Development and validation of a machine learning-based explainable predictive model for long-term outcomes in patients with high bleeding risk undergoing percutaneous coronary intervention

BackgroundP2Y12 inhibitor monotherapy is a promising novel strategy to reduce bleeding complications compared to dual antiplatelet therapy (DAPT) in patients undergoing percutaneous coronary intervention (PCI). In order to personalise treatment with DAPT based on patients' bleeding risk, we compared outcomes after PCI between P2Y12 inhibitor monotherapy and DAPT according to bleeding risk. MethodsA search for randomized clinical trials (RCTs) comparing P2Y12 inhibitor monotherapy after a short period of DAPT to standard DAPT after PCI was performed. Outcome differences between treatment groups regarding major bleedings, major adverse cardiac and cerebral events (MACCE) and net adverse clinical events (NACE) were assessed with hazard ratios (HRs) and corresponding credible intervals (CrI) according a Bayesian random effects model in patients with and without high bleeding risk (HBR). ResultsFive RCTs including 30,084 patients were selected. P2Y12 inhibitor monotherapy compared to DAPT reduced major bleedings in the total population (HR: 0.65, 95 % CrI: 0.44 to 0.92). The HRs of the HBR and non-HBR subgroups showed a similar reduction of bleedings for monotherapy (HBR: HR 0.66, 95 % CrI: 0.25 to 1.74; non-HBR: HR 0.63, 95 % CrI: 0.36 to 1.09). No notable differences between treatments on MACCE and NACE were observed in either sub-group or in the total population. ConclusionsRegardless of bleeding risk, P2Y12 inhibitor monotherapy is the favourable choice after PCI regarding major bleedings and does not increase ischemic events compared to DAPT. This suggests that bleeding risk is not decisive when considering P2Y12 inhibitor monotherapy.

Recurrent Events Analysis of MASTER DAPT: Total Ischemic and Bleeding Events After Abbreviated vs Prolonged DAPT in HBR Patients.

Background Previous studies have indicated that percutaneous coronary intervention (PCI) may significantly improve clinical outcomes in patients with chronic kidney disease (CKD) and non–ST-segment elevation myocardial infarction (NSTEMI). This study aimed to investigate how different CKD stages influence the prognostic effect of PCI. Objective To evaluate the impact of PCI on key clinical outcomes—including net adverse clinical events (NACE), cardiac death, myocardial infarction, and all-cause mortality—among NSTEMI patients with CKD. Additionally, we aimed to assess whether CKD staging confers a stratified effect on PCI efficacy and to explore any interaction between CKD stage and PCI on these outcomes. Methods This retrospective cohort analysis utilized data from 82 hospitals in the China Cardiovascular Association registry. We enrolled 22,473 adults with NSTEMI and concomitant CKD (KDIGO criteria) between January 2010 and March 2024. Participants were stratified into two cohorts: Early-stage CKD (stages 1-2, eGFR ≥60 mL/min/1.73m²); Advanced CKD (stages 3-5 or dialysis dependence). Propensity score matching (PSM) in a 1:1 ratio was applied to match patients who received PCI with those who did not. After matching, 13,469 patients in the CKD 1–2 group formed 3,179 matched pairs, while 9,004 patients in the CKD 3–5 + dialysis group formed 1,232 matched pairs. In the matched datasets, Cox proportional hazards regression models were used to estimate the independent effect of PCI on the primary endpoint (NACE) and its components (cardiac death, myocardial infarction, all-cause mortality, stroke, and bleeding). Interaction analyses were further performed to assess potential synergistic or antagonistic effects between CKD stage and PCI on NACE. Results In the CKD 3–5 group, patients undergoing PCI experienced significantly lower risks of NACE (HR = 0.577, p < 0.001), cardiac death (HR = 0.512, p < 0.001), myocardial infarction (HR = 0.534, p < 0.001), and all-cause mortality (HR = 0.529, p < 0.001), with no significant differences in bleeding (p = 0.461) or new-onset stroke (p = 0.639). In the CKD 1–2 group, PCI also conferred notable benefits regarding NACE (HR = 0.597, p < 0.001), cardiac death (HR = 0.423, p < 0.001), myocardial infarction (HR = 0.567, p < 0.001), and all-cause mortality (HR = 0.447, p < 0.001). Interaction analyses demonstrated statistically significant additive and multiplicative interactions for NACE (RERI = 0.052; AP = 0.085; S = 0.921; multiplicative interaction OR = 1.169; p < 0.05). Conclusion In NSTEMI patients with CKD, PCI is associated with reduced risks of NACE, cardiac death, myocardial infarction, and all-cause mortality, regardless of whether patients are in CKD stages 1–2 or 3–5. Interaction analyses suggest that PCI provides sustained and significant benefits across different CKD stages. Furthermore, as eGFR increases, the risk of NACE decreases accordingly.Kaplan-Meier curves for the l-year Multivariate Cox Regression Analysis

To construct machine learning prediction model for sepsis-associated encephalopathy (SAE), and analyze the application value of the model on early identification of SAE risk in elderly septic patients. Patients aged over 60 years with a primary diagnosis of sepsis admitted to intensive care unit (ICU) from 2008 to 2023 were selected from Medical Information Mart for Intensive Care-IV 2.2 (MIMIC-IV 2.2). Demographic variables, disease severity scores, comorbidities, interventions, laboratory indicators, and hospitalization details were collected. Key factors associated with SAE were identified using univariate Logistic regression analysis. The data were randomly divided into training and validation sets in a 7 : 3 ratio. Multivariable Logistic regression analysis was conducted in the training set and visualized using a nomogram model for prediction of SAE. The discrimination of the model was evaluated in the validation set using the receiver operator characteristic curve (ROC curve), and its calibration was assessed using calibration curve. Furthermore, multiple machine learning algorithms, including multi-layer perceptron (MLP), support vector machine (SVM), naive bayes (NB), gradient boosting machine (GBM), random forest (RF), and extreme gradient boosting (XGB), were constructed in the training set. Their predictive performance was subsequently evaluated on the validation set. Taking the XGB model as an example, the interpretability of the model through the SHapley Additive exPlanations (SHAP) algorithm was enhanced to identify the key predictive factors and their contributions. A total of 2 204 septic patients were finally enrolled, of whom 840 developed SAE (38.1%). A total of 21 variables associated with SAE were screened through univariate Logistic regression analysis. Multivariable Logistic regression analysis showed that endotracheal intubation [odds ratio (OR) = 0.40, 95% confidence interval (95%CI) was 0.19-0.88, P < 0.001], oxygen therapy (OR = 0.76, 95%CI was 0.53-0.95, P = 0.023), tracheotomy (OR = 0.20, 95%CI was 0.07-0.53, P < 0.001), continuous renal replacement therapy (CRRT; OR = 0.32, 95%CI was 0.15-0.70, P < 0.001), cerebrovascular disease (OR = 0.31, 95%CI was 0.16-0.60, P < 0.001), rheumatic disease (OR = 0.44, 95%CI was 0.19-0.99, P < 0.001), male (OR = 0.68, 95%CI was 0.54-0.86, P = 0.001), and maximum anion gap (AG; OR = 0.95, 95%CI was 0.93-0.97, P < 0.001) were associated with an decreased probability of SAE, and age (OR = 1.05, 95%CI was 1.03-1.06, P < 0.001), acute physiology score III (APSIII; OR = 1.02, 95%CI was 1.01-1.02, P < 0.001), Oxford acute severity of illness score (OASIS; OR = 1.04, 95%CI was 1.03-1.06, P < 0.001), and length of hospital stay (OR = 1.01, 95%CI was 1.01-1.02, P < 0.001) were associated with an increased probability of SAE. A nomogram model was constructed based on these variables. In the validation set, ROC curve analysis showed that the model achieved an area under the ROC curve (AUC) of 0.723, and the calibration curve showed good consistency between the predicted probability of the model and the observed probability. Among the machine learning algorithms, including MLP, SVM, NB, GBM, RF, and XGB, the SVM model and RF model demonstrated relatively good predictive performance, with AUC of 0.748 and 0.739, respectively, and the sensitivity was both exceeding 85%. The predictive performance of the XGB model was explained through SHAP analysis, and the results indicated that APSIII score (SHAP value was 0.871), age (SHAP value was 0.521), and OASIS score (SHAP value was 0.443) were important factors affecting the predictive performance of the model. The machine learning-based SAE prediction model exhibits good predictive capability and holds significant application value for the early identification of SAE risk in elderly septic patients.

Radial versus femoral access and bivalirudin versus unfractionated heparin in invasively managed patients with acute coronary syndrome (MATRIX): final 1-year results of a multicentre, randomised controlled trial

IntroductionThis study aimed to develop a diabetic retinopathy (DR) Prediction model using various machine learning algorithms incorporating the novel predictor Triglyceride-glucose index (TyG). Furthermore, the model was interpreted using the SHapley Additive exPlanations (SHAP) method.MethodReal-world data were collected from a general hospital in a major city and a county clinic, then divided into the DR Group (1392) and non-DR group (2358). Baseline data were collected, and variables were selected using Recursive Feature Elimination with Cross-Validation (RFECV). The performance of five machine learning algorithms, including Logistic Regression model (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and XGBoost (XGB), was assessed based on accuracy, sensitivity, specificity, and Area Under the Curve (AUC) of the Receiver Operating characteristic Curve (ROC). The optimal model was interpreted using SHAP.ResultLVM and LR demonstrated superior performance in both the test set and training set (ROC, 0.85 and 0.82, respectively). The top five predictors identified by SHAP analysis included TyG, Insulin therapy, HbA1c, Diabetes Course, HDL. HDL was identified as a protective factor, while the remaining factors were associated with retinopathy.ConclusionLR and SVM demonstrated the best performance. To our knowledge, this is the first machine learning-based DR prediction model integrating the triglyceride-glucose index (TyG) as a core predictor, overcoming limitations of insulin resistance (IR) assessment in resource-limited settings. TyG provides a cost-effective alternative to conventional IR biomarkers (e.g., HOMA-IR), enabling practical DR risk stratification in primary care.

Abbreviated or Standard Antiplatelet Therapy After PCI in Diabetic Patients at High Bleeding Risk

BACKGROUNDThe International Study Group of Pancreatic Surgery has established the definition and grading system for postpancreatectomy acute pancreatitis (PPAP). There are no established machine learning models for predicting PPAP following pancreaticoduodenectomy (PD).AIMTo explore the predictive model of PPAP, and test its predictive efficacy to guide the clinical work.METHODSClinical data from consecutive patients who underwent PD between 2016 and 2024 were retrospectively collected. An analysis of PPAP risk factors was performed, various machine learning algorithms [logistic regression, random forest, gradient boosting decision tree, extreme gradient boosting, light gradient boosting machine, and category boosting (CatBoost)] were utilized to develop predictive models. Recursive feature elimination was employed to select several variables to achieve the optimal machine algorithm.RESULTSThe study included 381 patients, of whom 88 (23.09%) developed PPAP. PPAP patients exhibited a significantly higher incidence of postoperative pancreatic fistula (55.68% vs 14.68%, P < 0.001), grade C postoperative pancreatic fistula (9.09% vs 1.37%, P = 0.001). The CatBoost algorithm outperformed other algorithms with a mean area under the receiver operating characteristic curve of 0.859 [95% confidence interval (CI): 0.814-0.905] in the training cohort and 0.822 (95%CI: 0.717-0.927) in the testing cohort. According to shapley additive explanations analysis, pancreatic texture, main pancreatic duct diameter, body mass index, estimated blood loss, and surgery time were the most important variables based on recursive feature elimination. The CatBoost algorithm based on selected variables demonstrated superior performance, with an area under the receiver operating characteristic curve of 0.837 (95%CI: 0.788-0.886) in the training cohort and 0.812 (95%CI: 0.697-0.927) in the testing cohort.CONCLUSIONWe developed the first machine learning-based predictive model for PPAP following PD. This predictive model can assist surgeons in anticipating and managing this complication proactively.

To assess the effectiveness and safety of bivalirudin compared with heparin monotherapy in elderly patients undergoing percutaneous coronary intervention (PCI). Bivalirudin is recommended for periprocedural use in patients undergoing PCI who are of high bleeding risk. However, its safe and efficacious use in elderly patients, a typical high bleeding risk cohort, in real world practice is yet to be reported. In this single center, real-world observational study, 4736 consecutive elderly patients who underwent PCI were enrolled. Of these, 1240 were treated with bivalirudin and 3496 with heparin according to the periprocedural anticoagulation strategies of PCI. The primary outcome was 12-month net adverse clinical events (NACE) defined as a composite of cardiac death, myocardial infarction, stroke, revascularization, or any bleeding. Propensity score matching (PSM) was used to balance baseline characteristics between groups. After PSM, bivalirudin was found to be associated with lower rates of NACE (19.1% vs. 24.7%, p = 0.002), cardiac death (2.7% vs. 4.3%, p = 0.038), and any bleeding (10.0% vs. 12.9%, p = 0.023) compared to heparin monotherapy. No differences were found in the incidences of myocardial infarction, stroke, revascularization, stent thrombosis (0.1% vs. 0.1%, p = 1.000), and major bleedings (0.5% vs. 0.5%, p = 1.000) between the two patient groups. In this real-world observational study, periprocedural use of bivalirudin in elderly patients who underwent PCI was associated with less cardiac death and any bleeding compared to heparin monotherapy, without increased risk of stent thrombosis.

Tree-based machine learning (ML) algorithms, such as Extra Trees (ET), Random Forest (RF), Gradient Boosting Machine (GBM), and XGBoost (XGB) are among the most widely used in early drug discovery, given their versatility and performance. However, models based on these algorithms often suffer from misclassification and reduced interpretability issues, which limit their applicability in practice. To address these challenges, several approaches have been proposed, including the use of SHapley Additive Explanations (SHAP). While SHAP values are commonly used to elucidate the importance of features driving models' predictions, they can also be employed in strategies to improve their prediction performance. Building on these premises, we propose a novel approach that integrates SHAP and features value analyses to reduce misclassification in model predictions. Specifically, we benchmarked classifiers based on ET, RF, GBM, and XGB algorithms using data sets of compounds with known antiproliferative activity against three prostate cancer (PC) cell lines (i.e., PC3, LNCaP, and DU-145). The best-performing models, based on RDKit and ECFP4 descriptors with GBM and XGB algorithms, achieved MCC values above 0.58 and F1-score above 0.8 across all data sets, demonstrating satisfactory accuracy and precision. Analyses of SHAP values revealed that many misclassified compounds possess feature values that fall within the range typically associated with the opposite class. Based on these findings, we developed a misclassification-detection framework using four filtering rules, which we termed "RAW", SHAP, "RAW OR SHAP", and "RAW AND SHAP". These filtering rules successfully identified several potentially misclassified predictions, with the "RAW OR SHAP" rule retrieving up to 21%, 23%, and 63% of misclassified compounds in the PC3, DU-145, and LNCaP test sets, respectively. The developed flagging rules enable the systematic exclusion of likely misclassified compounds, even across progressively higher prediction confidence levels, thus providing a valuable approach to improve classifier performance in virtual screening applications.

Machine learning provides a flexible technique to predict the survival of patients who are admitted to hospital as emergency admissions. Mortality prediction is a central component of emergency patient quality of care and this can act as an indicator of severity to determine who needs prioritized care. Machine learning-based models, as opposed to human-crafted severity score systems, allow for much more complex and updateable models to be developed based on a larger set of input data attributes. While various studies of machine learning-based predictive models for predicting inpatient mortality have been carried out there is little literature on performance maintenance of these models. Determining the performance maintenance of these models over time determines how reliably they can be utilized into the future and for how long. The best performing model in this study achieve&#x2019;s an AUC of 0.86 upon training and is able to maintain a similarly high AUC of 0.845 as of the end of the period of performance maintenance evaluation nine months later. This is the first paper that the authors are aware of to consider and measure relative performance maintenance of machine learning-based models for emergency admission mortality prediction.

Background: Bleeding complications following Percutaneous Coronary Interventions (PCI) occur in 2-6% of cases. Designed to provide rapid hemostasis, vascular closure devices (VCDs) have been found to have the largest benefit in high bleed risk cases but are utilized less often in these patients. Decision support tools and real-time feedback may impact physician practice patterns and increase utilization of VCDs in cases with high bleed risk. Methods: In May 2012, a real-time decision support tool was introduced to prospectively evaluate PCI bleed risk and stratify patients into high, intermediate, and low bleeding risk based on a validated model using data from the National Cardiovascular Data Registry. In January 2014, a group goal of 50% utilization of VCDs in high-bleed risk patients was set and weekly reports to individual interventional cardiologists performing PCI were provided. Group and individual physician practice patterns before and after the goal and real-time feedback were assessed. Results: From May 2012 to August 2015, 5,285 patients received PCI, including 1,399 (26.5%) who were classified as high bleed risk. Prior to the group utilization goal and real-time feedback being implemented, VCD use in high bleed risk patients was 40.3% (292 of 725) and utilization by individual provider ranged from 7.4% (5 of 68) to 83.1% (103 of 124) of cases. After implementation of the group utilization goal and real-time feedback, 74.2% (500 of 674) high bleed risk received VCDs and utilization by individual provider ranged from 61.4% (43 of 70) to 87.9% (87 of 99) of cases. Therefore, the physician group saw a relative increase of 84.1% in VCD utilization (40.3% vs. 74.2%; p&lt;0.001). Additionally, among individual providers, VCD use significantly increased for 7 of 8 physicians (see figure) and the group practice variation in use of VCDs decreased (standard deviation of VCD utilization across physicians decreased from 22.9% to 8.7%, p=0.021). Conclusions: Implementation of real-time feedback and clearly defined group goals for the use of VCDs in high bleed risk cases significantly increased VCD use in patients known to have the greatest benefit. These interventions also reduced the variation in care for high bleed risk patients.

BackgroundPrevious models for predicting delirium after cardiac surgery remained inadequate. This study aimed to develop and validate a machine learning-based prediction model for postoperative delirium (POD) in cardiac valve surgery patients.MethodsThe electronic medical information of the cardiac surgical intensive care unit (CSICU) was extracted from a tertiary and major referral hospital in southern China over 1 year, from June 2019 to June 2020. A total of 507 patients admitted to the CSICU after cardiac valve surgery were included in this study. Seven classical machine learning algorithms (Random Forest Classifier, Logistic Regression, Support Vector Machine Classifier, K-nearest Neighbors Classifier, Gaussian Naive Bayes, Gradient Boosting Decision Tree, and Perceptron.) were used to develop delirium prediction models under full (q = 31) and selected (q = 19) feature sets, respectively.ResultThe Random Forest classifier performs exceptionally well in both feature datasets, with an Area Under the Curve (AUC) of 0.92 for the full feature dataset and an AUC of 0.86 for the selected feature dataset. Additionally, it achieves a relatively lower Expected Calibration Error (ECE) and the highest Average Precision (AP), with an AP of 0.80 for the full feature dataset and an AP of 0.73 for the selected feature dataset. To further evaluate the best-performing Random Forest classifier, SHAP (Shapley Additive Explanations) was used, and the importance matrix plot, scatter plots, and summary plots were generated.ConclusionsWe established machine learning-based prediction models to predict POD in patients undergoing cardiac valve surgery. The random forest model has the best predictive performance in prediction and can help improve the prognosis of patients with POD.

This study aimed to develop and validate a machine learning-based predictive model for individualized assessment and management of chemotherapy-induced myelosuppression (CIM) in patients with colorectal cancer (CRC). A total of 450 patients with CRC undergoing chemotherapy were retrospectively enrolled in the training cohort, and an additional 150 patients were included for external validation. Feature selection was performed using the Least Absolute Shrinkage and Selection Operator (LASSO) regression. Three machine learning algorithms [Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)] were applied to construct predictive models. Model performance was assessed using multiple metrics, including accuracy, area under the receiver operating characteristic curve (AUC), F1 score, sensitivity, and specificity. The SHapley Additive exPlanations (SHAP) method was employed to rank and interpret the importance of predictive features. In the training cohort, 52.4% of patients developed CIM. The feature selection process identified 19 significant variables that were incorporated into the predictive models. Both LR and RF demonstrated optimal performance, with an AUC of 0.83 and an accuracy of 0.76 in the training set. In the test set, RF continued to outperform other models, achieving an AUC of 0.77 and an accuracy of 0.71. External validation confirmed the robustness of the RF model, which achieved an AUC of 0.93 (95% CI: 0.89-0.97), an accuracy of 0.89, a sensitivity of 0.86, and a specificity of 0.92. SHAP analysis revealed that the most important predictors included hematological parameters, nutritional risk score (NRS2002), and a history of radiotherapy. The RF-based machine learning model demonstrated high accuracy and strong external validation capability for predicting the risk of CIM in CRC patients.

Complete mesocolic excision (CME) is a technically complex and highly invasive surgical approach, and patients undergoing CME are consequently at elevated risk of postoperative acute kidney injury (AKI). Despite this vulnerability, reliable tools for individualized AKI risk prediction in this population remain unavailable. Moreover, the contributions of perioperative inflammatory responses and blood transfusion to AKI pathogenesis have not been comprehensively examined. Here, we sought to develop and validate a multicenter, machine learning-based model to predict AKI following CME and to delineate the relative impact of inflammation- and transfusion-related determinants. We retrospectively enrolled patients with colon cancer who underwent CME between 2010 and 2020 at five tertiary referral centers. Patients were allocated to an internal cohort or an external validation cohort according to hospital of origin. The internal cohort was randomly divided into training and validation subsets in a 7:3 ratio. Feature selection and model construction were performed using multivariable analyses and five machine learning algorithms: Extreme Gradient Boosting (XGBoost), Random Forest, Support Vector Machine, k-Nearest Neighbors (KNN), and Multilayer Perceptron (MLP). Model performance was assessed using receiver operating characteristic (ROC) curves, calibration plots, decision curve analysis (DCA), and k-fold cross-validation, followed by independent external validation. Model interpretability and the quantitative contribution of inflammatory and transfusion variables were evaluated using SHapley Additive exPlanations (SHAP). Among the evaluated models, XGBoost achieved the most favorable performance, exhibiting superior discrimination, calibration, clinical utility, and generalizability, with area under the curve (AUC) values of 0.92 in the training set, 0.88 in the validation set, and 0.922 in the external cohort. SHAP analysis highlighted tumor size, operative duration, preoperative anemia, postoperative neutrophil-to-lymphocyte ratio (NLR), intraoperative blood loss, C-reactive protein (CRP), intraoperative hypoxemia, and perioperative blood transfusion as the dominant predictors of AKI. Notably, inflammation-related markers (NLR and CRP) and transfusion-related factors exerted a substantial influence on AKI risk. We established an interpretable, multicenter machine learning-based model with high predictive accuracy, robustness, and clinical relevance for AKI following CME. Our findings identify perioperative inflammation and blood transfusion as key drivers of postoperative AKI, offering mechanistic insight and a foundation for early risk stratification and targeted preventive strategies in high-risk patients.