Cancer & Depression

ORIGINAL ARTICLE

Hospital Anxiety and Depression Scale (HADS) accuracy in cancer patients

Maria Antonietta Annunziata1 & Barbara Muzzatti1 & Ettore Bidoli2 & Cristiana Flaiban1 & Francesca Bomben1 & Marika Piccinin1 & Katiuscia Maria Gipponi1 & Giulia Mariutti1 & Sara Busato1 & Sara Mella1

Received: 5 April 2019 /Accepted: 11 December 2019 # Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract Purpose The Hospital Anxiety and Depression Scale (HADS) is a self-report questionnaire designed to screen anxious and depressive states in patients in non-psychiatric settings. In spite of its large use, no agreement exists in literature on HADS accuracy in case finding. The present research addresses the issue of HADS accuracy in cancer patients, comparing its two subscales (HADS-A and HADS-D) against tools not in use in psychiatry, which are able to detect prolonged negative emotional states. Methods 2121 consecutive adult cancer inpatients were administered the HADS together with the State Anxiety subscale of State-Trait Anxiety Inventory and the Center for Epidemiologic Studies Scale on Depression. Receiver operating characteristic (ROC) curves were computed to identify a cut-off for anxious and depressive states in cancer patients. All indicators were computed together with their corresponding 95% confidence interval (95% CI). Results Data of 1628 and 1035 participants were used to assess the accuracy in case finding of HADS-A and HADS-D, respectively. According to the ROC analysis, the optimal cut-off was > 9 units for the HADS-A and > 7 units for the HADS- D. The area under the ROC curve was 0.90 for HADS-A (95% CI 0.88–0.91) and 0.84 for HADS-D (95% CI 0.81–0.86). Conclusions This study suggested that risk scores of anxious and depressive states above specific HADS cut-offs are useful in identifying anxious and depressive states in cancer patients, and they may thus be applicable in clinical practice.

Keywords Accuracy . Cancer . Hospital Anxiety and Depression Scale . Oncology . Psychometrics . Receiver operating characteristic curve

Introduction

The Hospital Anxiety and Depression Scale (HADS) [1–3] is a well-known emotional distress self-report questionnaire, and it is one of the most frequently used in oncology [4–6] as well as in other physical health settings (e.g., cardiology, brain injury, general medicine). Originally, it was designed to screen emotional suffering of patients in non-psychiatric settings by

detecting the two most frequent distress components: anxiety and depression. Since it is specific to patients with organic diseases, HADS excludes somatic symptoms of emotional distress (e.g., headache, weight loss, insomnia) that could be caused by the illness itself (including its treatments) rather than being emotional distress expressions [1, 2, 4, 5]. Furthermore, to improve sensitivity to medical conditions, severely psychopathological symptoms are not covered by HADS [1, 2, 4]. Thus, HADS is a measure of prolonged state rather than trait [4], and it is not recommended in detecting psychopathological disorders.

In spite of this, no agreement exists in literature on HADS accuracy in case finding [7–9] also because of the large range of cut-off rankings used by different authors [6, 7, 9]. In 2011, Vodermaier and Millman [10] conducted a meta-analysis to identify optimal, empirically derived HADS thresholds for clinical decision-making. Analyzing data from 28 different studies comparing HADS (entirely and/or in its subscales) against a semi-structured or structured clinical interview as a

Maria Antonietta Annunziata and Barbara Muzzatti should be considered joint first author.

* Maria Antonietta Annunziata [email protected]

1 Unit of Oncological Psychology, Centro di Riferimento Oncologico di Aviano (CRO), IRCCS, Via F. Gallini, 2, 33081 Aviano, PN, Italy

2 Unit of Cancer Epidemiology, Centro di Riferimento Oncologico di Aviano (CRO), IRCCS, Aviano, Italy

https://doi.org/10.1007/s00520-019-05244-8

/ Published online: 19 December 2019

Supportive Care in Cancer (2020) 28:3921–3926

reference standard, they provided the most accurate (i.e., those with best sensitivity and specificity) HADS cut-offs for any mental disorder and depressive disorder alone.

The present research also addresses the issue of HADS accuracy, though with an approach that differs from those applied to previously published studies [7–10]. In fact, since HADS is considered inadequate as screening tool for diagnos- tic purposes in psychopathology, we compared it against other (validated but lengthy and time-consuming administration) screening tools able to detect prolonged negative emotional states rather than against psychiatric assessment tools. Moreover, since a previously published study [11], conducted in a similar setting and on a large sample like the present one, suggested to use HADS as a measure of anxious and depres- sive states rather than as a global measure of emotional dis- tress, we tested the accuracy of the two HADS subscales against two different gold standard tools (one able to detect anxious states, the other able to detect depressive states).

Methods

Participants

Study participants were consecutive adult cancer inpatients, admitted to the same cancer institute for cancer treatments. The eligibility criteria were the following: age ≥ 18 years old; good understanding of the Italian language; absence of mental disorders; absence of physical or sensory disabilities that would interfere with completing the questionnaires; and signed informed consent form.

Materials and procedure

The HADS together with the State Anxiety (Anxiety-S) sub- scale of the State-Trait Anxiety Inventory, Form Y (STAI-Y) [12], and the Center for Epidemiologic Studies Scale on Depression (CES-D) [13] were used in this study.

HADS consists of two subscales: HADS-A, designed to detect anxious states, and HADS-D, designed to detect depressive states. Each subscale consists of seven items with a 4-point ordinal response format. Scores ranges from 0 to 21 in each subscale, with higher scores indicating higher levels of anxious or depressive state. Participants answer each item thinking of how they felt and/or behaved during the past week.

The STAI-Y is a questionnaire widely used to assess anxiety in its trait and state components. Spielberger val- idated the Italian version [14–16]. STAI-Y consists of 40 items (20 for trait anxiety and 20 for state anxiety), which participants rate on a 4-point scale. Scores range from 20 to 80 for each scale (i.e., trait anxiety and state anxiety), with higher scores indicating higher levels of state and

trait anxiety. In this study, only state subscale (STAI-S) was administered and participants were requested to an- swer each item thinking of how they felt and/or behaved during the past week. Since it registers anxious states rather than general anxiety disorder (or other psychopath- ological disturbances), STAI-S was chosen as gold stan- dard for HADS-A.STAI-S scores over 1.5 standard devi- ation of the normative sample (depending on gender and age) in the Italian manual [14], which were used to dis- tinguish participants with anxious state from participants without anxious state. Since a hospital stay may induce, per sè, an anxiety state, we decided to raise the cut-off from one standard deviation to 1.5, over the normative score; concurrently, we fixed it a 1.5 standard deviation rather than at 2 standard deviations over the normative score to reduce false negatives.

The CES-D is a measure of depressive symptomatology that consists of 20 items, which participants rate on a 4-point scale, thinking of how they felt and/or behaved during the past week. The results are graded on a 0–60-point scale and are proportional to depressive state intensity. Italian validation data were provided by Pierfederici et al. [17] and Fava [18] for both the general population and general hospital inpatients. Since it was developed to screen for depression in the general population (therefore, primarily focused on depression affec- tive components, such as depressed mood and feelings of helplessness), and since it has adequate psychometric proper- ties and has been widely used as screening tool in oncology [6, 19], CES-D was chosen as gold standard for HADS-D. According to a previous research on depression in general hospital inpatients [17], cut-off of 28+ was used to distinguish participants in a depressive state from participants who were not.

Potential participants were selected by consulting clinical files. The three above-mentioned tools, together with the in- formed consent form to participate in the study, were illustrat- ed to each eligible participant by a psychologist. The forms were autonomously filled out by participants in one occasion and subsequently they were collected by the psychologist who also debriefed the participants. Participants’ socio- demographic and clinical data were collected by the psychol- ogist consulting clinical files.

All participants gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the pro- tocol was approved by the Ethics Committee of the Centro di Riferimento Oncologico di Aviano (CRO) IRCCS (CRO- 2011-27).

Sample size

For HADS-A, we estimated the minimum sample size by assuming the following parameters: a type I error of 5%; a

3922 Support Care Cancer (2020) 28:3921–3926

power of 90%; a hypothesized AUC of 0.8 for HADS-A; a null hypothesis AUC of 0.7 for STAI-Y; and a ratio of positive/negative patients of 1. The number of participants required was 290 (145 cases and 145 controls). For HADS- D, we estimated the minimum sample size by assuming the following parameters: a type I error of 5%; a power of 90%; a hypothesized AUC of 0.8 for HADS-D; a null hypothesis AUC of 0.7 for CES-D; and a ratio of positive/negative pa- tients of 1. The number of participants required was 290 (145 cases and 145 controls) [20].

A quota sampling method was used to gather the necessary number of cases. All consecutive inpatients filled in the STAI- Y and CES-D questionnaires in order to be classified as cases or non-cases. Due to the sampling method used and the rela- tively low number of detected cases by means of the two gold standard tests, an elevated number of controls were detected until the required number of cases was obtained. Thus, the final sample consisted of a number of non-cases higher than expected. According to this final sample size and to the ratio of positive/negative patients observed, we recomputed the power of the analysis a posteriori. The power was 99%, with a type I error of 1%, for HADS-A, and 98%, with a type I error of 1%, for HADS-D. Consequently, the power of the analysis was higher than initially declared by the sample size computation.

Statistical analyses

The ability of HADS-A and HADS-D to discriminate cancer patients with or without respectively anxious or depressive state was made by means of receiver operator characteristic (ROC) curves [21]. The following indicators were calculated: area under the curve (AUC), sensitivity and specificity, with their corresponding 95% confidence interval (95% CI). The Youden index was also calculated and, in conjunction with the ROC curve, allowed to select the optimal cut-off for each test [20]. In particular, the optimal cut-off of HADS-A and the optimal cut-off of HADS-D, located in the most superior top-left point on the ROC curve, were derived in each curve from the point with the maximum Youden index that repre- sented the maximized sensitivity and specificity [20]. All tests were two-tailed and a p value < 0.05 was considered statisti- cally significant. The statistical analyses were performed using the SAS language program (Version 9.4, SAS Institute Inc., Cary, NC).

Results

To reach the necessary sample size, we recruited 2121 inpa- tients, of whom 1628 (76.8%) provided complete subscale HADS-A and STAI-S (necessary to verify the accuracy of HADS_A) and 1035 (48.8%) provided complete both

HADS-D and CES-D (necessary to verify the accuracy of HADS-D). Table 1 summarizes the main socio-demographic and clinical characteristics of the final sample.

HADS-A accuracy in case finding

The performance of HADS-A scores was evaluated according to ROC curves. The optimal cut-off value of the HADS-Awas > 9 units (Fig. 1). The AUC was 0.90 (95% CI 0.88–0.91), p value<0.001, with a sensitivity of 83.2% (95% CI 76.6–88.5) and a specificity of 80.5% (95% CI 78.4–82.5).

When considering a cut-off value > 9 units of the HADS-A subscale, 423 participants (26% for the whole sample) result- ed to be in the anxious state (see Table 2).

HADS-D accuracy in case finding

The performance of HADS-D was evaluated according to ROC curves. The optimal cut-off value of the HADS-D was > 7 units (Fig. 2). The AUC was 0.84 (95% CI 0.81–0.86), p value < 0.001, with a sensitivity of 72.9% (95% CI 64.9–80.0) and a specificity of 79.0% (95% CI 76.2–81.6).

When considering a cut-off value > 7 units of the HADS-D subscale, 292 participants (28.2% of the whole sample) result- ed to be in the depressive state (see Table 2).

Discussion

HADS seems to be the tool of choice for detecting negative emotional states in cancer patients, thanks to its features (i.e., specific of medical settings; good psychometric properties; brief; rapid administration; and good compliance) [1–6]. Although it is necessary to identify the optimal thresholds to differentiate cases from non-cases (accuracy) for the HADS reliable and valid use in both clinical practice and research, no agreement exists in literature on this issue.

The present work contributes to define HADS accuracy in case finding. However, it addresses this issue in an original manner, i.e., by comparing it against other already validated screening tools able to detect prolonged negative emotional states rather than against psychiatric assessment tools useful in psychopathological diagnosis.

According to the present results, the optimal cut-off values were > 9 units for the HADS-A and > 7 units for the HADS-D. These thresholds are different from the cut- off provided for HADS in previous literature [6–10]. These dissimilarities find an explanation in the different psycho- logical constructs (emotional states vs. psychopathological disorders) assessed by the tools used as reference standard. For both established thresholds, sensibility and specificity were adequate. Indeed, given that HADS is an emotional state screening tool rather than a diagnostic one, a higher

3923Support Care Cancer (2020) 28:3921–3926

number of false positives (i.e., lower specificity) is accept- able and finds its balance in a well-recognized feasibility. In other words, with respect to the two employed gold standard tools (STAI-S e CES-D), the HADS subscales, anxiety, and depression, induce additional false positives, but they are quicker in terms of administration and scoring, more appropriate to capture patients emotional states in medical settings, as a consequence they can be usefully employed to detect anxious and depressive states in oncol- ogy settings.

The employment of gold standard tools developed to detect emotional states, rather than gold standard psychiatric

diagnostic tools, as well as the rigorous data analysis proce- dure and the adequate statistical power represent the major strengths of this study. Furthermore, our findings are easily reproducible as they were obtained in a non-selected consec- utive population of cancer inpatients. Mental disorders (i.e., documented presence and/or history of a psychiatric syn- drome) were an exclusion criterion for the present study to avoid their possible confusing role on detecting prolonged negative emotional states associated with cancer and its treatments.

A potential study limitation may consist of the non- negligible percentage of the enrolled sample that had provided

Table 1 Socio-demographic and clinical characteristics of sub- samples used to test HADS-A accuracy (N = 1628), and HADS- D accuracy (N = 1035)

HADS-A accuracy test (N = 1628)

HADS-D accuracy test (N = 1035)

N % N %

Gender

Male 460 28.3 294 28.4

Female 1168 71.7 741 71.6

Education

Compulsory 642 39.4 418 40.4

Secondary 756 46.4 476 46.0

Post-secondary 230 14.1 141 13.6

Occupational status

Employed 919 56.4 584 56.4

Unemployed/homemaker/student 599 36.8 383 37.0

Missing datum 110 6.8 68 6.6

Marital status

Partnered 1234 75.8 788 76.1

Non-partnered 393 24.1 247 23.9

Missing datum 1 0.1 0 –

Cancer diagnosis

Oro-pharyngeal 47 2.9 32 3.1

Digestive apparatus 246 15.1 126 12.2

Respiratory system and intrathoracic organs 65 4.1 55 5.3

Breast 520 31.9 329 31.8

Genito-urinary 391 24.0 265 25.6

Hematologic 204 12.5 134 12.9

Others 155 9.5 94 9.1

Anxious state (STAI-Y)

Non-case 1462 89.8 – –

Case 166 10.2 – –

Depressive state (CES-D)

Non-case – – 891 86.1

Case – – 144 13.9

Mdn Range Mdn Range

Age (years) 53 18–83 53 21–83

3924 Support Care Cancer (2020) 28:3921–3926

invalid data and were consequently dropped from the study. Unfortunately, reasons for providing incomplete or unfilled questionnaires were not recorded; consequently, no specula- tions on this point may be done.

Finally, the obtained results should be considered in the light of the ongoing debate on the HADS dimensional struc- ture [5, 6, 22–25], of the related item formulation, and of the translation aspects [9, 26–28]. Concerning the dimensional aspect of the HADS, a previous study [11], conducted on a large sample with similar characteristics to the present one, has shown that the bi-factorial structure is more appropriate than the mono-factorial one [29]. The satisfactory balance between sensitivity and specificity of both HADS-A and HADS-D subscales, emerging in the present study, supports their appropriateness in terms of content validity—therefore, an adequate formulation of each single item—at least for the

Italian context. Future studies mirroring the methodology herein reported, in linguistic-cultural contexts other than the Italian one and in clinical settings different from the oncologic one, will offer useful information for a further in-depth inves- tigation of this aspect.

In conclusion, this study suggests that risk scores of anx- ious and depressive states above specific cut-offs derived from HADS may be useful in identifying anxious and depressive states in cancer patients during clinical practice. In oncology, emotional distress (in its main components of anxiety and depression) is expected during the entire disease trajectory. It is a source of suffering on its own, but it may also interfere with treatment adherence, as well as with both health and well-being. Its reliable and valid detection and monitoring represent the first step toward a tailored comprehensive (bio- psycho-social) care of cancer patients.

Table 2 Comparison of HADS-A vs. STAI-Y (gold standard) and HADS-D vs. CES-D (gold standard)

State

Anxious Depressive

STAI-Y (gold standard) HADS-A CES-D (gold standard) HADS-D

≤ 9 (non-cases) > 9 (cases) ≤ 7 (non-cases) > 7 (cases) non-cases 1177 285 Non-cases 704 187

Cases 28 138 Cases 39 105

Sensitivity = 83.1 (95% confidence interval 76.6–88.5) Specificity = 80.5 (95% confidence interval 78.4–82.5)

Sensitivity = 72.9 (95% confidence interval 64.9–80.0) Specificity = 79.0 (95% confidence interval 76.2–81.6)

Fig. 2 Receiver operative characteristic (ROC) curve, corresponding area under curve (AUC), sensitivity, and specificity of depressive state risk score for distinguishing cases from non-cases (Aviano, Italy)

Fig. 1 Receiver operative characteristic (ROC) curve, corresponding area under curve (AUC), sensitivity, and specificity of anxious state risk score for distinguishing cases from non-cases (Aviano, Italy)

3925Support Care Cancer (2020) 28:3921–3926

Acknowledgments The authors wish to thank Ms. Luigina Mei for her editorial assistance.

Compliance with ethical standards

All participants gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the Centro di Riferimento Oncologico di Aviano (CRO) IRCCS (CRO-2011-27).

Conflict of interest The authors declare that they have no conflict of interest.

References

1. Zigmond AS, Snaith RP (1983) The hospital anxiety and depres- sion scale. Acta Psychiatr Scand 67:361–370

2. Snaith RP, Zigmond AS (1994) The Hospital Anxiety and Depression Scale manual. Nelson, Windsor, NFER

3. Snaith RP (2003) The Hospital Anxiety and Depression Scale. Health Qual Life Outcomes 1:29

4. Herrmann C (1997) International experiences with the Hospital Anxiety and Depression Scale. A review of validation data and clinical results. J Psychosom Res 42:17–41

5. Bjelland I, Dahl AA, Neckelmann D (2002) The validity of the Hospital Anxiety and Depression Scale: an updated literature re- view. J Psychosom Res 52:69–77

6. Vodermaier A, Linden W, Siu C (2009) Screening for emotional distress in cancer patients: a systematic review of assessment instru- ments. J Natl Cancer Inst 101(21):1464–1488

7. Mitchell AJ, Meader N, Symonds P (2010) Diagnostic validity of the hospital anxiety and depression scale (HADS) in cancer and palliative settings: a metaanalysis. J Affective Disord 126:335–348

8. Brennan C, Worrall-Davies A, McMillan D, Gilbody S, House A (2010) The hospital anxiety and depression scale: a diagnostic meta-analysis of case-finding ability. J Psychosom Res 69:371–378

9. Carey M, Noble N, Sanson-Fisher R, Mackenzie L (2012) Identifying psychological morbidity among people with cancer using the Hospital Anxiety and Depression Scale: time to revisit first principles? Psycho-Oncol. 21:229–238

10. Vodermaier A, Millman RD (2011) Accuracy of the Hospital Anxiety and Depression Scale as a screening tool in cancer patients: a systematic review and meta-analysis. Support Care Cancer 19: 1899–1908

11. Annunziata MA, Muzzatti B, Altoè G (2011) Defining Hospital Anxiety and Depression Scale (HADS) structure by confirmatory factor analysis: a contribution to validation for oncological settings. Ann Oncol 22:2330–2333

12. Spielberger CD (1983) State-trait anxiety inventory (form Y). Mind Garden, Palo Alto (CA)

13. Radloff L (1977) The CES-D Scale: a self-report depression scale for research in the general population. Appl Psych Meas 1:385–401

14. Spielberger CD (1989) S.T.A.I. (State-Trait-Anxiety Inventory). Inventario per l’ansia di stato e di tratto. Forma Y. Organizzazioni Speciali, Firenze

15. Pedrabissi L, Santinello M (1989) Verifica della validità dello STAI forma Y di Spielberger. BPA. 191:11–14

16. Macor A, Pedrabissi L, Santinello M. (190). Ansia di stato e di tratto: ulteriore contributo alla verifica della validità psicometrica e teorica dello S.T.A.I. forma Y di Spielberger. Psicologia e Società. 15: 67–74

17. Pierfederici A, Fava GA, Munari F et al (1982) Validazione italiana del CES-D per la misurazione della depressione. In: Canestrari R (ed) Nuovi metodi in psicoterapia. Organizzazioni Speciali, Firenze, pp 95–103

18. Fava GA (1983) Assessing depressive symptoms across cultures: Italian validation of the CES-D self-rating scale. J Clin Psychol 39: 249–251

19. Hann D, Winter K, Jacobsen P (1999) Measurement of depressive symptoms in cancer patients: evaluation of the Center for Epidemiological Studies Depression Scale (CES-D). J Psychosom Res 46:437–443

20. Hanley JA, McNeil BJ (1982) The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 143:29–36

21. Metz CE (1978) Basic principles of ROC analysis. Semin Nucl Med 8:283–298

22. Rodgers J, Martin C, Morse R et al (2005) An investigation into the psychometric properties of the Hospital Anxiety and Depression Scale in patients with breast cancer. Health Qual Life Outcomes 3:41

23. Cosco TD, Doyle F, Ward M, McGee H (2012) Latent structure of the Hospital Anxiety And Depression Scale: a 10-year systematic review. J Psychosom Res 72:180–184

24. Norton S, Cosco T, Doyle F, Done J, Sacker A (2013) The Hospital Anxiety and Depression Scale: a meta confirmatory factor analysis. J Psychosom Res 74:74–81

25. Straat JH, van der Ark LA, Sijtsma K (2013) Methodological arti- facts in dimensionality assessment of the Hospital Anxiety and Depression Scale (HADS). J Psychosom Res 74:116–121

26. Maters GA, Sanderman R, Kim AY, Coyne JC (2013) Problems in cross-cultural use of the Hospital Anxiety and Depression Scale: “no butterflies in the desert”. PLoS One 8:e70975

27. Cameron IM, Scott NW, Adler M, Reid IC (2014) A comparison of three methods of assessing differential item functioning (DIF) in the Hospital Anxiety Depression Scale: ordinal logistic regression, Rasch analysis and the Mantel chi square procedure. Qual Life Res 23:2883–2888

28. Verdam MG, Oort FJ, Sprangers MA (2017) Item bias detection in the Hospital Anxiety and Depression Scale using structural equa- tion modeling: comparison with other item bias detection methods. Qual Life Res 26:1439–1450

29. Saboonchi S, Wennman-Larsen A, Alexanderson K, Petersson ML (2013) Examination of the construct validity of the Swedish version of Hospital Anxiety and Depression Scale in breast cancer patients. Qual Life Res 22:2849–2856

Publisher’s note Springer Nature remains neutral with regard to jurisdic- tional claims in published maps and institutional affiliations.

3926 Support Care Cancer (2020) 28:3921–3926

Supportive Care in Cancer is a copyright of Springer, 2020. All Rights Reserved.