Prediction of complications of type 2 Diabetes: A Machine learning approach
Antonio Nicolucci1, Luca Romeo2, Michele Bernardini2
1Center for Outcomes Research and Clinical Epidemiology - CORESEARCH, Pescara, Italy.
Insights
Machine learning accurately predicts diabetes complications (DCs) within five years using electronic health records. This approach identifies high-risk patients, improving diabetes care and overcoming treatment delays.
Area of Science:
- Medical Informatics
- Machine Learning in Healthcare
- Diabetes Management
Background:
- Diabetes complications (DCs) pose a significant burden on patients and healthcare systems.
- Early identification of patients at risk for DCs is crucial for timely intervention.
- Electronic medical records (EMRs) contain vast data for predictive modeling.
Purpose of the Study:
- To develop and validate machine learning models for predicting the onset of six major diabetes complications.
- To assess the models' ability to predict complications within five years and differentiate early vs. late onset.
Main Methods:
- Utilized a supervised, tree-based learning algorithm (XGBoost) on a large EMR dataset (147,664 patients over 15 years).
- Developed models for six DC groups: eye, cardiovascular, cerebrovascular, peripheral vascular disease, nephropathy, and neuropathy.
- Performed external validation across five centers and evaluated models using accuracy, sensitivity, specificity, and AUC.
Main Results:
- Predictive models for all DCs achieved accuracy >70% and AUC >0.80 (up to 0.97 for nephropathy) in task 1.
- Task 2 models also showed accuracy >70% and AUC >0.85 for early (2-year) and late (3-5 year) complication prediction.
- Sensitivity for early complication detection ranged from 83.2% (peripheral vascular disease) to 88.5% (nephropathy).
Conclusions:
- Machine learning models effectively identify patients at high risk for diabetes complications.
- This predictive capability can help overcome clinical inertia and enhance the quality of diabetes care.
- Big data analytics in EMRs offers a powerful tool for proactive diabetes management.
Aim:
To construct predictive models of diabetes complications (DCs) by big data machine learning, based on electronic medical records.
Methods:
Six groups of DCs were considered: eye complications, cardiovascular, cerebrovascular, and peripheral vascular disease, nephropathy, diabetic neuropathy. A supervised, tree-based learning approach (XGBoost) was used to predict the onset of each complication within 5 years (task 1). Furthermore, a separate prediction for early (within 2 years) and late (3-5 years) onset of complication (task 2) was performed. A dataset of 147.664 patients seen during 15 years by 23 centers was used. External validation was performed in five additional centers. Models were evaluated by considering accuracy, sensitivity, specificity, and area under the ROC curve (AUC).
Results:
For all DCs considered, the predictive models in task 1 showed an accuracy > 70 %, and AUC largely exceeded 0.80, reaching 0.97 for nephropathy. For task 2, all predictive models showed an accuracy > 70 % and an AUC > 0.85. Sensitivity in predicting the early occurrence of the complication ranged between 83.2 % (peripheral vascular disease) and 88.5 % (nephropathy).
Conclusions:
Machine learning approach offers the opportunity to identify patients at greater risk of complications. This can help overcoming clinical inertia and improving the quality of diabetes care.
Related Concept Videos
01:22Diabetes Mellitus: Type 2 and Gestational
01:15Diabetes: Symptoms, Diagnosis, and Complications
01:22Diabetes Mellitus: Overview and Type I Subtype
Type 1 diabetes is an autoimmune disease in which the immune system mistakenly attacks and destroys the insulin-producing beta cells in the pancreas. As a result, the body is unable to produce sufficient insulin, and individuals with...
01:36Carbohydrate Metabolism
Starch accounts for approximately 60% of the carbohydrates consumed by humans. Since amylase enzymes cannot function in the stomach's acidic environment, starch can only be digested in the mouth and small intestine. Simple sugars are found naturally in milk and fruits in...
01:20Pathophysiology of Diabetes
Type 1 diabetes is characterized by autoimmune-mediated destruction of pancreatic β cells, with environmental factors potentially triggering this process in genetically susceptible individuals. Despite many not having a family history, certain genes increase susceptibility,...
01:15Diabetes: Management and Pharmacotherapy
Insulin remains the cornerstone of treatment for most patients with type 1 and many...