Acute kidney injury (AKI) is a complication with high morbidity and mortality after cardiac surgery. In order to predict the incidence of AKI after cardiac surgery, many risk prediction models have been established worldwide. We made a detailed introduction to the composing features, clinical application and predictive capability of 14 commonly used models. Among the 14 risk prediction models, age, congestive heart failure, hypertension, left ventricular ejection fraction, diabetes, cardiac valve surgery, coronary artery bypass grafting (CABG) combined with cardiac valve surgery, emergency surgery, preoperative creatinine, preoperative estimated glomerular filtration rate (eGFR), preoperative New York Heart Association (NYHA) score>Ⅱ, previous cardiac surgery, cadiopulmonary bypass (CPB) time and low cardiac output syndrome (LCOS) are included in many risks prediction models (>3 times). In comparison to Mehta and SRI models, Cleveland risk prediction model shows the best discrimination for the prediction of renal replacement therapy (RRT)-AKI and AKI in the European. However, in Chinese population, the predictive ability of the above three risk prediction models for RRT-AKI and AKI is poor.
ObjectiveTo construct a demand model for electronic medical record (EMR) data quality in regards to the lifecycle in machine learning (ML)-based disease risk prediction, to guide the implementation of EMR data quality assessment. MethodsReferring to the lifecycle in ML-based predictive model, we explored the demand for EMR data quality. First, we summarized the key data activities involved in each task on predicting disease risk with ML through a literature review. Second, we mapped the data activities in each task to the associated requirements. Finally, we clustered those requirements into four dimensions. ResultsWe constructed a three-layer structured ring to represent the demand model for EMR data quality in ML-based disease risk prediction research. The inner layer shows the seven main tasks in ML-based predictive models: data collection, data preprocessing, feature representation, feature selection and extraction, model training, model evaluation and optimization, and model deployment. The middle layer is the key data activities in each task; and the outer layer represents four dimensions of data quality requirements: operability, completeness, accuracy, and timeliness. ConclusionThe proposed model can guide real-world EMR data governance, improve its quality management, and promote the generation of real-world evidence.
ObjectiveProlonged mechanical ventilation (PMV) is a prognostic marker for short-term adverse outcomes in patients after lung transplantation.The risk of prolonged mechanical ventilation after lung transplantation is still not clear. The study to identify the risk factors of prolonged mechanical ventilation (PMV) after lung transplantation.Methods This retrospective observational study recruited patients who underwent lung transplantation in Wuxi People’s Hospital from January 2020 to December 2022. Relevant information was collected from patients and donors, including recipient data (gender, age, BMI, blood type, comorbidities), donor data (age, BMI, time of endotracheal intubation, oxygenation index, history of smoking, and any comorbidity with multidrug-resistant bacterial infections), and surgical data (surgical mode, incision type, operation time, cold ischemia time of the donor lung, intraoperative bleeding, and ECMO support), and postoperative data (multi-resistant bacterial lung infection, multi-resistant bacterial bloodstream infection, and mean arterial pressure on postoperative admission to the monitoring unit). Patients with a duration of mechanical ventilation ≤72 hours were allocated to the non-prolonged mechanical ventilation group, and patients with a duration of mechanical ventilation>72 hours were allocated to the prolonged mechanical ventilation group. LASSO regression analysis was applied to screen risk factors., and a clinical prediction model for the risk of prolonged mechanical ventilation after lung.ResultsPatients who met the inclusion criteria were divided into the training set and the validation set. There were 307 cases in the training set group and 138 cases in the validation set group. The basic characteristics of the training set and the validation set were compared. There were statistically significant differences in the recipient’s BMI, donor’s gender, CRKP of the donor lung swab, whether the recipient had pulmonary infection before the operation, the type of transplantation, the cold ischemia time of the donor lung, whether ECMO was used during the operation, the duration of ECMO assistance, CRKP of sputum, and the CRE index of the recipient's anal test (P<0.05). 2. The results of the multivariate logistic regression model showed that female recipients, preoperative mechanical ventilation in recipients, preoperative pulmonary infection in recipients, intraoperative application of ECMO, and the detection of multi-drug resistant Acinetobacter baumannii, multi-drug resistant Klebsiella pneumoniae and maltoclomonas aeruginosa in postoperative sputum were independent risk factors for prolonged mechanical ventilation after lung transplantation. The AUC of the clinical prediction model in the training set and the validation set was 0.838 and 0.828 respectively, suggesting that the prediction model has good discrimination. In the decision curves of the training set and the validation set, the threshold probabilities of the curves in the range of 0.05-0.98 and 0.02-0.85 were higher than the two extreme lines, indicating that the model has certain clinical validity.ConclusionsFemale patients, Preoperative pulmonary infection, preoperative mechanical ventilation,blood type B, blood type O, application of ECMO assistance, multi-resistant Acinetobacter baumannii infection, multi-resistant Klebsiella pneumoniae infection, and multi-resistant Stenotrophomonas maltophilia infection are independent risk factors for PMV (prolonged mechanical ventilation) after lung transplantation.
Objective To explore the risk factors for long-term death of patients with acute myocardial infarction (AMI) and reduced left ventricular ejection fraction (LVEF), and develop and validate a prediction model for long-term death. Methods This retrospective cohort study included 1013 patients diagnosed with AMI and reduced LVEF in West China Hospital of Sichuan University between January 2010 and June 2019. Using the RAND function of Excel software, patients were randomly divided into three groups, two of which were combined for the purpose of establishing the model, and the third group was used for validation of the model. The endpoint of the study was all-cause mortality, and the follow-up was until January 20th, 2021. Cox proportional hazard model was used to evaluate the risk factors affecting the long-term death, and then a prediction model based on those risk factors was established and validated. Results During a median follow-up of 1377 days, 296 patients died. Multivariate Cox regression analysis showed that age≥65 years [hazard ratio (HR)=1.842, 95% confidence interval (CI) (1.067, 3.179), P=0.028], Killip class≥Ⅲ[HR=1.941, 95%CI (1.188, 3.170), P=0.008], N-terminal pro-brain natriuretic peptide≥5598 pg/mL [HR=2.122, 95%CI (1.228, 3.665), P=0.007], no percutaneous coronary intervention [HR=2.181, 95%CI (1.351, 3.524), P=0.001], no use of statins [HR=2.441, 95%CI (1.338, 4.454), P=0.004], and no use of β-blockers [HR=1.671, 95%CI (1.026, 2.720), P=0.039] were independent risk factors for long-term death. The prediction model was established and patients were divided into three risk groups according to the total score, namely low-risk group (0-2), medium-risk group (4-6), and high-risk group (8-12). The results of receiver operating characteristic curve [area under curve (AUC)=0.724, 95%CI (0.680, 0.767), P<0.001], Hosmer-Lemeshow test (P=0.108), and Kaplan-Meier survival curve (P<0.001) showed that the prediction model had an efficient prediction ability, and a strong ability in discriminating different groups. The model was also shown to be valid in the validation group [AUC=0.758, 95%CI (0.703, 0.813), P<0.001]. Conclusions In patients with AMI and reduced LVEF, age≥65 years, Killip class≥Ⅲ, N-terminal pro-brain natriuretic peptide≥5598 pg/mL, no percutaneous coronary intervention, no use of statins, and no use of β-blockers are independent risk factors for long-term death. The developed risk prediction model based on these risk factors has a strong prediction ability.
Objective To develop and compare the predictive performance of five machine learning models for adverse postoperative outcomes in cardiac surgery patients, and to identify key decision factors through SHapley Additive exPlanations (SHAP) interpretability analysis. Methods A retrospective collection of perioperative data (including demographic information, preoperative, intraoperative, and postoperative indicators) with 88 variables was conducted from adult cardiac surgery patients at the First Affiliated Hospital of Xinjiang Medical University in 2023. Adverse postoperative outcomes were defined as the occurrence of acute kidney injury and/or in-hospital mortality during the postoperative hospitalization period following cardiac surgery. Patients were divided into an adverse outcome group and a favorable outcome group based on the presence of adverse postoperative outcomes. After screening feature variables using the least absolute shrinkage and selection operator (LASSO) regression method, five machine learning models were constructed: eXtreme gradient boosting (XGBoost), random forest (RF), gradient boosting machine (GBM), light gradient boosting machine (LightGBM), and generalized linear model (GLM). The dataset was randomly divided into a training set and a test set at a 7 : 3 ratio using stratified sampling, with postoperative outcome as the stratification factor. Model performance was evaluated using receiver operating characteristic curves, decision curve analysis, and F1 Score. The SHAP method was applied to analyze feature contribution. Results A total of 639 patients were included, comprising 395 males and 244 females, with a median age of 62 (55, 69) years. The adverse outcome group consisted of 191 patients, while the favorable outcome group included 448 patients, resulting in an adverse postoperative outcome incidence of 29.9%. Univariate analysis showed no significant differences between the two groups for any variables (P>0.05). Using LASSO regression, 16 feature variables were selected (including cardiopulmonary bypass support time, blood glucose on postoperative day 3, creatine kinase-MB isoenzyme, systemic inflammatory response index, etc.), and five machine learning models (GLM, RF, GBM, LightGBM, XGBoost) were constructed. Evaluation results demonstrated that the XGBoost model exhibited the best predictive performance on both the training set (n=447) and test set (n=192), with area under the curve values of 0.761 [95%CI (0.719, 0.800) ] and 0.759 [95%CI (0.692, 0.818) ], respectively. It also significantly outperformed other models in positive predictive value, and balanced accuracy in the test set. Decision curve analysis further confirmed its clinical utility across various risk thresholds. SHAP analysis indicated that variables such as cardiopulmonary bypass support time, blood glucose on postoperative day 3, creatine kinase-MB isoenzyme, and inflammatory markers (SIRI, NLR, CAR) had high contributions to the prediction. Conclusion The XGBoost model effectively predicts adverse postoperative outcomes in cardiac surgery patients. Clinically, attention should be focused on cardiopulmonary bypass support time, postoperative blood glucose control, and monitoring of inflammatory levels to improve patient prognosis.
Objective To systematically evaluate risk prediction models for acute exacerbation of chronic obstructive pulmonary disease (COPD), and provide a reference for early clinical identification. Methods The literature on the risk prediction models of acute exacerbation of COPD published by CNKI, VIP, Cochrane, Embase and Web of Science database was searched in Chinese and English from inception to April 2022, and relevant studies were collected on the development of risk prediction models for acute exacerbations of COPD. After independent screening of the literature and extraction of information by two independent researchers, the quality of the included literature was evaluated using the PROBASTA tool. Results Five prospective studies, one retrospective case-control study and seven retrospective cohort studies were included, totally 13 papers containing 24 models. Twelve studies (92.3%) reported the area under the receiver operator characteristic curve ranging 0.66 to 0.969. Only five studies reported calibrated statistics, and three studies were internally and externally validated. The overall applicability of 13 studies was good, but there was a high risk of bias, mainly in the area of analysis. Conclusions The existing predictive risk models for acute exacerbations of COPD are unsatisfactory, with wide variation in model performance, inappropriate and incomplete inclusion of predictors, and a need for better ways to develop and validate high-quality predictive models. Future research should refine the study design and study report, and continue to update and validate existing models. Secondly medical staff should develop and implement risk stratification strategies for acute exacerbations of COPD based on predicted risk classification results in order to reduce the frequency of acute exacerbations and to facilitate the rational allocation of medical resources.
Objective To scoping review the risk prediction models for sarcopenia in China was conducted, and provide reference for scientific prevention and treatment of the disease and related research. Methods We systematically searched PubMed, Web of Science, Cochrane Library, Embase, China Knowledge Network, China Biomedical Literature Database, Wanfang Database, and Weipu Database for literature related to myasthenia gravis prediction models in China, with a time frame from the construction of the database to April 30, 2024 for the search. The risk of bias and applicability of the included literature were assessed, and information on the construction of myasthenia gravis risk prediction models, model predictors, model presentation form and performance were extracted. Results A total of 25 literatures were included, the prevalence of sarcopenia ranged from 12.16% to 54.17%, and the study population mainly included the elderly, the model construction methods were categorized into two types: logistic regression model and machine learning, and age, body mass index, and nutritional status were the three predictors that appeared most frequently. Conclusion Clinical caregivers should pay attention to the high-risk factors for the occurrence of sarcopenia, construct models with accurate predictive performance and high clinical utility with the help of visual model presentation, and design prospective, multicenter internal and external validation methods to continuously improve and optimize the models to achieve the best predictive effect.
Objective To construct a risk prediction score model for serious adverse event (SAE) after cardiac catheterization in patients with adult congenital heart disease (ACHD) and pulmonary hypertension (PH) and verify its predictive effect. Methods The patients with PH who underwent cardiac catheterization in Wuhan Asian Heart Hospital Affiliated to Wuhan University of Science and Technology from January 2018 to January 2022 were retrospectively collected. The patients were randomly divided into a model group and a validation group according to the order of admission. The model group was divided into a SAE group and a non-SAE group according to whether SAE occurred after the catheterization. The data of the two groups were compared, and the risk prediction score model was established according to the results of multivariate logistic regression analysis. The discrimination and calibration of the model were evaluated using the area under the receiver operating characteristic (ROC) curve and the Hosmer-Lemeshow test, respectively. Results A total of 758 patients were enrolled, including 240 (31.7%) males and 518 (68.3%) females, with a mean age of 43.1 (18.0-81.0) years. There were 530 patients in the model group (47 patients in the SAE group and 483 patients in the non-SAE group) and 228 patients in the validation group. Univariate analysis showed statistical differences in age, smoking history, valvular disease history, heart failure history, N-terminal pro-B-type natriuretic peptide, and other factors between the SAE and non-SAE groups (P<0.05). Multivariate analysis showed that age≥50 years, history of heart failure, moderate to severe congenital heart disease, moderate to severe PH, cardiac catheterization and treatment, surgical general anesthesia, and N-terminal pro-B-type natriuretic peptide≥126.65 pg/mL were risk factors for SAE after cardiac catheterization for ACHD-PH patients (P<0.05). The risk prediction score model had a total score of 0-139 points and patients who had a score>50 points were high-risk patients. Model validation results showed an area under the ROC curve of 0.937 (95%CI 0.897-0.976). Hosmer-Lemeshow goodness-of-fit test: χ2=3.847, P=0.797. Conclusion Age≥50 years, history of heart failure, moderate to severe congenital heart disease, moderate to severe PH, cardiac catheterization and treatment, general anesthesia for surgery, and N-terminal pro-B-type natriuretic peptide≥126.65 pg/mL were risk factors for SAE after cardiac catheterization for ACHD-PH patients. The risk prediction model based on these factors has a high predictive value and can be applied to the risk assessment of SAE after interventional therapy in ACHD-PH patients to help clinicians perform early intervention.
Surgical risk prediction is to predict postoperative morbidity and mortality with internationally authoritative mathematical models. For patients undergoing high-risk cardiac surgery, surgical risk prediction is helpful for decision-making on treatment strategies and minimization of postoperative complications, which has gradually arouse interest of cardiac surgeons. There are many risk prediction models for cardiac surgery in the world, including European System for Cardiac Operative Risk Evaluation (EuroSCORE), Ontario Province Risk (OPR)score, Society of Thoracic Surgeons (STS)score, Cleveland Clinic risk score, Quality Measurement and Management Initiative (QMMI), American College of Cardiology/American Heart Association (ACC/AHA)Guidelines for Coronary Artery Bypass Graft Surgery, and Sino System for Coronary Operative Risk Evaluation (SinoSCORE). All these models are established from the database of thousands or ten thousands patients undergoing cardiac surgery in a specific region. As different sources of data and calculation imparities exist, there are probably bias and heterogeneities when the models are applied in other regions. How to decrease deviation and improve predicting effects had become the main research target in the future. This review focuses on the progress of risk prediction models for patients undergoing cardiac surgery.
Risk prediction models for postoperative pulmonary complications (PPCs) can assist healthcare professionals in assessing the likelihood of PPCs occurring after surgery, thereby supporting rapid decision-making. This study evaluated the merits, limitations, and challenges of these models, focusing on model types, construction methods, performance, and clinical applications. The findings indicate that current risk prediction models for PPCs following lung cancer surgery demonstrate a certain level of predictive effectiveness. However, there are notable deficiencies in study design, clinical implementation, and reporting transparency. Future research should prioritize large-scale, prospective, multi-center studies that utilize multiomics approaches to ensure robust data for accurate predictions, ultimately facilitating clinical translation, adoption, and promotion.