Enhancing the Capabilities of Continuous Glucose Monitoring With a Predictive App
Authors' conclusion
The consistency of the performance of the three predictive models when evaluated on different cohorts of subjects with T1D and T2D on different insulin therapies, including real-world data, offers reassurance for real-world efficacy.
Study context
- Population
- Subjects with T1D on MDI, T2D on MDI and T1D on insulin pump therapy; Accu-Chek SmartGuide Predict continuous glucose monitoring (CGM) appthree data sets · Subjects with T1D on MDI 21 · T2D on MDI 59 · T1D on insulin pump therapy 226
- Comparison
- Two-hour glucose forecast vs 30-minute low glucose detection vs Nighttime low glucose prediction
- Outcome
- Predictive performance of the three models (aggregated data set)
Methodological detailsmachine learning models, three data sets, includes real-world data
- Machine learning models
- Three data sets
- Includes real-world data
Performance
As reported in the abstract
| Test or model | Consensus Error Grid zones A+B, 30 min horizon | Consensus Error Grid zones A+B, 45 min horizon | Consensus Error Grid zones A+B, 60 min horizon | Consensus Error Grid zones A+B, 120 min horizon | Accuracy | Sensitivity | Specificity | Mean lead time | ROC AUC |
|---|---|---|---|---|---|---|---|---|---|
| Two-hour glucose forecast | 99.8% | 99.3% | 98.7% | 96.3% | |||||
| 30-minute low glucose detection | 98.9% | 95.2% | 98.9% | 16.2 minutes | 0.958 | ||||
| Nighttime low glucose prediction | 86.5% | 55.3% | 91.6% | 0.859 |
Consensus Error Grid zones A+B, 30 min horizon
Two-hour glucose forecast99.8%
30-minute low glucose detection— not stated in the abstract
Nighttime low glucose prediction— not stated in the abstract
Consensus Error Grid zones A+B, 45 min horizon
Two-hour glucose forecast99.3%
30-minute low glucose detection— not stated in the abstract
Nighttime low glucose prediction— not stated in the abstract
Consensus Error Grid zones A+B, 60 min horizon
Two-hour glucose forecast98.7%
30-minute low glucose detection— not stated in the abstract
Nighttime low glucose prediction— not stated in the abstract
Consensus Error Grid zones A+B, 120 min horizon
Two-hour glucose forecast96.3%
30-minute low glucose detection— not stated in the abstract
Nighttime low glucose prediction— not stated in the abstract
Accuracy
Two-hour glucose forecast— not stated in the abstract
30-minute low glucose detection98.9%
Nighttime low glucose prediction86.5%
Sensitivity
Two-hour glucose forecast— not stated in the abstract
30-minute low glucose detection95.2%
Nighttime low glucose prediction55.3%
Specificity
Two-hour glucose forecast— not stated in the abstract
30-minute low glucose detection98.9%
Nighttime low glucose prediction91.6%
Mean lead time
Two-hour glucose forecast— not stated in the abstract
30-minute low glucose detection16.2 minutes
Nighttime low glucose prediction— not stated in the abstract
ROC AUC
Two-hour glucose forecast— not stated in the abstract
30-minute low glucose detection0.958
Nighttime low glucose prediction0.859
Original publication
Enhancing the Capabilities of Continuous Glucose Monitoring With a Predictive App. J Diabetes Sci Technol. Published 19 August 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →