A machine learning approach to predict the parameters of COVID‐19 severity to improve the diagnosis protocol in Oman

Abstract

Abstract The purpose of this study is to utilize a Machine Learning-based methodology for predicting the key parameters contributing to severe COVID-19 cases among patients in Oman. To carry out the investigation, a comprehensive dataset of patient information, encompassing a range of blood parameters, was acquired from major government hospitals in Oman. Diverse machine learning algorithms were deployed to uncover underlying trends within the acquired dataset. The outcomes of this research delineated the determinants of severe cases into two categories: non-blood-related parameters and blood-related parameters. Among non-blood-related factors, advanced age, gender, and the presence of chronic kidney disease emerged as risk factors contributing to unfavorable prognoses, particularly in elderly patients. In the realm of blood parameters, male patients with blood types O-positive and A-positive exhibited heightened susceptibility to severe illness compared to their female counterparts. Additionally, deviations in Hemoglobin levels, Mean Cell Volume, and Eosinophil counts were identified as drivers of poor prognoses among elderly patients. The implications of these research findings extend to aiding healthcare decision-makers in quantifying the associated risks, health benefits, and cost-effectiveness pertaining to COVID-19. Furthermore, the acquired insights can empower decision-makers to refine the management of COVID-19, expediting treatment protocols and minimizing the risk of mortality. Interestingly, the study unveiled a correlation linking blood type to disease progression. A notable finding indicated that a staggering 96.5% of patients succumbed to the disease even when their blood sodium levels remained in the standard range of 136–145 mmol/L. These insights hold immense value for healthcare institution decision-makers, allowing a more in-depth evaluation of the risks, health benefits, and the cost-effectiveness related to COVID-19. Consequently, the findings offer a guiding light for implementing pivotal measures, optimizing treatment protocols, and substantially reducing mortality risks associated with the virus.

Keywords

• COVID-19
• SARS-CoV-2
• Pandemic
• Coronavirus
• Machine learning
• Treatment protocol