Machine Learning-Based Prediction of Acute Kidney Injury Following Pediatric Cardiac Surgery: Model Development and Validation Study

Machine Learning Models With Prognostic Implications for Predicting Gastrointestinal Bleeding After Coronary Artery Bypass Grafting and Guiding Personalized Medicine: Multicenter Cohort Study

BACKGROUND

Gastrointestinal bleeding is a serious adverse event of coronary artery bypass grafting and lacks tailored risk assessment tools for personalized prevention.

OBJECTIVE

This study aims to develop and validate predictive models to assess the risk of gastrointestinal bleeding after coronary artery bypass grafting (GIBCG) and to guide personalized prevention.

METHODS

Participants were recruited from 4 medical centers, including a prospective cohort and the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. From an initial cohort of 18,938 patients, 16,440 were included in the final analysis after applying the exclusion criteria. Thirty combinations of machine learning algorithms were compared, and the optimal model was selected based on integrated performance metrics, including the area under the receiver operating characteristic curve (AUROC) and the Brier score. This model was then developed into a web-based risk prediction calculator. The Shapley Additive Explanations method was used to provide both global and local explanations for the predictions.

RESULTS

The model was developed using data from 3 centers and a prospective cohort (n=13,399) and validated on the Drum Tower cohort (n=2745) and the MIMIC cohort (n=296). The optimal model, based on 15 easily accessible admission features, demonstrated an AUROC of 0.8482 (95% CI 0.8328-0.8618) in the derivation cohort. In external validation, the AUROC was 0.8513 (95% CI 0.8221-0.8782) for the Drum Tower cohort and 0.7811 (95% CI 0.7275-0.8343) for the MIMIC cohort. The analysis indicated that high-risk patients identified by the model had a significantly increased mortality risk (odds ratio 2.98, 95% CI 1.784-4.978; P<.001). For these high-risk populations, preoperative use of proton pump inhibitors was an independent protective factor against the occurrence of GIBCG. By contrast, dual antiplatelet therapy and oral anticoagulants were identified as independent risk factors. However, in low-risk populations, the use of proton pump inhibitors (χ21=0.13, P=.72), dual antiplatelet therapy (χ21=0.38, P=.54), and oral anticoagulants (χ21=0.15, P=.69) were not significantly associated with the occurrence of GIBCG.

CONCLUSIONS

Our machine learning model accurately identified patients at high risk of GIBCG, who had a poor prognosis. This approach can aid in early risk stratification and personalized prevention.

TRIAL REGISTRATION

Chinese Clinical Registry Center ChiCTR2400086050; http://www.chictr.org.cn/showproj.html?proj=226129.

Development and Validation of a Nomogram for Predicting Acute Kidney Injury in Pediatric Patients Undergoing Cardiac Surgery

Acute kidney injury (AKI) is a common complication after cardiac surgery and associated with adverse outcomes. The purpose of this study is to construct a nomogram to predict the probability of postoperative AKI in pediatric patients undergoing cardiac surgery. We conducted a single-center retrospective cohort study of 1137 children having cardiac surgery under cardiopulmonary bypass. We randomly divided the included patients into development and validation cohorts at a ratio of 7:3. The least absolute shrinkage and selection operator regression model was used for feature selection. We constructed a multivariable logistic regression model to select predictors and develop a nomogram to predict AKI risk. Discrimination, calibration and clinical benefit of the final prediction model were evaluated in the development and validation cohorts. A simple nomogram was developed to predict risk of postoperative AKI using six predictors including age at operation, cyanosis, CPB duration longer than 120 min, cross-clamp time, baseline albumin and baseline creatinine levels. The area under the receiver operator characteristic curve of the nomogram was 0.739 (95% CI 0.693-0.786) and 0.755 (95% CI 0.694-0.816) for the development and validation cohort, respectively. The calibration curve showed a good correlation between predicted and observed risk of postoperative AKI. Decision curve analysis presented great clinical benefit of the nomogram. This novel nomogram for predicting AKI after pediatric cardiac surgery showed good discrimination, calibration and clinical practicability.

Development and validation of a risk prediction model for acute kidney injury in coronary artery disease

BACKGROUND

Acute Kidney Injury (AKI) is a sudden and often reversible condition characterized by rapid kidney function reduction, posing significant risks to coronary artery disease (CAD) patients. This study focuses on developing accurate predictive models to improve the early detection and prognosis of AKI in CAD patients.

METHODS

We used Electronic Health Records (EHRs) from a nationwide CAD registry including 54 429 patients. Initially, univariate analysis identified potential predictors. Subsequently, a stepwise multivariate logistic model integrated clinical significance and data distribution. To refine predictor selection, we applied a random forest algorithm. The top 10 variables, including admission to the surgical department, EGFR, hemoglobin, and others, were incorporated into a logistic regression-based prediction model. Model performance was assessed using the area under the curve (AUC) and calibration analysis, and a nomogram was developed for practical application.

RESULTS

During hospitalization, 2,112 (3.88%) patients in the overall population of both the development and validation groups experienced AKI within 30 days. The final prediction model exhibited strong discrimination with an AUC of 0.867 (95% CI: 0.858 to 0.876) and well calibration capability in both the development and validation groups. Key predictors included surgical department admission, eGFR, hemoglobin, chronic kidney disease history, male sex, white blood cell count, age, left ventricular ejection fraction, acute myocardial infarction at admission, and congestive heart failure history. Bootstrap resampling confirmed model stability (Harrell's optimism-correct AUC = 0.866). The nomogram provided a practical tool for AKI risk assessment.

CONCLUSION

This study introduced a refined AKI risk prediction model for CAD patients. This model showed adaptability to subgroups and held the potential for early AKI alerts and personalized interventions, thereby enhancing patient care.

Major Adverse Kidney Events in Hospitalized Older Patients With Acute Kidney Injury: Machine Learning-Based Model Development and Validation Study

BACKGROUND

Acute kidney injury (AKI) is a common complication in hospitalized older patients, associated with increased morbidity, mortality, and health care costs. Major adverse kidney events within 30 days (MAKE30), a composite of death, new renal replacement therapy, or persistent renal dysfunction, has been recommended as a patient-centered endpoint for clinical trials involving AKI.

OBJECTIVE

This study aimed to develop and validate a machine learning-based model to predict MAKE30 in hospitalized older patients with AKI.

METHODS

A total of 4266 older patients (aged ≥ 65 years) with AKI admitted to the Second Xiangya Hospital of Central South University from January 1, 2015, to December 31, 2020, were included and randomly divided into a training set and an internal test set in a ratio of 7:3. An additional cohort of 11,864 eligible patients from the Medical Information Mart for Intensive Care Ⅳ database served as an external test set. The Boruta algorithm was used to select the most important predictor variables from 53 candidate variables. The eXtreme Gradient Boosting algorithm was applied to establish a prediction model for MAKE30. Model discrimination was evaluated by the area under the receiver operating characteristic curve (AUROC). The SHapley Additive exPlanations method was used to interpret model predictions.

RESULTS

The overall incidence of MAKE30 in the 2 study cohorts was 28.3% (95% CI 26.9%-29.7%) and 26.7% (95% CI 25.9%-27.5%), respectively. The prediction model for MAKE30 exhibited adequate predictive performance, with an AUROC of 0.868 (95% CI 0.852-0.881) in the training set and 0.823 (95% CI 0.798-0.846) in the internal test set. Its simplified version achieved an AUROC of 0.744 (95% CI 0.735-0.754) in the external test set. The SHapley Additive exPlanations method showed that the use of vasopressors, mechanical ventilation, blood urea nitrogen level, red blood cell distribution width-coefficient of variation, and serum albumin level were closely associated with MAKE30.

CONCLUSIONS

An interpretable eXtreme Gradient Boosting model was developed and validated to predict MAKE30, which provides opportunities for risk stratification, clinical decision-making, and the conduct of clinical trials involving AKI.

TRIAL REGISTRATION

Chinese Clinical Trial Registry ChiCTR2200061610; https://tinyurl.com/3smf9nuw.

Machine learning and decision making in aortic arch repair

BACKGROUND

Decision making during aortic arch surgery regarding cannulation strategy and nadir temperature are important in reducing risk, and there is a need to determine the best individualized strategy in a data-driven fashion. Using machine learning (ML), we modeled the risk of death or stroke in elective aortic arch surgery based on patient characteristics and intraoperative decisions.

METHODS

The study cohort comprised 1323 patients from 9 institutions who underwent an elective aortic arch procedure between 2002 and 2021. A total of 69 variables were used in developing a logistic regression and XGBoost ML model trained for binary classification of mortality and stroke. Shapely additive explanations (SHAP) values were studied to determine the importance of intraoperative decisions.

RESULTS

During the study period, 3.9% of patients died and 5.4% experienced stroke. XGBoost (area under the curve [AUC], 0.77 for death, 0.87 for stroke) demonstrated better discrimination than logistic regression (AUC, 0.65 for death, 0.75 for stroke). From SHAP analysis, intraoperative decisions are 3 of the top 20 predictors of death and 6 of the top 20 predictors of stroke. Predictor weights are patient-specific and reflect the patient's preoperative characteristics and other intraoperative decisions. Patient-level simulation also demonstrates the variable contribution of each decision in the context of the other choices that are made.

CONCLUSIONS

Using ML, we can more accurately identify patients at risk of death and stroke, as well as the strategy that better reduces the risk of adverse events compared to traditional prediction models. Operative decisions made may be tailored based on a patient's specific characteristics, allowing for maximized, personalized benefit.

A Machine Learning-Based Prediction Model for Acute Kidney Injury in Patients With Community-Acquired Pneumonia: Multicenter Validation Study

BACKGROUND

Acute kidney injury (AKI) is common in patients with community-acquired pneumonia (CAP) and is associated with increased morbidity and mortality.

OBJECTIVE

This study aimed to establish and validate predictive models for AKI in hospitalized patients with CAP based on machine learning algorithms.

METHODS

We trained and externally validated 5 machine learning algorithms, including logistic regression, support vector machine, random forest, extreme gradient boosting, and deep forest (DF). Feature selection was conducted using the sliding window forward feature selection technique. Shapley additive explanations and local interpretable model-agnostic explanation techniques were applied to the optimal model for visual interpretation.

RESULTS

A total of 6371 patients with CAP met the inclusion criteria. The development of CAP-associated AKI (CAP-AKI) was recognized in 1006 (15.8%) patients. The 11 selected indicators were sex, temperature, breathing rate, diastolic blood pressure, C-reactive protein, albumin, white blood cell, hemoglobin, platelet, blood urea nitrogen, and neutrophil count. The DF model achieved the best area under the receiver operating characteristic curve (AUC) and accuracy in the internal (AUC=0.89, accuracy=0.90) and external validation sets (AUC=0.87, accuracy=0.83). Furthermore, the DF model had the best calibration among all models. In addition, a web-based prediction platform was developed to predict CAP-AKI.

CONCLUSIONS

The model described in this study is the first multicenter-validated AKI prediction model that accurately predicts CAP-AKI during hospitalization. The web-based prediction platform embedded with the DF model serves as a user-friendly tool for early identification of high-risk patients.

The anesthesiologist's guide to critically assessing machine learning research: a narrative review

Artificial Intelligence (AI), especially Machine Learning (ML), has developed systems capable of performing tasks that require human intelligence. In anesthesiology and other medical fields, AI applications can improve the precision and efficiency of daily clinical practice, and can also facilitate a personalized approach to patient care, which can lead to improved outcomes and quality of care. ML has been successfully applied in various settings of daily anesthesiology practice, such as predicting acute kidney injury, optimizing anesthetic doses, and managing postoperative nausea and vomiting. The critical evaluation of ML models in healthcare is crucial to assess their validity, safety, and clinical applicability. Evaluation metrics allow an objective statistical assessment of model performance. Tools such as Shapley Values (SHAP) help interpret how individual variables contribute to model predictions. Transparency in reporting is key in maintaining trust in these technologies and to ensure their use follows ethical principles, aiming to reduce safety concerns while also benefiting patients. Understanding evaluation metrics is essential, as they provide detailed information on model performance and their ability to discriminate between individual class rates. This article offers a comprehensive framework in assessing the validity, applicability, and limitations of models, guiding responsible and effective integration of ML technologies into clinical practice. A balance between innovation, patient safety and ethical considerations must be pursued.

Relationship between intra-operative urine output and postoperative acute kidney injury in paediatric cardiac surgery: A retrospective observational study

BACKGROUND

Intra-operative urine output (UO) has been shown to predict postoperative acute kidney injury (AKI) in adults; however, its significance in children undergoing cardiac surgery remains unknown.

OBJECTIVE

To explore the association between intra-operative UO and postoperative AKI in children with congenital heart disease.

DESIGN

A retrospective observational study.

SETTING

A tertiary hospital.

PATIENTS

Children aged >28 days and <6 years who underwent cardiac surgery at Fuwai Hospital from 1 April 2022 to 30 August 2022.

MAIN OUTCOME MEASURES

AKI was identified by the highest serum creatinine value within postoperative 7 days using Kidney Disease Improving Global Outcomes (KDIGO) criteria.

RESULTS

In total, 1184 children were included. The incidence of AKI was 23.1% (273/1184), of which 17.7% (209/1184) were stage 1, 4.2% (50/1184) were stage 2, and others were stage 3 (1.2%, 14/1184). Intra-operative UO was calculated by dividing the total intra-operative urine volume by the duration of surgery and the actual body weight measured before surgery. There was no significant difference in median [IQR] intra-operative UO between the AKI and non-AKI groups (2.6 [1.4 to 5.4] and 2.7 [1.4 to 4.9], respectively, P  = 0.791), and multivariate logistic regression analyses showed that intra-operative UO was not associated with postoperative AKI [adjusted odds ratio (OR) 0.971; 95% confidence interval (CI), 0.930 to 1.014; P  = 0.182]. Regarding the clinical importance of severe forms of AKI, we further explored the association between intra-operative UO and postoperative moderate-to-severe AKI (adjusted OR 0.914; 95% CI, 0.838 to 0.998; P  = 0.046).

CONCLUSIONS

Intra-operative UO was not associated with postoperative AKI during paediatric cardiac surgery. However, we found a significant association between UO and postoperative moderate-to-severe AKI. This suggests that reductions in intra-operative urine output below a specific threshold may be associated with postoperative renal dysfunction.

TRIAL REGISTRATION

Clinicaltrials.gov identifier: NCT05489263.

Acute kidney disease in hospitalized pediatric patients: risk prediction based on an artificial intelligence approach

BACKGROUND

Acute kidney injury (AKI) and acute kidney disease (AKD) are prevalent among pediatric patients, both linked to increased mortality and extended hospital stays. Early detection of kidney injury is crucial for improving outcomes. This study presents a machine learning-based risk prediction model for AKI and AKD in pediatric patients, enabling personalized risk predictions.

METHODS

Data from 2,346 hospitalized pediatric patients, collected between January 2020 and January 2023, were divided into an 85% training set and a 15% test set. Predictive models were constructed using eight machine learning algorithms and two ensemble algorithms, with the optimal model identified through AUROC. SHAP was used to interpret the model, and an online prediction tool was developed with Streamlit to predict AKI and AKD.

RESULTS

The incidence of AKI and AKD were 14.90% and 16.26%, respectively. Patients with AKD combined with AKI had the highest mortality rate, at 6.94%, when analyzed by renal function trajectories. The LightGBM algorithm showed superior predictive performance for both AKI and AKD (AUROC: 0.813, 0.744). SHAP identified top predictors for AKI as serum creatinine, white blood cell count, neutrophil count, and lactate dehydrogenase, while key predictors for AKD included proton pump inhibitor, blood glucose, hemoglobin, and AKI grade.

CONCLUSION

The high incidence of AKI and AKD among hospitalized children warrants attention. Renal function trajectories are strongly associated with prognosis. Supported by a web-based tool, machine learning models can effectively predict AKI and AKD, facilitating early identification of high-risk pediatric patients and potentially improving outcomes.

Construction of a Clinical Prediction Model for Complications After Femoral Head Replacement Surgery

Background

While femoral head replacement is widely used with remarkable efficacy, the complexity and diversity of postoperative complications pose a serious prognostic challenge. There is an urgent need to develop a clinical prediction model that can integrate multiple factors and accurately predict the risk of postoperative complications to guide clinical practice and optimize patient management strategies. This study is dedicated to constructing a postoperative complication prediction model based on statistics and machine learning techniques, in order to provide patients with a safer and more effective treatment experience.

Methods

A total of 186 patients who underwent femoral head replacement in the Orthopedic Department of our hospital were collected in this study. Forty-two of the patients had at least one postoperative complication, and 144 had no complications. The preoperative and postoperative data of patients were collected separately and medical history was collected to study the correlation factors affecting the occurrence of postoperative complications in patients and to establish a prediction model.

Results

Possibly relevant factors were included in a one-way logistic regression, which included the patient's gender, age, body mass index, preoperative diagnosis of the mode of injury, osteoporosis or lack thereof, as well as medical history, surgical-related information, and laboratory indices. After analyzing the results, it was concluded that operation time, alanine transaminase (ALT), aspartate aminotransferase (AST), white blood cell count, serum albumin, and osteoporosis, were the risk factors affecting the development of complications after femoral head replacement in patients (P < 0.2). The data obtained were further included in a multifactorial regression, and the results showed that operation time, AST, white blood cell count, serum albumin, and osteoporosis were independent risk factors for complications after the patients underwent femoral head replacement (P < 0.05).

Conclusion

Based on the results of this study, five factors, including duration of surgery, AST, white blood cell count, serum albumin, and osteoporosis, were identified as independent risk factors for complications after patients underwent femoral head replacement. In addition, the prediction model developed in this study has a high scientific and clinical application value, providing clinicians and patients with an important tool for assessing the risk of complications after affected femoral head replacement.

Machine learning approach to predict acute kidney injury among patients undergoing multi-level spinal posterior instrumented fusion

Background

Acute kidney injury (AKI) after spinal fusion is a significant morbidity that can lead to poor post-surgical outcomes. Identifying AKI risk factors and developing a risk model can raise surgeons' awareness and allow them to take actions to mitigate the risks. The objective of the current study is to develop machine learning (ML) models to assess patient risk factors predisposing to AKI after posterior spinal instrumented fusion.

Methods

Data was collected from the IBM MarketScan Database (2009-2021) for patients >18 years old who underwent spinal fusion with posterior instrumentation (3-6 levels). AKI incidence (defined by the International Classification of Diseases codes) was recorded 90-day post-surgery. Risk factors for AKI were investigated and compared through several ML models including logistic regression, linear support vector machine (LSVM), random forest, extreme gradient boosting (XGBoost), and neural networks.

Results

Among the 141,697 patients who underwent fusion with posterior instrumentation (3-6 levels), the overall rate of 90-day AKI was 2.96%. We discovered that the logistic regression model and LSVM demonstrated the best predictions with area under the curve (AUC) values of 0.75. The most important AKI prediction features included chronic renal disease, hypertension, diabetes mellitus ± complications, older age (>50 years old), and congestive heart failure. Patients who did not have these five key risk factors had a 90-day AKI rate of 0.29%. Patients who had an increasing number of key risk factors subsequently had higher risks of postoperative AKI.

Conclusions

The analysis of the data with different ML models identified 5 key variables that are most closely associated with AKI: chronic renal disease, hypertension, diabetes mellitus ± complications, older age (>50 years old), and congestive heart failure. These variables constitute a simple risk calculator with additive odds ratio ranging from 3.38 (1 risk factor) to 91.10 (5 risk factors) over 90 days after posterior spinal fusion surgery. These findings can help surgeons risk-stratify their patients for AKI risk, and potentially guide post-operative monitoring and medical management.

Application of interpretable machine learning algorithms to predict acute kidney injury in patients with cerebral infarction in ICU

BACKGROUND

Acute kidney injury (AKI) is not only a complication but also a serious threat to patients with cerebral infarction (CI). This study aimed to explore the application of interpretable machine learning algorithms in predicting AKI in patients with cerebral infarction.

METHODS

The study included 3920 patients with CI admitted to the Intensive Care Unit and Emergency Medicine of the Central Hospital of Lishui City, Zhejiang Province. Nine machine learning techniques, including XGBoost, logistics, LightGBM, random forest (RF), AdaBoost, GaussianNB (GNB), Multi-Layer Perceptron (MLP), support vector machine (SVM), and k-nearest neighbors (KNN) classification, were used to develop a predictive model for AKI in these patients. SHapley Additive exPlanations (SHAP) analysis provided visual explanations for each patient. Finally, model effectiveness was assessed using metrics such as average precision (AP), sensitivity, specificity, accuracy, F1 score, precision-recall (PR) curve, calibration plot, and decision curve analysis (DCA).

RESULTS

The XGBoost model performed better in the internal validation set and the external validation set, with an AUC of 0.940 and 0.887, respectively. The five most important variables in the model were, in order, glomerular filtration rate, low-density lipoprotein, total cholesterol, hemiplegia and serum kalium.

CONCLUSION

This study demonstrates the potential of interpretable machine learning algorithms in predicting CI patients with AKI.

Predicting pediatric cardiac surgery-associated acute kidney injury using machine learning

BACKGROUND

Prediction of cardiac surgery-associated acute kidney injury (CS-AKI) in pediatric patients is crucial to improve outcomes and guide clinical decision-making. This study aimed to develop a supervised machine learning (ML) model for predicting moderate to severe CS-AKI at postoperative day 2 (POD2).

METHODS

This retrospective cohort study analyzed data from 402 pediatric patients who underwent cardiac surgery at a university-affiliated children's hospital, who were separated into an 80%-20% train-test split. The ML model utilized demographic, preoperative, intraoperative, and POD0 clinical and laboratory data to predict moderate to severe AKI categorized by Kidney Disease: Improving Global Outcomes (KDIGO) stage 2 or 3 at POD2. Input feature importance was assessed by SHapley Additive exPlanations (SHAP) values. Model performance was evaluated using accuracy, area under the receiver operating curve (AUROC), precision, recall, area under the precision-recall curve (AUPRC), F1-score, and Brier score.

RESULTS

Overall, 13.7% of children in the test set experienced moderate to severe AKI. The ML model achieved promising performance, with accuracy of 0.91 (95% CI: 0.82-1.00), AUROC of 0.88 (95% CI: 0.72-1.00), precision of 0.92 (95% CI: 0.70-1.00), recall of 0.63 (95% CI: 0.32-0.96), AUPRC of 0.81 (95% CI: 0.61-1.00), F1-score of 0.73 (95% CI: 0.46-0.99), and Brier score loss of 0.09 (95% CI: 0.00-0.17). The top ten most important features assessed by SHAP analyses in this model were preoperative serum creatinine, surgery duration, POD0 serum pH, POD0 lactate, cardiopulmonary bypass duration, POD0 vasoactive inotropic score, sex, POD0 hematocrit, preoperative weight, and POD0 serum creatinine.

CONCLUSIONS

A supervised ML model utilizing demographic, preoperative, intraoperative, and immediate postoperative clinical and laboratory data showed promising performance in predicting moderate to severe CS-AKI at POD2 in pediatric patients.

Specific patterns and potential risk factors to predict 3-year risk of death among non-cancer patients with advanced chronic kidney disease by machine learning

Chronic kidney disease (CKD) is a major public health concern. But there are limited machine learning studies on non-cancer patients with advanced CKD, and the results of machine learning studies on cancer patients with CKD may not apply directly on non-cancer patients. We aimed to conduct a comprehensive investigation of risk factors for a 3-year risk of death among non-cancer advanced CKD patients with an estimated glomerular filtration rate < 60.0 mL/min/1.73m2 by several machine learning algorithms. In this retrospective cohort study, we collected data from in-hospital and emergency care patients from 2 hospitals in Taiwan from 2009 to 2019, including their international classification of disease at admission and laboratory data from the hospital's electronic medical records (EMRs). Several machine learning algorithms were used to analyze the potential impact and degree of influence of each factor on mortality and survival. Data from 2 hospitals in northern Taiwan were collected with 6565 enrolled patients. After data cleaning, 26 risk factors and approximately 3887 advanced CKD patients from Shuang Ho Hospital were used as the training set. The validation set contained 2299 patients from Taipei Medical University Hospital. Predictive variables, such as albumin, PT-INR, and age, were the top 3 significant risk factors with paramount influence on mortality prediction. In the receiver operating characteristic curve, the random forest had the highest values for accuracy above 0.80. MLP, and Adaboost had better performance on sensitivity and F1-score compared to other methods. Additionally, SVM with linear kernel function had the highest specificity of 0.9983, while its sensitivity and F1-score were poor. Logistic regression had the best performance, with an area under the curve of 0.8527. Evaluating Taiwanese advanced CKD patients' EMRs could provide physicians with a good approximation of the patients' 3-year risk of death by machine learning algorithms.

Machine-learning predictions for acute kidney injuries after coronary artery bypass grafting: a real-life muticenter retrospective cohort study

BACKGROUND

Acute kidney injury (AKI) after coronary artery bypass grafting (CABG) surgery is associated with poor outcomes. The objective of this study was to apply a new machine learning (ML) method to establish prediction models of AKI after CABG.

METHODS

A total of 2,780 patients from two medical centers in East China who underwent primary isolated CABG were enrolled. The dataset was randomly divided for model training (80%) and model testing (20%). Four ML models based on LightGBM, Support vector machine (SVM), Softmax and random forest (RF) algorithms respectively were established in Python. A total of 2,051 patients from two other medical centers were assigned to an external validation group to verify the performances of the ML prediction models. The models were evaluated using the area under the receiver operating characteristics curve (AUC), Hosmer-Lemeshow goodness-of-fit statistic, Bland-Altman plots, and decision curve analysis. The outcome of the LightGBM model was interpreted using SHapley Additive exPlanations (SHAP).

RESULTS

The incidence of postoperative AKI in the modeling group was 13.4%. Similarly, the incidence of postoperative AKI of the two medical centers in the external validation group was 8.2% and 13.6% respectively. LightGBM performed the best in predicting, with an AUC of 0.8027 in internal validation group and 0.8798 and 0.7801 in the external validation group. The SHAP revealed the top 20 predictors of postoperative AKI ranked according to the importance, and the top three features on prediction were the serum creatinine in the first 24 h after operation, the last preoperative Scr level, and body surface area.

CONCLUSION

This study provides a LightGBM predictive model that can make accurate predictions for AKI after CABG surgery. The LightGBM model shows good predictive ability in both internal and external validation. It can help cardiac surgeons identify high-risk patients who may experience AKI after CABG surgery.

Predicting Colorectal Cancer Survival Using Time-to-Event Machine Learning: Retrospective Cohort Study

BACKGROUND

Machine learning (ML) methods have shown great potential in predicting colorectal cancer (CRC) survival. However, the ML models introduced thus far have mainly focused on binary outcomes and have not considered the time-to-event nature of this type of modeling.

OBJECTIVE

This study aims to evaluate the performance of ML approaches for modeling time-to-event survival data and develop transparent models for predicting CRC-specific survival.

METHODS

The data set used in this retrospective cohort study contains information on patients who were newly diagnosed with CRC between December 28, 2012, and December 27, 2019, at West China Hospital, Sichuan University. We assessed the performance of 6 representative ML models, including random survival forest (RSF), gradient boosting machine (GBM), DeepSurv, DeepHit, neural net-extended time-dependent Cox (or Cox-Time), and neural multitask logistic regression (N-MTLR) in predicting CRC-specific survival. Multiple imputation by chained equations method was applied to handle missing values in variables. Multivariable analysis and clinical experience were used to select significant features associated with CRC survival. Model performance was evaluated in stratified 5-fold cross-validation repeated 5 times by using the time-dependent concordance index, integrated Brier score, calibration curves, and decision curves. The SHapley Additive exPlanations method was applied to calculate feature importance.

RESULTS

A total of 2157 patients with CRC were included in this study. Among the 6 time-to-event ML models, the DeepHit model exhibited the best discriminative ability (time-dependent concordance index 0.789, 95% CI 0.779-0.799) and the RSF model produced better-calibrated survival estimates (integrated Brier score 0.096, 95% CI 0.094-0.099), but these are not statistically significant. Additionally, the RSF, GBM, DeepSurv, Cox-Time, and N-MTLR models have comparable predictive accuracy to the Cox Proportional Hazards model in terms of discrimination and calibration. The calibration curves showed that all the ML models exhibited good 5-year survival calibration. The decision curves for CRC-specific survival at 5 years showed that all the ML models, especially RSF, had higher net benefits than default strategies of treating all or no patients at a range of clinically reasonable risk thresholds. The SHapley Additive exPlanations method revealed that R0 resection, tumor-node-metastasis staging, and the number of positive lymph nodes were important factors for 5-year CRC-specific survival.

CONCLUSIONS

This study showed the potential of applying time-to-event ML predictive algorithms to help predict CRC-specific survival. The RSF, GBM, Cox-Time, and N-MTLR algorithms could provide nonparametric alternatives to the Cox Proportional Hazards model in estimating the survival probability of patients with CRC. The transparent time-to-event ML models help clinicians to more accurately predict the survival rate for these patients and improve patient outcomes by enabling personalized treatment plans that are informed by explainable ML models.

Construction and validation of machine learning models for sepsis prediction in patients with acute pancreatitis

BACKGROUND

This study aimed to construct predictive models for the risk of sepsis in patients with Acute pancreatitis (AP) using machine learning methods and compared optimal one with the logistic regression (LR) model and scoring systems.

METHODS

In this retrospective cohort study, data were collected from the Medical Information Mart for Intensive Care III (MIMIC III) database between 2001 and 2012 and the MIMIC IV database between 2008 and 2019. Patients were randomly divided into training and test sets (8:2). The least absolute shrinkage and selection operator (LASSO) regression plus 5-fold cross-validation were used to screen and confirm the predictive factors. Based on the selected predictive factors, 6 machine learning models were constructed, including support vector machine (SVM), K-nearest neighbour (KNN), multi-layer perceptron (MLP), LR, gradient boosting decision tree (GBDT) and adaptive enhancement algorithm (AdaBoost). The models and scoring systems were evaluated and compared using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and the area under the curve (AUC).

RESULTS

A total of 1, 672 patients were eligible for participation. In the training set, 261 AP patients (19.51%) were diagnosed with sepsis. The predictive factors for the risk of sepsis in AP patients included age, insurance, vasopressors, mechanical ventilation, Glasgow Coma Scale (GCS), heart rate, respiratory rate, temperature, SpO2, platelet, red blood cell distribution width (RDW), International Normalized Ratio (INR), and blood urea nitrogen (BUN). The AUC of the GBDT model for sepsis prediction in the AP patients in the testing set was 0.985. The GBDT model showed better performance in sepsis prediction than the LR, systemic inflammatory response syndrome (SIRS) score, bedside index for severity in acute pancreatitis (BISAP) score, sequential organ failure assessment (SOFA) score, quick-SOFA (qSOFA), and simplified acute physiology score II (SAPS II).

CONCLUSION

The present findings suggest that compared to the classical LR model and SOFA, qSOFA, SAPS II, SIRS, and BISAP scores, the machine learning model-GBDT model had a better performance in predicting sepsis in the AP patients, which is a useful tool for early identification of high-risk patients and timely clinical interventions.