Longitudinal Analysis of Diabetes-Respiratory Distress Connections in Multimodal QBB Data Using Artificial Intelligence

Type 2 Diabetes Mellitus (T2DM) and Respiratory Distress Syndrome (RDS) are complex disorders that potentially share biological mechanisms. This study presents a longitudinal analysis of multimodal data from the Qatar Biobank (QBB) comprising 1,533 participants (1,220 healthy, 313 diabetic), integrating clinical and respiratory features to investigate associations between these conditions. Spirometry, a commonly utilized pulmonary function test, serves as a critical diagnostic tool for assessing airway obstruction and lung restriction, aiding in early detection of respiratory issues and monitoring treatment effectiveness. Recent evidence suggests that reductions in spirometric measurements, particularly forced vital capacity (FVC) and forced expiratory volume in 1 second (FEV1), are emerging as novel risk indicators for T2DM, with abnormalities often detectable before clinical diabetes diagnosis. Despite this association, spirometric evaluation remains uncommon in standard diabetic care. Leveraging state-of-the-art artificial intelligence techniques-including XGBoost, LightGBM, random forest, and deep learning models-we identified meaningful patterns between diabetes and respiratory data, with XGBoost achieving superior performance (76.22% accuracy, 76.48% AUROC, 76% average recall). Moreover, we applied the SHAP value method to investigate the association of identified risk factors with diabetes onset. Our multimodal approach enables comprehensive understanding of disease trajectories, highlighting potential biomarkers and risk factors common to both conditions while demonstrating the value of biobank data for advancing precision medicine in chronic disease management.