Commentary: Considerations for the use of registry data to study adolescent eating disorders

Background: Clinical anecdote suggests that rates of eating disorders (ED) vary between schools. Given their high prevalence and mortality, understanding risk factors is important. We hypothesised that rates of ED would vary between schools, and that school proportion of female students and proportion of parents with post-high school education would be associated with ED, after accounting for individual characteristics. Method: Multilevel analysis of register-based, record-linkage data on 55 059 females born in Stockholm County, Sweden, from 1983, finishing high school in 2002-10. Outcome was clinical diagnosis of an ED, or attendance at a specialist ED clinic, aged 16-20 years. Results: The 5-year cumulative incidence of ED diagnosis aged 16-20 years was 2.4%. Accounting for individual risk factors, with each 10% increase in the proportion of girls at a school, the odds ratio for ED was 1.07 (1.01 to 1.13), P = 0.018. With each 10% increase in the proportion of children with at least one parent with post-high school education, the odds ratio for ED was 1.14 (1.09 to 1.19), P < 0.0001. Predicted probability of an average girl developing an ED was 1.3% at a school with 25% girls where 25% of parents have post-high school education, and 3.3% at a school with 75% girls where 75% of parents have post-high school education. Conclusions: Rates of ED vary between schools; this is not explained by individual characteristics. Girls at schools with high proportions of female students, and students with highly educated parents, have higher odds of ED regardless of individual risk factors.