Short Form 36 and Hospital Anxiety and Depression Scale
A comparison based on patients with testicular cancer
S. D. Fosså, , a and A. A. Dahlb
a Department of Medical Oncology and Radiotherapy, The Norwegian Radium Hospital, University of Oslo, Montebello, 0310 Oslo, Norway
b Department of Psychiatry, Aker Hospital, University of Oslo, Oslo, Norway
Received 24 April 2001; accepted 27 September 2001 Available online 30 January 2002.
Abstract
Background: The aim of this study was to compare the scorings of anxiety and depression assessed by the Hospital Anxiety and Depression Scale (HADS-A [Anxiety] and HADS-D [Depression]) with the scorings on the eight subscales of Short Form 36 (SF-36) and the Physical (PCS) and Mental Component Summary (MCS) assessed by the same patients. Method: In a cross-sectional study 736 long-term survivors after treatment for testicular cancer (TC) completed HADS and SF-36. Pearson's correlation coefficients were calculated on item and scale level to assess the associations between the HADS and the SF-36 scales and, in particular, between HADS and PCS and MCS, respectively. Independent predictors for PCS and MCS were identified by linear regression analysis. Results: HADS-A and HADS-D were significantly associated with the SF-36 summary scales. HADS-A explained 5% of the variance of PCS and 49% of the variance of MCS. The comparable figures for HADS-D were 10% and 45%, respectively. In the multivariate analysis the HADS-D scoring independently predicted the level of PCS together with the patients' educational level, long-lasting working disability and age (variance: 30%). Both HADS-D and HADS-A remained independent parameters for MCS (variance: 58%) together with the patient's civil status. HADS-D item D4 ("slowed down") was similarly associated with both PCS and MCS. Conclusion: In univariate analyses HADS-D and HADS-A were statistically associated with PCS and MCS. The highest r values were observed for the associations between HADS and MCS, in particular between HADS-A and MCS. In the multivariate analyses HADS-D, but not HADS-A, contributed to PCS, whereas both HADS-A and HADS-D were associated with MCS. This pattern of different predictions of the summary scales of SF-36 supports a clinical practice that anxiety and depression should be assessed separately. Additional use of a self-rating instrument for depression and anxiety, such as HADS, is recommended when SF-36 is used for quality of life (QL) assessment.
Author Keywords: Anxiety; Depression; Hospital Anxiety and Depression Scale; Mental Functioning Scales—Short Form 36
Article Outline
• Introduction
• Patients and methods
• Patients
• Eligibility and rating of scales
• Reliability and validity
• Statistics
• Ethics
• Results
• Compliance
• Reliability and validity
• Associations
• Discussion
• Conclusion
• Acknowledgements
• References
Introduction
Testicular cancer (TC) is the most common malignancy in males under 40 years of age, and the mean age at diagnosis is 33 years. The cure rate of TC in the Western world is about 90% when radiotherapy, chemotherapy and/or surgery are used. During recent years increasing attention has been paid to the patients' long-term posttreatment morbidity such as hypogonadism, cardiovascular disorders, and psychological distress [1, 2, 3 and 4]. Several projects have recently been started in order to identify truly long-term somatic and psychological sequelae and quality of life (QL) in such patients [5, 6 and 7].
Most QL assessments in cancer patients have been done with validated self-rating instruments such as the Short Form 36 (SF-36) [8], the General Health Questionnaire (GHQ) [9], the Quality of Life Questionnaire of the European Organisation for Treatment of Cancer [10] (EORTC QLQ-C30), the General version of the Functional Assessment of Cancer Therapy scale (FACT-G) [11], or the Rotterdam Symptom Check List (RSCL) [12]. The SF-36 comprises eight scales: Physical Functioning (PF), Role Physical (RP), Bodily Pain (BP), General Health (GH), Vitality (VT), Social Functioning (SF), Role Emotional (RE) and Mental Health (MH) (Table 1). Based on these eight scales, two summary scales have been constructed: the Physical Component Summary (PCS) and the Mental Component Summary (MCS) scales. Though there exists some overlap between these summary scales, PF, RP, BP and GH are principally associated with the PCS scale, whereas the MCS scale predominantly comprises VT, SF, RE and MH. SF-36 has been extensively validated, and most of the psychometric evaluations have been done in the American general population and in American patients with the most common somatic and mental disorders [13 and 14]. The international IQOLA project has published summarised SF-36 norm data from other countries [14], showing that the mean scale scores and mean summary component scores of SF-36 for the Norwegian general population are very similar to those of the US population [13 and 14]. The Norwegian version of SF-36 has been used in patients cured for Hodgkin's disease, and the results have been compared to those from the Norwegian general population [15 and 16]. No reports are available comparing MCS and PCS scores from Norwegian oncological patients with those of the general population.
Table 1. Summary of the head words of SF-36 (A) and HADS (B)
The Hospital Anxiety and Depression Scale (HADS) was developed to screen for anxiety (HADS-A) and depressive (HADS-D) disorders, particularly in somatically ill patients [17]. HADS has been extensively used and has been proven useful in cancer patients before and during treatment as well as for evaluation of mental health in long-term survivors [18, 19 and 20]. HADS has also been used in the general population, primarily in Europe, including Norway [21]. Results obtained by HADS have been compared to those derived from responses to RSCL and GHQ: The concordance between HADS and RSCL was only 43% as to the identification of cases with affective disorders [22]. Ibbotson et al. [20] have concluded that HADS did best in disease-free cancer patients, whereas RSCL should be selected if progressing patients are to be evaluated.
Though HADS and SF-36 have been used concomitantly in several studies, and some comparisons have been made between HADS and several of the eight SF-36 subscales, extensive comparisons between HADS and the summary scales of SF-36 have not previously been reported. Ruta et al. [23] demonstrated a high correlation between HADS and the MH scale of SF-36 in patients with rheumatoid arthritis (correlation coefficient: −.80), but the MH scale was much less responsive to change than HADS. Morriss and Wearden [24] compared HADS with the mental functioning scales of SF-36 as screening instruments for mental disorders diagnosed by a structured psychiatric interview. HADS was found to be superior compared to the SF-36 scales. These latter two studies did not separate HADS-D from HADS-A. Tedman et al. [25] reported that the correlation coefficient between HADS-D and MH was significantly lower than between HADS-A and MH (.63 vs. .72).
As clinical depression is frequent in cancer patients, adequate evaluation of depression as a component of QL is obviously important, but is not always guaranteed. The suitability of the available instruments has been investigated in this respect. For example, the performance of HADS-A was found to be superior to that of HADS-D when compared to the results of an independent interview of patients with advanced breast cancer [26]. Recognising possible limitations of SF-36 in this respect, McHorney et al. [27] suggested that studies that focus on mental health should supplement the MH scale with instruments that evaluate mental disorders more specifically.
With the simultaneous use of HADS and SF-36 for the evaluation of psychological well-being and QL in cancer patients, the question arises whether both instruments measure the same aspects of mental health and to the same degree. If this is the case, future studies should avoid duplication in the use of instruments. The application of several questionnaires is, however, justifiable and even desirable, if different instruments assess different aspects of QL or mental health.
The principal aim of the present study was to compare scores achieved by HADS with those obtained by the primary SF-36 scales and the component summary scales in patients cured for TC. Our hypothesis was that clinically important differences exist between HADS-D and HADS-A as to their association with SF-36.
Patients and methods
Patients
The patients were taken from an ongoing multicenter project studying all surviving Norwegian TC patients, aged 18–75 years, who were successfully treated between 1980 and 1994. The patients were identified by the five oncological university departments in Norway and they were cross-checked against the National Cancer Registry. Eligible patients were invited to complete and return a mailed questionnaire with 219 items, which contained SF-36 and HADS. The questionnaire also asked for the patients' marital status and educational level. The patients, furthermore, recorded if they ever had used psychopharmacological drugs or narcotic substances before or after the diagnosis of TC, and whether they at any time had consulted a psychiatrist or clinical psychologist. "Prior psychological distress" was rated as present if a patient responded positively to any of these six questions. In addition, the patient answered whether his working capacity had been reduced during the last 12 months due to mental or somatic health problems. The present analysis was based on returned questionnaires from patients who had been contacted during the first 18 months of the study. All had been treated at the Department of Medical Oncology and Radiotherapy at The Norwegian Radium Hospital, University of Oslo.
Eligibility and rating of scales
To be eligible for the present analysis, the patient should have completed at least half of the items of both HADS and SF-36 MF scales. Among eligible patients, missing item scores were substituted by a patient-specific estimate (mean of the nonmissing items of the scale).
HADS-A consists of seven items rating anxiety and HADS-D of seven items rating depression. Each item is scored from 0 to 3, and the HADS-A and HADS-D scores are the sum of the relevant item scores. According to Zigmond and Snaith [17], each patient may subsequently be allocated to one of three caseness categories for anxiety and depression, based on the individual final scores: 0–7, non-case; 8–10, borderline case; and ≥11, definite case. Good psychometric properties of HADS have been demonstrated previously [28], also of the Norwegian translation [29].
The item scores of the SF-36 dimensions were, if necessary, recoded, summarised and transformed to scales ranging from 0 (worst situation) to 100 (best situation) according to the published recommendations [30]. Following the guidelines [31], PCS and MCS scales were calculated based on published means, standard deviations and factor score coefficients from the US general population. "PCS cases" were identified by a summary scale value of <50. The analogue cut-off value of "MCS cases" was 42. (The American data are very similar to the published values for means and standard deviations from the Norwegian general population and are comparable to unpublished values of the factor score coefficients obtained in Norway [Loge, personal communication].)
Reliability and validity
The internal consistency of the SF-36 and HADS scales was evaluated by the Cronbach α coefficient. "Known groups" validation was based on the following assumptions: Depression was expected to increase with increasing age, as shown in the general Norwegian population [21]. Living alone, low educational level, prior psychological distress and reduced working capacity during the preceding year would be negatively associated with mental health. As all patients were without evidence of disease and had completed their treatment 5 years or more prior to the assessment, the interval since orchiectomy would probably be of minor importance for psychological health, based on the overall clinical experience with these patients. On the other hand, the type of treatment could have some impact on somatic and probably psychological well-being.
Statistics
Standard methods were used for descriptive statistics (mean, median, standard deviation [S.D.]) together with a backward linear regression analysis to determine independent predictors of PCS and MCS. Differences between mean values of 2 or ≥3 groups were evaluated by, respectively, the Student t test or one-way ANOVA (with testing ad modum Bonferroni in case of statistical significance). Pearson's correlation coefficients (r) assessed the association between scales. The difference between two r values was regarded as significant if it exceeded two times the standard error (S.E.). The S.E. of the correlation coefficient was calculated as and was 0.037 in the present series. Backward linear regression analysis identified independent predictors of PCS and MCS. A P value <.05 was regarded as statistically significant. Significant post hoc comparisons included the Bonferroni modification (P<.01). The data were analysed using SPSS for Windows, Version 9.0.
Ethics
The study design and the invitation to the patients were accepted by the Regional Committee of Medical Research Ethics in Health Region II of Norway.
Results
Compliance
By January 31, 2000, 840 patients had been addressed, and by April 1 745 have returned completed questionnaires, the response rate thus being 89%. There was no difference as to age and treatment between the responding and nonresponding patients. Nine of the returned questionnaires were from ineligible patients, leaving us with questionnaires from 736 patients for analyses. A total of 63 patients had missing responses (9%) to individual questions, and these scores were substituted according the published guidelines. The demographics for patients with substituted responses were similar to those with a complete set of answered items, as were the mean values of the scales of HADS and SF-36, thus indicating missingness at random.
For the 736 patients the mean interval since orchiectomy was 12 years and was >8 years in 72% of the patients (Table 2). Radiotherapy for nonmetastatic seminoma and chemotherapy with or without radiotherapy or surgery for metastatic TC were the most frequent treatments except that unilateral orchiectomy had been done in all patients. A total of 176 patients (24%) recorded that they had experienced "prior psychological distress."
Table 2. Demographics and disease characteristics
Reliability and validity
All Cronbach α coefficients were above .70 (Table 3). Overall, the patients' physical health was comparable to that of the general population (PCS: 50.5) whereas their mental health was superior (MCS: 52.5) [1]. Two hundred forty-three patients (33%) were identified as "PCS cases" and 128 patients (17%) were "MCS cases." HADS identified 100 patients (14%) as borderline cases as to anxiety, and 42 men (6%) were allocated to the anxiety case group. The comparable figures for depression were 43 (6%) and 22 (3%), respectively (Table 4). HADS cases displayed the lowest PCS scores, and the reduction of mean PCS per HADS caseness category was more pronounced for HADS-D than for HADS-A. Seventy-seven percent of the HADS-D cases and 69% of the HADS-A cases were PCS cases.