Nutritional markers of undiagnosed type 2 diabetes in adults: Findings of a machine learning analysis with external validation and benchmarking

Abstract

Objectives Using a nationally-representative, cross-sectional cohort, we examined nutritional markers of undiagnosed type 2 diabetes in adults via machine learning. Methods A total of 16429 men and non-pregnant women ≥ 20 years of age were analysed from five consecutive cycles of the National Health and Nutrition Examination Survey. Cohorts from years 2013–2016 (n = 6673) was used for external validation. Undiagnosed type 2 diabetes was determined by a negative response to the question “Have you ever been told by a doctor that you have diabetes?” and a positive glycaemic response to one or more of the three diagnostic tests (HbA1c > 6.4% or FPG >125 mg/dl or 2-hr post-OGTT glucose > 200mg/dl). Following comprehensive literature search, 114 potential nutritional markers were modelled with 13 behavioural and 12 socio-economic variables. We tested three machine learning algorithms on original and resampled training datasets built using three resampling methods. From this, the derived 12 predictive models were validated on internal- and external validation cohorts. Magnitudes of associations were gauged through odds ratios in logistic models and variable importance in others. Models were benchmarked against the ADA diabetes risk test. Results The prevalence of undiagnosed type 2 diabetes was 5.26%. Four best-performing models (AUROC range: 74.9%-75.7%) classified 39 markers of undiagnosed type 2 diabetes; 28 via one or more of the three best-performing non-linear/ensemble models and 11 uniquely by the logistic model. They comprised 14 nutrient-based, 12 anthropometry-based, 9 socio-behavioural, and 4 diet-associated markers. AUROC of all models were on a par with ADA diabetes risk test on both internal and external validation cohorts (p>0.05). Conclusions Models performed comparably to the chosen benchmark. Novel behavioural markers such as the number of meals not prepared from home were revealed. This approach may be useful in nutritional epidemiology to unravel new associations with type 2 diabetes.

The methodology of food design. Part 2. Digital nutritiology in personal food

“…But they cannot interpret the large amount of data collected during the formation of personal dietary recommendations, when analyzing the relation between intake of nutrients and the health of a consumer. Therefore it is necessary to supplement traditional statistical methods with machine-aided processing of data [44,45]. Machine-aided processing of datarefers to a computer system capable of describing the solution to a given problem and creating an algorithm based on this solution.…”

Predicting the 2-Year Risk of Progression from Prediabetes to Diabetes Using Machine Learning among Chinese Elderly Adults

“…The insufficiency of features might also be one of the reasons why our XGBoost model did not perform as well as in other research [ 28 ], whose model included 300 features. After all, in addition to demographic and lifestyle, nutrition intake has also been found to be an important predictor of incident diabetes [ 29 ]. However, high-dimensional features generally bring about information redundancy and overfitting problem.…”

Recent applications of machine learning and deep learning models in the prediction, diagnosis, and management of diabetes: a comprehensive review

“…To determine the undiagnosed T2DM in adults, the nutritional markers were found by ANN, LR, and RF models [ 55 ]. To overcome the impact of the class imbalance, resampling algorithms containing Random Oversampling Examples (ROSE), minority class over-sampling, and Synthetic Minority Oversampling Technique (SMOTE) were applied.…”

“…The publicly available dataset are mentioned in bold Sample number ML models Refs. Early diagnosis and prediction of diabetes T2DM 15,005 subjects with age ≥ 3 XGBoost, DNN, and RF [ 63 ] 1512 subjects LR, RF, Naive Bayes (NB), SVM, XGBT, ANN, K-nearest neighbor (KNN), DT, XceptionResNet 50, DenseNet121, Vgg16, Vgg19, and InceptionV3, Stacking model of non-invasive variables and the Resnet50 model [ 53 ] 530 participants: 272 were diabetic patients and 258 were non-diabetic patients Deep autoencoder learning algorithm with CNN networks and deep radial basis function neural network (RBFNN) classifier [ 52 ] 217 participants with diabetes, prediabetes and normal conditions SVM, K-nearest neighbors, RF, XGBoost, hybrid feature selection-XGBoost [ 91 ] 2371 T1-weighted whole-body MRI data sets DenseNet architecture [ 54 ] 8454 subjects over five years of follow- up XGBoost, SVM, LR, RF, and ensemble algorithms [ 64 ] 16,429 men and non-pregnant women ≥ 20 years of age ANN, LR, and RF models [ 55 ] 453,487 T2DM patients Reverse engineering and forward simulation (REFS) [ 124 ] 82 obese women (40 non-diabetic and 42 diabetes) Separability-correlation measure (SCM) and ANN [ 57 ] 13,309 Canadian patients GBM and LR [ 92 ] Kaggle diabetes dataset RF ...…”