INTRODUCTION: There are several studies showing the presence of Differential Item Functioning (DIF) in some items of the Beck Depression Inventory (BDI), when comparing men and women. The presence of a large number of items with DIF in BDI is a severe threat to the validity of measurement of the intensity of depressive symptoms obtained by Item Response Theory (IRT) and to the conclusions based on the scores derived from the items with or without DIF. OBJECTIVE: The objectives of this study were to identify these items from the BDI, adjust the IRT model for embarrassing items (model 2), which accommodates items with the presence of DIF, and compare these results with the fit of the traditional two-parameter logistic IRT model (model 1). METHODS: The results obtained with the both models were compared. RESULTS: Items with DIF were: sadness, feeling of failure, dissatisfaction, guilty, punishment, crying, fatigability and loss of libido. The results of the adjustment of the two models are similar in discrimination, gravity (except for items with DIF), and in the calculation of scores for individuals. Nevertheless, model 2 is beneficial because it shows the differences in gravity of depressive symptoms for groups evaluated, thus providing more information to the researcher on the study population. CONCLUSION: This model, which has a broader scope in terms of target population, may be a good alternative to the identification and follow-up of individuals with potential depression.
INTRODUCTION: There are several studies showing the presence of Differential Item Functioning (DIF) in some items of the Beck Depression Inventory (BDI), when comparing men and women. The presence of a large number of items with DIF in BDI is a severe threat to the validity of measurement of the intensity of depressive symptoms obtained by Item Response Theory (IRT) and to the conclusions based on the scores derived from the items with or without DIF. OBJECTIVE: The objectives of this study were to identify these items from the BDI, adjust the IRT model for embarrassing items (model 2), which accommodates items with the presence of DIF, and compare these results with the fit of the traditional two-parameter logistic IRT model (model 1). METHODS: The results obtained with the both models were compared. RESULTS: Items with DIF were: sadness, feeling of failure, dissatisfaction, guilty, punishment, crying, fatigability and loss of libido. The results of the adjustment of the two models are similar in discrimination, gravity (except for items with DIF), and in the calculation of scores for individuals. Nevertheless, model 2 is beneficial because it shows the differences in gravity of depressive symptoms for groups evaluated, thus providing more information to the researcher on the study population. CONCLUSION: This model, which has a broader scope in terms of target population, may be a good alternative to the identification and follow-up of individuals with potential depression.
Authors: Anzalee Khan; Lora Liharska; Philip D Harvey; Alexandra Atkins; Daniel Ulshen; Richard S E Keefe Journal: Innov Clin Neurosci Date: 2017-12-01