Toward Detection of Nocturnal Hypoglycemia in People With Diabetes Using Consumer-Grade Smartwatches and a Machine Learning Approach
Authors' conclusion
Consumer-grade smartwatches demonstrate promise for detecting nocturnal hypoglycemia, particularly during early sleep.
Study context
- Population
- Adults with insulin-treated diabetes wearing two consumer-grade smartwatches alongside continuous glucose monitoring (CGM) systems351 nights
- Participants
- 36 adults
- Comparison
- Level 1 hypoglycemia vs Level 2 hypoglycemia
- Outcome
- ML detection of hypoglycemia in nights with hypoglycemia during early sleep
- Follow-up
- Early sleep
Methodological detailsmachine learning
- Machine learning
Performance
As reported in the abstract
| Test or model | AUC | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|
| Level 1 hypoglycemia | 0.78 | 0.78 | 0.64 | 0.25 | 0.94 |
| Level 2 hypoglycemia | 0.83 | 0.89 | 0.64 | 0.13 | 0.99 |
AUC
Level 1 hypoglycemia0.78
Level 2 hypoglycemia0.83
Sensitivity
Level 1 hypoglycemia0.78
Level 2 hypoglycemia0.89
Specificity
Level 1 hypoglycemia0.64
Level 2 hypoglycemia0.64
PPV
Level 1 hypoglycemia0.25
Level 2 hypoglycemia0.13
NPV
Level 1 hypoglycemia0.94
Level 2 hypoglycemia0.99
Further findings
Nights with spontaneous hypoglycemia
- Nights with spontaneous hypoglycemia recorded
- 66
Sleep and vital signs, hypoglycemic vs non-hypoglycemic nights · early and late sleep
- Hypoglycemic nights showed increased wake periods, heart rate, stress levels and activity during early sleep, with weaker effects during late sleep
Original publication
Toward Detection of Nocturnal Hypoglycemia in People With Diabetes Using Consumer-Grade Smartwatches and a Machine Learning Approach. J Diabetes Sci Technol. Published 25 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →