Article Abstract

Online ISSN: 1099-176X    Print ISSN: 1091-4358
The Journal of Mental Health Policy and Economics
Volume 19, Issue 2, 2016. Pages: 79-89
Published Online: 1 June 2016

Copyright © 2016 ICMPE.

Mapping the Beck Depression Inventory to the EQ-5D-3L in Patients with Depressive Disorders

Thomas Grochtdreis,¹* Christian Brettschneider,² Andr‚ Hajek,³ Katharina Schierz,⁴ Jürgen Hoyer,⁵ Hans-Helmut König⁶

¹MSc, Research Associate, Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center Hamburg-Eppendorf Hamburg, Germany
²PhD, Dipl Ges-Ök, Research Associate, Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
³PhD, MA, Research Associate, Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
⁴Dipl Psych, Research Associate, Institute of Clinical Psychology and Psychotherapy, TU Dresden, Dresden, Germany
⁵PhD, Dipl Psych, Professor for behavioral psychotherapy, Institute of Clinical Psychology and Psychotherapy, TU Dresden, Dresden, Germany
⁶MD, MPH, Professor for health services research and health economics, Director of Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Background: For cost-utility analyses, data on health state utilities, as provided by the EQ-5D-3L, is needed but not always available. This study specified mapping algorithms from the Beck Depression Inventory (BDI) index to the EQ-5D-3L index adjusted for specific socio-demographic variables for patients with depressive disorders.

Aims of the Study: The objective of this study was to specify mapping algorithms from the BDI index to the preference-based EQ-5D index for patients with depressive disorders, adjusting for specific socio-demographic variables.

Methods: A sample of 1,074 consecutive patients with depressive disorders from a psychotherapeutic outpatient clinic was included in the study. Standardized clinical interviews were applied to establish reliable diagnoses. For the prediction of the EQ-5D-3L index from the BDI index and selected patient socio-demographic characteristics, ordinary least squares regression with robust standard errors was used. Model prediction properties were tested using the root mean squared error and repeated random sub-sampling cross-validation.

Results: The BDI index predicted the EQ-5D-3L index with a significant proportion of variance explained. The highest model goodness of fit was estimated for models with the BDI index and age as independent variables. The root mean squared error of the predicted EQ-5D-3L index in the validation samples was 0.23 for all models.

Discussion: The mean observed EQ-5D-3L index values and the mean predicted EQ-5D-3L index values seemed not to differ between models. However, a reduction of variability using cross-validation led to those (rather) accurate mean predicted values. One limitation of this study was the restricted generalizability. Moreover, some uncertainty was introduced in model predictive performance by usage of a dependent estimation sample for validation.

Implications for Further Research: The specified mapping algorithms from the BDI index to the EQ-5D-3L index for patients with depressive disorders are acceptable as approximation in cost-utility analyses. A further validation in independent samples is necessary to obtain more confidence in their performance.