SUMMARY/ABSTRACTCarvalho, LSL. Risk factors for dropping out of inpatient treatment for eating disorers [Dissertação]. São Paulo, "Faculdade de Medicina, Universidade de São Paulo"; 2013.INTRODUCTION: Although is a frequent event, there is no consensus on what factors would be involved in the dropout of treatment for eating disorders. This study aimed to identify risk factors for the dropout of an inpatient treatment for anorexia nervosa and bulimia nervosa. Aspects previous to the hospitalization, clinical features present at the admission and factors related to the period of the inpatient treatment of patients who dropped out and completed treatment were compared. METHODS: We performed a retrospective study, case-control, with the dropout of the treatment being considered the interest condition. Medical records of female patients admitted to a specialized ward in treating anorexia nervosa and bulimia nervosa between the years 2005 and 2010 were reviewed. RESULTS: A total of 259 admissions and 178 patients, research subjects were 103 women, of which 83 completed treatment and 20 dropped out, resulting in a dropout rate of treatment of 19.42%. For the statistical treatment of the data was performed univariate analysis with subsequent multivariate logistic regression with backward stepwise selection analysis of the variables considered significant. In univariate analysis, professional removal due to illness severity, existence of affective relationship, lower BMI at admission, admissions motivated by low weight, diagnosis of anorexia nervosa and existence of purging behaviors (present in all patients who dropped out the treatment) were considered as factors related to dropout treatment. After conducting the logistic regression, we found that patients at admission who were removed from the job due to the illness severity, maintained a romantic relationship and were hospitalized due to low weight were more likely to dropout the treatment. CONCLUSIONS: In this sample, the dropout of treatment was influenced by sociodemographic factors (professional removal due to illness severity and existence of a romantic relationship), for diagnostic variables (diagnosis of anorexia nervosa and purging symptoms) and aspects of symptom severity (lower BMI and hospitalization motivated by low weight). These results lead us to the egosyntonic character of eating disorders, especially anorexia nervosa, and bring a paradox: patients that on admission showed more severe symptoms of the Eating Disorder (with lower weight and disabled to work) were more likely to dropout of the treatment. Although it has several limitations, mainly due to the retrospective design, the present research is the first study of dropout of inpatient treatment for eating disorders held in Brazil. It is suggested to run with prospective investigations to confirm or refute our findings.