Explainable Machine-Learning Models to Predict Weekly Risk of Hyperglycemia, Hypoglycemia, and Glycemic Variability in Patients With Type 1 Diabetes Based on Continuous Glucose Monitoring
Authors' conclusion
Prediction models based on real-world CGM data can be used to predict the risk of unstable glycemic control in the forthcoming week. The models showed good performance in both internal and external validation cohorts.
Study context
- Population
- Patients with T1D with long-term continuous glucose monitoring (CGM) monitoring; validation in two independent cohorts of varying ages+223 validation patients
- Participants
- 187 patients
- Comparison
- Logistic regression vs XGBoost vs TabNet prediction models
- Outcome
- Prediction of weekly hyperglycemia (TAR ≥5%), hypoglycemia (TBR ≥4%) and glycemic variability (CV ≥36%)
- Follow-up
- Forthcoming week
Methodological detailsinternal testing, external validation in two independent cohorts, binary classification
- Internal testing
- External validation in two independent cohorts
- Binary classification
Performance
As reported in the abstract
ROC-AUC, internal validation
| Test or model | AUC hyperglycemia | AUC hypoglycemia | AUC glycemic variability |
|---|---|---|---|
| XGBoost | 0.9 (95% CI 0.89-0.91) | 0.89 (95% CI 0.88-0.9) | 0.8 (95% CI 0.79-0.81) |
XGBoost
AUC hyperglycemia0.9 (95% CI 0.89-0.91)
AUC hypoglycemia0.89 (95% CI 0.88-0.9)
AUC glycemic variability0.8 (95% CI 0.79-0.81)
Further findings
ROC-AUC, external validation cohorts
- AUC hyperglycemia
- 0.88 to 0.95
- AUC hypoglycemia
- 0.84 to 0.89
- AUC glycemic variability
- 0.80 to 0.82
Original publication
Explainable Machine-Learning Models to Predict Weekly Risk of Hyperglycemia, Hypoglycemia, and Glycemic Variability in Patients With Type 1 Diabetes Based on Continuous Glucose Monitoring. J Diabetes Sci Technol. Published 8 October 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →