What is the clinical evidence from this study?

Study design: Cross-Sectional. Population: Type 2 Diabetes Mellitus (n=834). Intervention: Deep Neural Network (DNN) CVD risk prediction model. Primary outcome: Area Under Curve (AUC) for CVD risk prediction.

April 28, 2021Open Access

Roles of Anxiety and Depression in Predicting Cardiovascular Disease Among Patients With Type 2 Diabetes Mellitus: A Machine Learning Approach

Key Result

A Deep Neural Network model incorporating bio-psycho-social factors, including anxiety and depression, predicted cardiovascular disease risk in patients with type 2 diabetes with an AUC of 0.91.

Study Design

Type

Cross-Sectional (n=834)

Multicenter

Structured PICO

Does a machine learning model incorporating bio-psycho-social factors accurately predict cardiovascular disease risk in patients with type 2 diabetes mellitus?

Population

834 adults with type 2 diabetes mellitus (T2DM) recruited from the Department of Endocrinology, Affiliated Hospital of Harbin Medical University, China. Mean age 52.58, 58.6% male. Exclusion criteria: type 1 diabetes mellitus, voluntary problems, and severe diseases affecting safety.

Intervention

Deep Neural Network (DNN) machine learning model incorporating bio-psycho-social factors (including anxiety and depression scores) for cardiovascular disease risk prediction

Outcome

Cardiovascular disease (composite of coronary heart disease, peripheral vascular disease, myocardial infarction, and heart failure events)composite

A machine learning model incorporating psychological factors like anxiety and depression alongside traditional clinical metrics accurately predicts cardiovascular disease risk in patients with type 2 diabetes.

Limitations

Study participants were recruited from one hospital, leading to a potential sampling bias.
Cross-sectional design prevents drawing causative conclusions about the associations between factors.
Use of self-rating scales for anxiety and depression may lead to underestimation due to possible alexithymia traits.
Limited number of input variables in the CVD risk prediction model.

Abstract

Cardiovascular disease (CVD) is a major complication of type 2 diabetes mellitus (T2DM). In addition to traditional risk factors, psychological determinants play an important role in CVD risk. This study applied Deep Neural Network (DNN) to develop a CVD risk prediction model and explored the bio-psycho-social contributors to the CVD risk among patients with T2DM. From 2017 to 2020, 834 patients with T2DM were recruited from the Department of Endocrinology, Affiliated Hospital of Harbin Medical University, China. In this cross-sectional study, the patients' bio-psycho-social information was collected through clinical examinations and questionnaires. The dataset was randomly split into a 75% train set and a 25% test set. DNN was implemented at the best performance on the train set and applied on the test set. The receiver operating characteristic curve (ROC) analysis was used to evaluate the model performance. Of participants, 272 (32.6%) were diagnosed with CVD. The developed ensemble model for CVD risk achieved an area under curve score of 0.91, accuracy of 87.50%, sensitivity of 88.06%, and specificity of 87.23%. Among patients with T2DM, the top five predictors in the CVD risk model were body mass index, anxiety, depression, total cholesterol, and systolic blood pressure. In summary, machine learning models can provide an automated identification mechanism for patients at CVD risk. Integrated treatment measures should be taken in health management, including clinical care, mental health improvement, and health behavior promotion.

اسأل الذكاء الاصطناعي

Bookmark

View Full Paper