Bayesian prediction of treatment outcome in anorexia nervosa: A preliminary study
Sammanfattning
Background: Knowledge of the prognostic factors predicting treatment outcome in anorexia nervosa
(AN) measured with health-related quality of life (HRQoL) is limited. Aims: We performed a novel
statistical analysis to identify factors predicting treatment outcome in AN. Methods: 39 patients
entering treatment of an ICD-10-defined AN completed the 15D HRQoL survey, the Eating Disorder
Inventory (EDI) and a questionnaire evaluating self reported health status and eating habits before
and 2 years after the start of treatment. The analysis was based on a Bayesian approach, which
allows analyses of small data sets, and was performed using a naïve Bayes classifier. Results: An
impaired follow-up HRQoL score was associated with three baseline risk factors: low self-reported
vitality, high scores in eating control and a poor reported health status. Low baseline body mass
index (BMI) and a high score in the eating dimension of the 15D predicted low follow-up BMI.
Conclusions: In our preliminary study, we identified a set of variables predicting poor HRQoL in AN.
An effort to treat these symptoms effectively in the beginning of AN treatment may influence the
outcome.