Predicting Response to Bolus Insulin Therapy in Patients With Type 2 Diabetes
Authors' conclusion
Predictive modeling using routine health care data reasonably accurately classified patients initiating bolus insulin who would achieve and maintain HbA1c goals, but less so for differentiation between patients who never met and who did not maintain goals. Prior HbA1c was a major contributing parameter for the predictions.
Study context
- Population
- People with T2D initiating bolus insulin, US nationally representative insurance claims database (dNHI, 2007–2017); mean age 53.0 years (SD 8.7)
- Participants
- 15,331 people
- Exposure
- Class 0 vs Class 1 vs Class 2
- Class 0n = 4,510
- Class 1n = 7,800
- Class 2n = 3,021
- Outcome
- Model determining at individual level whether people with T2D are likely to achieve and maintain HbA1c goals (HbA1c <8.0% or reduction of baseline HbA1c by >1.0%)
Methodological detailsmachine learning, boosted decision tree ensembles (XGBoost), insurance claims database
- Machine learning
- Boosted decision tree ensembles (XGBoost)
- Insurance claims database
Main finding
As reported in the abstract
Model prediction performance
| Group | Value | ROC | Area under precision-recall curve |
|---|---|---|---|
| Class 0 | 29.4% | 0.71 | 0.46 |
| Class 1 | 50.9% | 0.62 | 0.58 |
| Class 2 | 19.7% | 0.92 | 0.71 |
Value
Class 029.4%
Class 150.9%
Class 219.7%
ROC
Class 00.71
Class 10.62
Class 20.92
Area under precision-recall curve
Class 00.46
Class 10.58
Class 20.71
- Overall model ROC
- 0.79
Observational data: associations, not effects.
Original publication
Predicting Response to Bolus Insulin Therapy in Patients With Type 2 Diabetes. J Diabetes Sci Technol. Published 20 May 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →