Use of Machine Learning and Routine Laboratory Tests for Diabetes Mellitus Screening

Predicting diabetic retinopathy based on routine laboratory tests by machine learning algorithms

OBJECTIVES

This study aimed to identify risk factors for diabetic retinopathy (DR) and develop machine learning (ML)-based predictive models using routine laboratory data in patients with type 2 diabetes mellitus (T2DM).

METHODS

Clinical data from 4259 T2DM inpatients at Beijing Tongren Hospital were analyzed, divided into a model construction data set (N = 3936) and an external validation data set (N = 323). Using 39 optimal variables, a prediction model was constructed using the eXtreme Gradient Boosting (XGBoost) algorithm and compared with four other algorithms: support vector machine (SVM), gradient boosting decision tree (GBDT), neural network (NN), and logistic regression (LR). The Shapley Additive exPlanation (SHAP) method was employed to interpret the XGBoost model. External validation was performed to assess model performance.

RESULTS

DR was present in 47.69% (N = 1877) of T2DM patients in the model construction data set. Among the models tested, the XGBoost model performed best with an AUC of 0.831, accuracy of 0.757, sensitivity of 0.754, specificity of 0.759, and F1-score of 0.752. SHAP explained feature importance for XGBoost model and identified key risk factors for DR. External validation yielded an accuracy of 0.650 for the XGBoost model.

CONCLUSIONS

The XGBoost-based prediction model effectively assesses DR risk in T2DM patients using routine laboratory data, aiding clinicians in identifying high-risk individuals and guiding personalized management strategies, especially in medically underserved areas.

Automated sample annotation for diabetes mellitus in healthcare integrated biobanking

Healthcare integrated biobanking describes the annotation and collection of residual samples from hospitalized patients for research purposes. The central idea of the current work is to establish an automated workflow for sample annotation, selection and storage for diabetes mellitus. This is challenging due to incomplete data at the time of sample selection. The study evaluates a machine learning (ML) and natural language processing (NLP) based two-step procedure for timely and precise sample annotation for diabetes mellitus. Electronic health record data of 785 persons were extracted from the hospital information system. In the first step, a conditional inference forest (CIF) model was trained and tested based on laboratory values from the first 72 h of the hospital stay using test- (n = 550) and training data sets (n = 235). Performance was compared with a simple laboratory cut-off classifier (LCC) and a logistic regression (LR) model. Algorithms based on laboratory values, ICD-10 codes or information from discharge summaries extracted by a natural language processing software (NLP-DS) were evaluated as a second (review) step designed to increase the precision of annotations. For the first step, recall/precision/F1-score/accuracy were 71 %/86 %/0.78/0.82 for CIF and 77 %/70 %/0.74/0.75 for LR compared to 73 %/68 %/0.70/0.72 for LCC. NLP-DS was the best-performing second (review) step (93 %/100 %/0.97/0.97). Combining first-step models with NLP-DS increased precision to 100 % for all procedures (66 %/100 %/0.80/0.85 for CIF&NLP-DS, 72 %/100 %/0.84/87.2 for LR&NLP-DS and 66 %/100 %/0.80/0.85 for LCC&NLP-DS). The number of samples removed by NLP-DS was higher for LR&NLP-DS and LCC&NLP-DS (removal rate 35 % and 38 % of initially selected samples) compared to CIF&NLP-DS (removal rate of 20 %). The developed two-step procedure is an efficient implementable method for timely and precise annotation of samples from diabetic hospitalized patients.

Performance of different machine learning algorithms in identifying undiagnosed diabetes based on nonlaboratory parameters and the influence of muscle strength: A cross-sectional study

AIMS/INTRODUCTION

Machine learning algorithms based on the artificial neural network (ANN), support vector machine, naive Bayesian or logistic regression model are commonly used to identify diabetes. This study investigated which approach performed the best and whether muscle strength provided any incremental benefit in identifying undiagnosed diabetes in Chinese adults.

METHODS

This cross-sectional study enrolled 4,482 eligible participants from eight provinces in China, who were randomly divided into the training dataset (n = 3,586) and the testing dataset (n = 896). Muscle strength was assessed by handgrip strength and the number of chair stands in the 30-s chair stand test. An oral glucose tolerance test was used to ascertain undiagnosed diabetes. The areas under the curve (AUCs) were calculated accordingly and compared with each other.

RESULTS

Of the included participants, 233 had newly diagnosed diabetes. All the four machine learning algorithms, which were developed based on nonlaboratory parameters, showed acceptable discriminative ability in identifying undiagnosed diabetes (all AUCs >0.70), with the ANN approach performing the best (AUC 0.806). Adding handgrip strength or the 30-s chair stand test to this approach did not increase the AUC further (P = 0.39 and 0.26, respectively). Furthermore, compared with the New Chinese Diabetes Risk Score, the ANN approach showed a larger AUC in identifying undiagnosed diabetes (Pcomparison < 0.01), regardless of the addition of handgrip strength or the 30-s chair stand test.

CONCLUSIONS

The ANN approach performed the best in identifying undiagnosed diabetes in Chinese adults; however, the addition of muscle strength might not improve its efficacy.

Metabolic and Genetic Markers Improve Prediction of Incident Type 2 Diabetes: A Nested Case-Control Study in Chinese

CONTEXT

It is essential to improve the current predictive ability for type 2 diabetes (T2D) risk.

OBJECTIVE

We aimed to identify novel metabolic markers for future T2D in Chinese individuals of Han ethnicity and to determine whether the combined effect of metabolic and genetic markers improves the accuracy of prediction models containing clinical factors.

METHODS

A nested case-control study containing 220 incident T2D patients and 220 age- and sex- matched controls from normoglycemic Chinese individuals of Han ethnicity was conducted within the Wuxi Non-Communicable Disease cohort with a 12-year follow-up. Metabolic profiling detection was performed by high-performance liquid chromatography‒mass spectrometry (HPLC-MS) by an untargeted strategy and 20 single nucleotide polymorphisms (SNPs) associated with T2D were genotyped using the Iplex Sequenom MassARRAY platform. Machine learning methods were used to identify metabolites associated with future T2D risk.

RESULTS

We found that abnormal levels of 5 metabolites were associated with increased risk of future T2D: riboflavin, cnidioside A, 2-methoxy-5-(1H-1, 2, 4-triazol-5-yl)- 4-(trifluoromethyl) pyridine, 7-methylxanthine, and mestranol. The genetic risk score (GRS) based on 20 SNPs was significantly associated with T2D risk (OR = 1.35; 95% CI, 1.08-1.70 per SD). The area under the receiver operating characteristic curve (AUC) was greater for the model containing metabolites, GRS, and clinical traits than for the model containing clinical traits only (0.960 vs 0.798, P = 7.91 × 10-16).

CONCLUSION

In individuals with normal fasting glucose levels, abnormal levels of 5 metabolites were associated with future T2D. The combination of newly discovered metabolic markers and genetic markers could improve the prediction of incident T2D.