Evaluation of Machine Learning Methods Developed for Prediction of Diabetes Complications: A Systematic Review
Authors' conclusion
Random forest was found to have the overall best prediction performance. Current ML prediction models remain largely exploratory, and external validation studies are required before their clinical implementation.
Study context
- Population
- Studies developing or validating machine learning (ML) prediction models for microvascular or macrovascular complications in adults with T2D32 studies · 87 ML models
- Comparison
- ML prediction models vs non-ML methods
- Outcome
- Prediction performance by AUC (AUC >0.75 = clearly useful discrimination; positive mean relative AUC difference = better comparative performance)
Methodological detailsPRISMA, 13 606 articles screened
- PRISMA
- 13 606 articles screened
Pooled result
As reported in the abstract
Discrimination performance (AUC)
Models with clearly useful discrimination, across predicted outcomes
36%
Further pooled results
Model types and predictors
- Neural networks most frequently utilized
- 15
- Common predictors: age, diabetes duration, BMI
Risk of bias
- Studies with high risk of bias
- 31
Original publication
Evaluation of Machine Learning Methods Developed for Prediction of Diabetes Complications: A Systematic Review. J Diabetes Sci Technol. Published 3 November 2021. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →