Predicting the outcome of eating disorders using structural equation modeling.

OBJECTIVE There is a need for models that predict accurately the course of mental disorders. METHOD Eating-disordered female inpatients were assessed longitudinally at the beginning of treatment (t1), at the end of treatment (t2), at 2 or 3-year follow-up (t3), and at 6-year follow-up (t4). The sample consisted of 196 women with bulimia nervosa (BN) purging type, 103 women with anorexia nervosa (AN), and 68 women with binge eating disorder (BED; N=367). Confirmatory factor analysis and path analysis were used to predict the women's status at 6-year follow-up. RESULTS The results for BN and BED show that the specific eating disorder pathology was influenced mainly by specific eating disorder pathology at earlier time points and not by non-eating-specific (general) psychopathology. Similarly, general psychopathology was influenced mainly by general psychopathology at earlier time points. For AN patients, both categories of psychopathology (eating specific and general) were relevant for the 6-year outcome. The potential impact of 14 factors on the level of pathology was estimated (a) at baseline (at the beginning of treatment), (b) during the course of illness (baseline controlled), and (c) on the 6-year outcome of eating disorders (baseline and course controlled). Although there were many correlations between potential factors and baseline pathology, there was only a limited number of significant correlations with the 6-year outcome. This effect was mediated largely by the level of general psychopathology. DISCUSSION The models for outcome prediction based on structural equation modeling techniques were very similar for BN and BED. For both BN and BED, there were almost entirely separate predictions for the specific eating disorder on the one hand and non-eating-related (general) psychopathology on the other hand. This was true to a lesser degree for AN. CONCLUSIONS The use of refined path analytic methods in follow-up studies on larger general populations will be helpful to increase our understanding of the course of illness of psychiatric disorders.