ArticlesEmpirical Test of the Factor Structure of the West Haven-Yale Multidimensional Pain InventoryRiley, Joseph L. III Ph.D.*†; Zawacki, Tricia M. M.A.†; Robinson, Michael E. Ph.D.†; Geisser, Michael E. Ph.D.‡ Author Information *Claude Pepper Center for Research of Oral Health in Aging, College of Dentistry, †Department of Clinical and Health Psychology, University of Florida, Gainesville, Florida, U.S.A.; and Department of Physical Medicine and Rehabilitation, University of Michigan Medical Center, Ann Arbor, Michigan, U.S.A. Manuscript submitted February 26 1998; revision received September 18, 1998; accepted for publication December 4, 1998. Address correspondence and reprint requests to Dr. Joseph L. Riley III, P.O. Box 100415 HSC, University of Florida, Gainesville, FL 32610-0415, U.S.A. E-mail: [email protected] The Clinical Journal of Pain: March 1999 - Volume 15 - Issue 1 - p 24-30 Buy Abstract Objective: Although the West Haven-Yale Multidimensional Pain Inventory (MPI) is frequently used in clinical evaluation and research with chronic pain patients, few studies have reported item-level factor analyses. After performing such an analysis, Bernstein et al. (Spine 1995;20:956-63) reported lack of independence between the solicitous and distracting response scales in section II as well as the activities away from home and social activities scales in section III. They suggested that the combination of these scales would improve the internal structure of the MPI. The purpose of this study was to perform a confirmatory factor analysis testing whether the MPI would be improved by the consolidation of these scales. In addition, a third, empirical model was generated for comparison with the West Haven-Yale and Bernstein models. Design: This study used exploratory and confirmatory factor analysis on two independent samples of chronic pain patients (n = 472 and n = 346) to test hypotheses regarding the factor structure of the MPI. Results and Conclusion: Principal axis factor analysis resulted in an empirical model that suggested that the primary psychometric problem of the MPI was lack of item-factor discrimination for several items. When the three models were tested using confirmatory factor analysis, improvement in model fit occurred when cross loading items were excluded. Nevertheless, the goodness of fit of original factor structure was adequate, suggesting it would be premature to suggest changes in this instrument. © 1999 Lippincott Williams & Wilkins, Inc.