Community-acquired pneumonia (CAP) is a major cause of pediatric morbidity and mortality. Currently, there is no common approach to determination of CAP severity in children, which hampers early diagnosis and treatment of the disease. The study was aimed to determine clinical and laboratory predictors of severe CAP in children under 4 years of age. Analysis of clinical data, parameters of complete blood count (CBC), C-reactive protein (CRP) using nonparametric methods for hypothesis testing, univariate correlation analysis, cross-tabulation (Statistica 10.0), logistic regression, and ROC analysis (SPSS Statistics 20.0) was performed in 72 children aged 1 month to 3 years 11 months admitted to hospital due to CAP. Severe CAP was diagnosed in 16.7% of children. Causes of severe CAP included respiratory distress (moderate — 58.3%, severe — 16.7% of cases) and sepsis (25%). We identified significant clinical predictors of severe CAP: vomiting (OR 4.2), tachypnea (OR 28.3), chest wall retractions (OR 6), wheezing (OR 4), and the absence of rhinitis (OR 0.21). Isolated assessment of the CBC and CRP did not allow to predict CAP severity. We have developed a prediction model predicting severe CAP in children under 4 years of age based on the presence of rhinitis, tachypnea, as well as leukocyte count (sensitivity and specificity 91.7%). Thus, currently the main cause of severe CAP in children under 4 years of age is respiratory distress, in which wheezing predominates. Physical examination with an emphasis on detection of rhinitis and respiratory distress is essential for diagnosing severe CAP. The use of a pneumonia severity prediction model may contribute to improvement of management of CAP in patients under 4 years of age.