The factor structure of the Hospital Anxiety and Depression Scale in individuals with traumatic brain injury

Abstract

There is a lack of validated scales for screening for anxiety and depression in individuals with traumatic brain injury (TBI). The purpose of this study was to examine the factor structure of the Hospital Anxiety and Depression Scale (HADS) in individuals with TBI. A total of 294 individuals with TBI (72.1% male; mean age 37.1 years, S.D. 17.5, median post-traumatic amnesia (PTA) duration 17 days) completed the HADS 1 year post-injury. A series of confirmatory factor analyses was conducted to examine the fit of a one-, two- and three-factor solution, with and without controlling for item wording effects (Multi-Trait Multi-Method approach). The one-, two- or three-factor model fit the data only when controlling for negative item wording. The results are in support of the validity of the original anxiety and depression subscales of the HADS and demonstrate the importance of evaluating item wording effects when examining the factor structure of a questionnaire. The results would also justify the use of the HADS as a single scale of emotional distress. However, even though the three-factor solution fit the data, alternative scales should be used if the purpose of the assessment is to measure stress symptoms separately from anxiety and depression.

Introduction

Elevated rates of depression and anxiety in individuals with traumatic brain injury (TBI) have been documented in several studies (Jorge and Robinson, 2003, Ashman et al., 2004, Dikmen et al., 2004, Jorge et al., 2004, Draper et al., 2007, Ponsford et al., 2008, Whelan-Goodinson et al., 2008, Whelan-Goodinson et al., 2009a, Whelan-Goodinson et al., 2009b). Both primary and secondary injury to the brain (Jorge and Robinson, 2003, Jorge et al., 2004, Jorge et al., 2007) and the lifestyle changes resulting from TBI for those injured (Pagulayan et al., 2008) as well as their relatives (Anderson et al., 2002) have been shown to be predictive of anxiety and depression in individuals with TBI. However, most of the measures of anxiety and depression that have been used in the TBI population have not been evaluated for use in this group.

One of the measures frequently used of such measures in TBI individuals is the Hospital Anxiety and Depression Scale (HADS; Snaith and Zigmond, 1983). This brief and widely used dimensional measure assesses the extent of anxiety and depression symptoms experienced throughout the preceding week. Only a few studies have examined the factor structure of the HADS in individuals with an acquired brain injury. Dawkins et al. (2006) used exploratory factor analysis (EFA) in a sample of 140 individuals with acquired brain injury of mixed aetiology and found support for the original two-factor solution. Johnston et al. (2000) confirmed the two-factor solution in stroke patients using both EFA and confirmatory factor analysis (CFA) but, due to the high correlation of the anxiety and depression scales, also discussed the possibility of lumping the anxiety and depression scales together. Both articles attribute the low factor loadings of some items to brain-injury-related problems other than anxiety and depression that may have impacted the participants' answers. For example, Dawkins et al. (2006) argued that answers to HADS item 8 (‘I feel slowed down’) may in some cases be influenced by head-injury-related, slowed speed of information processing rather than depression. Similarly, the inability to enjoy things that one used to enjoy (item 2) or the inability to enjoy a good radio or TV programme (item 14) may be a consequence of TBI-related cognitive problems (e.g., being unable to focus on the TV programme over a longer time period).

However, both Dawkins et al. (2006) and Johnston et al. (2000) included individuals with acquired brain injuries of mixed aetiologies. TBI individuals have different demographic and injury characteristics and partly different symptoms than for example stroke survivors. Therefore, it is important to examine the factor structure of the HADS in a sample comprising only individuals with TBI. Such studies are lacking.

The factor structure of the HADS has been extensively investigated in other populations. The majority of studies taking an EFA approach have found support for the two-factor (anxiety and depression) structure (Bjelland et al., 2002, Martin, 2005). A one-factor structure has also been proposed (Razavi et al., 1990, Johnston et al., 2000). However, a number of recent studies employing CFA have found best support for three-factor solutions (Dunbar et al., 2000, Caci et al., 2003, Martin and Newell, 2004, Martin et al., 2004a, Martin et al., 2004b, Desmond and MacLachlan, 2005, Martin, 2005). Dunbar et al.'s (2000) three-factor model stands out because it consistently showed a comparatively good model fit in several studies, and because it has a theoretical foundation, namely the Tripartite Model of Depression and Anxiety developed by Clark and Watson (1991). The Tripartite Model assumes that anxiety and depression are separate constructs that overlap because both share a general Negative Affect (NA) component that is an inherent and important aspect of each mood state. Hence, in Dunbar et al.'s (2000) three-factor solution, the anxiety subscale of the HADS is split into an autonomous anxiety factor and a negative affectivity factor. The latter factor is causally related to both anxiety and depression. However, in the discussion of the factor structure of the HADS, the potential impact of item wording has been largely overlooked. While the inclusion of both positively and negatively formulated items in a questionnaire controls for response biases, it also creates a source of variance that can reduce the reliability of a measure (Schrisheim, 1981). Several studies have demonstrated that the fit of a CFA model can be improved by controlling for item wording effects (Tomas and Oliver, 1999, Motl and DiStefano, 2002, Greenberger et al., 2003, Abbott et al., 2006, Quilty et al., 2006, Aluja et al., 2007, Wu, 2008, DiStefano and Motl, 2009b). It could therefore be argued that before the two-factor solution of the HADS is abandoned on the basis of CFA studies, it should be examined whether a better CFA model fit can be achieved by controlling the two-factor model for item wording effects. The current study aimed to examine the factor structure of the HADS specifically in individuals with TBI, taking into consideration both the original two-factor solution, Dunbar et al.'s (2000) three-factor solution and the one-factor solution (Razavi et al., 1990), as well as the potential impact of item wording on the factor analysis. Furthermore, since some HADS items might be more valid than others, the study aimed to provide clinicians with an improved formula for HADS scale score computation that takes the respective factor loading of each individual item into account in order to increase scale reliability.