Sign in Subscribe
Machine learning

Machine Learning Algorithm Accurately Predicts Risk of Type 2 Diabetes in Children with Obesity

Strategic Insights

— 1 min read

A groundbreaking study by researchers from Tongji University in Shanghai, People's Republic of China, has developed and internally validated a machine learning (ML)-based model to predict the risk of type 2 diabetes mellitus (T2DM) in children with obesity. The research, published in Frontiers in Endocrinology, found that the Support Vector Machine (SVM) algorithm was the most effective predictor of T2DM, with an accuracy rate of 93.2% and an area under the receiver operating characteristic curve (AUC) of 0.98. The study followed 292 children with obesity and T2DM between July 2023 and February 2024, with 49 diagnoses made during the follow-up period.

Key Takeaways:

  • The SVM algorithm identified eight predictors of T2DM, including BMI, creatinine, prealbumin, glucose (180 min), glycosylated hemoglobin A1c, thyrotropin, total thyroxine (T4), and free T4 concentrations.
  • The machine learning-based prediction model accurately identifies children with obesity at high risk of T2DM, with a potential to facilitate early, personalized interventions to prevent the disease.
  • The rising prevalence of obesity in childhood is associated with an increased risk of early-onset T2DM, highlighting the need for early identification and prevention strategies.
  • The study found that the SVM algorithm was the best predictor of T2DM among eight ML algorithms compared, with a larger AUC and accuracy rate compared to the other algorithms.
  • The research team identified eight key clinical and laboratory characteristics of T2DM in children, which can be used to create a risk prediction model.
  • The study's findings suggest that a machine learning-based prediction model could be a valuable tool in the early detection and prevention of T2DM in at-risk children.

Statistics:

  • 292 children with obesity were enrolled in the study between July 2023 and February 2024.
  • 49 children were diagnosed with T2DM during the follow-up period.
  • The SVM algorithm had an accuracy rate of 93.2% and an AUC of 0.98.
  • The machine learning-based prediction model identified eight predictors of T2DM, including BMI, creatinine, prealbumin, glucose (180 min), glycosylated hemoglobin A1c, thyrotropin, total thyroxine (T4), and free T4 concentrations.

Sources:

  • Development and internal validation of a machine learning algorithm for the risk of type 2 diabetes mellitus in children with obesity. Frontiers in Endocrinology, 2025, 16.
  • NewsRx. Researchers from Tongji University Describe Findings in Obesity and Diabetes (Development and internal validation of a machine learning algorithm for the risk of type 2 diabetes mellitus in children with obesity). Obesity & Diabetes Week. August 25, 2025; p 158.

Read more