HADS

HADS in Trials and Economic Evaluation

The Hospital Anxiety and Depression Scale was published by Zigmond and Snaith as a self-assessment scale for picking up states of anxiety and depression among patients attending a hospital medical outpatient clinic. Its authors argued that bringing the scale into general hospital practice would help with detecting and managing emotional disorder in patients under investigation and treatment in medical and surgical departments, and the HADS is now widely used as a mood outcome in trials of treatments for physical conditions. For a health economist the HADS raises two practical questions. The first is how to read a score or a caseness rate taken from a trial, since the scale gives separate anxiety and depression scores that different studies read against different cut points. The second is how to bring HADS results into a cost-utility analysis, because the scale has no preference weights and cannot be turned into QALYs on its own terms.

Origins of the HADS in hospital medical clinics

Zigmond and Snaith built the HADS for general hospital practice, for patients under investigation and treatment for physical illness in medical and surgical departments. They reported that the scale was a reliable instrument for detecting depression and anxiety in a hospital medical outpatient clinic, and that the anxiety and depressive subscales were also valid measures of severity of the emotional disorder.

Use spread quickly. In a 1997 review, Herrmann counted more than 200 published studies, from most medical settings and many countries, reporting experience with a scale developed for use with physically ill patients. Herrmann concluded that the HADS gives clinically meaningful results as a psychological screening tool and is sensitive to changes during the course of diseases and in response to treatment, which is the property a trial outcome needs.

The word "hospital" in the name does not restrict where the scale can be used. Writing in 2003, Snaith noted that the name might suggest the HADS is valid only in hospital, but that studies conducted throughout the world had confirmed its validity in community settings and primary care medical practice. Snaith also stressed that self-assessment scales are valid only for screening and that a definitive diagnosis must rest on clinical examination.

Subscales, score bands and the meaning of caseness

The HADS has 14 items in total, seven of which form the depression subscale (HADS-D) and the rest the anxiety subscale (HADS-A), and each subscale is read on its own. The developers' bands describe how far a subscale score suggests a mood disorder, and later reviews added cut points chosen for screening accuracy.

Snaith described a score of 0 to 7 on either subscale as in the normal range, a score of 8 to 10 as just suggestive of the respective state, and a score of 11 or higher as indicating probable presence, or caseness, of the mood disorder. These bands give three groups, so a trial can report the proportion scoring 8 or more or the proportion scoring 11 or more, and the two proportions answer different questions.

Brennan and colleagues described the lower threshold, which falls between 7 and 8, as the recommended clinical cut point and the higher one, between 10 and 11, as the research cut point. A trial that defines a responder or a case at 8 counts more people as cases than one that uses 11, and any cost per case or proportion of cases in a model inherits that choice.

Screening accuracy reported in later reviews

Screening accuracy decides how many people a HADS-based case-finding programme sends for clinical assessment and how many of them have a disorder. The figures come from studies that set HADS scores against a psychiatric diagnostic interview and report sensitivity and specificity at each threshold.

Bjelland and colleagues reviewed the literature in 2002 and found that in most studies the best balance between sensitivity and specificity came when caseness was defined by a score of 8 or above on both HADS-A and HADS-D. Sensitivity and specificity were then approximately 0.80 for both subscales. They also found that the HADS performed well in assessing symptom severity and caseness of anxiety disorders and depression in somatic, psychiatric and primary care patients and in the general population.

Brennan and colleagues pooled 41 studies that compared the HADS with a standardised psychiatric interview. For major depressive disorder, a cut point of 8 or more gave a pooled sensitivity of 0.82 and specificity of 0.74, while 11 or more gave a sensitivity of 0.56 and specificity of 0.92. They concluded that the HADS is useful for detecting emotional distress in non-psychiatric patients but is not superior to other screening tools for identifying specific mental disorders in physical health settings.

Wu and colleagues later analysed individual participant data on the depression subscale and reported in 2021 that combined sensitivity and specificity for major depression was maximised at a HADS-D cut-off of 7 or higher, for semi-structured interviews, fully structured interviews and the MINI alike. The best threshold therefore depends on the reference interview and on the trade-off an analyst accepts, which a ROC curve displays across all thresholds at once.

Factor structure and what a HADS score represents

Whether the HADS measures two separate things or one general distress decides whether the two subscale scores, or a single total, should carry the analysis. Factor analysis of item responses is the usual test, and its results have varied.

Bjelland and colleagues found that most factor analyses supported a two-factor solution in good accord with the anxiety and depression subscales. The subscales were moderately correlated, with correlations between 0.40 and 0.74 and a mean of 0.56. For internal consistency, Cronbach's alpha ranged between 0.68 and 0.93 for HADS-A (mean 0.83) and between 0.67 and 0.90 for HADS-D (mean 0.82).

A later systematic review by Cosco and colleagues looked at 50 studies published between May 2000 and May 2010 that applied latent variable methods to the HADS. Twenty-five found a two-factor structure, five a single dimension, 17 a three-factor structure and two a four-factor structure. The result tracked the method, since exploratory factor analyses mostly produced two factors, confirmatory factor analyses mostly three and item response theory studies mostly one.

For trial analysis this disagreement has a practical consequence. A summed total of both subscales assumes that anxiety and depression items measure one underlying quantity, which only some of the studies support, while separate subscale scores follow the developers' bands and the two-factor solution that half of the reviewed studies found. For depression screening, Wu and colleagues reported in 2023 that the 14-item total HADS was no more accurate than the 7-item HADS-D for detecting major depression, and they concluded that the shorter HADS-D would be preferred in most settings.

The HADS as a trial outcome in economic evaluation

A cost-utility analysis values health states on a scale anchored at full health and death so that time spent in each state can be summed into QALYs. The HADS is a patient-reported outcome measure without preference weights or a value set, so a subscale score grades symptom severity without valuing it, and nothing in its construction makes a two-point change worth the same at every level of the scale.

Drummond and colleagues, and Brazier and colleagues, describe the standard routes for a measure of this kind. The first is to collect a preference-based measure such as the EQ-5D in the same trial and calculate QALYs from that, keeping the HADS as a clinical outcome. The second, used when only HADS data exist, is to predict utilities with a mapping function estimated in a dataset where both measures were collected. Mapping depends on overlap in content, and the EQ-5D has one anxiety or depression dimension beside mobility, self-care, usual activities and pain or discomfort, so much of what the HADS records has no direct counterpart in the target measure.

Several choices change the economic result when HADS data are used. The outcome can be a continuous subscale score, the proportion above a cut point of 8 or 11, or a combined total, and a cost per case avoided depends heavily on which threshold defines a case. A cost per point of improvement on HADS-D cannot be compared with a cost-effectiveness threshold, and a change large enough to matter to patients is a separate question of minimal clinically important difference. When the HADS is a secondary outcome in a trial of a treatment for physical illness, a preference-based measure collected alongside it carries the QALY calculation, as in other mental health economics work.

Illustrative worked example of screening at two HADS cut points

This example is illustrative. The sensitivity and specificity values are the pooled estimates for major depressive disorder reported by Brennan and colleagues, while the population, prevalence and costs are invented to show the arithmetic. A clinic screens 1,000 patients with a long-term physical condition, and 15 per cent, or 150 patients, have major depression. Everyone who screens positive receives a clinical assessment.

1. Cut point of 8 or more. True positives are 0.82 × 150 = 123 and true negatives are 0.74 × 850 = 629, so false positives are 850 − 629 = 221 and 123 + 221 = 344 patients screen positive.

2. Cut point of 11 or more. True positives are 0.56 × 150 = 84 and true negatives are 0.92 × 850 = 782, so false positives are 850 − 782 = 68 and 84 + 68 = 152 patients screen positive.

3. Positive predictive value.

$$\text{PPV} = \frac{TP}{TP + FP}$$

where $TP$ is the number of true positives and $FP$ the number of false positives. At a cut point of 8 the value is 123 ÷ 344 = 0.358, and at a cut point of 11 it is 84 ÷ 152 = 0.553.

4. Cost per true case found. Suppose the questionnaire costs GBP 5 per patient and each assessment costs GBP 120. At a cut point of 8 the programme costs 1,000 × 5 + 344 × 120 = 46,280, or GBP 46,280, which is 46,280 ÷ 123 = 376.3, about GBP 376 per true case. At a cut point of 11 it costs 1,000 × 5 + 152 × 120 = 23,240, or GBP 23,240, which is 23,240 ÷ 84 = 276.7, about GBP 277 per true case.

5. Incremental comparison. Moving the cut point down to 8 finds 123 − 84 = 39 more cases at an extra 46,280 − 23,240 = 23,040, or GBP 23,040, which is 23,040 ÷ 39 = 590.8, about GBP 591 per additional case found.

The lower cut point finds more cases but sends 221 people without major depression for assessment, against 68 at the higher one. The incremental cost per additional case shows the price of that extra sensitivity, but a case found is an intermediate outcome. Judging value for money needs a model that follows detected and missed cases through treatment to costs and QALYs, measured with a preference-based instrument.

How the HADS differs from nearby measures

The HADS sits close to several outcome concepts and instruments, and the differences affect how its results are pooled or compared. Most confusion arises between the scale and the outcomes it measures.

Anxiety symptoms and depression symptoms are clinical outcomes, and the HADS is one instrument for measuring them, so a review that pools anxiety outcomes across trials needs to convert between instruments. The PHQ-9 and the GAD-7 are separate questionnaires for depression and anxiety, each with its own items, scoring and cut points, so a HADS-D of 8 and a PHQ-9 score of 8 do not mean the same thing.

A disease-specific instrument targets the symptoms of one condition, whereas the HADS was designed for mood disorder across physical illnesses and has since been used in primary care and community samples. It also differs from health-related quality of life measures, which cover physical, psychological and social functioning, because the HADS covers only anxiety and depression. As a non-preference-based measure it describes symptoms without valuing them, which is why a QALY calculation needs a preference-based measure or a mapping function beside it.

Limitations and common misreadings of HADS results

Several errors recur when HADS results are used in economic work. Most come from treating a screening score as more than the scale was built to provide.

• Reading caseness as diagnosis. Snaith stressed that self-assessment scales are valid only for screening, and at the 8 or more cut point in the worked example only about a third of positive screens had major depression.
• Comparing rates across cut points. Proportions of cases defined at 8 and at 11 differ greatly, and Wu and colleagues found the best balance for HADS-D at 7 or higher, so cost per case results are comparable only when the threshold matches.
• Summing the subscales without checking structure. Cosco and colleagues found two, three, one and four factor solutions in different studies, so a single total needs its own justification.
• Treating the score as a utility. A HADS subscale score has no anchor at full health or death and cannot be multiplied by time to give QALYs.
• Overlooking the limits of mapping. A HADS-to-EQ-5D mapping is limited by the narrow overlap between the two instruments, and its estimation sample should match the population being modelled.

Sources

• Bjelland I, Dahl AA, Haug TT, Neckelmann D. The validity of the Hospital Anxiety and Depression Scale: an updated literature review. Journal of Psychosomatic Research. 2002;52(2):69-77. https://pubmed.ncbi.nlm.nih.gov/11832252/
• Brazier J, Ratcliffe J, Salomon JA, Tsuchiya A. Measuring and Valuing Health Benefits for Economic Evaluation. 2nd ed. Oxford University Press; 2017.
• Brennan C, Worrall-Davies A, McMillan D, Gilbody S, House A. The Hospital Anxiety and Depression Scale: a diagnostic meta-analysis of case-finding ability. Journal of Psychosomatic Research. 2010;69(4):371-378. https://pubmed.ncbi.nlm.nih.gov/20846538/
• Cosco TD, Doyle F, Ward M, McGee H. Latent structure of the Hospital Anxiety And Depression Scale: a 10-year systematic review. Journal of Psychosomatic Research. 2012;72(3):180-184. https://pubmed.ncbi.nlm.nih.gov/22325696/
• Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. 4th ed. Oxford University Press; 2015.
• Herrmann C. International experiences with the Hospital Anxiety and Depression Scale: a review of validation data and clinical results. Journal of Psychosomatic Research. 1997;42(1):17-41. https://pubmed.ncbi.nlm.nih.gov/9055211/
• Snaith RP. The Hospital Anxiety And Depression Scale. Health and Quality of Life Outcomes. 2003;1:29. doi:10.1186/1477-7525-1-29
• Wu Y, Levis B, Sun Y, et al. Accuracy of the Hospital Anxiety and Depression Scale Depression subscale (HADS-D) to screen for major depression: systematic review and individual participant data meta-analysis. BMJ. 2021;373:n972. https://pubmed.ncbi.nlm.nih.gov/33972268/
• Wu Y, Levis B, Sun Y, et al. Comparison of the accuracy of the 7-item HADS Depression subscale and 14-item total HADS for screening for major depression: a systematic review and individual participant data meta-analysis. Psychological Assessment. 2023;35(2):95-114. doi:10.1037/pas0001181
• Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatrica Scandinavica. 1983;67(6):361-370. doi:10.1111/j.1600-0447.1983.tb09716.x