A Machine Learning Model to Predict COVID-19 Infection Risk in the United Arab Emirates

Background: One of the biggest challenges that public health officials face is to identify the infection risk among the population and the same scenario was repeated during the COVID-19 period. It was crucial toidentify the patients at high risk who needed immediate care, and accordingly, an optimal machine learning model has been developed for the prediction of infection risk which can be modified and reused.Objective: The main objective of this article is to predict the COVID-19 infection risk of a patient. The predictive model would help in developing comprehensive plans to respond to both clinical treatments and controlof the spread.Method: A retrospective study was carried out on 23,996 encounters in a nationwide cohort in the United Arab Emirates. Patient information was collected retrospectively from respective physicians and was uploaded in a patient under investigation (PUI) form filled during the COVID-19 screening. The variables were age, gender, body temperature, comorbidities and patient symptoms. This data (age, gender, body temperature, comorbidities and patient symptoms) was fed into AIbased machine learning algorithms to come up with a COVID-19 infection prediction model.Results: The study found that 3818 (15.9%) cases were COVID-19 positive. Based on model performance, parsimony, and explainability, the Gradient Boosting Model was finally selected. The area under thecurve (AUC) for this model was 0.746 on the training dataset and 0.736 on the validation dataset.Conclusion: A highly interpretable machine learning model comprising multiple patient characteristics and symptoms was developed to predict COVID-19 infection in patients. How to cite this article:AlShaya S, Jaffar MS, Mohideen SAJ, Gopakumar A, Mathur V, Saeed M. A Machine Learning Model to Predict COVID-19 Infection Risk in the United Arab Emirates. J Commun Dis. 2023;55(4):71-79. DOI: https://doi.org/10.24321/0019.5138.202358