Table 2 General characteristics of the included studies in the systematic review of cardiovascular diabetes prediction models
References | Study designs | Population | Predictors | Model types | Outcome |
|---|---|---|---|---|---|
[31] | Cohort | Greece | Body mass index (BMI), Hba1c, fasting blood glucose (FBG), lipid profile (LP), age, smoking habit, hypertension (HPT), pulse pressure, lipid-lowering therapy, parental history of diabetes | XGBoost | Development of prediction model for fatal or non-fatal incidence in T2DM individuals |
[32] | Cohort | Denmark | Disease codes, prescription of insulin and analogues, and prescription of blood glucose lowering drugs | Logistic ridge regression, random forest, decision tree gradient boosting | Prediction of individuals at elevated risk of developing T2DM comorbidities |
[29] | Cohort | Greece | Age, diabetes duration, Hba1c, blood pressure (BP), FBG, LP, smoking habit, sex, HPT, lipid-lowering therapy | HWNN, SOM, BLR, FFN, CART, RF, NB | Development of prediction model for fatal or non-fatal incidence in T2DM individuals |
[30] | Cohort | United States | BMI, BP, age, sex, hypertensions, heart and diabetic complications, other nosology, insulin, sugar-lowering drugs, other drugs | XGBoost, DT, RF, LR, Dummy, kNN, multinominal, complement and Bernoulli’s NB | Prediction of individuals at elevated risk of developing T2DM comorbidities |
[13] | Cohort | United States, Greece | Age, diabetes duration, BMI, BP, Hba1c, FBG, LP, HPT, ACE inhibitor, sex, diabetic parents, retinopathy, calcium antagonists, diuretics, B-blockers, smoking habit, proteinuria, hypolipid diet, aspirin, diet, sulphonyl urea, diguanide, insulin | ANN, binary logistic model, logistic model tree, Bayes net, DT, naïve Bayes | Assess the ability and performance of six machine learning models in prediction T2DM and CVD complications |
[33] | Cohort | China | Sex, age, race, total cholesterol, high density lipoprotein (HDL), systolic BP, anti-HPT treatment, diabetes, and smoking habit | Knowledge learning symbiosis (KLS) | Development of prediction model for CVD risk in T2DM individuals |
[34] | Cohort | Sweden | Sex, systolic BP, BMI, smoking habit, diagnosis of atrial fibrillation, myocardial and stroke history, HbA1c, HDL, total cholesterol, duration of type 2 diabetes, microalbuminuria, macroalbuminuria | Cox gradient boosting machine learning (GBM) | Assess eighty cardiovascular and inflammatory proteins for biomarker discovery and the prediction of major cardiovascular events in T2DM |
[28] | Cross-sectional | China | Sex, age, marital status, educational level, monthly income, diabetes duration, insulin treatment, HbA1c, FBG, LP, BP, BMI, anxiety, depression, smoking habit, and drinking habit | Deep neural network | Development of a CVD risk prediction model based on the bio-psycho-social contributors in T2DM patients |
[35] | Cohort | Australia | Age, sex, admission episode, discharge dates and disease codes | LR, SVM, DT, RF, NB, kNN | Development of prediction model for CVD risk in T2DM individuals |
[36] | Cohort | United States | BMI, age, and fasting plasma glucose | SVM, kNN | Development of prediction model for CVD risk in T2DM individuals |