A Hartmann1, A Zeeck, A J van der Kooij. 1. Department of Psychosomatic Medicine and Psychotherapy, University of Freiburg, Freiburg, Germany.
Abstract
AIMS: In order to identify the most important components of the severity of bulimia nervosa (as well as identifying clinical cases), we explored the relation between dimensional and categorical assessment. This was achieved by studying the performance of variables from standard instruments (measuring specific and general psychopathology) in predicting an expert rating of overall syndrome severity. METHOD: In total, 213 cases were selected (across the whole range of severity). We applied regression with optimal scaling to model nonlinear relations in the data, and the lasso method with bootstrapping for predictor selection. The best model contained 2 scales of the Eating Disorders Inventory ('bulimia' and 'drive for thinness') and the frequency of the binges. The sensitivity and specificity of case classification using the obtained model was determined. RESULTS: The model can predict the probability of being a clinical case at a rate of 88%. The presented statistical methods are innovative and promising approaches that can help researchers and clinicians to better define sets of variables for treatment evaluation and outcome studies. CONCLUSION: The results indicate that severity and outcome in bulimia nervosa should be determined by measuring both cognitive and behavioral aspects of the symptoms. (c) 2008 S. Karger AG, Basel.
AIMS: In order to identify the most important components of the severity of bulimia nervosa (as well as identifying clinical cases), we explored the relation between dimensional and categorical assessment. This was achieved by studying the performance of variables from standard instruments (measuring specific and general psychopathology) in predicting an expert rating of overall syndrome severity. METHOD: In total, 213 cases were selected (across the whole range of severity). We applied regression with optimal scaling to model nonlinear relations in the data, and the lasso method with bootstrapping for predictor selection. The best model contained 2 scales of the Eating Disorders Inventory ('bulimia' and 'drive for thinness') and the frequency of the binges. The sensitivity and specificity of case classification using the obtained model was determined. RESULTS: The model can predict the probability of being a clinical case at a rate of 88%. The presented statistical methods are innovative and promising approaches that can help researchers and clinicians to better define sets of variables for treatment evaluation and outcome studies. CONCLUSION: The results indicate that severity and outcome in bulimia nervosa should be determined by measuring both cognitive and behavioral aspects of the symptoms. (c) 2008 S. Karger AG, Basel.
Authors: Kim A Celone; Heather Thompson-Brenner; Robert S Ross; Elizabeth M Pratt; Chantal E Stern Journal: Neuroimage Date: 2011-03-23 Impact factor: 6.556
Authors: Kathryn E Smith; Jo M Ellison; Ross D Crosby; Scott G Engel; James E Mitchell; Scott J Crow; Carol B Peterson; Daniel Le Grange; Stephen A Wonderlich Journal: Int J Eat Disord Date: 2017-06-17 Impact factor: 4.861