Results
Socio-demographic characteristics
A total of 392 parents of children with hematological malignancies who completed the questionnaires were eligible for this survey. The socio-demographic characteristics of parents and children are described in Table 1.
Latent class analysis
Mode fit statistics/indices are shown in Table 2. The single-class model had the largest AIC, BIC, and ABIC, compared to all other models. In addition, the LR tests of the 2-class model were all statistically significant (LMR P <0.0001, aLMRP <0.0001, and BLRT P <0.0001), indicating the population is not homogeneous, but heterogeneous. Comparing the 2-class and 3-class models, the latter had smaller information criterion indices (AIC, BIC, and ABIC) and all significant LR tests (LMR P =0.0021, aLMR P =0.0021, and BLRTP <0.0001). We, therefore, rejected the 2-class model in favor of 3 or more classes. Comparing the 3-class and the 4-class models, BIC (9705.262), LMR (P =0.7261), and aLMR (P =0.7265) favored the 3-class model, but AIC, aBIC, and BLRT favored the 4-class model. On the balance of model fit, model parsimony, and clinical interpretability of the results, the three-class model was selected. Figure 1 displays the results of the 3-class solution. The attribution probabilities of the model were 0.923, 0.948 and 0.958, respectively, and entropy=0.872 indicating good quality of class classification. The plot of item response probabilities endorsing ”no difficulty” for the 23 caregiving ability indicators in each of the classes shows explicit patterns of caregiving ability in the parents of children. The probabilities of endorsing ”no difficulty” in Class 1 (n=131, 33.4%) are all high across six domains. We, therefore, labeled this class as “high caregiving ability” class. The other two classes (Classes 2 and 3) were labeled “medium caregiving ability” (n=170, 43.3%) and “low caregiving ability” (n=91, 23.2%) classes, respectively, based on the item response probabilities in the classes.