Classification of childhood onset eating disorders: A latent class analysis

This study tested the hypothesis that latent class analysis (LCA) would successfully classify eating disorder (ED) symptoms in children into categories that mapped onto DSM-5 diagnoses and that these categories would be consistent across countries. Childhood onset ED cases were ascertained through prospective active surveillance by the Australian Paediatric Surveillance Unit, the Canadian Paediatric Surveillance Program, and the British Paediatric Surveillance Unit for 36, 24, and 14 months, respectively. Pediatricians and child psychiatrists reported symptoms of any child aged ≤ 12 years with a newly diagnosed restrictive ED. Descriptive analyses and LCA were performed separately for all three countries and compared. Four hundred and thirty-six children were included in the analysis (Australia n = 70; Canada n = 160; United Kingdom n = 206). In each country, LCA revealed two distinct clusters, both of which presented with food avoidance. Cluster 1 (75%, 71%, 66% of the Australian, Canadian, and United Kingdom populations, respectively) presented with symptoms of greater weight preoccupation, fear of being fat, body image distortion, and over exercising, while Cluster 2 did not (all p < .05). Cluster 1 was older, had greater mean weight loss and was more likely to have been admitted to an inpatient unit and have unstable vital signs (all p < .01). Cluster 2 was more likely to present with a comorbid psychiatric disorder (p < .01). Clusters 1 and 2 closely resembled the DSM-5 criteria for anorexia nervosa and avoidant/restrictive food intake disorder, respectively. Symptomatology and distribution were remarkably similar among countries, which lends support to two separate and distinct restrictive ED diagnoses.