Factor structure of the Beck Depression inventory-II in patients with chronic pain

Objectives: To examine the factor structure of the Beck Depression Inventory version II (BDI-II) in patients seeking treatment for chronic pain, using exploratory and confirmatory factor analysis and provide comparative data for use with similar patient populations. In addition, to consider the utility of using BDI-II subscale scores to further inform the management of patients with chronic pain. Methods: Phase I involved analysis of responses to the BDI-II by 1227 patients assessed for a pain management program. The sample data was split into 2 halves. A series of exploratory factor analyses on the first half suggested 2 factors. Confirmatory factor analysis was then used on the second half to confirm goodness-of-fit for this 2-factor solution and compare with a 1-factor solution and factor models presented in the BDI-II handbook derived on student and psychiatric outpatient populations. Comparison was also made of derived factor scores between this pain clinic sample and the normative psychiatric outpatient and student samples. In the second phase, longitudinal data on a further 269 patients who had completed the pain management program was used to examine the amount of variance in pain and disability outcomes accounted for by total BDI-II and 2-factor subscale scores at assessment. Results: Two correlated factors incorporating 18 items from the BDI-II gave good goodness-of-fit (0.916). Factor 1 loaded heavily onto negative cognitions about the self plus mood symptoms, factor 2 onto changes in behavior and activity plus low mood. Summed scores on factor 1 from pain clinic patients were very significantly lower than for psychiatric outpatients, indicating less negative cognitions about the self, and very significantly higher than for a student sample. Pain clinic patient scores on factor 2 were very significantly higher than those for both psychiatric outpatients and students, indicating more reporting of behavior change and affective symptoms. Subscale scores accounted for a small but significant amount of variance in both pain and disability at follow-up, with each scale predicting in opposite directions. Total BDI-II scores predicted similar amounts of variance in disability at follow-up, but were not significantly associated with pain at follow-up. Discussion: Results are consistent with studies using previous versions of the BDI in suggesting that 2-factor scores may be more clinically useful in the assessment of patients referred with chronic pain.