2.4-Statistical analysis
Data were presented as median (interquartile ranges [IQR]) or number (%). The categorical variables were compared between groups using the χ2 or Fisher Exact test. The differences in the median of the continuous variables were tested with the 2–sided Mann–Whitney or Kruskal–Wallis test. With a priori adjustment for age (categoric variables), binary logistic regression was used to predict the relationship between viral coinfection and severe LRTI. Age was entered separately into the model as a continuous and categorical variable and categorized as 1-3 months, 4-6 months, 7-12 months, 13-24 months, and 25-60 months (reference level). In the table, potential risks were presented by using categorized age. As a binary dependent variable, we assessed the severity of the disease according to the site of the hospital stay: the ICU (severe LRTI) or the short-term unit (mild LRTI). The predictors associated with the dependent variable on univariate analyses (P< .05) were included in the model. As independent variables, the weight-for-height SDS, history of atopy, exposure to tobacco smoke, underlying conditions (categorized as; healthy [reference level], CLD, CHD, NMD, IS, and previous history of LRTI), neutrophil (categorized as; normal range [reference level], neutropenia, and neutrophilia) and lymphocyte (categorized as; normal range [reference level], lymphopenia, and lymphocytosis) status, and CRP value were included. Also, dummy variables to represent each level of qualitative independent variables included in the regression equation. First, virus infection type in the model was categorized as; i) single virus infection (reference level); ii) viral coinfection; iii) no-virus infection. Secondly, subgroup analyses by each of the viruses were performed to detect whether a specific virus coinfection led to severe LRTI than this specific-virus infection alone. The 13 models of comparing single and coinfection for each of the viruses were categorized as; i) specific-virus single infection (reference level); ii) specific virus and any virus coinfection; iii) single infections not including this specific virus; iv) coinfections not including this specific virus; v) no-virus infection. Finally, we conducted subgroup analyses by age for viral coinfections. The Hosmer-Lemeshow test was used to assess the goodness of fit for the model. Results were expressed as odds ratios (ORs) with 95% confidence intervals (CIs), and P< .05 was considered statistically significant. Statistical analyses were performed using the statistical program package SPSS (Version 22.0; IBM Corp., Armonk, NY).