Machine Learning–Driven Prediction of Comorbidities and Mortality in Adults With Type 1 Diabetes
Authors' conclusion
The developed models showed good performance for predicting CVD and mortality, suggesting they could help in the early identification of these outcomes in individuals with T1D. The importance of SES in individual prediction within diabetes remains uncertain.
Study context
- Population
- Adults (≥21 years) with T1D, nationwide Danish registry data; features age, sex, personal income and education
- Participants
- 6,572 adults
- Comparison
- Cardiovascular disease (CVD) model vs diabetic kidney disease (DKD) model vs Mortality model
- Outcome
- Prediction of CVD, DKD and mortality (AUROC and PR-AUC)
- Follow-up
- Within five years of T1D diagnosis
Methodological detailsnationwide registry, logistic regression, five-fold cross-validation
- Nationwide registry
- Logistic regression
- Five-fold cross-validation
Performance
As reported in the abstract
Model performance (AUROC ±SD; PR-AUC) for 5-year prediction, five-fold cross-validation
- AUROC (±SD), CVD
- 0.79 (±0.03)
- AUROC (±SD), DKD
- 0.61 (±0.03)
- AUROC (±SD), mortality
- 0.87 (±0.01)
- Precision-recall area under the curve (PR-AUC), CVD
- 0.18 (±0.01)
- PR-AUC, DKD
- 0.15 (±0.03)
- PR-AUC, mortality
- 0.49 (±0.02)
Further findings
Incidence within five-year follow-up
- Developed CVD
- 379 (6%)
- Developed DKD
- 668 (10%)
- Died
- 921 (14%)
Importance of socioeconomic status (SES)
- SES was the most important feature in the DKD model but had minimal impact on models for CVD and mortality
Original publication
Machine Learning–Driven Prediction of Comorbidities and Mortality in Adults With Type 1 Diabetes. J Diabetes Sci Technol. Published 2 August 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →