Development of a model for the differential diagnosis of community-acquired bacterial pneumonia and viral lung injury in hospitalized adult patients

Abstract

Relevance. During and after the COVID-19 pandemic, viruses have become a more common cause of pulmonary infections in adults; therefore, the distinction between viral lung injury and community-acquired bacterial pneumonia is of increasing importance. Aim. Development of a model for differentiating community-acquired bacterial pneumonia and viral lung injury, including COVID-19. Materials and methods. This retrospective case–control study included 300 adult patients with viral lung injury and 100 adult patients with community-acquired bacterial pneumonia. Clinical, laboratory, and instrumental data were analyzed, significant factors were selected by which the samples differed, and a model was developed using logistic regression to distinguish between community-acquired bacterial pneumonia and viral lung damage, including COVID-19. Results. The developed model included the following parameters: total protein level, neutrophil/lymphocyte index, heart rate, unilateral infiltration on CT or chest x-ray, vasopressor prescription in the first 24 h of hospitalization, altered level of consciousness, chills, and fatigue. The model had the following characteristics: AUC = 0.94 (0.92–0.96), AUC_PR = 0.84 (0.76 to 0.92), prediction accuracy — 90%, sensitivity — 76%, specificity — 95%, positive predictive value — 83 %. Conclusion. The use of this model can facilitate the differential diagnosis of community-acquired bacterial pneumonia and viral lung injury, including COVID-19, in adults in general wards and intensive care units.