Machine Learning Prognostics for the Obstructive Sleep Apnea Disorder Following Long COVID

In the aftermath of the COVID-19 pandemic, the phenomenon of Long COVID has emerged as a profound health concern. It presents enduring symptoms that significantly overlap with those of Obstructive Sleep Apnea Disorder (OSAD), such as inconsistent breathing patterns, sleep disturbances, and cardiovascular complications. This paper presents an analysis of Long COVID using healthcare data from the Froedtert Health Medical System in Wisconsin. By leveraging advanced Machine Learning (ML) methodologies, we have formulated predictive models aimed at assessing the risk of OSAD onset in individuals diagnosed with Long COVID. Additionally, our study reveals critical factors influencing the incidence of OSAD. Considering recent research that underscores the increased risk of Long COVID in patients with pre-existing OSADs, this innovative research inversely investigates the likelihood of developing OSAD post-Long COVID diagnosis. We utilized the Recursive Feature Elimination (RFE) approach to extract salient features that substantially impact OSAD from our dataset. To counter the dataset's underlying imbalance, we implemented the Synthetic Minority Over-sampling Technique and Edited Nearest Neighbors (SMOTEEN) strategy. We experimented with multiple ML models and validated them using cross-validation techniques. The results indicate that the Gaussian Naive Bayes (GNB) classifier exhibits superior performance, with an area under the ROC curve (AUC) of 0.967, precision of 0.942, and recall of 0.929. Random Forest (RF) classifier also demonstrates robust predictive capabilities for OSAD risk prediction, achieving an AUC of 0.964, precision of 0.936, and recall of 0.918. The Support Vector Classifier (SVC) similarly achieves commendable results with an AUC of 0.969, precision of 0.947, and recall of 0.926. The pivotal features identified by our predictive models are instrumental in recognizing individuals at an elevated risk of OSAD post Long Covid diagnosis, thus paving the way for targeted preventive interventions and the allocation of essential healthcare resources.