Part of this study was presented (in poster session) at the 2nd Annual Conference of the American Psychosocial Oncology Society, Phoenix, AZ, USA, January 2005
The HADS and the DT for screening psychosocial distress of cancer patients in Taiwan †
Article first published online: 15 MAR 2011
Copyright © 2011 John Wiley & Sons, Ltd.
Special Issue: Screening for Distress, the 6th Vital Sign
Volume 20, Issue 6, pages 639–646, June 2011
How to Cite
Wang, G.-L., Hsu, S.-H., Feng, A.-C., Chiu, C.-Y., Shen, J.-F., Lin, Y.-J. and Cheng, C.-C. (2011), The HADS and the DT for screening psychosocial distress of cancer patients in Taiwan . Psycho-Oncology, 20: 639–646. doi: 10.1002/pon.1952
- Issue published online: 27 MAY 2011
- Article first published online: 15 MAR 2011
- Manuscript Accepted: 25 JAN 2011
- Manuscript Revised: 13 JAN 2011
- Manuscript Received: 3 APR 2010
Purpose: This study assesses the sensitivity and specificity of Mandarin versions of two psychosocial screening tools for adjustment, anxiety and depressive disorders: the Hospital Anxiety and Depression Scale (HADS), and the Distress Thermometer (DT).
Methods: The two scales were used to screen 103 consecutive cancer patients seen for psychiatric evaluation at KF-SYSCC between May and November 2004 prior to their psychiatric interviews. Each scale was tested against clinical psychiatric diagnoses based on the Diagnostic and Statistical Manual of Mental Disorders, 4th edition for their sensitivity and specificity.
Results: For the Mandarin version of the DT, receiver operating characteristic (ROC) analyses identified a DT score of 4 as the optimal cut-off, with sensitivity and specificity of 98 and 73%, respectively. For the Mandarin version of the HADS, ROC identified a score of 9 and 8 as the optimal cutoffs for the respective anxiety and depression subscales (HADS-a and HADS-d), with sensitivities and specificities of 84 and 73, 72 and 86%, respectively. For the full scale of the HADS (HADS-t), 15 was identified as the optimal cutoff, which yielded sensitivity and specificity of 84 and 68%, respectively. Using the frequency table, the concordance rate of the two scales was found to be 72–80% based on the above optimal cut-offs.
Conclusion: The Mandarin versions of the HADS and the DT are efficacious for screening anxiety and depression for our population. Compared with the HADS-t, the DT appears to have not only higher sensitivity, but also higher specificity. Copyright © 2011 John Wiley & Sons, Ltd.
The research data on psycho-oncology have shown that up to 45% of cancer patients demonstrated a significant level of distress 1–3. However, a very small proportion of these patients were actually referred to proper resources for psychosocial assistance 4, 5. In order to identify those who are in need of psychosocial intervention, screening with validated tools has been recommended 4. The Hospital Anxiety and Depression scale (HADS) and the Distress Thermometer (DT) have been widely used for these purposes.
The HADS was initially designed to assess the psychological states of physically ill patients 6. It has been accepted as an effective screening tool for anxiety and depression and has been widely used in psycho-oncology researches and practices 7–9. It is a 14-item self-report questionnaire composed of 7 items used to identify anxiety and 7 items used to identify depression. Each item has a 4-point (0–3) Likert-type scale, totaling from 0 to 21 for each subscale, and from 0 to 42 for the combined scales. Higher scores indicate greater anxiety and/or depression. In the original report, the cutoff score was set at 8 for doubtful cases and 11 for definite cases for both anxiety (HADS-a) and depression subscales (HADS-d) 6. However, most subsequent studies have identified the optimal cutoffs for both HADS-a and HADS-d as ≥8, and for HADS-t, ≥15 10–12.
For Asian patients, the Japanese version of HADS was tested in a cancer population by using psychiatric interviews to define cases. It revealed that the optimal cutoff of the HADS-t for screening adjustment disorders and/or major depression was 14/15 with sensitivity and specificity of 76 and 86%, respectively, whereas the optimal cutoff for the detection of major depression alone would be 16/17, with sensitivity and specificity of 77 and 74%, respectively 13. For Korean patients, a cutoff of 8 was adopted as it yielded good sensitivity and specificity (89.2 and 82.5%, respectively) in patients having anxiety and depression as well as in the general population 14.
Cantonese and Mandarin versions of the HADS
A number of different dialects are spoken by ethnic Chinese people, with Cantonese being the predominant one in Southeast China and Hong Kong, whereas Mandarin in the other parts of Mainland China, Taiwan and Singapore. The HADS was translated into Cantonese and specifically worded for interviewers administering to a sample of Hong Kong elderly patients seen in general medical practice, of which more than 40% of the patients were illiterate 15. Its validity was tested by the Clinical Interview Schedule. ROC analyses showed that the optimal cut-off points were 3 and 6 for anxiety and depression, respectively. Another Cantonese version of the HADS, translated for self-reporting, was tested among a sample of medical students in Hong Kong 16. It demonstrated favorable linguistic, structural and scale equivalence as compared with the English original. It was further tested with the diagnoses according to DSM-III-R criteria through a semi-structured psychiatric interview, the 17-item Hamilton Rating Scale of Depression and the Hamilton Rating Scale of Anxiety 17. It was found to have good internal consistency and external validity, with favorable sensitivity for screening psychiatric disorders. The optimal cut-offs of HADS-a, HADS-d and HADS-t were identified as 5/6, 8/9 and 15/16, respectively. It has been widely used among health-care professionals in Hong Kong 18.
Using DSM–IV diagnoses derived from the Mini-International Neuro-psychiatric Interview as the gold standard, the validity of the Mandarin version of the HADS was tested to screen depression in cancer patients in Taiwan 19. It revealed that the Mandarin version of HADS achieved a sensitivity rate of 81.0% and specificity of 63.3%, with the cut-off point set at 7 for depression. However, its validity in screening anxiety was not tested.
In order to de-stigmatize the psychological aspects of patient care and to screen patients rapidly for distress in cancer patients, Roth et al. designed the single-item DT 1. Analogous to the pain scale, they ask patients, ‘How is your distress in the past week including today, on a scale of 0 to 10?’ The single-item DT has been incorporated into the Clinical Practice Guidelines for Distress Management by the National Comprehensive Cancer Network (NCCN) 20. NCCN further developed the problem list (PL), which consists of 34 problems that are commonly experienced by cancer patients. The problems are grouped into five categories: practical, physical, family, emotional and spiritual. Under this screening process, patients are asked to answer the single-item DT and identify any of the problem items in the PL they may have experienced in the past week.
Initially, the NCCN Clinical Practice Guidelines for Distress Management has recommended a cutoff of 5 on the DT as indicative of significant distress that warrants a referral to appropriate supportive services. When both the DT and the HADS were assessed, using the DT score at or above 5 and the HADS total score at 15, the concordance of scores on the two scales was 74.4% 1.
Owing to the simplicity and the non-stigmatizing wordings, the DT with the PL has been studied and proven to be an effective screening tool for detecting distress in patients having different types of cancer 21–25. Most of these studies have consistently shown a DT cutoff score of 4 as optimal for identifying clinically significant distress. However, Hoffman et al. has revealed in their study that there does not appear to be a single cutoff score that clearly maximizes sensitivity and specificity. Therefore, they suggest that the use of multiple cutoffs is more effective. A score of 4–6 could indicate possible distress and a score of 7–10 would indicate definite distress 26. The DT with the PL has also been translated into several languages and used for cancer patients in Jordan 27, Turkey 28, Japan 13, and Korea 29.
At KF-SYSCC in Taiwan, using the HADS, with 8 as the cut-off for both anxiety and depression, we have found that about 55% of newly diagnosed cancer patients in the outpatient clinics have significant anxiety and/or depression 30. We have observed that these cancer patients, while transferred through different departments in the hospital for procedures and tests, have neither the time nor the patience to answer complicated questionnaires. Additionally, some patients or their families were unwilling to address psychological issues. For those who were unable to promptly complete the questionnaires, only 40% submitted completed questionnaires later. We, therefore, turned our focus to the simplified DT screening tool.
Goals and hypotheses of the study
In our current study, we proposed to:
- 1.Evaluate the sensitivity and specificity of the Mandarin version of the HADS and the DT in screening anxiety and depression, using psychiatric diagnosis through psychiatric interview as the gold standard to define the ‘caseness’.
- 2.Identify the proper cutoff points for both scales at which to make referrals for clinical interventions.
- 3.Assess the concordance of the scores of the Mandarin version of the HADS and the DT.
We hypothesized that both these scales would have high sensitivities and satisfactory specificities, with the DT having higher sensitivity and lower specificity. We also anticipate that the optimal cutoffs of the HADS-a, HADS-d and HADS-t would be similar to those reported in the previous international studies.
This research met the ethical requirement and was approved by the Institution Review Board of KF-SYSCC. The DT was translated into Mandarin by two of this study's authors (G. L. W. and S. H. S.). The Mandarin version of DT was reviewed by the remainder of our research team for editorial input. Back translation was performed by another professional staff who had not read the English DT scale. Subsequent discussions were held to reconcile any discrepancies before a final version was set. The final version was given to 20 outpatients at the psychiatric clinic of KF-SYSCC to test its feasibility and language clarity. Further modifications were made as needed.
One hundred and twenty-eight consecutive cancer patients who were seen for psychiatric evaluation at KF-SYSCC between May and November 2004 were solicited to be included in this study. After signing an informed consent, the study patients were given the DT and HADS before their psychiatric interview. For patients who were illiterate or physically unable to read, which comprised about 10% of the patients, the psychiatric nurse read and completed the DT and HADS for them.
At the time of psychiatric interview, the two psychiatrists were blind to the results of the scores of the DT and the HADS. They interviewed the first 20 study patients together to arrive at a consensus of diagnoses based on the diagnostic criteria of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM IV) 31. Patients receiving psychiatric diagnoses of adjustment disorder, anxiety disorder or depressive disorder were identified as ‘cases’.
The patients in this study were mostly referred by the attending physicians or nursing staff for clinical symptoms or concerns. However, there were some who came on their own, and some who were sent by friends/families. It was anticipated that some of them would be classified as non-cases on psychiatric interview.
Statistical analyses were performed using Statistical Analysis Software (SAS, version 9.1, SAS Institute, Cary, NC, USA). Each of the patient's characteristics was analyzed using descriptive statistics (mean/range and frequency/percent).
The DT and HADS were used to evaluate the anxiety and/or depression among the study patients. The sensitivity and specificity of both screening tools were determined in comparison to the psychiatric diagnosis results made by the interviewing psychiatrists in identifying ‘cases’. Since reliability of specificity is low with small sample without psychiatric disorders, 95% confidence intervals were also calculated by using standard methods for proportions 32.
ROC curve analyses were performed to identify cutoff scores for the DT, HADS-a, HADS-d and HADS-t. The value of the c statistic represents the area under the ROC curve (AUC) and shows the diagnostic accuracy of the method 33.
Chi-square test of the Fisher's exact test was used to examine the significance of the association between the HADS-a (≥9 vs <9), HADS-d (≥8 vs <8), HADS-t (≥15 vs <15) and DT (≥4 vs <4) and the PL.
Out of the 128 patients, 25 were of diagnoses other than adjustment, anxiety or depressive disorders, and, therefore, excluded from this analysis. Of the remaining 103, 22 had no psychiatric diagnosis and 10 had more than one diagnosis. The mean age was 48 (range 18–71) (SD = 11.9). Of the 103 patient, 82 (80%) were female. The oncology diagnoses were: breast (44%), head and neck, mostly nasopharyngeal (16%), colorectal (11%), lung (10%), gynecological (5%) and others (14%). The psychiatric diagnoses were led by Adjustment Disorders (61%), followed by Anxiety Disorders (15%) and Depressive Disorders (13%). The mean number of months since diagnosis was 5.4 (SD = 16.4) (Table 1).
|Characteristics||N (%)||N (%)|
|Age mean (range)||48 (18–71)|
|Head and neck (NPC 6, oral 9, thyroid 1)||16 (16)|
|GYN (cervix uteri 2 corpus uteri 1 ovarian 1 ovary 1)||5 (5)|
|Others (leukemia 1 lymphoma 3)||15 (14)|
|Adjustment disorders||63 (61)|
|With anxiety||19 (18)|
|With depressed mood||19 (18)|
|With both anxiety and depressed mood||25 (24)|
|Anxiety disorders||15 (15)|
|Anxiety disorder NOS||11 (11)|
|Panic disorder||4 (4)|
|Depressive disorders||13 (13)|
|Major depression||5 (5)|
|Dysthymic disorder||3 (3)|
|Depressive disorder NOS||5 (5)|
|No diagnosis||22 (21)|
The optimal cut-off score
Using the DSM-IV diagnoses of adjustment disorders, anxiety disorders and depressive disorders as the standard for identifying cases, the DT and HADS scores were tested for their sensitivity and specificity through ROC curve analysis. The AUC was used to estimate the discriminative accuracy of cutoff scores in relation to a criterion, with a range from 1 (perfect discriminative accuracy) to 0.5 (poor discriminative accuracy). For the Mandarin version of the DT, the ROC analysis identified a DT score of 4 as the optimal cut-off, which yielded an AUC of 0.89, with a sensitivity and specificity of 98% (95% CI, 91–100%) and 73% (95% CI, 63–80%), respectively (Figure 1).
For HADS-a scores, utilizing 9 as the cut-off yielded an AUC of 0.83 with sensitivity and specificity of 84% (95% CI, 74–91%) and 73% (95% CI, 63–80%), respectively; whereas utilizing 11 as the cutoff resulted in a sensitivity and specificity of 68% (95 CI, 57–78%) and 82% (95% CI, 73-88%), respectively. For that of HADS-d scores, utilizing 8 as the cutoff yielded an AUC of 0.82, with a sensitivity and specificity of 72% (95% CI, 61–81%) and 86% (95% CI, 78–92%), respectively; whereas utilizing 11 as the cutoff resulted in a sensitivity and sensitivity of 51% (95% CI, 40–62%) and 95% (95% CI, 88–99%), respectively. Using HADS-t, the cutoff was set at 15, which yielded an AUC of 0.86, with a sensitivity and specificity of 84% (95% CI, 74–91%) and 68% (95% CI, 61–75%), respectively; whereas 17 as the cutoff resulted in a sensitivity and specificity of 80% (95% CI, 70–88%) and 86% (95% CI, 78–92%), respectively (Figures 2–4).
Relationships between the DT cutoff score of 4 and the HADS scores, gender, age, and cancer type are listed in Table 2. A DT cutoff of 4 correlate significantly with the following cutoffs: HADS-t cutoff of 15, HADS-a cutoff of 9 and HADS-d cutoff of 8. The concordance rate is 72–80%. DT scores are not significantly related to gender, age and/or cancer type.
|N (%)||N (%)||p values|
|HADS-t≥15||70 (93)||5 (7)|
|HADS-t<15||15 (54)||13 (46)|
|HADS-a≥9||71 (96)||3 (4)|
|HADS-a<9||14 (48)||15 (52)|
|HADS-d≥8||59 (97)||2 (3)|
|HADS-d<8||26 (62)||16 (38)|
|Female||68 (83)||14 (17)|
|Male||17 (81)||4 (19)|
|Age⩽48||45 (82)||10 (18)|
|Age>48||40 (84)||8 (17)|
|Breast||38 (83)||8 (17)|
|H&N||12 (75)||4 (25)|
|Colorectal||9 (82)||2 (18)|
|GYN||4 (80)||1 (20)|
|Lung||10 (100)||0 (0)|
|Others||12 (80)||3 (20)|
Distress and the PL
All items of the PL were used in this study. The items in the emotional problems were endorsed most often, while those of spiritual/religious the least. Among the practical problems, the frequency of items checked in descending order were child care, work/school, housing, transportation and insurance/financial, with the transportation problem being significantly associated with DT≥4 (p = 0.02), and transportation and work/school significantly associated with HADS-t≥15 (p = 0.04). In the category of family problems, dealing with a partner and then with children was identified, with the former being significantly associated with HADS-t≥15 (p = 0.005). In the emotional problem group, worry, depression, nervousness, fears and sadness were all significantly associated with the DT and HADS, except that worry was not associated with HADS scores. The spiritual/religious concerns were checked by 15% of patients and were not more frequently endorsed by patients with DT≥4. For the physical problems, the most frequently checked items were, in descending order, sleep/insomnia, tired, eating, pain, breath, swelling, fatigue, action, appearance and hand/foot. Reporting hand/foot problems was significantly associated with DT≥4, whereas those of tired and action were significantly associated with HADS-t≥15.
This study demonstrated the efficacy of the Mandarin versions of the DT and HADS as screening tools for our cancer patients.
For the HADS, we found that the optimal cutoffs are 9 for HADS-a, 8 for HADS-d and 15 for HADS-t, with sensitivities and specificities of 84% (95% CI, 74–91%) and 73% (95% CI, 63–80%), 72% (95% CI, 61–81%) and 86% (95%CI, 78–92%) and 84% (95% CI, 74–91%) and 68% (95% CI, 61–75%), respectively. This may be acceptable for general screening of psychosocial distress with lower specificity for HADS-a and HADS-t. For more serious cases, taking the cutoffs of 11 for HADS-a and HADS-d, and 17 for HADS-t will achieve higher specificity, with sensitivity and specificity of 68% (95% CI, 57–78%) and 82% (95% CI, 73–88%), 51% (95% CI, 40–62%) and 95% (95% CI, 88–99%) and 80% (95% CI, 70–88%) and 86% (95% CI, 78–92%), respectively. However, the sensitivity will drop significantly for HADS-a and HADS-d.
Many previous studies have set optimal cutoffs for HADS-a and HADS-d at≥8 11. Our results are comparable to these reports except that HADS-a cutoff of 9 is higher but with satisfactory sensitivity and specificity. The Cantonese version of the HADS, tested with the diagnoses according to DSM-III-R criteria through a semi-structured psychiatric interview, identified the optimal cut-off of HADS-a, HADS-d and HADS-t as 5/6, 8/9 and 15/16, respectively, which yielded sensitivities and specificities of 81 and 61%, 74 and 78%, and 79 and 80%, respectively 17. Similar findings have also been reported in a Korean study, with a cutoff of 8 for both HADS-a and HADS-d with sensitivity and specificity of 89.2 and 82.5%, respectively 29. With respect to the HADS-t, we found the optimal cutoff of 15, and for more serious cases, 17. This concurs with many previous reports 10, 12, 13.
For the DT, we detected that a cutoff score of 4 yielded optimal sensitivity (98%; 95% CI, 91–100%) and specificity (73%; 95% CI, 63–80%) relative to psychiatric diagnoses of adjustment, anxiety and depressive disorders as defined by psychiatric interview based on DSM VI diagnoses. Patients who scored at or above the cutoff were significantly more likely to report predominantly emotional problems, a few practical and physical problems, but very few family problems or spiritual/religious concerns.
The optimal cutoff of the DT was initially set at 5 by NCCN 20. Subsequent reports have consistently found the optimal cutoff to be 4 25–29, 34–36. Our findings have also confirmed the optimal cutoff to be 4. Our results yielded an AUC of 0.89, indicating very good discriminative accuracy. This is somewhat higher than some previous reports (e.g. Jacobsen et al. 25: 0.80, Gil et al. 34: 0.77 and Shim et al.29: 0.75). As described above, the DT is designed to be non-stigmatizing and covering more general concerns of the patients, while the HADS is more specified for screening anxiety and depression. We have hypothesized that both these scales will have high sensitivities and satisfactory specificities, with higher sensitivity and lower specificity for the DT. Yet, in our study, the screening by the DT cutoff of 4 and the HADS-t cutoff of 15 yielded higher sensitivity and specificity for the former: (98 and 73% vs 84 and 68%). Chochinov et al. have also reported a similar paradoxical result, finding that an one-item questionnaire outperformed Beck Depression Inventory—Short Form and A Visual Analog Scale for Depression in screening depression in terminally ill patients 37.
If the DT scores are tested against the HADS cutoff of 15 as the standard from our data for identifying cases, it identifies 4 as the optimal cutoff with sensitivity of 93%, and specificity of 47%. This specificity is comparable to one previous study: Ozalp et al. 28: 49%; but lower than those reported by most of previous studies: Jacobsen et al.25: 68%, Shim et al.29: 59%, Akizuki et al.13: 61%.
Regarding the PL, we have used all of the items. The items of the emotional problems were endorsed most often and significantly associated with DT≥4 and HADS≥15. Those of spiritual/religious problems and physical problems were less frequently endorsed except for items of tiredness and sleep/insomnia. This may be related to cultural factors that, at a time of crisis, many of our patients might have already sought spiritual support from traditional faith or ritual practices. Some may receive counseling from Buddhist masters or Christian priests. It may also be related to the fact that our sample consisted of patients who were referred for psychiatric evaluation. They would be focusing mostly on the emotional aspects of their difficulties.
All patients who agreed to participate in this study completed DT, PL and HADS, with reminders from our nursing staff to complete any missed parts of the questionnaires before psychiatric interview. In the PL, there were items omitted because they were unrelated to that particular patient. For example, patients who had no children would not check the item ‘dealing with children’, etc. Further research will be needed to assess the completion rate of these instruments in the actual screening of general cancer patients in our population.
Our study has several limitations. First, our sample is from patients who were referred for psychiatric evaluation. Referrals were mostly made by the attending physicians or nursing staff. Some patients were self referred or sent by families/friends. Among the 103 eligible candidates, 22 had not received any psychiatric diagnoses. Nevertheless, most of them might have been highly aware of their psychological distress or symptoms. This may lead to overestimation of sensitivity and the predictive power of these instruments in our study.
The second limitation relates to the use of psychiatric interview to arrive at diagnosis based on the DSM-VI as the ‘gold standard’. Although time consuming (on average, 45 min per interview), it is still not as robust as the structured diagnostic interview. However, the psychiatrists are blind to the screening results. In addition, as described above, the two psychiatrists in our study had interviewed the first 20 patients together to clarify the diagnostic criteria. This may have contributed to an improved validity in the diagnoses. The identified cases may be more specific than the cases used in most of the previous studies, where cases were defined by HADS total scores or subscale scores. This may have, in turn, increased the specificity achieved in our study.
Our sample represented a higher proportion of female and breast cancer patients. It has been reported that patients who scored at or above the cutoff of 4 on the DT were significantly more likely to be female and to have a poorer performance status 25. Moreover, it has been reported that different types of cancer may result in varied prevalence of distress 3. However, as shown in Table 2, the DT cutoff of 4 was not significantly related to gender or cancer type in our study. Notwithstanding, variations among cancer types will need to be taken into account when the prevalence is considered in future studies. Between 1 May 2003 and 31 July 2004, we used the DT for 705 consecutive, newly diagnosed cancer patients at KF-SYSCC. Utilizing 4 as the cut-off, we found 65% of the cancer patients presented with significant distress.
The modest sample size of this study is another limitation. Future studies will need to address the issue of sample size and subgroup differences in screening efficacy levels of the DT and HADS.
The Mandarin versions of both these two screening tools are efficacious for screening adjustment disorders, anxiety disorders and depressive disorders for our cancer patients. Compared with the HADS-t, the DT appears to have higher sensitivity and specificity in screening anxiety and depression. The optimal cutoff scores of DT, HADS-a, HADS-d and HADS-t are similar to those reported in the previous international studies.
- 4Update: NCCN practice guidelines for the management of psychosocial distress. Oncology 1999;13(11A):459–507..
- 14A study on the standardization of the Hospital Anxiety and Depression Scale for Koreans—a comparison of normal, depressed and anxious groups. J Korean Neuropsychiatr Assoc 1999;38:289–296., , .
- 19Screening for depression in cancer inpatients. Master's Thesis, The Graduate Institute of Nursing Science, Chang Gung University, Taiwan, 2000..
- 20National Comprehensive Cancer Network, Inc. NCCN Practice Guidelines in Oncology—V. I. 2003, Distress Management, 2003.
- 27Distress in cancer in-patients in King Hussein Cancer Center (KHCC): a study using the Arabic-modified version of the Distress Thermometer. Psycho-Oncology 2004;13(S1):S42., , .
- 30Feasibility study of 2 screening tools for psychosocial distress for cancer patients in Taiwan. Psycho-Oncology 2005;14(S1):S49., , , , , .
- 31American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (4th edn), DSM IV. American Psychiatric Association: Washington, DC, 1994.
- 32Calculating confidence intervals for proportions and their differences. In Statistics with Confidence, GardnerMJ, AltmanDG (eds). BMJ Publishing Group: London, 1989; 28–33., .
- 35Psychometric properties and correlates of Distress Thermometer Scores. Psycho-Oncology 2004; 3(Suppl):S16., , et al.