Machine learning-based reproducible prediction of type 2 diabetes subtypes
Authors' conclusion
The new ML model for predicting Ahlqvist's subtypes of type 2 diabetes has great potential for application in clinical practice and cohort studies because it can classify individuals with missing HOMA2 indices and predict glycaemic control, diabetic complications and treatment outcomes with long-term consistency by using readily available variables. Future studies are needed to assess whether our approach is applicable to research and/or clinical practice in multiethnic populations.
Study context
- Population
- Japanese individuals with T2D (cohort 1, training/test 7:3) and external validation cohort with T2D (cohort 2)cohorts of 619 and 597
- Comparison
- Machine learning (ML) model vs k-means clustering for Ahlqvist's T2D subtypes
- Outcome
- Accuracy of T2DRF15 for predicting T2Dkmeans T2D subtypes (SIDD, SIRD, MOD, MARD)
Methodological detailsmachine learning, random forest, external validation
- Machine learning
- Random forest
- External validation
Performance
As reported in the abstract
| Test or model | Value | Accuracy | AUC | F1 score | Accuracy, external validation |
|---|---|---|---|---|---|
| T2DRF15 | 94% | 94% | ≥0.99 | ≥0.9 | 86.3% |
T2DRF15
Further findings
- Accuracy in putative subset with missing insulin-related variables (with imputation) — T2DRF15
- 82.9%
Complications (Kaplan-Meier) · median 11.6 (4.5-18.3) years
- T2DRF15 clusters showed distinct accumulation risks of DR in severe insulin-deficient diabetes (SIDD) and CKD in severe insulin-resistant diabetes (SIRD), similarly to T2Dkmeans
Consistency of subtypes over time, after excluding undecidable individuals
- SIDD
- 100%
- SIRD
- 68.6%
- MOD
- 94.4%
- Mild age-related diabetes (MARD)
- 97.9%
- Consistency higher with T2DRF15 than T2Dkmeans
Original publication
Composed by AI from the published abstract and checked against it. How Briefs are made →