Table 1 Characteristics and main results of included studies (n = 88).
Author, year and title | Location | Sample characteristics | Predictor variables | Outcome | Models | AUC results and best performing model |
|---|---|---|---|---|---|---|
Banerjee et al., 202149 | United Kingdom | 1706 patients with schizophrenia, average age not specified | Antidepressants, second-generation antipsychotics, alcohol/substance abuse, delirium, dementia, cardiovascular disease, diabetes, other physical and mental health issues, social factors like family support | All-cause mortality in schizophrenia patients | Logistic regression, random forests, deep learning models | 0.80 Random forest model |
Meredith et al., 200250 | United States | 76,871 incidents, 72,827 survivors, 4044 deceased | Scores based on AIS and ICD-9, including ISS, NISS, APS, maxAIS and their mapped ICD to AIS versions, and the ICD-9 based ICISS score | All-cause mortality | Scoring algorithms based on AIS and ICD, logistic regression used for calibration | 0.89 ICISS |
Li et al., 202328 | China | 10,311 patients with heart failure | 66 predictors including demographics, pre-existing conditions, treatments received, and other clinical and laboratory details | All-cause mortality within 30 days, 180 days, 365 days, and after 365 days | Deep Learning System based on Multi-head Self-attention Mechanism (DLS-MSM) | 0.75 DLS-MSM |
Barsasella et al., 202227 | Taiwan | 58,618 patients, including 25,868 with T2DM, 32,750 with HTN, and 6,419 with both conditions, average age 75.12 ± 13.65 years | 67 predictor variables including hospital cost, vital signs and symptoms, comorbidities, demographic characteristics | All-cause mortality | Logistic Regression, Ridge Classifier, Random Forest, K-Neighbors Classifier, Bagging Classifier, Gradient Boosting Classifier | 0.97 Logistic Regression |
Wang Y. et al., 202151 | China | 1,200 patients with ESRD undergoing HD, using 36 continuous HD sessions to predict 90-day mortality | 64 variables related to hemodialysis sessions, including patient blood pressure recorded multiple times per session | Mortality within 90/180/365 days | LSTM Autoencoder, Logistic Regression, Support Vector Machine, Random Forest, LSTM Classifier, Isolation Forest, Stacked Autoencoder | 0.73 LSTM |
Díez-Sanmartín et al., 202325 | Spain | 44,663 adults on kidney transplant waiting list, dialysis < 15 years | Sociodemographic factors (e.g., age, gender, etc.) | Mortality on kidney transplant waiting list | XGBoost, K-Means, Agglomerative Clustering | 0.99 XGBoost |
Xiong J et al., 202310 | China | 579 tumor samples from endometrial cancer patients | PCD-related genes | All-cause mortality | LASSO | 0.92 LASSO |
Lin et al., 201952 | Taiwan | 48,153 ESRD patients aged ≥ 65 years | Age, sex, urbanization level, occupation, comorbidities (including diabetes, hypertension, etc.) | One-year all-cause mortality | Random Forest, Artificial Neural Networks | 0.68 Artificial Neural Networks |
Liu et al., 202253 | Taiwan, Japan | 28,745 patients, 20–60 years old | ECG recordings, AI-predicted LVH | All-cause mortality | Deep learning | 0.74 Deep learning |
Shi et al., 201254 | Taiwan | 22,926 patients undergoing surgery for hepatocellular carcinoma | Initial clinical data, surgeon volume, hospital volume | 5-year mortality | Artificial Neural Networks and Logistic Regression | 0.89 Artificial Neural Networks |
L. Shi, X.C. Wang, Y.S. Wang, 201355 | China | 2,150 elderly patients with intertrochanteric fractures, mean age 81.6 years | Age, gender, nursing home, New Mobility Score, dementia or cognitive impairment, diabetes, cancer, cardiac disease | 1-year mortality | Artificial Neural Networks and Logistic Regression | 0.87 Artificial Neural Networks |
Puddu and Menotti, 201256 | Italy | 1591 men aged 40–59 years, enrolled in 1960 from rural communities in Crevalcore and Montegiorgio with cardiovascular disease | Age, father life status, mother life status, family history of CVD, job-related physical activity, smoking, BMI, arm circumference, blood pressure, heart rate, forced expiratory volume, serum cholesterol, corneal arcus, diagnoses of CVD, cancer, diabetes, minor ECG abnormalities | 45-year all-cause mortality | Cox proportional hazards models, Multilayer Perceptron, AND Neural Networks | 0.842 Neural Networks |
Harris et al., 201957 | United States | 107,792 patients undergoing nonemergency primary total hip arthroplasties (THAs) and total knee arthroplasties (TKAs) from the ACS-NSQIP dataset 2013–2014 | Demographic and clinical variables such as American Society of Anesthesiologists classification, comorbidities, age, and gender | 30-day mortality | LASSO Regression | 0.73 LASSO |
Takahama et al., 20238 | Japan | 987 heart failure patients, hospitalized, data from 2013 to 2016 | 15 variables including troponin levels, blood pressure, BMI, hematocrit, BNP levels, CRP, LDL cholesterol, white blood cell count, diastolic blood pressure, creatinine levels, BUN, LVEF | One-year mortality in heart failure patients | Light Gradient Boosting Machine (LightGBM) | 0.87 LightGBM |
Arostegui et al., 20189 | Spain | 1,945 patients with colon cancer who had surgery | Residual tumor, ASA Physical Status, pathologic tumor staging, Charlson Comorbidity Index, intraoperative complications, adjuvant chemotherapy, tumor recurrence | 1-year mortality post-surgery | Random Forest, Genetic Algorithms, Classification and Regression Trees (CART) | 0.90 Multimodels |
Jing et al., 202258 | United States | 124,360 veterans aged > 50; 93.9% male; mean age 68.2 years | 924 predictors including demographics, vital signs, medication classes, disease diagnoses, laboratory results, healthcare utilization | 10-year all-cause mortality | Gradient Boosting, Random Forest, Neural Networks, SuperLearner ensemble, LASSO | 0.84 Gradient Boosting |
Tedesco et al., 202129 | Sweden | 2,291 healthy older adults aged 70 | Anthropometric variables, physical and lab exams, questionnaires, lifestyle, wearable data | All-cause mortality | Logistic Regression, Decision Tree, Random Forest, AdaBoost | 0.51 AdaBoost |
Sakr et al., 201759 | United States | 34,212 patients, age 54 ± 13 years, 55% male | Fitness data from exercise treadmill stress testing, demographic and clinical variables | All-cause mortality | Decision Tree, Support Vector Machine, Artificial Neural Networks, Naïve Bayesian Classifier, Bayesian Network, K-Nearest Neighbor, Random Forest | 0.97 Random Forest |
Jones et al., 202160 | United States | 297,498 encounters, median age 68 years, 95% male, including demographics, vital signs, and 21 laboratory values | Demographic characteristics, 38 comorbid conditions, five vital signs, 21 laboratory values, utilizing PSI variables and excluding mental status and chest imaging | 30-day all-cause mortality | Logistic regression, spline models, Extreme Gradient Boosting (XGBoost) incorporating various subsets of predictor variables | 0.88 XGBoost |
Singh et al., 202261 | United States | 4,735 patients referred for PET between 2010–2018, median follow-up of 4.15 years | Polar maps of stress and rest perfusion, myocardial blood flow, myocardial flow reserve, spill-over fraction, cardiac volumes, singular indices, sex | All-cause mortality | Deep Learning | 0.82 Deep Learning |
Lu et al., 201962 | United States | PLCO: 10,464 participants, mean age 62.4; NLST: 5,493 participants, mean age 61.7; both included smokers and nonsmokers aged 55–74 years | Chest radiographs | All-cause mortality | Convolutional Neural Network | 0.75 and 0.68 Convolutional Neural Network |
Siegersma et al., 202226 | Netherlands | 1,136,113 ECGs from 249,262 individuals, ages 18–85, from UMC Utrecht | 12-lead resting ECGs | All-cause mortality | Deep Neural Network | 0.96 Deep Neural Network |
Ulloa Cerna et al., 202163 | United States | 34,362 individuals, 812,278 echocardiographic videos | Raw pixel data from echocardiographic videos | All-cause mortality | Convolutional Neural Network | 0.84 Convolutional Neural Network |
Wang et al., 202164 | United States | Patients aged 50 + from a large healthcare system with documented cognitive decline | Clinical notes from EHRs | All-cause mortality | Deep Learning | 0.94 Deep Learning |
Mohammad et al., 202265 | Sweden | 139,288 patients admitted for myocardial infarction from the SWEDEHEART registry | Demographic info, medical history, hospital characteristics, lab results | 1-year all-cause mortality | Artificial Neural Network | 0.85 Artificial Neural Network |
Valsaraj et al., 202366 | India and Canada | 6,083 echos from Taiwan; 997 echos from Alberta, Canada | Echocardiographic data, demographic and clinical data | All-cause mortality (1-year, 3-year, 5-year) | ResNet, CatBoost | 0.92 CatBoost |
Li et al., 202367 | China | Two cohorts: CHNS (8,355 adults > 18 years) and CHARLS (12,711 adults > 45 years) | 159 variables from demographics, family, community, socioeconomic status, lifestyle, health conditions, etc | All-cause mortality | Cox Regression, LASSO, Survival Tree, Random Survival Forest, Conditional Inference Forest, glmBoost, Gradient Boosting | 0.86 Random Survival Forest |
Zhou et al., 202268 | Hong Kong | 2,560 patients with pulmonary hypertension; median age 63.4 years | Age, average readmission interval, cumulative hospital stay, anti-hypertensive drugs, total bilirubin | All-cause mortality | Random Survival Forest | 0.95 Random Survival Forest |
Giang et al., 202169 | Sweden | 71,941 patients with congenital heart disease, mean follow-up time of 16.47 years | Congenital heart disease classifications, comorbidities | All-cause mortality | Neural Networks and Logistic Regression | 0.92 Neural Networks |
Forte et al., 202170 | Netherlands | 8,241 patients undergoing valve or CABG operations, mean follow-up of 5 years | Peri-operative clinical parameters | 5-year mortality | Super Learner, GLM, XGBoost | 0.81 Super Learner |
Bergquist et al., 202124 | United States | 1,264,000 patients | Various clinical predictors | All-cause mortality | Boosted methods, logistic regression, neural networks | 0.95 LightGBM |
Hernesniemi et al., 201971 | Finland | 9,066 patients with acute coronary syndrome | Extensive clinical data, GRACE score | Six-month mortality | Logistic regression and XGBoost | 0.89 XGBoost |
Heyman et al., 202123 | Sweden | 148 patients discharged from ED, ages 23–106, general ED population | Age, sex, comorbidity score, arrival mode, discharge time, triage priority, imaging during visit | All-cause mortality within 30 days | Logistic Regression, Random Forest, Support Vector Machine | 0.95 Logistic Regression |
Qiu et al., 202272 | United States | 47,261 participants, ages across various ranges (data from NHANES 1999–2014) | Demographic, laboratory, examination, questionnaire predictors | All-cause mortality | Gradient Boosted Trees and TreeExplainer | 0.92 Gradient Boosting |
Niedziela et al., 202173 | Poland | 17,793 patients treated with primary percutaneous coronary intervention for anterior STEMI | Age, gender, blood pressure, heart rate, Killip class, medical history, in-hospital treatment | 6-month all-cause mortality | Logistic Regression, Neural Network | 0.81 Neural Network |
Mostafaei et al., 202374 | Sweden | 28,023 dementia-diagnosed patients from the Swedish Registry for Cognitive/Dementia Disorders (SveDem). Median follow-up time was 1053 days for surviving and 1125 days for deceased patients | Age, sex, BMI, MMSE score, dementia type, comorbidities, medications | All-cause mortality | Logistic Regression, Support Vector Machine, Neural Networks | 0.74 Support Vector Machine |
Cui et al., 202275 | United States | 19,887 lung cancer patients with bone metastases | Age, primary site, histology, race, sex, T stage, N stage, brain metastasis, liver metastasis, cancer-directed surgery, radiation, chemotherapy | 3-month mortality | Logistic Regression, XGBoosting Machine, Random Forest, Neural Network, Gradient Boosting Machine, Decision Tree | 0.82 Gradient Boosting Machine |
Parikh et al., 201976 | United States | 26,525 adult patients who had outpatient oncology or hematology/oncology encounters at a large academic cancer center and affiliated community practices | Demographic variables, Elixhauser comorbidities, laboratory data | 180-day mortality | Logistic Regression, Random Forest, Gradient Boosting | 0.88 Random Forest |
Tong et al., 202177 | China | 578 liver cancer patients undergoing RFA | Platelet count (PLT), Alpha-fetoprotein (AFP), age, tumor size, total bilirubin | All-cause mortality | Logistic Regression, DecisionTree, gbm, Gradient Boosting, Forest | 0.74 Gradient Boosting model |
de Capretz et al., 202378 | Sweden | 9519 ED chest pain patients, avg. age 59, 47.3% female | Age, sex, ECG, hs-cTnT, glucose, creatinine, hemoglobin | AMI or all-cause death within 30 days | Convolutional neural network, ANN, Logistic Regression | 0.94 Convolutional neural network |
Tian et al., 202379 | China | 424 patients with HFmrEF, median follow-up of 1008 days | Age, NYHA class, LVEF, eGFR, NT-proBNP, various lab results | All-cause mortality | XGBoost, Random Forest, SVM | 0.92 XGBoost |
Mamprin et al., 202180 | Netherlands | 1,931 TAVI procedures (1,300 at AMC and 631 at CZE) | Clinical predictors including age, health history, and procedural data | One-year mortality | Logistic Regression, Random Forest, CatBoost | 0.68 CatBoost |
Motwani et al., 201681 | Multiple countries | 10,030 patients with suspected CAD, 5-year follow-up | 25 clinical and 44 CCTA parameters | All-cause mortality within 5 years | Boosted ensemble | 0.79 Boosted ensemble |
Santos et al., 201982 | Brazil | 2,808 elderly participants, mean age not specified, general population | 37 demographic, socioeconomic, and health profile variables | All-cause mortality within five years | Logistic regression, neural networks, gradient boosted trees, random forest | 0.80 Neural Network |
Feng et al., 202383 | China | 8,943 participants, mean age 61.1 years, 79.6% male, all with three-vessel coronary artery disease | 18 selected predictors including demographics, medical history, blood tests, and cardiac function assessments | All-cause mortality over 4 years | Random forest | 0.81 Random Forest |
Yu et al., 202284 | China | 7,368 patients, age older than 18, post-cardiac surgery including CABG, valvular operations | 25 selected predictors including demographics, comorbidities, vital signs, and lab results | 4-year all-cause mortality | Logistic regression, neural networks, naïve bayes, gradient boosting, adapting boosting, random forest, bagged trees, extreme gradient boosting | 0.80 Adapting boosting |
Tamminen et al., 202185 | Finland | 2,853 unselected prehospital patients encountered in June 2015 | NEWS parameters, blood glucose | 30-day mortality | Random Forest | 0.76 Random Forest |
Xu et al., 202386 | United States | 630 patients with advanced cancer, mean age 59.1 years, 56.19% female | Demographics, clinical data, patient-reported outcomes (PROs) from the Edmonton Symptom Assessment System (ESAS) | 180-day mortality | GLM with elastic net, XGBoost, SVM, Neural Network | 0.69 XGBoost |
Katsiferis et al., 202387 | Denmark | 48,944 Danish citizens aged 65 and older, spousal bereavement | Age, sex, healthcare expenditures, sociodemographics | All-cause mortality | XGBoost | 0.81 XGBoost |
Kanda et al., 202288 | Japan | 24,949 adults with hyperkalemia, multifactorial conditions, aged ≥ 18 | Clinical data, lab results, prescription history | All-cause mortality | XGBoost and Logistic Regression | 0.82 XGBoost |
Lu et al., 202189 | Australia | 68,889 patients, mean age 76 years, 54% female | Age, sex, medication history, disease history | All-cause mortality | Gradient Boosting Machine, Multi-Layer Neural Network, Support Vector Machine | 0.75 Gradient Boosting |
Scrutinio et al., 202090 | Italy | 1,207 patients with severe stroke, average follow-up 988 days, 15.7% 3-year mortality rate | Age, comorbidities (e.g., diabetes, CAD), functional measures | 3-year mortality | Random Forest, Logistic Regression | 0.93 Random Forest |
Li et al., 202091 | China | 1,244 patients, mean age 63.8, 78.4% male, 75.18% received reperfusion therapy | Comprehensive clinical dataset including demographics and treatment details | 1-year mortality after anterior STEMI | GaussianNB, Logistic Regression, KNN, Decision Tree, Random Forest, XGBoost | 0.94 XGBoost |
Guo et al., 202292 | China | 751 patients diagnosed with spontaneous intracerebral hemorrhage at West China Hospital | Clinical presentations, laboratory data (e.g., monocyte and lymphocyte levels), radiographic data, Glasgow Coma Scale, hematoma volume, location, age | 90-day mortality | Logistic Regression, Category Boosting, Support Vector Machine, Random Forest, Extreme Gradient Boosting | 0.84 Logistic Regression |
Wang et al., 202311 | United States | 1229 patients from the MIMIC-IV database, including critical pulmonary embolism patients with or without septic or other cardiopulmonary complications | Age, gender, DVT, VTE history, hematocrit, hemoglobin, anion gap, heart rate, blood pressure, respiratory rate, congestive heart failure, hypertension, atrial fibrillation, vasopressor | All-cause mortality within 30 days | Logistic Regression and XGBoost | 0.82 XGBoost |
Li et al., 202393 | China | 451 older patients with CAD, IGT, and DM from a hospital cohort, split into a training group (308) and a validation group (143). Median age 86 years. Majority were males | Demographics, comorbidities (e.g., CHF, DM, hypertension), laboratory tests (e.g., glucose, NT-proBNP), and medications used during hospitalization (e.g., statins, beta blockers) | All-cause mortality | Logistic Regression, Gradient Boosting, Random Forest, Decision Tree | 0.84 Gradient Boosting |
Asrian et al., 202494 | United States | 3751 hip fracture patients from the MIMIC-IV database. Includes demographics like age and basic lab tests. Average age was 73 | Age, glucose, red blood cell distribution width, mean corpuscular hemoglobin concentration, white blood cells, urea nitrogen, prothrombin time, platelet count, calcium levels, and partial thromboplastin time | 1-, 5-, and 10-year mortality | LightGBM, Random Forest, Logistic Regression | 0.79 LightGBM |
Ivanics et al., 202395 | Canada, United Kingdom, United States | Adults who underwent primary liver transplants from Jan 2008 to Dec 2018: Canada (n = 1214), UK (n = 5287), US (n = 59,558) | Harmonized pre-transplant variables like recipient BMI, donor age, MELD score, etc., across three national registries | 90-day mortality | LASSO, Ridge, ElasticNET, LightGBM | 0.74 Ridge |
Behnoush et al., 202396 | Iran | 8,493 hypertensive patients undergoing CABG, mean age 68.27 years, 63.86% male, 46.84% with diabetes, 38.61% with a family history of CAD | Age, total ventilation hours, ejection fraction, hemoglobin, total cholesterol, LDL-C, HDL-C, triglycerides, fasting blood glucose, creatinine, BMI, and several perioperative factors like ICU hours and cardiopulmonary pump use | 1-year mortality | Logistic Regression, Extreme Gradient Boosting, Naïve Bayes, Random Forest, Artificial Neural Network | 0.82 Logistic Regression |
Kampaktsis et al., 202397 | United States | 1033 recipients (median age 34 years, 61% male) of isolated heart transplants analyzed from the UNOS database, 2000–2020 | Variables included recipient, donor, procedural characteristics, post-transplant predictors, selected using SHapley Additive exPlanations (SHAP) | 1-year mortality | CatBoost | 0.80 CatBoost |
Lin et al., 202098 | Taiwan | 1,903 patients from Wan Fang Hospital and Taipei Medical University Hospital, diagnosed with chronic liver diseases such as cirrhosis and hepatic coma | Variables included clinical scores (e.g., MELD score), laboratory data (e.g., bilirubin, creatinine), and demographic data | All-cause mortality | Random Forest, Adaptive Boosting, other machine learning models | 0.85 Random Forest |
Lin et al., 202399 | United States | 63,215 asymptomatic patients from four centers, median age 54, 68% male, median follow-up 12.6 years | Age, sex, race, CAD risk factors, CAC score, CAC density, number of calcified vessels | 10-year all-cause mortality | XGBoost | 0.82 XGBoost |
Liu et al., 2022100 | China | 340 patients with sepsis-induced CRS at Shanghai Tongji Hospital, aged ≥ 18, data from 2015–2020 | Age, SOFA score, myoglobin levels, vasopressor use, mechanical ventilation | 1-year mortality | Random Forest, Support Vector Machine, Gradient Boosted Decision Tree | 0.85 Random Forest |
Forssten et al., 2021101 | Sweden | 124,707 traumatic hip fracture cases, aged 18 or older, between 2008 and 2017 | Age, sex, ASA classification, CCI, RCRI, various comorbidities, type of fracture, surgical procedure | 1-year postoperative mortality | Logistic Regression, Naive Bayes, Support Vector Machine, Random Forest | 0.74 Random Forest |
Alimbayev et al., 2023102 | Kazakhstan | 472,950 patients diagnosed with diabetes mellitus, collected between 2014–2019 from the Unified National Electronic Health System of Kazakhstan | Age, duration of diabetes, hypertension, sex, and other comorbidities like coronary heart disease and cerebrovascular accident | 1-year mortality | Gaussian Naïve Bayes, K-nearest neighbors, Logistic Regression, Random Forest, AdaBoost, Gradient Boosting, XGBoost, Linear Discriminant Analysis, Perceptron | 0.80 XGBoost |
El-Bouri et al., 2023103 | United Kingdom | 2,183 venous thromboembolism patients; 1235 pulmonary embolism cases; mean age 69 years | Neutrophils, white blood cell counts, C-reactive protein, haemoglobin, age, O2 saturation, heart rate, blood pressure | 30-day, 90-day, and 365-day mortality | Random Forest, XGBoost, logistic regression | 0.73 Random Forest |
Park et al., 2022104 | South Korea | 4,312 patients with acute heart failure, median age 73, 55% male, median left ventricular ejection fraction 38% | 19 clinical predictors (e.g., age, sex, blood pressure) and 8 echocardiographic parameters (e.g., LV ejection fraction) | All-cause mortality at 3 years | CoxBoost | 0.76 CoxBoost |
Penso et al., 2021105 | Italy | 471 patients with severe AS, undergoing TAVI, median age 81 | 83 pre-TAVI clinical and echocardiographic variables | All-cause mortality at 5 years | Random forest, XGBoost, Multilayer perceptron, Logistic regression | 0.79 Multilayer perceptron |
Guo et al., 2021106 | United States | 34,575 patients with liver cirrhosis | 41 health variables including demographic and laboratory data | All-cause mortality at 90, 180, and 365 days | Deep Neural Networks (DNN), Random Forest (RF), Logistic Regression (LR) | 0.86 Random Forest |
Abedi et al., 2021107 | United States | 7,144 patients post-stroke | 37 predictors including demographics, medical history, laboratory data | All-cause mortality within 1, 3, 6, 12, 18, and 24 months | Logistic Regression, Extreme Gradient Boosting, Random Forest | 0.82 Random Forest |
Zhou et al., 2023108 | China | 706 patients, median age 66, 57% male, diagnosed with mitral regurgitation | Age, blood pressure, P-wave duration, lab values (e.g., albumin, creatinine), echocardiographic measurements (e.g., LVEF, LADs) | All-cause mortality | Gradient Boosting Machine, Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Artificial Neural Networks | 0.80 Gradient Boosting Machine |
Rauf et al., 2023109 | Pakistan | 2,184 patients with mitral stenosis and atrial flutter, 81.85% females, median age 65 | Mitral valve area, right ventricular systolic pressure, pulmonary artery pressure, left ventricular ejection fraction, NYHA class, surgery | All-cause mortality | Gradient Boosting Machine, Decision Tree, Support Vector Machine, Random Forest, Artificial Neural Network | 0.84 Gradient Boosting Machine |
Shi et al., 2022110 | China, United Kingdom | 2,846 patients with acute pancreatitis, median age 46 years | Demographic info, lab tests on admission (WBC, creatinine, etc.), clinical severity scores (APACHE II, SOFA) | Mortality | Random Forest | 0.89 Ranfom Forest |
Zhou et al., 2021111 | China | 381 heart transplant recipients, average age 43.8 years | Albumin, recipient age, left atrium diameter, red blood cells, hemoglobin, lymphocyte%, smoking history, use of rhBNP, Levosimendan, hypertension, cardiac surgery history, malignancy, endotracheal intubation history | 1-year mortality | Logistic Regression, Support Vector Machine, Random Forest, Extreme Gradient Boosting, Adaptive Boosting, Gradient Boosting Machine, Artificial Neural Network | 0.80 Random Forest |
Lee et al., 2021112 | South Korea | 22,182 AMI patients, mean age 64 years, 71.8% male | Demographics, clinical predictors, lab results, history of heart disease, interventions, medications | 1-year all-cause mortality | Decision Trees, Logistic Regressions, Deepnets, Random Forest | 0.92 Random Forest |
Tran et al., 2023113 | France | 534 stage 4–5 CKD patients, median age 72 years, 55% male | Age, ESA usage, cardiovascular history, smoking status, vitamin D levels, PTH levels, ferritin levels | 2-year all-cause mortality | Bayesian Network, Deep Learning, Logistic Regression, Random Forest | 0.81 Bayesian Network |
Raghunath et al., 2020114 | United States | 253,397 patients with 1,169,662 12-lead resting ECGs from a large regional health system, 99,371 events occurred over a 34-year period | 12-lead ECG voltage–time traces, age, sex | 1-year all-cause mortality | Deep Neural Network | 0.88 Deep Neural Network |
Kawano et al., 2022115 | Japan | 116,749 participants from health checkups; age, sex, and other health data collected | Age, sex, smoking, AST levels, alcohol consumption, and other health checkup data | 5-year all-cause mortality | Gradient Boosting Decision Tree (XGBoost), Neural Network, Logistic Regression | 0.81 XGBoost |
Weng et al., 2019116 | United Kingdom | 502,628 participants aged 40–69, recruited from the general population | 60 predictor variables including demographics, lifestyle factors, and clinical measures | Premature all-cause mortality | Deep Learning, Random Forest, Cox regression | 0.79 Deep Learning |
Zhou et al., 2021117 | China | 1,241 patients with end-stage renal disease (ESRD) on peritoneal dialysis; age range 18 + , routine follow-up for 12 + months | Age, sex, chronic heart disease, diabetes, malignancy, systolic and diastolic blood pressure, cholesterol levels, serum albumin, hemoglobin, blood urea nitrogen, serum creatinine | Premature all-cause mortality | Logistic Regression, Classic Artificial Neural Network, Mixed Artificial Neural Network | 0.79 Artificial Neural Network |
Huang et al., 2017118 | Taiwan, China | 3,632 breast cancer patients from a retrospective cohort study, underwent surgery between 1996 and 2010 | Age, Charlson Comorbidity Index (CCI), chemotherapy, radiotherapy, hormone therapy, surgery volumes of hospital and surgeon | 5-year mortality after breast cancer surgery | Artificial Neural Network, Multiple Logistic Regression, Cox Regression | 0.72 Artificial Neural Network |
Sheng et al., 2020119 | China | 5,351 patients in training cohort and 5,828 in testing cohort from 97 renal centers, all new hemodialysis patients | Demographic info, disease diagnoses, comorbidities, and lab results collected at dialysis start and 0–3 months post-start | First-year all-cause mortality | XGBoost, Random Forest, and Logistic Regression | 0.85 XGBoost |
Zachariah et al., 2022120 | United States | 2,041 patients with advanced cancer, median age 62.6 years | Demographic data, lab results, diagnostic codes, past medical history | 3-month mortality | XGBoost | 0.81 XGBoost |
Unterhuber et al., 2021121 | Germany, Italy | 1,998 patients from LIFE-Heart Study (Germany), 772 from PLIC Study (Italy), aged 41–85, at increased cardiovascular risk | 92 plasma proteins measured, alongside clinical risk scores like SCORE and Framingham, and other clinical data | All-cause mortality | XGBoost, Neural Network | 0.94 Neural Network |
Wu et al., 2024122 | China | 11,894 patients aged ≥ 65 years who underwent non-cardiac surgery across 20 tertiary hospitals | Preoperative risk factors including medical history (stroke, chronic diseases), laboratory data (mononuclear cell ratio, total cholesterol), and preoperative assessments | 6-month mortality | Random Forests, Support Vector Machine, Decision Tree, Naive Bayes | 0.97 Random Forest |
Hwangbo et al., 2022123 | South Korea | 19,435 patients assessed from the IST-1 dataset, with 8,787 included after exclusions for various reasons such as treatment type and missing data | 18 variables including age, sex, level of consciousness, underlying conditions, and imaging findings | 6-month all-cause mortality | Stacking Ensemble, Extreme Gradient Boosting, KNN, SVM, Naive Bayes, Random Forest, Logistic Regression | 0.78 Stacking Ensemble |
Ross et al., 2016124 | United States | 1,755 patients undergoing elective coronary angiography at Stanford University Medical Center or Mount Sinai Medical Center from 2004 to 2008 | Demographics, clinical comorbidities, medications, lab tests, physical exam variables, socioeconomic variables, genomic markers | All-cause mortality | Elastic Net, Random Forest | 0.76 Random Forest |
Wang et al., 2019125 | United States | 40,711 completed cases from the Cardiovascular Disease Life Risk Pooling Project, which includes various demographics, physiological test results, medication status, socio-behavioral factors, and mortality indicators | Demographics, physiological tests, medication status, socio-behavioral factors | All-cause mortality | Naïve Bayes, Logistic Regression, Support Vector Machine, Random Forest | 0.89 Random Forest |
