Prolonged Fatigue,Anxiety and Depression: Exploring Relationships in a Primary Care Sample

Recruitment of subjects

Consecutive adult and ambulatory patients of selected family practices were asked to participate in a longitudinal study of medical and psychological symptoms. The specific practices were approached on the basis of their geographical locations and practice characteristics so that a range of sociodemographic, socioeconomic and cultural variables could be examined [11].

Evaluation of psychological distress

The 30-item General Health Questionnaire (GHQ) has been used extensively in community and general medical settings to detect cases of psychological disorder [14]. A score of five or more symptoms is used to define patients as ‘psychological’ (PSYCH) cases (using the GHQ-scoring: 00–11). For the principal component analysis of this study, the items were scored as 0-1-2-3 (Likert scoring) [14, pp.19–22]. The factor structure of the GHQ-30 suggests that it is underpinned essentially by the two moderately correlated constructs of anxiety and depression [15].

Evaluation of fatigue

The Schedule of Fatigue and Anergia, Community version (SOFA/GP-[11]) has been developed to identify cases of prolonged and disabling fatigue in community and general practice settings. The scale consists of 10 items assessing prolonged and disabling fatigue, malaise, musculoskeletal symptoms, hypersomnia and neurocognitive symptoms. Patients are asked to rate the frequency of each of 10 items in the past few weeks on a four-point scale (‘1’, none or a little to ‘4’, most of the time). For the purposes of case identification, an individual item is scored as positive, if the subject has experienced it a ‘good part of the time (3)’ or ‘most of the time (4)’. A cut-off score of more than two positive items best identifies cases of prolonged fatigue (PF). Of the fatigue cases identified by the SOFA in primary care, two-thirds will also be identified as psychological cases by the GHQ-30 [11],[12].

For the principal component analyses of this study, the continuous scoring was kept and the items were scored as 1-2-3-4.

Statistical methods

Principal component analysis (SPSS FACTOR) was used to analyse the factor structure of SOFAand GHQ.

To achieve a simple structure, varimax rotation was applied to the factor matrix to enhance the inter-pretability of the factors. This permits the factors to be differentiated from each other without affecting the communalities or the percentage of total variance explained.

Ethical approval

Ethical approval was obtained from the University of New South Wales Ethics Committee.

Sample characteristics

The sample of 1593 patients was predominantly female (75%), middle-aged (mean = 37.8 years, median = 35.0 years, range = 18–89) and of Anglo-Celtic background (88%). Complete results for SOFA and GHQ were available for 1335 patients.

There was no significant difference between means scores of the GHQ for men and women (Z = −1.0, NS) but SOFA scores were higher in women (Z = −2.9, p < 0.01, Wilcoxon two-sample test).

Principal component analysis

Principal component analysis (PCA varimax rotation) was used to investigate the factor structure of patient response to the SOFA and GHQ scales. Inspections of the correlation matrix, eigenvalues, scree plots and the degree of loading of individual items across factors were used to indicate the acceptability of a two-factor, three-factor or four-factor solution. Increasing the number of factors beyond four did not result in the formation of additional meaningful factors.

Four-factor solution

In this solution the first factor explained 34% of variance and consisted of 15 GHQ items for anxiety and depression (Cronbach's α = 0.94). The second factor, which explained 7% of the variance, then consisted of the rest of the GHQ items (Cronbach's α = 0.88). That is, the first two factors consisted of overt psychological symptoms with the first factor describing conventional notions of anxiety and depression and the second factor largely indicating symptoms of global distress. The third factor consisted of seven SOFA items for physical fatigue (Cronbach's α = 0.78) and explained 5% of the variance. The fourth factor comprised three SOFA items which referred to the symptoms of mental fatigue (Cronbach's α = 0.71), explained 4% of the variance. No substantial cross-loading was observed between factors three and four or between factors one and two and these ‘fatigue’ factors. The only individual item which showed a tendency to cross-load was the GHQ item ‘able to concentrate’.

Three-factor solution

The three-factor solution resulted in the formation of two factors which consisted of the 30 GHQ items (Cronbach's α = 0.94 and 0.88). However, six of the depression and one of the anxiety items showed significant cross-loadings between these two factors. A third factor then incorporated all of the SOFA items (Cronbach's α = 0.81). There was no cross-loading between factor three and the first two factors.

Two-factor solution

The degree of cross-loading between the first two factors in the three-factor solution indicated that a two-factor solution may be more appropriate. This reduction in the number of factors resulted into a first factor consisting of all the GHQ items (Cronbach's α = 0.95), and a second factor comprising all of the SOFA items (Cronbach's α = 0.81). There was generally little cross-loading between the two factors though the differences in factor loadings for two of the common items (‘lost sleep over worry’ and ‘restless disturbed nights’) were quite small (0.44 vs 0.42 and 0.47 vs 0.43, respectively). The varimax-rotated factor structures for the three solutions are reported in Tables 1 2 3.

Discussion

This study, which applies principal component analyses to self-report data obtained from patients attending primary care, is consistent with earlier studies in a community based sample [13], which also demonstrated that prolonged fatigue could be differentiated from conventional notions of anxiety and depression. While those items related to sleep disturbance are shared by the two constructs, the remaining symptoms show a surprising degree of separation. If, as has been argued traditionally, somatic symptoms represent simply the somatic features of anxiety and depression [15], then the factor solutions should have shown that both the somatic and psychological symptoms were incorporated largely within the one dimension. The factor structure obtained here can be usefully contrasted with that obtained by latent trait analysis of the GHQ [16], where a two-factor solution of separate anxiety and depression factors was obtained from data collected in a primary care sample. That is, unless one uses an instrument designed to record key somatic symptoms, as well as the usual psychological symptoms, then it is unlikely that a separate dimension of somatic distress will be identified. Additionally, when somatic distress is measured, anxiety is less likely to be differentiated from depression [17–26].

While such multivariate statistical studies provide a useful conceptual model, the validation of prolonged fatigue syndromes requires a number of other approaches. The longitudinal stability of such syndromes is now supported by longitudinal studies in the community [27] and primary care [28] and for such syndromes when they develop secondary to glandular fever [29]. Recent modelling of aetiological factors in twins also confirms that prolonged fatigue has its own unique genetic and environmental determinants. That is, while one genetic factor appeared to contribute to the development of both prolonged fatigue and psychological distress, a second genetic factor was specific for prolonged fatigue alone [30],[31]. Such syndromes result in more disability than conventional depressive disorders and preferentially respond to cognitive-behavioural rather than pharmacological treatment approaches [1], [32–34].

While ‘classical’ anxiety and depressive disorders do present in general practice settings, more commonly the clinician will be assessing patients with an admixture of somatic and psychological symptoms, patients who are just below or close to the threshold for psychiatric ‘caseness’ and patients who do not fit readily into the dichotomous anxiety and depression categories [35–37]. In primary care in Australia [11], we have demonstrated that prolonged fatigue syndromes are common (comorbid form, 16%; pure form, 8%) but that general practitioners have great difficulty in recognising them as a form of neuro-psychiatric disorder [38]. A recent selected community survey in Australia noted that while 95% of patients with neurasthenia visited a general practitioner during the last 12 months, specific psychological interventions were not frequently provided unless neurasthenia occurred in association with other psychiatric disorders. Only 4.8% of neurasthenia patients without comorbid major depression visited a psychologist or psychiatrist in the last 12 months [39].

These results, therefore, have wider implications for the recognition and management of psychological disorders, particularly in primary care and other general medical environments. If the diagnostic and education systems promoted by psychiatrists do not readily match the symptoms profiles presented, practitioners will continue to underdiagnose psychological disorders [35],[37],[40],[41]. There is also a need to focus efforts on developing relevant treatment approaches for fatigue syndromes in primary care.