Differentiate Clinical Characteristics Between Viral Pneumonia and Mycoplasma pneumoniae and Nomograms for Predicting Mycoplasma pneumoniae: A Retrospective Study in Primary Hospitals

Abstract

Objective: To identify the difference in clinical characteristics between viral pneumonia and Mycoplasma pneumoniae, providing cues on their differential diagnosis for primary hospitals with the insufficient pathogen detection capacity. Methods: We retrospectively reviewed the medical records of hospitalized children with acute respiratory tract infections, and pathogenic microbes test results were analyzed. Clinical characteristics, routine blood parameters and hospitalization duration and fee were compared between M. pneumoniae and viral pneumonia. We used in the multivariable logistic regression to predict the probability of children with M. pneumoniae and graphically represented by a dynamic nomogram. The discrimination and clinical utility of the model were confirmed by receiver operating characteristic and decision curve analysis curves. Result: A total of 375 children with community-acquired pneumonia were included. Mycoplasma infection accounted for the largest proportion (22.13%). The incidence of both hypothermia and vomiting was lower in M. pneumoniae compared to viral pneumonia (hypothermia: 10.50% vs. 0.00%; vomiting: 7.90% vs. 0.00%). The prevalence of hyperthermia was higher in M. pneumoniae (hyperthermia: 89.5% vs. 100%). Procalcitonin, peripheral blood white blood cell count and lymphocyte levels were higher in the viral pneumonia group, and eosinophil levels were conversely lower. As for the duration of illness, the mean length of stay was 5.20 ± 2.12 (viral pneumonia) and 6.27 ± 2.48 days (M. pneumoniae). Children with M. pneumoniae had higher overall hospital costs and required more medical treatment. The above were all statistically significant with a P < 0.05. The scoring system was established based on the above results. Receiver operating characteristic curves showed good model-discrimination ability with 0.844 of the area under the curve in the training set and 0.778 in the test set. Decision curve analysis curves demonstrated the discriminative superiority of this model. The web-based dynamic nomogram calculator is accessible at https://zhxylxy0160128.shinyapps.io/Nomogram/. Conclusion: Nomograms have satisfactory discrimination, and clinical utility may benefit in predicting the probability of developing M. pneumoniae in children. Children with M. pneumoniae have a higher burden than those with viral pneumonia and may require more intensive in-hospital monitoring.