Early prediction of suboptimal glycemic control in children and adolescents with type 1 diabetes

Objectives: Children and adolescents with type 1 diabetes (T1D) face significant challenges in managing glycemic control due to developmental, psychological, and social factors. Failure to achieve optimal glycemic control can persist into adulthood, leading to the development of long-term complications. Early prediction of future suboptimal glycemic control among children and adolescents with T1D would be beneficial. Although machine learning (ML) models have shown promise for such predictions, few studies have explored their application to predicting glycemic control in children and adolescents. Thus, this study aimed to develop and validate ML models designed for early prediction of glycemic control in children and adolescents with T1D. Material and Methods: Children and adolescents aged 19 years or younger who were diagnosed with T1D between January 2020 and December 2022 were included. A glycosylated hemoglobin cut-off score ≥7% (53 mmoL/moL) was classified as suboptimal glycemic control. Two logistic regression models were developed using a forward feature selection. The performance was assessed using the area under the receiver operating characteristic area under the curve (ROC AUC). Furthermore, a correlation analysis between the most informative feature and future glycemic control was conducted. Results: A total of 100 participants were included. The models yielded an ROC AUC of 0.893 and 0.835, respectively. Age at diagnosis was the most informative feature, showing a significant nonlinear correlation with future glycemic control ( p < 0.0001). Conclusion: The models achieved acceptable performance for predicting future suboptimal glycemic control in children and adolescents with T1D, with age at diagnosis identified as the most informative feature.