Prediction of Incident Diabetic Retinopathy in Adults With Type 1 Diabetes Using Machine Learning Approach: An Exploratory Study
Authors' conclusion
Machine learning approaches using CGM data may have potential to aid in identifying adults with T1D at risk of DR.
Study context
- Population
- Adults with T1D with incident diabetic retinopathy (DR) or no retinopathy, June 2018–March 2022; Continuous glucose monitoring (CGM) data up to seven years before incident DR or no retinopathy
- Participants
- 30 vs 30
- Comparison
- Classifiers using CGM glycemic features (scenario 1) vs two principal components of those features (scenario 2)
- Incident DRn = 30
- Controln = 30reference
- Outcome
- Classification performance (AUC-ROC) for predicting incident DR
Methodological detailsretrospective CGM data, 10-fold cross-validation
- Retrospective CGM data
- 10-fold cross-validation
Performance
As reported in the abstract
AUC-ROC of classifiers
Scenario 2 (two PCs: exposure to hyperglycemia and hypoglycemia risk) classifiers outperformed scenario 1; average AUC-ROC for two of three models
0.92
Further findings
| Outcome | Incident DR | Control |
|---|---|---|
| Baseline characteristics | mean 21.2 (SD 9.4) years (age) | 41.8 (SD 14.7) years (age) |
| HbA1c | 8.6%±1.0% | 7.0%±0.9% |
| BMI | 24.5±4.8 kg/m² | 26.2±3.6 kg/m² |
Original publication
Prediction of Incident Diabetic Retinopathy in Adults With Type 1 Diabetes Using Machine Learning Approach: An Exploratory Study. J Diabetes Sci Technol. Published 28 October 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →