into OA risk prediction and support its potential for targeted prevention strategies in high-risk groups.

BACKGROUND Heart failure (HF) is a leading cause of morbidity and mortality among patients in intensive care units (ICUs), particularly those with chronic critical illness (CCI). OBJECTIVE We aimed to develop and validate a machine learning (ML) model to predict in-hospital mortality for in CCI patients with CCI and HF. METHODS Retrospective data encompassing medical records from over 200 hospitals were obtained from the Medical Information Mart for Intensive Care III (MIMIC-III), MIMIC-IV, and eICU Collaborative Research Database (eICU-CRD). Patients diagnosed with CCI and HF at their first ICU admission were included. The MIMIC-III and -IV datasets were used as a derivation cohort, while that from eICU-CRD was employed as a validation cohort. Key predictive features were identified utilizing the recursive feature elimination with 10-fold cross-validation method. Subsequently, multiple ML algorithms were evaluated, including Random Forest, K-Nearest Neighbors, Support Vector Machine (SVM), Extreme Gradient Boosting, Naive Bayes, Light Gradient Boosting Machine, and Adaptive Boosting. The performance of the models was assessed via metrics such as area under the receiver operating characteristic curve (AUROC), decision curve analysis, accuracy, sensitivity, specificity, and F1 score. Furthermore, model interpretability was enhanced by applying the SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) methods, providing insights into the contribution of individual features to the predictive outcomes. RESULTS A total of 780 (males: 451 [57.8%]) and 610 (males: 343 [56.2%]) patients with CCI and HF were allocated to the derivation and validation cohorts, respectively. Eleven features were selected to develop the prediction models. Among all models, the SVM algorithm-based model demonstrated high predictive accuracy (derivation cohort: AUROC, 0.781; sensitivity, 0.739; specificity, 0.691; and F1 score, 0.613; validation cohort: AUROC, 0.683; accuracy, 0.645; sensitivity, 0.607; specificity, 0.656; and F1 score, 0.44). The SHAP and LIME analyses evaluated the feature contributions, highlighting Sequential Organ Failure Assessment score, oxyhemoglobin saturation, diastolic blood pressure, and systolic blood pressure as significant predictors of in-hospital mortality. CONCLUSIONS The SVM model developed in this study effectively predicts in-hospital mortality in patients with CCI and HF and can serve as a reliable tool for early intervention and improved patient management. Furthermore, this ML model combines high accuracy with interpretability, thereby substantially contributing to clinical predictive analytics.

BackgroundThe risk of developing atherosclerotic cardiovascular disease (ASCVD) varies among individuals and is related to a variety of lifestyle factors in addition to the presence of chronic diseases.ObjectiveWe aimed to assess the predictive accuracy of machine learning (ML) models incorporating lifestyle risk behaviors for ASCVD risk using the Korean nationwide database.MethodsUsing data from the Korea National Health and Nutrition Examination Survey, 5 ML algorithms were used for the prediction of high ASCVD risk: logistic regression (LR), support vector machine, random forest, extreme gradient boosting, and light gradient boosting models. ASCVD risk was assessed using the pooled cohort equations, with a high-risk threshold of ≥7.5% over 10 years. Among the 8573 participants aged 40‐79 years, propensity score matching (PSM) was used to adjust for demographic confounders. We divided the dataset into a training and a test dataset in an 8:2 ratio. We also used bootstrapping to train the ML model with the area under the receiver operating characteristics curve score. Shapley additive explanations were used to identify the models’ important variables in assessing high ASCVD risks. In sensitivity analysis, we additionally performed binary LR analysis, in which the ML model’s results were consistent with the conventional statistical model.ResultsOf the 8573 participants, 41.7% (n=3578) had high ASCVD risk. Before PSM, age and sex differed significantly between groups. PSM (1:1) yielded 1976 patients with balanced demographics. After PSM, the high ASCVD risk group had higher alcohol or tobacco use, lower omega-3 intake, higher BMI, less physical activity, and spent less time sitting. In 5 ML models, the extreme gradient boosting model showed the highest area under the receiver operating characteristics curve, indicating superior overall discrimination between high and low ASCVD risk groups. However, the light gradient boosting model demonstrated better performance in accuracy, recall, and F1-score. Variable importance analysis using Shapley additive explanations identified smoking and age as the strongest predictors, while BMI, sodium or omega-3 intake, and low-density lipoprotein cholesterol also had significant variables. Sensitivity analysis using multivariable LR analysis also confirmed these findings, showing that smoking, BMI, and low-density lipoprotein cholesterol increased ASCVD risk, whereas omega-3 intake and physical activity were associated with lower risk.ConclusionsAnalyzing lifestyle behavioral factors in ASCVD risk with an ML model improves the predictive performance compared to traditional models. Personalized prevention strategies tailored to an individual’s lifestyle can effectively reduce ASCVD risk.

Breast cancer continues to be a leading cause of death among women in the world. The prediction of survival outcomes based on treatment modalities, i.e., chemotherapy, hormone therapy, surgery, and radiation therapy is an essential step towards personalization in treatment planning. However, Machine Learning (ML) models may improve these predictions by investigating intricate relationships between clinical variables and survival. This study investigates the performance of several ML models to predict survival rate in patients undergoing diverse breast cancer treatments i.e., chemotherapy, hormone therapy, surgery and radiation using multiple clinical parameters. The dataset consisted of 5000 samples and turned into downloaded from Kaggle. The models assessed blanketed Support Vector Machines (SVM), K-Nearest Neighbor (KNN), AdaBoost, Gradient Boosting, Random Forest, Gaussian Naive Bayes, Logistic Regression, Extreme Gradient Boosting (XG boost), and Decision tree. Performance of the models was assessed using parameters such as Accuracy, Precision, Recall, F1-Score and Area under the Receiver Operating Characteristic Curve (AUC-ROC). SHAP (SHapley Additive exPlanations) values analysis was done to provide an explanation for the impact of a feature on model predictions using Waterfall and Beeswarm plots. Anticipated baseline (E(f(x))) were in comparison to the predictions (f(x)) for each therapy group. Performance of Gradient Boosting was shown to be the best with an Accuracy: 0.972, Precision: 0.973, Recall: 0.972, F1-Score: 0.973, and AUC-ROC Score: 0.997. Chemotherapy had a notably bad impact on survival, with an f(x) of -0.274 and an E(f(x)) of -0.025. Hormone therapy showed the maximum detrimental effect on survival, with an f(x) of -0.408. Surgical operation had an extraordinarily impartial impact (f(x) = -0.041), even as radiation therapy positively impacted survival results with an f(x) of 0.22. Gradient Boosting was the most predictive algorithm for breast cancer survival outcomes. This SHAP-primarily based analysis provides a complete knowledge of ways one-of-a-kind treatments have an effect on survival predictions in breast cancer patients. Radiation therapy indicates the most tremendous effect on survival, whilst hormone therapy reveals the maximum poor effect. Future studies need to explore personalized treatment strategies that comprise these insights to enhance patient effects.

BackgroundDepressive disorder, particularly major depressive disorder (MDD), significantly impact individuals and society. Traditional analysis methods often suffer from subjectivity and may not capture complex, non-linear relationships between risk factors. Machine learning (ML) offers a data-driven approach to predict and diagnose depression more accurately by analyzing large and complex datasets.MethodsThis study utilized data from the National Health and Nutrition Examination Survey (NHANES) 2013–2014 to predict depression using six supervised ML models: Logistic Regression, Random Forest, Naive Bayes, Support Vector Machine (SVM), Extreme Gradient Boost (XGBoost), and Light Gradient Boosting Machine (LightGBM). Depression was assessed using the Patient Health Questionnaire (PHQ-9), with a score of 10 or higher indicating moderate to severe depression. The dataset was split into training and testing sets (80% and 20%, respectively), and model performance was evaluated using accuracy, sensitivity, specificity, precision, AUC, and F1 score. SHAP (SHapley Additive exPlanations) values were used to identify the critical risk factors and interpret the contributions of each feature to the prediction.ResultsXGBoost was identified as the best-performing model, achieving the highest accuracy, sensitivity, specificity, precision, AUC, and F1 score. SHAP analysis highlighted the most significant predictors of depression: the ratio family income to poverty (PIR), sex, hypertension, serum cotinine and hydroxycotine, BMI, education level, glucose levels, age, marital status, and renal function (eGFR).ConclusionWe developed ML models to predict depression and utilized SHAP for interpretation. This approach identifies key factors associated with depression, encompassing socioeconomic, demographic, and health-related aspects.

ObjectivePostoperative delirium (POD) is a prevalent neurological complication linked to adverse clinical outcomes. The underlying mechanisms of POD remain unclear. This study aimed to investigate the association between POD and frailty and determine whether frailty influences POD incidence. Furthermore, machine learning algorithms were utilized to identify key predictors of POD in patients undergoing hip or knee replacement.MethodsA total of 625 Han Chinese patients were recruited between September 2021 and May 2023. Preoperative frailty was assessed using the Frailty Scale and Frailty Phenotype criteria. The Mini-Mental State Examination (MMSE) evaluated preoperative cognitive function, while the Confusion Assessment Method (CAM) diagnosed POD. The severity of POD was additionally quantified using the Memorial Delirium Assessment Scale (MDAS). Receiver Operating Characteristic (ROC) curve analysis explored the association between preoperative frailty and POD, and the mediating effect of cerebrospinal fluid (CSF) biomarkers was analyzed. Ten machine learning algorithms—including Logistic Regression (LR), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Artificial Neural Network (ANN), Random Forest (RF), XGBoost, K-Nearest Neighbors (KNN), AdaBoost, LightGBM, and CatBoost—were implemented to develop predictive models. The dataset was randomly split into training (70%) and testing (30%) subsets. Ten-fold cross-validation was incorporated during model training and validation to mitigate overfitting and enhance generalizability. Model performance was evaluated using multiple metrics, such as accuracy, sensitivity, specificity, precision, Brier score, area under the ROC curve (AUC), and F1 score. Furthermore, graphical analyses—including calibration curves, decision diagrams, clinical impact curves, and confusion matrices—were applied to assess model robustness and clinical utility. Finally, SHAP (Shapley Additive Explanations) analysis elucidated the model’s decision-making process, emphasizing the pivotal role of preoperative frailty in POD prediction.ResultsThe incidence of POD was 14.7%. The study identified frailty, Tau, and P-tau as significant risk factors for POD (OR = 67.229, 95% CI: 34.649–130.444, p < 0.001; OR = 1.020, 95% CI: 1.016–1.024, p < 0.001; OR = 1.018, 95% CI: 1.010–1.027, p < 0.001). ROC curve analysis (AUC = 0.983) demonstrated that combining frailty with CSF biomarkers had strong predictive power for distinguishing POD. The direct effect of frailty on POD was 0.504878, the total effect was 0.6547619, and the mediating effect of Tau accounted for 22.89%. Using Lasso regression for variable selection, we subsequently identified eight predictors—frailty, Tau, Aβ42/Tau, Aβ40, age, Aβ42, P-tau, and drinking history—from the training set via logistic regression. Based on these factors, we constructed 10 machine learning models. Among all machine learning algorithms, GBM performed the best, achieving an AUC of 0.973 (95% CI, 0.973–1.000) in the test set. Furthermore, SHAP analysis confirmed that frailty and Tau were the key determinants influencing the machine learning model’s predictions.ConclusionPreoperative frailty is an independent risk factor for POD. A machine learning model for predicting POD in patients undergoing hip or knee replacement was developed, with GBM demonstrating superior performance among all models. The GBM-based model enabled early identification of patients at high risk of delirium.

BackgroundAlveolar bone loss (ABL) is common in modern society. Heavy metal exposure is usually considered to be a risk factor for ABL. Some studies revealed a positive trend found between urinary heavy metals and periodontitis using multiple logistic regression and Bayesian kernel machine regression. Overfitting using kernel function, long calculation period, the definition of prior distribution and lack of rank of heavy metal will affect the performance of the statistical model. Optimal model on this topic still remains controversy. This study aimed: (1) to develop an algorithm for exploring the association between heavy metal exposure and ABL; (2) filter the actual causal variables and investigate how heavy metals were associated with ABL; and (3) identify the potential risk factors for ABL.MethodsData were collected from National Health and Nutrition Examination Survey (NHANES) between 2015 and 2018 to develop a machine learning (ML) model. Feature selection was performed using the Least Absolute Shrinkage and Selection Operator (LASSO) regression with 10-fold cross-validation. The selected data were balanced using the Synthetic Minority Oversampling Technique (SMOTE) and divided into a training set and testing set at a 3:1 ratio. Logistic Regression (LR), Support Vector Machines (SVM), Random Forest (RF), K-Nearest Neighbor (KNN), Decision Tree (DT), and XGboost were used to construct the ML model. Accuracy, Area Under the Receiver Operating Characteristic Curve (AUC), Precision, Recall, and F1 score were used to select the optimal model for further analysis. The contribution of the variables to the ML model was explained using the Shapley Additive Explanations (SHAP) method.ResultsRF showed the best performance in exploring the association between heavy metal exposure and ABL, with an AUC (0.88), accuracy (0.78), precision (0.76), recall (0.83), and F1 score (0.79). Age was the most important factor in the ML model (mean| SHAP value| = 0.09), and Cd was the primary contributor. Sex had little effect on the ML model contribution.ConclusionIn this study, RF showed superior performance compared with the other five algorithms. Among the 12 heavy metals, Cd was the most important factor in the ML model. The relationship of Co & Pb and ABL are weaker than that of Cd. Among all the independent variables, age was considered the most important factor for this model. As for PIR, low-income participants present association with ABL. Mexican American and Non-Hispanic White show low association with ABL compared to Non-Hispanic Black and other races. Gender feature demonstrates a weak association with ABL. In the future, more advanced algorithms should be developed to validate these results and related parameters can be tuned to improve the accuracy of the model.Clinical trial numbernot applicable.

Cystitis glandularis (CG) is a rare urological condition characterized by glandular metaplasia of the bladder mucosa. Recurrence following transurethral resection (TUR) is a significant clinical challenge. Traditional predictive models often fail to capture the complexity of the data, resulting in insufficient accuracy. In contrast, machine learning (ML) has demonstrated substantial potential in medical prediction by identifying and analyzing complex patterns that are undetectable by conventional methods. This study aims to develop and evaluate an interpretable ML model to predict recurrence after TUR for CG, thereby improving clinical decision-making and patient outcomes. We analyzed predictors of recurrence using the least absolute shrinkage and selection operator (LASSO) and multivariate logistic regression. We developed and tested seven ML-based models: Cox proportional hazards model (CoxPH), LASSO regression, decision tree (rpart), random survival forest (RSF), gradient boosting machine (GBM), support vector machine (SVM), and extreme gradient boosting (XGBoost). Participants were diagnosed with CG by pathology following TUR and treated from 2012 to 2018. Model discrimination was assessed using the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC), while model preference was evaluated through the Brier score (BS). Decision curve analysis (DCA) was used for model comparison. The SHapley Additive exPlanations (SHAP) method was employed for interpretation, providing insights into recurrence prediction and prevention strategies. Finally, user-friendly platform was developed, allowing users to predict CG recurrence by entering feature values into designated text boxes on the webpage. The RSF model demonstrated the best performance in predicting recurrence, as indicated by superior ROC, DCA, and BS metrics. In SHAP, postoperative regular instillation (PRI) contributed the most to model construction. The RSF model effectively predicts CG recurrence, offering a framework for individualized treatment strategies. PRI was identified as the most significant risk factor influencing recurrence.

Deciphering the environmental chemical basis of muscle quality decline by interpretable machine learning models

Background/Objectives: Postoperative deep vein thrombosis (DVT) is a common and serious complication after tibial fracture surgery. This study aimed to develop and evaluate machine learning (ML) models to predict the occurrence of DVT following tibia fracture surgery. Methods: A retrospective analysis was conducted on patients who had undergone surgery for isolated tibial fractures. A total of 42 predictive models were developed using combinations of six ML algorithms-logistic regression, support vector machine, random forest, extreme gradient boosting, Light Gradient Boosting Machine (LightGBM), and neural networks-and seven feature selection methods, including SHapley Additive exPlanations (SHAP), Least Absolute Shrinkage and Selection Operator (LASSO), Boruta, recursive feature elimination, univariate filtering, and full-variable inclusion. Model performance was assessed based on discrimination, quantified by the area under the receiver operating characteristic curve (AUC-ROC), and calibration, measured using Brier scores, with internal validation performed via bootstrapping. Results: Of 471 patients, 80 (17.0%) developed postoperative DVT. The ML models achieved high overall accuracy in predicting DVT. Twenty-four models showed similarly excellent discrimination (pairwise AUC comparisons, p > 0.05). The top-performing model (random forest with RFE) attained an AUC of ~0.99, while several others (including LightGBM and SVM-based models) also reached AUC values in the 0.97-0.99 range. Notably, support vector machine models paired with Boruta or LASSO feature selection demonstrated the best calibration (lowest Brier scores), indicating reliable risk estimation. The final selected SVM models achieved high specificity (≥95%) with moderate sensitivity (~75-80%) for DVT detection. Conclusions: ML models demonstrated high accuracy in predicting postoperative DVT following tibial fracture surgery. Support vector machine-based models showed particularly favorable discrimination and calibration. These results suggest the potential utility of ML-based risk stratification to guide individualized prophylaxis, warranting further validation in prospective clinical settings.

Anemia is one of the common adverse reactions after hip fracture surgery. The traditional method to solve anemia is allogeneic transfusion. However, the transfusion may lead to some complications such as septicemia and fever. So far, few studies have reported roles of machine learning in predicting whether blood transfusion is needed or not after hip fracture surgery. Therefore, the purpose of this study is to develop machine learning models to predict the likelihood of postoperative blood transfusion in patients undergoing hip fracture surgery. This study enrolled 1355 patients who underwent hip fracture surgery at the Affiliated Hospital of Qingdao University from January 2016 to December 2021. Among all patients, 210 cases received postoperative blood transfusion. All patients were randomly divided into a training group and a testing group at a ratio of 7:3. In the training group, univariate and multivariate logistic regression analyses were used to determine independent risk factors for the postoperative transfusion. Then, based on these independent risk factors, tenfold cross-validation method was utilized to develop five machine learning models, including logistic, multilayer perceptron (MLP), extreme gradient boosting (XGBoost), random forest (RF), and support vector machine (SVM). The receiver operating characteristic (ROC) curve, area under ROC curve (AUC), and Matthews correlation coefficient (MCC) were generated to evaluate the performance of the models. Calibration plot and decision curve analysis (DCA) were used to test the performance, stability, and clinical applicability of the models. The models were validated using the testing group; and the ROC curve, MCC, calibration plot, and DCA curves were also generated to validate the performance, stability, and clinical applicability of the models. To further verify the robustness of the model, we randomly grabbed 70% of the samples in the testing set, performed 1000 iterations, and calculated the AUC and confidence interval of the five models. Finally, we used SHapley Additive exPlanations (SHAP) to explain these models. Multivariate logistic regression analysis showed that there were 8 independent risk factors, including age, blood transfusion history, albumin (ALB), globulin (GLO), total bilirubin (TBIL), indirect bilirubin (IBIL), hemoglobin (HB), and blood loss > 200ml. We finally selected five independent risk factors including HB, GLO, age, IBIL, and blood loss > 200ml. Based on these five independent risk factors, we generated six characteristic variables, namely HB, HB × HB, HB × blood loss, GLO × HB, age, age × IBIL, and established five machine learning models using a tenfold cross-validation method. In the training group, the AUC values of logistic, RF, MLP, SVM, and XGB were 0.9320, 0.8911, 0.9327, 0.9225, and 0.8825, respectively, and the average AUC was 0.9122 ± 0.0212. The MCC values were 0.65, 0.77, 0.65, 0.66, and 0.68, respectively, and the calibration plot and DCA performed well. In the testing group the AUC values of logistic, RF, MLP, SVM, and XGB were 0.8483, 0.7978, 0.8576, 0.8598, and 0.8216, respectively. The average AUC was 0.8370 ± 0.0238, and the MCC values were 0.41, 0.35, 0.40, 0.41, and 0.41, respectively. The calibration plot and DCA in the testing group also showed good performance. The AUC values and confidence intervals of the 1000-iteration model were: logistic (AUC, min confidence interval [CI]-max confidence interval [CI] 0.848, 0.804-0.903), RF (AUC, minCI-maxCI 0.797, 0.734-0.857), MLP (AUC, minCI-maxCI 0.858, 0.812-0.902), SVM (AUC, minCI-maxCI 0.859, 0.819-0.910), and XGB (AUC, minCI-maxCI 0.821, 0.764-0.894). The model performed well. Finally, according to SHAP, among all five models, HB played the most important role in model prediction and interpretation. The five models we developed all performed well in predicting the likelihood of blood transfusion after hip fracture surgery. Therefore, we believed that the prediction model based on machine learning had great application prospects in clinical practice, which could help clinicians better predict the risk of blood transfusion after hip fracture surgery.

Background Chronic Kidney Disease (CKD) affects approximately 697.5 million people worldwide. Volatile organic compounds (VOCs) are emerging as potential risk factors, but their complex relationships with CKD may be underestimated by traditional linear methods. This study explores the association between urinary VOC metabolites and CKD risk using a combination of epidemiological and interpretable machine learning approaches. Methods Data from the National Health and Nutrition Examination Survey (2011–March 2020 pre-pandemic) were analyzed to examine 15 urinary VOC metabolites. Analytical methods included multivariable logistic regression, LASSO regression, and five machine learning models: Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), and Multilayer Perceptron (MLP). SHapley Additive exPlanations (SHAP) analysis was used to enhance model interpretability. Results Significant associations were observed for metabolites including CEMA (N-Acetyl-S-(2-carboxyethyl)-L-cysteine) (OR = 1.66, 95% CI: 1.17–2.37), DHBMA (N-Acetyl-S-(3,4-dihydroxybutyl)-L-cysteine) (OR = 1.95, 95% CI: 1.38–2.76), HMPMA (N-Acetyl-S-(3-hydroxypropyl-1-methyl)-L-cysteine) (OR = 2.18, 95% CI: 1.53–3.10), and PGA (Phenylglyoxylic acid) (OR = 1.66, 95% CI: 1.22–2.27). The XGBoost model demonstrated strong predictive performance, with SHAP analysis highlighting DHBMA as a key predictor. Inverse associations were observed for AAMA (N-Acetyl-S-(2-carbamoylethyl)-L-cysteine) and CYMA (N-Acetyl-S-(2-cyanoethyl)-L-cysteine) in their highest quartiles. Conclusions This integrated approach identified significant associations between specific urinary VOC metabolites and CKD risk, particularly DHBMA. These findings underscore the role of environmental VOC exposure in CKD pathogenesis and may inform targeted prevention strategies.

12074 Background: Skeletal-related events (SREs) are detrimental clinical events in bone metastasis from advanced cancer, defined by pathologic fracture, spinal cord compression, and inevitable surgical or radiational intervention to the bone. Given their negative impact on quality of life and prognosis and interpatient heterogeneity in the SRE risk, accurate identification of patients with high SRE risk is critical. Methods: The patient-level data from three randomized clinical trials that administered zoledronic acid (ZA) to patients with bone-metastatic breast cancer, castration-resistant prostate cancer (CRPC), and other types of cancer were analyzed. (N = 460, 452, and 315, respectively) Machine learning (ML) models to predict SREs within 18 months (breast cancer), 12 months (CRPC), and 9 months (other cancers) were developed based on more than 40 baseline clinical and laboratory data. Seven ML algorithms and five feature selection methods were utilized to develop multiple models. The ML model with the best performance was identified based on the F1 score and the area under the receiver operating characteristic curve (AUC-ROC) for each cancer type and interrogated for important features with Shapley additive explanations. Lastly, the ML models’ ability to stratify patients by the cumulative SRE risk was evaluated by calculating hazard ratio (HR) with Cox-proportional hazards models. Results: Among the multiple ML models developed with different algorithms and feature selection methods, the model developed utilizing the random forests algorithm and the Boruta method for selecting features demonstrated the best performance in all types of cancer (F1 0.70, 0.67, and 0.67, and AUC-ROC 0.72, 0.68, and 0.73 for breast cancer, CRPC, and other cancers, respectively). In the ML model for breast cancer, performance status (PS), history of SRE, serum alkaline phosphatase (ALP), history of anti-neoplastic surgery, radiation therapy, and pathologic fracture were included as important features. Serum ALP, albumin, sodium, Gleason scores, and geographic regions were shown to be relevant in the CRPC model. For other cancers, serum ALP, albumin, total protein, phosphorus, red blood cell count, white blood cell count, visceral metastases, and history of arthritis were incorporated in the ML model. The ML model prediction successfully stratified the patients for cumulative SRE risk in all three cohorts. (HR with 95% confidence interval: 2.43 [1.86 – 3.18], 1.92 [1.51 – 2.45], and 3.06 [2.29 – 4.09] for breast cancer, CRPC, and other types of cancer, respectively). Conclusions: ML models incorporating baseline clinical and laboratory data can identify patients with bone metastasis on ZA harboring a high SRE risk.