A Machine Learning Model for Week-Ahead Hypoglycemia Prediction From Continuous Glucose Monitoring Data
Authors' conclusion
We developed a machine learning algorithm to estimate the probability of a clinically significant hypoglycemic event within one week. Predictive algorithms may provide diabetes care providers using RPM with additional context when prioritizing T1D patients for review.
Study context
- Population
- Patients with T1D with continuous glucose monitoring (CGM) data from three cohorts (REPLACE-BG, Juvenile Diabetes Research Foundation (JDRF), Tidepool)3 cohorts
- Participants
- 226 / 355 / 120
- Comparison
- REPLACE-BG vs JDRF vs Tidepool
- REPLACE-BGn = 226
- JDRFn = 355
- Tidepooln = 120
- Outcome
- Prediction of clinically significant hypoglycemic event (CGM <54 mg/dL for ≥15 consecutive minutes) the following week from one week of CGM data; performance by ROC-AUC
- Follow-up
- Within one week
Methodological detailsmachine learning, 10-fold cross-validation, external validation
- Machine learning
- 10-fold cross-validation
- External validation
Performance
As reported in the abstract
Average ROC-AUC, 10-fold cross-validation
REPLACE-BG
0.77
average · SD 0.0233
JDRF
0.74
average · SD 0.0188
Tidepool
0.76
average · SD 0.02
Further findings
Average ROC-AUC across the three cohorts, external validation
- Average ROC-AUC (SD 0.0262)
- 0.74
Original publication
A Machine Learning Model for Week-Ahead Hypoglycemia Prediction From Continuous Glucose Monitoring Data. J Diabetes Sci Technol. Published 6 March 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →