Latent structure of the Hospital Anxiety And Depression Scale: A 10-year systematic review

https://doi.org/10.1016/j.jpsychores.2011.06.008Get rights and content

Abstract

Objective

To systematically review the latent structure of the Hospital Anxiety and Depression Scale (HADS).

Methods

A systematic review of the literature was conducted across Medline, ISI Web of Knowledge, CINAHL, PsycInfo and EmBase databases spanning articles published between May 2000 and May 2010. Studies conducting latent variable analysis of the HADS were included.

Results

Twenty-five of the 50 reviewed studies revealed a two-factor structure, the most commonly found HADS structure. Additionally, five studies revealed unidimensional, 17 studies revealed three-factor, and two studies revealed four-factor structures. One study provided equal support for two- and three-factor structures. Different latent variable analysis methods revealed correspondingly different structures: exploratory factor analysis studies revealed primarily two-factor structures, confirmatory factor analysis studies revealed primarily three-factor structures, and item response theory studies revealed primarily unidimensional structures.

Conclusion

The heterogeneous results of the current review suggest that the latent structure of the HADS is unclear, and dependent on statistical methods invoked. While the HADS has been shown to be an effective measure of emotional distress, its inability to consistently differentiate between the constructs of anxiety and depression means that its use needs to be targeted to more general measurement of distress.

Introduction

Anxiety and depression are two of the most common psychological disorders [1], existing comorbidly with other psychological disorders, somatic disorders, and each other. Due to the high levels of comorbidity, the degree of symptom overlap, and the inextricable links between the symptoms of these disorders, they are often very difficult to differentiate [1], [2], [3]. However, the Hospital Anxiety and Depression Scale (HADS) was created specifically to accomplish this task and to assess possible and probable cases of anxiety and depression in non-psychiatric hospital outpatients [4]. The HADS is an important psychometric tool in the assessment of individuals with somatic illnesses, notably for coronary heart disease patients, predicting cardiovascular morbidity and mortality [5], [6]. The 14-item HADS is composed of two 7-item subscales, the HADS-A and HADS-D, intended to measure mutually exclusive levels of anxiety and depression, respectively. Although the HADS has been used prolifically, a considerable level of controversy has arisen regarding the validity of the original anxiety and depression bidimensional structure. While a latent variable analysis was not conducted during the creation of the HADS, numerous studies have subsequently examined the validity of the originally proposed bidimensional anxiety–depression structure.

Although other psychometric aspects of the HADS have been shown to be consistently satisfactory, i.e. sensitivity, specificity, reliability, [7], [8], [9], the proposed bidimensional factor structure has come under significant scrutiny. In spite of the robustness of all other psychometric properties of the HADS, if Zigmond and Snaith's original bidimensional structure is shown to be erroneous it cannot be conclusively deduced that the HADS is accurately measuring, and differentiating between, anxiety and depression [10]. While two previous systematic reviews [7], [8] have supported the original bidimensional structure, the last of these was published over 10 years ago, and more recent studies have been adopting more sophisticated analyses.

There is a great degree of variance in the statistical robustness of the methods used to determine the latent structure of a psychometric measure. Exploratory factor analysis (EFA) methods summarise patterns of correlations amongst observed variables and reduce these observed variables into a smaller set of underlying variables, using largely arbitrary and subjective criteria to select the appropriate number of factors [10], [11], namely Kaiser criterion (Eigenvalues > 1) and Scree plots (extraction of factors above an inflection point on a graph of plotted Eigenvalues) [12]. Confirmatory factor analysis (CFA), however, is an advanced model of classical test theory (CTT) factor analysis, allowing for the fitting of established factor models to the data, comparing and contrasting models for best-fit [10], [11], [13], [14]. Based on a non-linear function created from item and ability parameters, item response theory (IRT) models, for example Rasch analysis, provide many advantages over CTT methods [15].

The main advantages of IRT are centered on the scale (or item) and group (or examinee) independence [16]. This item and ability parameter invariance is due to the incorporation of item information into the ability-estimation process and conversely the incorporation of examinee ability into item-parameter estimation [15]. In CTT, examinee and scale characteristics cannot be separated; the ability of the examinee and test can only be interpreted within the context of one another. Therefore, results derived from CTT method can only be interpreted within the context of the original sample population [15]. In contrast, IRT results can be generalised to populations outside of the scope of the original study [15]. Therefore, the strength and robustness of evidence provided by each of these methods increase from EFA to CFA to IRT.

The current study conducts a systematic review of studies examining the latent structure of the HADS, aiming to summarise evidence of extant HADS structures and of the existence of the bidimensional anxiety–depression structure, updating existing reviews.

Section snippets

Search strategy

A systematic review of the literature was conducted across Medline, ISI Web of Knowledge, CINAHL, PsycInfo and EmBase databases spanning articles published between May 2000 (the cut-off date of the most recent systematic review [7]), but inclusive of articles not identified in the 2002 review, and May 2010. The words “Hospital Anxiety And Depression Scale,” “hospital anxiety and depression,” “HADS,” and “HAD scale” were combined with the Boolean operator “OR”. The HADS related search terms were

Results

The literature search identified 1666 unique studies, 199 of which were identified as an appropriate article conducting an analysis of the HADS; 50 articles met inclusion criteria (see Fig. 1).

The latent structure of the HADS was examined using CTT methods such as principal components analysis, EFA, CFA, as well as using IRT methods, notably Differential Item Functioning (DIF) and Rasch analysis. CTT methods were utilized in all but four studies; 36 studies employed EFA and 24 CFA, 14 employed

Discussion

The HADS is a prolifically administered self-report psychometric tool; however, despite its popularity, the latent structure remains unclear. The current study conducts a systematic review of HADS studies, published after the most recent systematic review [7], in order to examine the latent structure of the HADS and the existence of Zigmond and Snaith's originally proposed bidimensional anxiety–depression structure. The 50 extracted studies revealed a variety of methods and structures providing

References (31)

  • American Psychiatric Association

    Diagnostic and statistical manual of mental disorders: DSM-IV-TR

    (2000)
  • A Zigmond et al.

    The Hospital Anxiety and Depression Scale

    Acta Psychiatr Scand

    (Jun 1983)
  • C Martin

    What does the Hospital Anxiety and Depression Scale (HADS) really measure in liaison psychiatry settings?

    Curr Psychiatry Rev

    (Jan 2005)
  • B Tabachnick et al.

    Using multivariate statistics

    (2001)
  • R Cattell

    The scree test for the number of factors

    Multivariate Behav Res

    (1966)
  • Cited by (330)

    View all citing articles on Scopus
    View full text