Using inductive machine learning to identify risk factors for healthcare workers to get infected with highly contagious viruses (based on COVID-19 model)

Epidemic and pandemic spread of highly contagious viruses (SARS-CoV, influenza A virus, Ebola virus, MERS-CoV, and SARS-CoV-2) has been a trend observed in the first two decades of the 21st century. The predominant impact made by the biological occupational factor on healthcare workers determines their high occupational risk of infection, a severe disease course and a fatal outcome. Epidemiological data mining based on machine learning algorithms is successfully used in epidemiological practice to identify factors (predictors) contributing to infection in various risk populations. In this study, the database generated from a survey of 1312 healthcare workers was analyzed intelligently. A total of 6912 machine learning models were implemented. SARS-CoV-2 infection was found to be facilitated by providing medical care to a COVID-19 patient, using a full set of PPE after direct contact with a COVID-19 patient, direct contact with items in the external (hospital) environment, vaccination against COVID-19 after direct contact with a COVID-19 patient, acting as nursing staff (cleaners) and being present during aerosol-generating procedures. The study identified four groups of predictors determining SARS-CoV-2 infection in healthcare workers: contact with a COVID-19 patient and environmental items, PPE quality and complexity, occupational affiliation of healthcare workers and their BMI values. One predictor was found in 56.2 % of healthcare workers; two, in 19.2 %; three, in 16.4 %; four, in 5.5 %; and five predictors, in 2.7 %. Thus, epidemiological data mining is a modern stage in epidemiological analysis. The use of machine learning methods allows for multifactorial assessment of SARS-CoV-2 infection risks in healthcare workers and enables identifying and reliably estimating the most significant predictors. Intelligent data analysis has flexible architecture, which allows adjusting the model under study and supplementing new data to the existing database, detecting changes in an epidemiological situation and accomplishing relevant preventive and anti-epidemic activities.