Development and Validation of Binary Classifiers to Predict Nocturnal Hypoglycemia in Adults With Type 1 Diabetes
Authors' conclusion
Our model shows state-of-the-art performance, generalizability, and robustness in sensor devices from different manufacturers. We believe it is a potential viable approach to inform people with T1D about their risk of NH before it occurs.
Study context
- Population
- Adults with T1D using MDI; development cohort from 6-month free-living study with continuous glucose monitor (CGM); external population of 20 adults with T1D on MDI wearing CGM and flash glucose monitoring sensors for two 8-week periodsplus 20 external validation
- Participants
- 37 adults
- Comparison
- SVM vs Random Forests (RF) classifiers for predicting nocturnal hypoglycemia
- Outcome
- Prediction of nocturnal hypoglycemia (blood glucose <70 mg/dL) from daytime features (ROC-AUC)
Methodological detailsmachine learning, free-living conditions
- Machine learning
- Free-living conditions
Performance
As reported in the abstract
ROC-AUC for nocturnal hypoglycemia (NH) prediction, population-level
| Test or model | Value | Receiver operating characteristic-area under curve (ROC-AUC) |
|---|---|---|
| SVM | 79.36% (CI 76.86%–81.86%) | 79.36% |
SVM
Value79.36% (CI 76.86%–81.86%)
Receiver operating characteristic-area under curve (ROC-AUC)79.36%
Further findings
- SVM ROC-AUC in unseen external population
- 77.06%
- SVM ROC-AUC between the two different glucose sensors
- 77.74%
Original publication
Development and Validation of Binary Classifiers to Predict Nocturnal Hypoglycemia in Adults With Type 1 Diabetes. J Diabetes Sci Technol. Published 11 July 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →