Predicting poor glycemic control during Ramadan among non-fasting patients with diabetes using artificial...

Area of Science:

• Endocrinology and Metabolism
• Artificial Intelligence in Healthcare
• Diabetes Management

Background:

• Ramadan fasting presents unique challenges for glycemic control in diabetic patients.
• Non-fasting diabetic individuals also experience glycemic variability during Ramadan.
• Predictive models are needed to identify patients at risk of poor glycemic control.

Purpose of the Study:

• To develop and validate machine learning models for predicting poor glycemic control in non-fasting diabetic patients during Ramadan.
• To identify key clinical and metabolic factors associated with glycemic deterioration.
• To provide a tool for proactive management of diabetes during Ramadan.

Main Methods:

• Prospective study with data collection at three time points: before, during, and after Ramadan.
• Inclusion of demographic, diabetes history, caloric intake, anthropometric, and metabolic parameters.
• Training and evaluation of multiple machine learning algorithms including Extra Trees Classifier, Logistic Regression, SVM, Naive Bayes, KNN, Decision Tree, Random Forest, and Catboost.

Main Results:

• The prevalence of poor glycemic control was 52.6%.
• Extra Trees Classifier achieved the highest performance (accuracy = 0.87, AUC = 0.87) in predicting glycemic deterioration.
• Significant predictors included caloric intake evolution, gender, baseline caloric intake, baseline weight, BMI variation, waist circumference evolution, and post-Ramadan Total Cholesterol levels.

Conclusions:

• Machine learning models, particularly Extra Trees Classifier, can effectively predict poor glycemic control in non-fasting diabetic patients during Ramadan.
• Identifying and targeting key risk factors like caloric intake, weight changes, and cholesterol levels can improve patient outcomes.
• Clinical application of these predictive insights can enhance diabetes management strategies during the Ramadan period.