Statistical analysis
We calculated the RTI incidence per week and for the total study period. We performed a descriptive analysis of all survey variables, stratifying them by respiratory-related absence or absence due to other causes. Additionally, we ran a univariate logistic model to study the association between the type of absence and each described variable to obtain the corresponding odds ratios, 95% confidence intervals and p-values.
Furthermore, we conducted a Latent Class Analysis (LCA) to explore potential groups of students with similar symptomatology. We started exploring the optimal number of latent classes, trying two to five classes, and we selected the best model using the entropy criterion, which indicates the accuracy of the latent classes, combined with other goodness of fit criteria such as BIC, cAIC and likelihood ratio. All these measures suggested that the optimal number of classes was two. However, we also examined the rest of number of classes to see if their classification had more clinical significance. All statistical analyses were performed in R (version 4.2.2).