Latent classes of children with AR
Figure 1 shows assignment of children into AR phenotypes over a sequence of latent class analysis models with two− to five− classes based on most likely membership class. We selected a four−class solution as the optimal model based on statistical fit (BIC: two−class=6552, three−class=6512, four−class=6496, five−class=6568). Figure 2 shows the distribution of variables across classes. Based on this and clinical interpretation, the latent classes were labeled as: (1) AR with grass mono-sensitization and conjunctivitis (n=361, 70.8%); (2) AR with house dust mite sensitization and asthma (n=75, 14.7%); (3) AR with pet and grass polysensitization and conjunctivitis (n=35, 6.9%) and (4) AR among children with tonsils and adenoids removed (n=39, 7.6%).