Creation of a prognostic model of outcomes of viral lung lesions on a sample of COVID-19

Identifying characteristic changes in a comprehensive set of laboratory tests for the current infection can provide diagnostic and prognostic value. By analyzing this data, it is possible to develop prognostic models to determine treatment strategies.

Objective. To identify prognostic criteria for the development and outcomes of viral lung damage, using the example of a COVID-19, to stratify patient management in a hospital setting.

Material and methods. The study included 294 adult patients (aged 58.6±13.7 years; 39% male) hospitalized with lung damage caused by a COVID-19 in 2020. The assessment of disease severity and patient treatment was conducted in accordance with interim guidelines for the prevention, diagnosis, and treatment of the novel coronavirus infection. Based on outcomes, two groups were formed: Group 1 included 77 patients with fatal outcomes; Group 2 included 217 patients who recovered. To detect SARS-CoV-2 RNA, swab samples were analyzed using the polymerase chain reaction method.

Results. Based on the obtained data, a formula for predicting unfavorable outcomes was derived using clinical, anamnestic, and laboratory predictors through logit regression followed by logit transformation. The logit model identified eight of the most significant factors associated with a high risk of mortality in patients. Using these factors, a computer program was developed using HTML and JavaScript technologies.

Conclusion. To stratify patient management in cases of widespread viral diseases (using the COVID-19 pandemic as an example) upon hospital admission, the logit model can be employed to predict disease progression. In cases of high risk for unfavorable outcomes, it can guide treatment strategies, including the potential need for transfer to the intensive care unit.