Clinical characteristics analysis and construction of a predictive diagnostic model of community-acquired pneumonia in adults requiring hospitalization in Fujian provincial hospital

Abstract

Abstract BACKGROUND: This study aimed to compare the clinical characteristics of infections caused by different pathogens and then establish a viral/bacterial infection prediction model to guide early clinical identification of pathogens in inpatients with community-acquired pneumonia (CAP). METHODS: A total of 687 patients who were diagnosed with CAP in our hospital between March 2012 and December 2018 were studied. Basic data, clinical symptoms, laboratory examinations, and imaging examinations of patients were collected, and a virus/bacteria prediction equation was established. In the prediction model, the relevant variables were screened according to a univariate logistic regression analysis, and then, a multivariate logistic regression analysis was performed to establish the prediction equation. RESULTS: The proportions of patients with muscle soreness and headaches were significantly higher in the viral infection group than in the bacterial infection group. Procalcitonin (PCT) concentrations, the erythrocyte sedimentation rate (ESR), and the neutrophil alkaline phosphatase (NAP) score were significantly higher in the bacterial infection group than in the viral infection group. Creatine kinase concentrations were significantly higher in the viral infection group than in the bacterial infection group (P < 0.05). A higher proportion of patients had lung degeneration in the atypical pathogen infection group than in other groups (P = 0.005). Patchy shadows were more common in the viral infection group than in the other groups. A binary logistic regression equation was obtained that could predict the probability of viral infection (sensitivity: 57.5%, specificity: 67.7%, and area under the receiver operating characteristics curve: 0.651). CONCLUSIONS: Adult patients with CAP and viral infection are more likely to have headaches and muscle soreness than those with bacterial infection. An elevated PCT concentration, NAP score, and ESR indicate a high possibility of bacterial infection. We successfully established a viral and bacterial infection prediction model.