Forecasting Type 2 Diabetes Complications in Kuwait (Dr. Mohammad Alfailakawi)

A Bayesian Approach with Electronic Health Records and Severity Analysis Applications.
This research proposes a Bayesian Network model to predict the onset and progression of complications in type 2 diabetes mellitus (T2DM) patients, using electronic health records (EHR) and codes from the 10th revision of the International Classification of Diseases (ICD-10). The model is evaluated using data, including complication-related features and risk factors, from Kuwait's Dasman Diabetes Institute (DDI).
The methodology used in the developed model uses ICD-10 codes to establish a three-level scale indicating the severity of complications, as well as to identify new complications. The Bayesian model effectively integrates this information to predict the possibility of new complications developing or of existing complications becoming more severe, thus providing physicians with the tool needed to offer a nuanced management of T2DM. Feature selection methods, particularly Complications as Feature Models CF-T¬all, showed consistent superior performance, achieving balanced precision, recall, and F1-scores. The results highlight the reliability and versatility of the CF-T¬all model in predicting complications; it achieved an accuracy of 75% and 97% for ophthalmic and kidney complications, respectively. 

Supervisor: Dr. Mohammad Alfailakawi
Co-Supervisor: Dr. Ameer Mohammed
Convener: Prof. Imtiaz Ahmad
Examination Committee: Dr. Sa'ed Abed