Advertisement

Analyzing the Pain/Discomfort and Anxiety/Depression Composite Domains and the Meaning of Discomfort in the EQ-5D: A Mixed-Methods Study

Open AccessPublished:August 13, 2022DOI:https://doi.org/10.1016/j.jval.2022.06.012

      Highlights

      • Several preference-accompanied measures have composite domains that join 2 or more different but related domains of health in a single domain. The EQ-5D has 2 such domains: pain/discomfort and anxiety/depression. Little is known about how respondents use the composite domains to self-report their own health. Furthermore, there is conflicting evidence on the extent to which the current operationalization of pain/discomfort is able to capture discomfort.
      • This study identified 6 different response behaviors in the EQ-5D composites: “uniform,” “most severe,” “least severe,” “average,” “synergistic,” and “inconsistent.” Discomfort is not solely pain related; it is an umbrella term for approximately 100 different nonpain physical sensations, mental problems, or feelings. Composites were found to be sources of measurement error, including under- and inconsistent reporting of health problems, ordering effect, potential differential item functioning, and interdomain dependency.
      • This study sheds light on potential issues surrounding the composite domains in generic preference-accompanied measures. These findings may contribute to the development of new and further refinement of existing measures used to evaluate health outcomes in terms of quality-adjusted life-years and to facilitate decision making in healthcare.

      Abstract

      Objectives

      The EQ-5D has 2 composite domains: pain/discomfort (PD) and anxiety/depression (AD). This study aims to explore how respondents use the composites to self-report health and what the meaning of discomfort is in the EQ-5D for the general public.

      Methods

      Both qualitative and quantitative data were collected in an online cross-sectional survey involving a nationally representative general population sample in Hungary (n = 1700). Respondents completed the 5-level version of EQ-5D, followed by the composites split into individual subdomains. Open-ended questions were asked to explore respondents’ interpretations and experiences of discomfort.

      Results

      Six different response behaviors were identified in the composites: “uniform” (21%-32%), “most severe” (30%-34%), “least severe” (16%-23%), “average” (2%-4%), “synergistic” (4%-5%), and “inconsistent” (13%-15%). Compared with the individual subdomains, many respondents under-reported their problems on both composites (PD 16%-22% and AD 6%-13%, P < .05). In respondents who scored differently in the 2 separate domains, mainly problems with the first subdomain determined responses in the composites (PD 66% and AD 61%). The discomfort subdomain in the EQ-5D captured more than 100 different problems, including pain, nonpain physical discomfort (eg, tiredness, dizziness, and nausea), and psychological discomfort (eg, anxiety, nervousness, and sadness). Women, older adults, and those in worse general health status more often considered discomfort as pain (P < .05).

      Conclusions

      We found empirical evidence of measurement error in the composite responses on the EQ-5D, including under- and inconsistent reporting, ordering effects, potential differential item functioning, and interdomain dependency. Our findings contribute new knowledge to the development of new and refinement of existing self-reported health status instruments, also beyond the EQ-5D.

      Keywords

      Introduction

      There are a growing number of preference-accompanied generic health status measures that can be used to inform resource allocation decisions.
      • Brazier J.
      • Ara R.
      • Rowen D.
      • Chevrou-Severac H.
      A review of generic preference-based measures for use in cost-effectiveness models.
      These instruments typically comprise 2 parts, a self-completed questionnaire that describes health status along different domains of health and response levels within these (ie, the descriptive system) and a value set that enables to score the questionnaire based on societal preferences, the latter enabling the calculation of quality-adjusted life-years. Composite domains, combining multiple different but related domains of health into a single domain, are a commonly used approach to keeping the potential number of health states feasible for valuation. Several instruments, including the EQ-5D, SF-6D, and 15D, use composite domains to capture physical (eg, having pain, discomfort, or symptoms) and mental symptoms (eg, being anxious, depressed, tense, downhearted, or low).
      • Richardson J.
      • Khan M.A.
      • Iezzi A.
      • Maxwell A.
      Comparing and explaining differences in the magnitude, content, and sensitivity of utilities predicted by the EQ-5D, SF-6D, HUI 3, 15D, QWB, and AQoL-8D multiattribute utility instruments.
      The available methodological literature on the use and interpretation of composite domains to report one’s own health is scarce and mainly limited to the EQ-5D.
      • McDonald R.
      • Mullett T.L.
      • Tsuchiya A.
      Understanding the composite dimensions of the EQ-5D: an experimental approach.
      • Tsuchiya A.
      • Bansback N.
      • Hole A.R.
      • Mulhern B.
      Manipulating the 5 dimensions of the EuroQol instrument: the effects on self-reporting actual health and valuing hypothetical health states [published correction appears in Med Decis Making. 2020;40(2):115].

      Bryan S., Jowett S., Hardyman W., Bentham P. Does the EQ-5D anxiety/depression item measure anxiety, depression, both or neither? 21st Plenary Meeting of the EuroQol Group. Chicago, US; 16-18 September 2004.

      Macran S., Kind P. EQ-5D valuations from a British national postal survey. In: 17th Plenary Meeting of the EuroQoL Group, Pamplona, Spain; September 18-29, 2000.

      • Engel L.
      • Haagsma J.A.
      • Janssen B.
      • Whitehurst D.G.T.
      • Mulhern B.
      An exploratory analysis of the pain/discomfort dimension of the EQ-5D-5L in people living with physical and mental health conditions.
      • Spronk I.
      • Bonsel G.J.
      • Polinder S.
      • van Baar M.E.
      • Janssen M.F.
      • Haagsma J.A.
      Exploring the relation between the EQ-5D-5L pain/discomfort and pain and itching in a sample of burn patients.
      • Feng Y.
      • Herdman M.
      • van Nooten F.
      • et al.
      An exploration of differences between Japan and 2 European countries in the self-reporting and valuation of pain and discomfort on the EQ-5D.
      The EQ-5D is one of the most frequently used preference-accompanied health status measures that is recommended in several national pharmacoeconomic guidelines worldwide.
      • Kennedy-Martin M.
      • Slaap B.
      • Herdman M.
      • et al.
      Which multi-attribute utility instruments are recommended for use in cost-utility analysis? A review of national health technology assessment (HTA) guidelines.
      ,
      • Rencz F.
      • Gulácsi L.
      • Drummond M.
      • et al.
      EQ-5D in Central and Eastern Europe: 2000-2015.
      It classifies health into 5 domains, 2 of which are composite domains, pain/discomfort (PD) and anxiety/depression (AD).
      • Brooks R.
      EuroQol: the current state of play.
      The composites in the EQ-5D may be considered a special form of double-barreled questions. Questions using the conjunction “or” pose a variety of challenges for both respondents completing the questionnaire and researchers interpreting the responses. Undoubtedly, the complexity of these questions may cause respondents to not know which component of the question they are supposed to answer.
      • Peasgood T.
      • Mukuria C.
      • Carlton J.
      • Connell J.
      • Brazier J.
      Criteria for item selection for a preference-based measure for use in economic evaluation.
      • Menold N.
      Double barreled questions: an analysis of the similarity of elements and effects on measurement quality.
      • Grant Levy S.
      Deconstructing a double-barreled alternative: evolution and creationism.
      This increased cognitive burden on respondents may also result in longer completion time and higher number of missing responses.
      • Bassili J.N.
      • Scott B.S.
      Response latency as a signal to question problems in survey research.
      A further apparent difficulty with such questions is that it is impossible to disentangle which part of the question (eg, pain or discomfort) was answered.
      Existing qualitative evidence suggests that respondents demonstrate a good understanding of the concepts of pain, anxiety, and depression in the EQ-5D.
      • Keeley T.
      • Al-Janabi H.
      • Lorgelly P.
      • Coast J.
      A qualitative assessment of the content validity of the ICECAP-A and EQ-5D-5L and their appropriateness for use in health research.
      • Whalley D.
      • Globe G.
      • Crawford R.
      • et al.
      Is the EQ-5D fit for purpose in asthma? Acceptability and content validity from the patient perspective.
      • Matza L.S.
      • Boye K.S.
      • Stewart K.D.
      • Curtis B.H.
      • Reaney M.
      • Landrian A.S.
      A qualitative examination of the content validity of the EQ-5D-5L in patients with type 2 diabetes.
      • Rencz F.
      • Mukuria C.
      • Bató A.
      • Poór A.K.
      • Finch A.P.
      A qualitative investigation of the relevance of skin irritation and self-confidence bolt-ons and their conceptual overlap with the EQ-5D in patients with psoriasis [published online ahead of print, 2022 Apr 26].
      By contrast, there is a lack of conceptual clarity surrounding the term “discomfort.”
      • Gudex C.
      The descriptive system of the EuroQol instrument.
      • Yang F.
      • Jiang S.
      • He X.-N.
      • et al.
      Do rural residents in China understand EQ-5D-5L as intended? Evidence from a qualitative study.
      • Ashkenazy S.
      • DeKeyser Ganz F.
      The differentiation between pain and discomfort: a concept analysis of discomfort.
      • Tighe P.J.
      • Sannapaneni B.
      • Fillingim R.B.
      • et al.
      Forty-two million ways to describe pain: topic modeling of 200,000 PubMed pain-related abstracts using natural language processing and deep learning-based text generation.
      Some generic preference-accompanied measures give examples for discomfort in their descriptive systems; for example, the discomfort and symptoms domain of 15D mentions “pain, ache, nausea, itching, etc,” whereas the EQ-5D does not mention specific examples. Nevertheless, without any specification, it is not clear what respondents think of discomfort when they describe their health or value health states. A recent concept analysis found that individuals attribute not only physical but also psychological meaning to discomfort.
      • Ashkenazy S.
      • DeKeyser Ganz F.
      The differentiation between pain and discomfort: a concept analysis of discomfort.
      So far, no studies have been conducted to qualitatively explore the meaning of discomfort in any generic health status measure among members of the general population.
      This study aims to explore (1) how respondents use the PD and AD composites to self-report health status on the EQ-5D, (2) what the meaning of discomfort is in the EQ-5D for the general public, and (3) whether sociodemographic and health-related characteristics of people affect what they consider as discomfort.

      Methods

      Study Design and Data Collection

      Our study adopted a convergent mixed-methods design, where quantitative and qualitative data were collected in parallel. The integration of quantitative and qualitative data was achieved at the methods level through connecting the sampling frame and at the interpretation and reporting level through data transformation.
      • Fetters M.D.
      • Curry L.A.
      • Creswell J.W.
      Achieving integration in mixed methods designs-principles and practices.
      In November 2020, an online cross-sectional survey was conducted involving a large general population sample in Hungary. Nonprobabilistic quota-based sampling was applied to recruit respondents aged 18 years or older from members of an existing online panel. The study sought to attain an approximate representativeness applying “soft targets” for age, gender, education, place of living, and region. Upon completion of the questionnaire, online panelists earned survey points. Data collection was approved by the Research Ethics Committee of the Corvinus University of Budapest (no. KRH/343/2020). Respondents were asked to read a brief information sheet and to provide their informed consent before starting the survey.

      5-Level Version of EQ-5D

      The 5-level version of EQ-5D (EQ-5D-5L) health status measure aims to capture the respondent’s current health status (recall period: today).
      • Herdman M.
      • Gudex C.
      • Lloyd A.
      • et al.
      Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L).
      It consists of 2 parts, a descriptive system and a visual analogue scale (EQ VAS) with endpoints of 0 (“the worst health you can imagine”) and 100 (“the best health you can imagine”). The descriptive system assesses health status across 5 domains (mobility, self-care, usual activities, PD, and AD) with 5 response levels in each (1 = no problems, 2 = slight problems, 3 = moderate problems, 4 = severe problems, and 5 = extreme problems/unable to), therefore defining a total of 3125 health profiles.

      Survey Instrument

      Respondents were asked to rate their own health at the time of the survey (ie, current health) on the EQ-5D-5L.
      • Herdman M.
      • Gudex C.
      • Lloyd A.
      • et al.
      Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L).
      In addition, 3 modified versions of the EQ-5D-5L were also used to collect data on self-reported health. The selected modifications and some of their combinations have been used in earlier studies.
      • McDonald R.
      • Mullett T.L.
      • Tsuchiya A.
      Understanding the composite dimensions of the EQ-5D: an experimental approach.
      ,
      • Tsuchiya A.
      • Bansback N.
      • Hole A.R.
      • Mulhern B.
      Manipulating the 5 dimensions of the EuroQol instrument: the effects on self-reporting actual health and valuing hypothetical health states [published correction appears in Med Decis Making. 2020;40(2):115].
      ,
      • Devlin N.J.
      • Shah K.K.
      • Mulhern B.J.
      • Pantiri K.
      • van Hout B.
      A new method for valuing health: directly eliciting personal utility functions.
      After completing the EQ-5D-5L and EQ VAS, the first adaptation asked respondents to fill in the composites split into 4 individual subdomains (the first 3 domains were not used). The second modification changed the recall period of the EQ-5D-5L and EQ VAS to the time when respondents felt the worst because of their health (ie, worst recalled health), and the third one combined the first 2 modifications; thus, the recall period was also changed for the 4 individual subdomains. The inclusion of worst recalled health was expected to increase variability in responses in the composite domains. The sequencing of EQ-5D-5L and EQ VAS modifications within the questionnaire is presented in Appendix 1 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012.
      Rather than imposing our preconceptions about the meaning of discomfort, we opted to explore the respondents’ own interpretations of discomfort by open-ended questions. To avoid priming participants to focus on discomfort when filling in the EQ-5D-5L, all open-ended questions were asked after completing the EQ-5D-5L and the individual subdomains. The exact wording and placement of these questions are presented in Appendix 1 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012. All respondents were asked 2 open-ended questions about the theoretical meaning of discomfort. Only those respondents who experienced any problems in the individual discomfort subdomain for current health or worst recalled health received open-ended questions about their experienced discomfort, respectively. Those reporting any pain in the individual pain subdomain were asked about their pain to further understand what the PD domain captures (eg, type, duration, and possible cause of pain).
      Respondents were asked to identify their age, gender, level of education, place of living, geographical region, employment status, marital status, self-perceived health status (excellent to poor scale), and the presence of any chronic conditions or chronic consequences of acute conditions.

      Data Analysis

      Quantitative and qualitative data were first analyzed separately. Then, respondents’ answers to the open-ended questions were transformed into numeric counts using content analysis and merged with the quantitative data for the mixed-methods analyses.

      Quantitative analyses

      Descriptive statistics were used to provide an overview of the sample’s demographic and health-related characteristics. All analyses on EQ-5D-5L responses were first conducted for current health and then were repeated for worst recalled health. A series of cross-tabulations were created to explore the distribution of responses across the individual subdomains and composites. Chi-square test was used to compare (1) the proportion of respondents reporting problems in the composite versus problems in individual subdomains and (2) the proportion of respondents under-reporting problems for PD and AD (ie, did not score the composite but did score an individual subdomain).
      To explore variations in response behavior for each composite, we restricted the main analysis sample to those respondents who reported any health problems either in the composite or in any of the 2 individual subdomains. For respondents scoring at different levels in the 2 individual subdomains, we examined the presence of any ordering effect (ie, if the responses in the composite were mainly driven by the first mentioned [pain or anxiety] or the second mentioned [discomfort or depression] subdomain). Two sensitivity analyses were conducted to evaluate the impact of potential inconsistencies in the data (Appendix 2 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012). Correlations among EQ-5D-5L domains, the 4 individual subdomains, and EQ VAS were analyzed by Spearman’s rank-order correlations. Correlation coefficients were interpreted as very weak (< 0.20), weak (0.20-0.39), moderate (0.40-0.59), strong (0.60-0.79), and very strong (≥ 0.80).
      • Evans J.D.
      Straightforward Statistics for the Behavioral Sciences.
      Statistical analyses were performed using Stata 14 (StataCorp. 2015, College Station, TX).

      Qualitative content analysis

      Text responses on the 4 open-ended questions about discomfort were analyzed using inductive content analysis in Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA).
      • Elo S.
      • Kyngäs H.
      The qualitative content analysis process.
      Four separate analyses were conducted: (1) theoretical examples for discomfort, (2) theoretical descriptions of discomfort, (3) examples for discomfort experienced on the day of the survey, and (4) examples for discomfort for worst recalled health.
      For the first, third, and fourth parts of the analysis, the unit of analysis was a single word or a few words, whereas in the second analysis a clause or a sentence. First, we read all the 1700 respondents’ text answers multiple times. Then, we constructed an initial coding framework using open coding. For coding examples for discomfort, subcategories of related content were grouped together as generic categories that were further grouped to form main categories. Descriptions of discomfort were categorized based on a shared meaning. Few examples that did not fit into any of these main categories were placed in an “other” category. Examples and quotes were selected to support each subcategory. Coding was performed by one researcher, and any uncertainties were resolved through discussion with the other researcher.

      Mixed-methods analyses

      Three sets of mixed-methods analyses were conducted. First, the sample was divided into subgroups based on how participants interpreted discomfort (considering their responses to all open-ended questions), and then we examined whether respondents’ demographic characteristics and health status were associated with belonging to any of these particular groups. Second, we analyzed whether there was an association between the type of discomfort reported for current health and worst recalled health (considering responses on each open-ended question separately) and self-reporting it in the PD composite. Finally, we explored whether there was an association between respondents’ overall interpretations of discomfort (considering their responses to all open-ended questions) and self-reporting discomfort in the individual subdomain. Proportions were compared using chi-square tests.

      Results

      Characteristics of the Study Population

      Overall, 2502 individuals initiated the survey, 423 of whom (17%) were screened out because they did not consent and 379 (15%) did not finish the questionnaire. A total of 1700 respondents (68%) completed the survey. The sample was roughly representative of the Hungarian adult general population for age, gender, employment and marital status, place of residence, geographical region, and the presence of any chronic illness (Table 1).
      Table 1Characteristics of the study population.
      VariablesGeneral population reference (%)
      Hungarian Central Statistical Office (KSH), Microcensus 2016.
      Total sampleConsidered discomfort “pain”Considered discomfort “nonpain, physical”Considered discomfort “psychological”P value (χ2 test)
      Comparison across the 3 groups.
      n%n%n%n%
      Total
      Sixty-nine respondents are not included in these groups who indicated “don’t know” or “refused to answer” in all open-ended questions or their responses did not allow to include them in any of the 3 groups.
      1001700100112266396231137-
      Gender
       Female539575668661192485549< .001
       Male477434443639204525851
      Age (years)
       18-546110546267360255648273.016
       55+396463844940141363127
      Highest level of education
       Primary or secondary7911506874967267678172.566
       College/university215503237333129333228
      Marital status
       Married/domestic partnership5910786371964254646658.479
       Single/widowed/divorced/other416223740336142364742
      Employment status
       Employed538655158052189485650.388
       Not employed478354954248207525750
      Place of residence
       Capital18380222562386222320.939
       Other town538204853348196495549
       Village305002933330114293531
      Geographical region
       Central305723438334138353027.052
       Western304932933130140355750
       Eastern406353740836118302623
      Self-perceived health
      Reference population: Rencz et al.30
       Excellent/very good/good8312227277769299769181.005
       Fair/poor17478283453197242219
      History of chronic illness
      Don’t know or refused to answer n = 144.
      ,
      Hungarian Central Statistical Office (KSH), Health at a Glance 2019.
       Yes4811467481378256715151< .001
       No524102623322103294949
      Hungarian Central Statistical Office (KSH), Microcensus 2016.
      Comparison across the 3 groups.
      Sixty-nine respondents are not included in these groups who indicated “don’t know” or “refused to answer” in all open-ended questions or their responses did not allow to include them in any of the 3 groups.
      § Reference population: Rencz et al.
      • Rencz F.
      • Brodszky V.
      • Gulácsi L.
      • et al.
      Parallel valuation of the EQ-5D-3L and EQ-5D-5L by time trade-off in Hungary.
      ǁ Don’t know or refused to answer n = 144.
      Hungarian Central Statistical Office (KSH), Health at a Glance 2019.

      Self-Reported Health in the Composites and Individual Subdomains

      The distribution of responses across the composites and individual subdomains is presented in Table 2. In self-reporting current health, 44% of respondents reported problems in the composite PD, whereas 53% reported problems in either of the 2 individual subdomains (P < .001). In contrast, there was only a very small difference between self-reporting problems in the composite AD (34%) and in its 2 individual subdomains (36%) (P = .350). For worst recalled health, 78% reported any problems in the composite PD, whereas 87% separately and 52% reported problems in the AD composite and 60% separately (P < .001 for both).
      Table 2Self-reported health in the EQ-5D-5L composites and individual subdomains.
      ResponsesPain/discomfortAnxiety/depression
      Current healthWorst recalled healthCurrent healthWorst recalled health
      n%n%n%n%
      Self-reported responses (A)1700100170010017001001700100
      Any problems in the composite (B, proportion B/A)756441333785853489252
      Any problems in the 1st or 2nd subdomain (C, proportion C/A)
      1st subdomain, pain or anxiety; 2nd subdomain, discomfort or depression.
      9085314718761136102160
      Difference composite vs subdomains (B vs C, χ2 test P value)< .001< .001.350< .001
      Any problems in both the 1st and 2nd subdomains (D, proportion D/A)424251271753572173443
      Any problems in the 1st subdomain (E, proportion E/A)728431382815633398758
      Not reported in the composite (F, proportion F/E)120161118761314815
      Any problems in the 2nd subdomain (G, proportion G/A)604361360804052476845
      Not reported in the composite (H, proportion H/G)1302212792468111
      Under-reporting 1st vs 2nd subdomain (F vs H, χ2 test P value).019.224< .001.006
      Response behavior in the composite (I)96810014831006741001060100
      Uniform responses (J, proportion J/I)20521448302133230529
      Nonuniform responses (K, proportion K/I)763791035704616875571
      Most severe problem (L, proportion L/I)32834408281993026825
      Most severe problem (1st subdomain) (M, proportion M/L)24274297731648224391
      Most severe problem (2nd subdomain) (N, proportion N/L)8626111273518259
      Least severe problem (O, proportion O/I)22723240161091620820
      Least severe problem (1st subdomain) (P, proportion P/O)122541365723213014
      Least severe problem (2nd subdomain) (Q, proportion Q/O)1054610443867917886
      Average (R, proportion R/I)434775132565
      Synergistic (S, proportion S/I)
      Problems in the 2 subdomains strengthen each other.
      404634375505
      Inconsistent (T, proportion T/I)
      Paired responses (composite vs the nearest individual subdomain) not belonging to any other nonuniform category and differing at least one level were defined as “inconsistent,” with a size of inconsistency ranging from 1 to 4, for example, moderate pain, slight discomfort, and no problems in the composite (one-level inconsistency).
      12513247171031517316
      Average size of inconsistency
      Paired responses (composite vs the nearest individual subdomain) not belonging to any other nonuniform category and differing at least one level were defined as “inconsistent,” with a size of inconsistency ranging from 1 to 4, for example, moderate pain, slight discomfort, and no problems in the composite (one-level inconsistency).
      1.22-1.53-1.23-1.43-
      Ordering effect (U)
      Computed only for those respondents where this could be unambiguously assessed.
      613100804100343100567100
      Composite response driven by the 1st subdomain (V, proportion V/U)
      1st subdomain, pain or anxiety; 2nd subdomain, discomfort or depression.
      40466507632096131255
      Composite response driven by the 2nd subdomain (W, proportion W/U)
      1st subdomain, pain or anxiety; 2nd subdomain, discomfort or depression.
      20934297371343925545
      EQ-5D-5L indicates 5-level version of EQ-5D.
      1st subdomain, pain or anxiety; 2nd subdomain, discomfort or depression.
      Problems in the 2 subdomains strengthen each other.
      Paired responses (composite vs the nearest individual subdomain) not belonging to any other nonuniform category and differing at least one level were defined as “inconsistent,” with a size of inconsistency ranging from 1 to 4, for example, moderate pain, slight discomfort, and no problems in the composite (one-level inconsistency).
      § Computed only for those respondents where this could be unambiguously assessed.
      In the composite, more respondents under-reported discomfort than pain (22% vs 16%, P = .019) and anxiety than depression (13% vs 6%, P < .001). In contrast, for worst recalled health, almost the same proportion of participants under-reported discomfort as pain (9% and 8%, P = .224), whereas again more respondents under-reported anxiety than depression (15% vs 11%, P = .006). The sensitivity analyses confirmed the robustness of these findings (Appendix 2 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012).

      Analysis of Response Behaviors

      When describing current health, 968 and 674 subjects reported any problems in either the individual or corresponding composite domains of PD and AD, respectively. Six different response behaviors were identified that are summarized and defined in Box 1. Overall, 21% and 32% of the participants reported the same level of problems in both individual subdomains and the corresponding composites of PD and AD, respectively (“uniform responses”) (Table 2). Among respondents who scored at different levels on the 2 individual subdomains (“nonuniform responses”), for PD 34% and for AD 30% used the composite to report their most severe problem across the subdomains. These proportions were lower for those scoring according to their least severe problem (PD 23% and AD 16%). A minority of respondents reported the average rating across the 2 subdomains in the composite (PD 4% and AD 2%). Some respondents considered that problems of the subdomains could strengthen each other (“synergistic”) in the composite (PD 4% and AD 5%). Proportion of inconsistent responses was 13% for PD and, slightly higher, 15% for AD. The average size of inconsistency in terms of difference in levels was generally low (1.22 for PD and 1.23 for AD). In most respondents scoring different levels in the 2 individual subdomains, responses in the composite were driven by their problems in the first subdomain (PD 66% and AD 61%). The sensitivity analyses reduced the proportion of inconsistent responses, but otherwise showed little change in results (Appendix 2 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012).
      Response behaviors in the composite domains of EQ-5D.
      • 1.
        “Uniform”: respondent reports the same level of problems in both individual subdomains and the composite, for example, slight pain, slight discomfort, slight problems in the composite;
      • 2.
        “Nonuniform”: respondent reports different levels of problems in the 2 individual subdomains or same level of problems but a different level in the composite;
        • A
          “The most severe problem”: respondent reports the most severe problem across the subdomains in the composite, for example, no pain, slight discomfort and slight problems in the composite;
        • B
          “The least severe problem”: respondent reports the least severe problem across the subdomains in the composite, for example, no pain, slight discomfort and no problems in the composite;
        • C
          “Average”: respondent reports the average of problems across the subdomains in the composite, for example, no pain, moderate discomfort and slight problems in the composite;
        • D
          “Synergistic”: respondent reports more severe problems in the composite than in the subdomains, for example, slight pain, slight discomfort and moderate problems in the composite;
        • E
          “Inconsistent”: paired responses (composite vs the nearest individual subdomain) not belonging to any other nonuniform category and differing at least one level were defined as “inconsistent,” with a size of inconsistency ranging from 1 to 4, for example, moderate pain, slight discomfort and no problems in the composite (one-level inconsistency).

      Self-Reported Pain

      In total, 728 respondents reported pain in the individual subdomain, 91% of whom had physical pain (Appendix 3 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012). Notably, 37% of respondents with any pain reported to have psychological pain (9% without coexisting physical pain).

      Content Analysis of Examples for Discomfort

      Respondents provided overall 10 428 examples for discomfort (6193 theoretical, 1196 for current health, and 3069 for worst recalled health). Responses were condensed into 108 subcategories and 16 categories. These were then consolidated under 3 main categories: physical discomfort due to pain, nonpain physical discomfort, and psychological discomfort (Table 3).
      Table 3Content analysis of the examples for discomfort.
      CategoriesExamplesTheoretical (n = 1700)Current health (n = 604)Worst recalled health (n = 1360)Total responses (n = 3664)
      n%n%n%n%
      (A) Physical discomfort due to pain92354.323739.254339.9170346.5
       (1) Pain
      Pain (in general)Pain, something hurts, physical pain40423.8447.327420.172219.7
      HeadacheHeadache, migraine50029.49315.4896.568218.6
      Musculoskeletal painLow back pain, muscle pain, joint pain20111.88814.61047.639310.7
      Abdominal painStomach ache, abdominal pain865.1213.5503.71574.3
      ToothacheToothache261.5101.7181.3541.5
      Pain in other body partEar pain, eye pain130.881.3261.9471.3
      Chest painChest pain, tightness in the chest60.481.3272.0411.1
      Sore throatSore throat70.450.8120.9240.7
      Kidney painKidney stones, kidney stone attack00.000.0151.1150.4
      (B) Nonpain physical discomfort114767.526644.089365.7230662.9
       (2) Illness1659.7315.118313.537910.3
      Illness (in general)Illness, being sick1378.1264.317112.63349.1
      Health problemHealth problem402.450.840.3491.3
      Start of illnessAs if something is hiding in me, sickening for something, the start of a disease321.900.000.0320.9
       (3) Homeostasis-related1207.1162.61168.52526.9
      FeverFever, low-grade fever684.030.5967.11674.6
      ShiveringShivering181.100.0110.8290.8
      Cold/warmTo be cold, to be warm140.871.240.3250.7
      SweatingSweating60.400.0151.1210.6
      Hunger or thirstHunger, thirst70.440.710.1120.3
      Hot flushHot flush80.510.210.1100.3
       (4) Gastrointestinal38222.5274.516111.857015.6
      NauseaNausea23513.8101.7826.03278.9
      Loss of appetiteLoss of appetite925.461.0483.51464.0
      Digestive problems (unspecified)Bad or upset stomach, digestive problems523.171.2100.7691.9
      VomitingVomiting30.200.0423.1451.2
      Diarrhea or constipationDiarrhea, constipation191.100.0251.8441.2
      BloatednessBloatedness221.350.810.1280.8
      HeartburnHeartburn, acid reflux symptoms110.600.010.1120.3
      OvereatingEating too much, overeating80.510.230.2120.3
       (5) Respiratory824.8244.0503.71564.3
      Common coldHaving a common cold, runny nose462.7111.8100.7671.8
      Breathing problemsBreathlessness, rapid breathing221.350.8332.4601.6
      Coughing or sneezingCoughing, sneezing130.871.2120.9320.9
      Nasal congestionStuffy nose80.530.500.0110.3
       (6) Neurological32018.8427.01017.446312.6
      DizzinessDizziness25114.8284.6725.33519.6
      Concentration or memory problemsDifficulty concentrating, deconcentrated533.130.550.4611.7
      NumbingNumbing140.830.5151.1320.9
      TremblingTrembling, shaking hands130.840.7120.9290.8
      Ringing in the earsRinging in the ears, tingling ears40.250.820.1110.3
      DisorientationDisorientated, confused, dopey90.510.210.1110.3
      Feeling imbalancedFeeling imbalanced70.400.010.180.2
       (7) Cardiovascular432.550.8282.1762.1
      Blood pressureLow/high/fluctuating blood pressure362.140.7161.2561.5
      ArrhythmiaPalpitation, arrhythmia100.610.2161.2270.7
       (8) Tiredness68440.214724.317813.1100927.5
      TirednessFatigue, tiredness47828.19115.1624.663117.2
      WeaknessWeak18811.1193.11007.43078.4
      Sleep problemsDifficulty falling asleep, insomnia885.2315.1282.11474.0
      VitalityLack of vitality, lack of energy, laziness, slowness895.2152.5110.81153.1
      SleepinessSleepiness, drowsiness613.6223.6100.7932.5
       (9) Women's health100.661.0322.4481.3
      Pregnancy-relatedPregnancy, morning sickness, in labor, after labor, miscarriage, cesarean section10.120.3292.1320.9
      Menstruation-relatedMenstruation, premenstrual syndrome90.540.750.4180.5
       (10) Other physical symptoms1156.8201.2442.61794.9
      Physical malaiseMalaise, light-headedness, fainting342.010.2141.0491.3
      Sensory symptomsVision problems, decreased hearing100.610.2131.0240.7
      Sensitivity to weather changeSensitivity to weather change181.130.500.0210.6
      Skin symptomsSensitive skin, itch, having a skin rash70.400.050.4120.3
      Other physical sensationsPounding in the head, itching throat, hangover, hygiene problems, noise513.0152.5201.5862.3
       (11) Other physical problems915.4201.219311.43048.3
      Mobility problemsDifficulty moving, disabled, not able to walk462.7122.013710.11955.3
      Problems with usual activitiesProblems with work or leisure402.461.0231.7691.9
      Confined to bedBedridden, confined to bed10.110.2423.1441.2
      Problems with self-careProblems with/not able to take care oneself20.110.2251.8280.8
      Problems with physical fitnessLack of fitness, sedentary lifestyle, gaining weight, being overweight110.600.030.2140.4
      Problems related to agingGetting old60.410.210.180.2
       (12) Specific diseases342.0244.018513.62436.6
      Herniated diskSlipped disk, spinal disk herniation10.120.3211.5240.7
      CancerCancer, tumor, malignant disease00.010.2221.6230.6
      InfluenzaInfluenza70.420.360.4150.4
      PneumoniaPneumonia00.010.2120.9130.4
      DiabetesDiabetes60.420.330.2110.3
      Heart attackHeart attack00.000.0100.7100.3
      StrokeStroke00.010.280.690.2
      Other diseaseAllergy, asthma, COPD, hemorrhoids, ulcer241.4162.61178.61574.3
       (13) Injury, accidentInjury, accident40.220.31299.51353.7
       (14) Medical interventions100.640.718913.92035.5
      Surgery/anesthesia(Post-)surgery, (post-)anesthesia70.420.315811.61674.6
      Hospital careStaying in hospital, intensive care00.000.0302.2300.8
      MedicationsSide effects of treatments30.220.390.7140.4
      (C) Psychological discomfort125974.133956.170051.5229862.7
       (15) Mental problems, emotions or feelings1 15167.725141.642931.5183150.0
      Mood problemsListless, in a bad mood, downhearted70641.58714.4664.985923.4
      AnxietyAnxious17910.5396.5624.62807.6
      NervousnessNervous, irritated, irritable, tense1669.8274.5272.02206.0
      SadnessSad1277.5122.0100.71494.1
      DepressionDepressed824.861.0382.81263.4
      FearFear, fear of tomorrow/the future573.471.2493.61133.1
      StressStress663.9233.8100.7992.7
      WorryWorrying573.4233.8181.3982.7
      Tired of lifeApathy, lethargy, indifference, tired of life653.8111.8131.0892.4
      InsecurityInsecure472.861.0151.1681.9
      Mental problemsMental pain, mental fatigue321.9111.8221.6651.8
      Negative thoughtsNegative thoughts472.871.250.4591.6
      HelplessnessHelpless140.820.3433.2591.6
      IrritabilityIrritated513.040.730.2581.6
      VulnerabilityVulnerable10.130.5513.8551.5
      LonelinessLoneliness, neglect, lack of love281.6122.0141.0541.5
      Lack of prospectsLack of prospects, aimlessness271.6101.7171.3541.5
      Reference to deathThoughts related to death80.510.2342.5431.2
      Lack of interestUninterested392.310.220.1421.1
      GriefGrief80.550.8292.1421.1
      BoredomBored331.940.740.3411.1
      Motivation problemsMotivation problems, lack of motivation352.130.510.1391.1
      FretfulFretful251.530.500.0280.8
      PanicPanic attack, panic disorder30.220.3191.4240.7
      FrustrationFrustrated140.810.230.2180.5
      Reference to cryingCrying, being on the verge of crying90.530.560.4180.5
      Other specific mental health conditionObsessions, paranoia, alcohol/drug withdrawal symptoms, bipolar disorder30.230.5100.7160.4
      UnhappinessUnhappy110.610.210.1130.4
      ImpatienceImpatient90.510.200.0100.3
      FailureFailure80.500.000.080.2
      Other feelingsDisappointed, hopeless, desperate, missing a loved one, anger, hatred, disgust, guilt1478.6406.6624.62496.8
       (16) Source of psychological discomfort19911.714423.8594.340211.0
      Weather, time of dayWeather, time of day, seasons663.9528.630.21213.3
      COVID-19Corona, quarantine, lockdown, restrictions402.4609.920.11022.8
      Work(place) problemsWorkplace problems, one’s boss291.7203.3221.6711.9
      Financial problemsLack of money, unemployment392.3183.070.5641.7
      Family problemsFamily problems291.7162.6131.0581.6
      Private life problemsFight, divorce, cheating, bad company231.481.3171.3481.3
      Politics and societyNews, politics, healthcare system372.291.510.1471.3
      Bad newsGetting bad news90.510.210.1110.3
      (D) Other1388.1335.51229.02938.0
       Refused to answer or nonsensical response-, xx, “pass,” “none,” “I am not afraid of anything,” “0”543.2223.6483.51243.4
       Don't know“I don’t know”, “I don’t know this notion,” “I’ve never had it”171.040.7695.1902.5
       ParaphraseDiscomfort, lack of comfort, uncomfortable feeling/sensation673.971.250.4792.2
      Total6163-1196-3069-10 428-
      COPD indicates chronic obstructive pulmonary disease.
      Overall, 66% of respondents referred to pain as a form of discomfort, of which unspecified pain was the most common example provided (35%), followed by headache (32%) and musculoskeletal pain (18%). More than four-fifth of participants (81%) mentioned forms of nonpain physical discomfort including tiredness (31%), illness (in general) (18%), dizziness (17%), nausea (16%), and weakness (15%). Many respondents (78%) referred to psychological discomfort, of which mood problems (43%), anxiety (15%), nervousness (11%), sadness (8%), and depression (7%) were the most prevalent.
      When interpreting discomfort theoretically, 204 respondents (12%) used an adjective to describe the level of discomfort in a total of 239 examples (Appendix 4 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012). Among these, 101 examples provided by 71 respondents (4%) described discomfort as “mild pain.” Other examples included “mild headache,” “small dizziness,” and “strong tiredness.”

      Content Analysis of Descriptions for Discomfort

      Overall, 595 respondents (35%) provided a description about the theoretical meaning of discomfort. These responses were synthesized in 20 subcategories and the following 5 categories: (1) the relation between pain and discomfort (ie, pain vs not pain), (2) source of discomfort, (3) time or duration of discomfort, (4) feeling/sensation of discomfort, and (5) other (Table 4).
      Table 4Content analysis of the descriptions for discomfort.
      No.CategorySubcategoryn%Example quote
      1The relation of discomfort and pain
      When considering both descriptions and examples provided, there were 14 respondents who responded that discomfort can be both pain and “not pain.”
      Pain153I have discomfort due to pain. When someone is in pain, they have discomfort at any level of pain.
      When something hurts that bad that I have to take medications.
      When even sounds are painful.
      Not pain
      When considering both descriptions and examples provided, there were 14 respondents who responded that discomfort can be both pain and “not pain.”
      224An unnatural, tight, pressing or itchy sensation in the body which is not yet pain.
      It is a sensation when one cannot precisely tell, but feels unwell. Pain does not belong to this notion.
      When I have no particular pain, but still I don’t feel well.
      2Source of discomfortRather physical132It is mostly of physical origin.
      To me, discomfort covers bodily symptoms.
      Rather psychological153Who has such a problem [discomfort], they have problems with their nerves.
      It is a mental state. It is like having a cold, once you’ve got it, you don’t have the mood to do anything, you are fretful, you have discomfort.
      It is a mood characterized by dissatisfaction, restlessness, depression and anxiety. It is an unpleasant or uncomfortable feeling.
      Physical and psychological7312An unpleasant physical or mental condition or state of mind, or the combination of these.
      It is a mind-body imbalance.
      It is the coexistence of larger emotional and smaller physical problems.
      Societal/environmental71Nevertheless, discomfort might arise from political or public life reasons. I, myself, am very sensitive to the (current bad) state of the world or our country too.
      I don’t feel myself. This originates from societal, workplace and private life problems.
      No (medical) or unknown reason/ inexplicable468Inexplicable symptoms that cannot be diagnosed by routine medical examinations.
      An inexplicable, unpleasant inner sensation.
      An insecure feeling of unknown source.
      3Time or duration of discomfortShort term (moments or hours)41In my opinion, [by discomfort] we mean one’s momentary mood and health status.
      Reference to the morning or getting up366In the morning I don’t want to open my eyes, I barely want to get up. I don’t want to start the day.
      A bad day203… on that day everything goes wrong.
      I am lost all day, I am tired despite sleeping a lot and not in the mood to get out of bed, even the smallest thing that is not how I want it to be annoys me.
      Long term/constant386A prolonged, unpleasant feeling, weakness.
      4Feeling of discomfortNot feeling well10718I am not feeling well.
      Not feeling good in one's skin458I am not feeling good in my skin.
      No mood to do anything15626One is not in the mood to do anything.
      Not finding one’s place102I cannot find my place.
      Something is wrong427I cannot put it into words, something is wrong.
      … when I don’t know what exactly the problem is, because I have no pain, I just feel that something is not good.
      I don’t feel 100% today.
      Everything is wrong346When nothing is good.
      Everything annoys me, I am tired of everything, I don’t want to live, I find myself hopeless and helpless.
      Not as usual356Somehow I don’t feel like I used to, I am not well.
      I have a feeling/sensation that I don’t experience on normal days. For example, when I am ill, I don’t feel well, I easily get tired…
      Withdrawal from social relationships427When one does not want to do anything or see anyone.
      5OtherAny other description that does not fit to the categories above142When one shares their discomfort with others, they [others] will also develop discomfort.
      It is indefinable that is why we use the word 'discomfort' for it.
      Total595100-
      When considering both descriptions and examples provided, there were 14 respondents who responded that discomfort can be both pain and “not pain.”

      Determinants of What is Considered Discomfort

      Based on the examples and descriptions provided on all open-ended questions about discomfort, respondents were classified into 3 mutually exclusive groups. The first group comprised respondents who mentioned pain with or without other forms of discomfort (n = 1222, 66%), the second group included respondents who mentioned other forms of physical discomfort than pain with or without any psychological discomfort (n = 369, 29%), and the third group consisted of those respondents who mentioned only psychological discomfort (n = 113, 7%). Women, respondents aged 55 years or older, and those in worse health status or having chronic conditions tended to consider discomfort as pain more than others (Table 1). Level of education, marital and employment status, region, and place of residence had no significant impact on what meaning participants attributed to discomfort.

      The Type of Discomfort and Self-Reported Problems in the PD Composite Domain

      Respondents with nonpain physical discomfort (20%) and psychological discomfort (30%) more frequently under-reported their current health problems in the composite PD domain than respondents having discomfort due to pain (16%) (P = .002). This trend was also detected for worst recalled health; nevertheless, the difference was smaller and insignificant (Appendix 5 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012).

      Respondents’ Overall Interpretations of Discomfort and Self-Reported Problems in the Individual Subdomain

      For current health, no significant difference was observed in the overall interpretation of discomfort between respondents with and without discomfort (Appendix 6 in Supplemental Materials found at https://doi.org/10.1016/j.jval.2022.06.012). Nevertheless, respondents reporting discomfort for their worst recalled health more frequently considered discomfort as pain and respondents without discomfort more commonly considered discomfort as psychological (P < .001).

      Correlations Between the Composites and Individual Subdomains and EQ VAS

      Correlation between the individual pain and discomfort subdomains was moderate (rs = 0.47) (Table 5). Pain showed a stronger correlation with the composite PD than discomfort (rs = 0.72 vs rs = 0.56). Discomfort was more strongly correlated with anxiety than with pain (rs = 0.59 vs rs = 0.47). Among all (sub)domains, the individual discomfort subdomain demonstrated the strongest correlation with EQ VAS (rs = −0.51). With few exceptions, correlations across domains strengthened when participants reported worst recalled health.
      Table 5Spearman’s correlations among the 5 domains, 4 individual subdomains, and EQ VAS.
      DomainsWorst recalled health (n = 1700)—above the diagonal
      EQ VASMOSCUAPDADPainDiscAnxDep
      Current health (n = 1700)—below the diagonalEQ VAS−0.40−0.45−0.56−0.60−0.47−0.50−0.59−0.46−0.40
      MO−0.410.660.590.560.330.470.330.260.22
      SC−0.290.390.720.580.350.450.380.290.22
      UA−0.420.510.440.710.440.530.500.360.31
      PD−0.550.500.360.520.490.720.610.420.35
      AD−0.390.210.200.330.440.340.550.800.74
      Pain−0.490.500.350.470.720.340.620.410.30
      Disc−0.510.250.260.380.560.600.470.650.54
      Anx−0.390.200.190.320.400.790.330.590.78
      Dep−0.370.200.210.310.380.740.320.540.69
      Note. All correlation coefficients were statistically significant (P < .05).
      AD indicates anxiety/depression; Anx, anxiety; Dep, depression; Disc, discomfort; MO, mobility; PD, pain/discomfort; SC, self-care; UA, usual activities; VAS, visual analogue scale.

      Discussion

      In this study, we examined how respondents use the PD and AD composite domains in the EQ-5D to self-report own health. Our findings highlight several problems with the current form of the composite domains and wording of discomfort in the EQ-5D, implications of which are outlined in the discussion.
      The first problem our study revealed is that respondents under- and inconsistently report their health using the composite domains. These results coincide with those of 2 previous studies that found more participants reporting problems in the individual subdomains than in the composite among the UK general public.
      • McDonald R.
      • Mullett T.L.
      • Tsuchiya A.
      Understanding the composite dimensions of the EQ-5D: an experimental approach.
      ,
      • Tsuchiya A.
      • Bansback N.
      • Hole A.R.
      • Mulhern B.
      Manipulating the 5 dimensions of the EuroQol instrument: the effects on self-reporting actual health and valuing hypothetical health states [published correction appears in Med Decis Making. 2020;40(2):115].
      This may be symptomatic of confusion among respondents about how the response levels of the composites ought to be interpreted or of not accurately reading the response levels (eg, not taking the conjunction “or” into account).
      Second, both for PD and AD, we detected signs of systematic order effects: mainly problems with the first subdomain determined responses in the composite in respondents who scored differently in the 2 separate concepts. In support to these findings, the first subdomains correlated more strongly with the composites. The primacy of the first subdomain was more pronounced for PD than for AD. It may be possible that, in the current form of the composite, the effects of the higher relative importance of pain in health than discomfort and the order in which it is mentioned add up. These results suggest the need for further studies testing how the order of the 2 subdomains influences respondents’ answers, for example, through subdomain randomization.
      The third issue is the ambiguous meaning of discomfort that could lead to differential item functioning in the composite PD domain. The variables identified to systematically influence interpretations were age, gender, and health status. This might introduce bias into direct comparisons of self-reported health across subgroups given that differences stemming from differential item functioning may be attributed to differences in underlying health.
      • Knott R.J.
      • Black N.
      • Hollingsworth B.
      • Lorgelly P.K.
      Response-scale heterogeneity in the EQ-5D.
      Supporting this, 2 previous studies found differential item functioning for the PD domain of EQ-5D.
      • Smith A.B.
      • Cocks K.
      • Parry D.
      • Taylor M.
      A differential item functioning analysis of the EQ-5D in Cancer.
      ,
      • Whynes D.K.
      • Sprigg N.
      • Selby J.
      • Berge E.
      Bath PM; ENOS Investigators
      Testing for differential item functioning within the EQ-5D.
      The more than 100 different forms of discomfort identified imply that the EQ-5D may be able to capture a broad spectrum of forms of discomfort. Nevertheless, previous clinical studies concluded that the EQ-5D was less efficient in capturing several symptoms that this study uncovered as forms of discomfort, including respiratory symptoms (eg, shortness of breath and coughing) or gastrointestinal symptoms (eg, nausea, constipation, and diarrhea).
      • Spronk I.
      • Bonsel G.J.
      • Polinder S.
      • van Baar M.E.
      • Janssen M.F.
      • Haagsma J.A.
      Exploring the relation between the EQ-5D-5L pain/discomfort and pain and itching in a sample of burn patients.
      ,
      • Spronk I.
      • Bonsel G.J.
      • Polinder S.
      • van Baar M.E.
      • Janssen M.F.
      • Haagsma J.A.
      The added value of extending the EQ-5D-5L with an itching item for the assessment of health-related quality of life of burn patients: an explorative study.
      • Swinburn P.
      • Lloyd A.
      • Boye K.S.
      • Edson-Heredia E.
      • Bowman L.
      • Janssen B.
      Development of a disease-specific version of the EQ-5D-5L for use in patients suffering from psoriasis: lessons learned from a feasibility study in the UK.
      • van Dongen-Leunis A.
      • Redekop W.K.
      • Uyl-de Groot C.A.
      Which questionnaire should be used to measure quality-of-life utilities in patients with acute leukemia? An evaluation of the validity and interpretability of the EQ-5D-5L and preference-based questionnaires derived from the EORTC QLQ-C30.
      • Hoogendoorn M.
      • Oppe M.
      • Boland M.R.S.
      • Goossens L.M.A.
      • Stolk E.A.
      • Rutten-van Mölken M.
      Exploring the impact of adding a respiratory dimension to the EQ-5D-5L.
      In addition, many of the existing EQ-5D “bolt-ons” (additional domains to the EQ-5D) target these specific areas of discomfort, such as breathlessness or tiredness.
      • Geraerds A.J.L.M.
      • Bonsel G.J.
      • Janssen M.F.
      • Finch A.P.
      • Polinder S.
      • Haagsma J.A.
      Methods used to identify, test, and assess impact on preferences of bolt-ons: a systematic review.
      A direction for future improvement could be to provide supportive examples for discomfort in the composite descriptor, similarly to usual activities.
      Although this study explored the PD and AD composite domains and possible meanings of discomfort in the context of self-reporting own health, valuation implications of our findings remain unclear; for example, which of these examples for discomfort (if any) are considered by respondents when valuing health states. Earlier results indicated that, in valuation, respondents interpret the composite PD as pain and AD is interpreted to lie between anxiety and depression.
      • McDonald R.
      • Mullett T.L.
      • Tsuchiya A.
      Understanding the composite dimensions of the EQ-5D: an experimental approach.
      Furthermore, problems in the pain subdomain alone are considered worse than the same level of discomfort, and similar to this, depression is perceived worse than anxiety.
      • McDonald R.
      • Mullett T.L.
      • Tsuchiya A.
      Understanding the composite dimensions of the EQ-5D: an experimental approach.
      If respondents self-report discomfort in the composite but in valuation, PD is mainly interpreted as pain, then disutilities of health states actually containing discomfort and without pain might be overestimated. Moreover, for both PD and AD, the order effects in self-reporting health might also mismatch with the values attached to the composites.
      The fourth problem our study revealed concerns the independence of the questionnaire domains. Our findings indicate that discomfort, to some extent, covers mental functioning and this could imply an overlap between PD and AD. Nevertheless, this observation may be subject to considerable variations across different languages. The Hungarian translation of discomfort is “rossz közérzet,” whereby “rossz” means “bad” and “közérzet” refers to general state, sensation, or feeling. Furthermore, 37% of respondents reported to experience psychological pain. Therefore, the current form of the composite PD somewhat contradicts the developers’ intentions that were, in fact, to capture physical pain and other forms of physical discomfort.
      • Gudex C.
      The descriptive system of the EuroQol instrument.
      A further refinement of the PD domain could consist of focusing solely on physical symptoms; adding the term “bodily” or “physical” to the domain heading could be a solution. Interestingly, the German version of the EQ-5D already uses the expression of “körperliche Beschwerden” (bodily discomfort). Harmonization of discomfort in other languages has also been conducted; for instance, the Dutch EQ-5D-3L used the word “klachten” (complaint) that was revised as “ongemak” (discomfort) in the EQ-5D-5L. These findings provide supportive evidence for the recently developed EQ Health and Wellbeing
      • Brazier J.E.
      • Peasgood T.
      • Mukuria C.
      • et al.
      The EQ-HWB: overview of the development of a measure of health and well-being and key results.
      instrument that asks about physical pain and physical discomfort in 2 separate questions, with the latter being supplemented by the examples of “feeling sick, breathless, itching (not including pain).”
      There are some limitations to our study. First, the study population included respondents from the general public, and it might have missed certain forms of discomfort that are relevant only to patients with specific health conditions. Second, the survey was conducted in Hungarian that might restrict the generalizability of the interpretations of discomfort to other populations. Third, positioning the individual subdomains after the composites might have enhanced the under-reporting of problems in the composites, as being repeatedly exposed to the same item might lead respondents to reconsider their previous response. Furthermore, questions about pain and discomfort preceded the EQ-5D and individual subdomains for worst recalled health in the questionnaire that might have caused overstating pain and discomfort for the worst recalled health. The association between respondents’ demographic characteristics and health status and discomfort interpretation subgroups was only tested using univariate analyses, and therefore, the independence of these associations could not be tested. Finally, all coding was performed by one researcher and therefore may be prone to errors.
      Some of the issues identified in this study about the composites may also arise with regard to the self-care and usual activities domains of the EQ-5D that use 2 or more examples to clarify the meaning of a single health domain. It is possible, for example, that someone has moderate problems washing oneself, while no problems with dressing. Nevertheless, given that these examples serve to clarify a single health domain, one may anticipate fewer problems, including under- or inconsistent reporting.
      In spite of the limitations revealed about the PD and AD composites, the EQ-5D shows overall good measurement properties, including validity and responsiveness across different populations and settings.
      • Feng Y.-S.
      • Kohlmann T.
      • Janssen M.F.
      • Buchholz I.
      Psychometric properties of the EQ-5D-5L: a systematic review of the literature.
      ,
      • Buchholz I.
      • Janssen M.F.
      • Kohlmann T.
      • Feng Y.S.
      A systematic review of studies comparing the measurement properties of the three-level and five-level versions of the EQ-5D.
      It may be questioned whether creating an EQ-7D by splitting the composites would be able to outperform the current 5-domain version.
      • McDonald R.
      • Mullett T.L.
      • Tsuchiya A.
      Understanding the composite dimensions of the EQ-5D: an experimental approach.
      Although longer questionnaires may yield a wealth of information about the respondents’ health, the larger the descriptive system, the more unique health states are defined that may require more complex valuation designs placing a higher level of cognitive demand on respondents.
      • Bansback N.
      • Hole A.R.
      • Mulhern B.
      • Tsuchiya A.
      Testing a discrete choice experiment including duration to value health states for large descriptive systems: addressing design and sampling issues.
      Therefore, composite domains seem to have a rightful place in generic preference-based accompanied measures. Nevertheless, our findings put forward potential future directions for improving the structure and wording of the composite domains in the EQ-5D and beyond.

      Conclusions

      Analyzing a mixture of quantitative and qualitative data, this study found empirical evidence of measurement error in the PD and AD composite responses on the EQ-5D in a large general population sample in Hungary including under-reporting of health problems, systematic ordering effect, potential differential item functioning, and interdomain dependency. Our findings contribute new knowledge to the future development of new and refinement of existing self-reported health status instruments.

      Article and Author Information

      Author Contributions: Concept and design: Rencz, Janssen
      Acquisition of data: Rencz
      Analysis and interpretation of data: Rencz, Janssen
      Drafting of the manuscript: Rencz
      Critical revision of the paper for important intellectual content: Rencz, Janssen
      Statistical analysis: Rencz
      Provision of study materials or patients: Rencz
      Obtaining funding: Rencz, Janssen
      Conflict of Interest: Drs Rencz and Janssen reported receiving grants from the EuroQol Research Foundation during the conduct of this study. Dr Rencz reported receiving grants from the Hungarian Academy of Sciences during conduct of this study. Drs Rencz and Janssen are active members of the EuroQol Group. Views expressed in the article are those of the authors and are not necessarily those of the EuroQol Research Foundation.
      Funding/Support: This work was supported by a grant from the EuroQol Research Foundation (no. 240-2020RA). Data collection was supported by the Hungarian Academy of Sciences (MTA-BCE PPD 462025).
      Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

      Acknowledgment

      The authors thank Brendan Mulhern for his comments on the manuscript.

      References

        • Brazier J.
        • Ara R.
        • Rowen D.
        • Chevrou-Severac H.
        A review of generic preference-based measures for use in cost-effectiveness models.
        Pharmacoeconomics. 2017; 35: 21-31
        • Richardson J.
        • Khan M.A.
        • Iezzi A.
        • Maxwell A.
        Comparing and explaining differences in the magnitude, content, and sensitivity of utilities predicted by the EQ-5D, SF-6D, HUI 3, 15D, QWB, and AQoL-8D multiattribute utility instruments.
        Med Decis Making. 2015; 35: 276-291
        • McDonald R.
        • Mullett T.L.
        • Tsuchiya A.
        Understanding the composite dimensions of the EQ-5D: an experimental approach.
        Soc Sci Med. 2020; 265113323
        • Tsuchiya A.
        • Bansback N.
        • Hole A.R.
        • Mulhern B.
        Manipulating the 5 dimensions of the EuroQol instrument: the effects on self-reporting actual health and valuing hypothetical health states [published correction appears in Med Decis Making. 2020;40(2):115].
        Med Decis Making. 2019; 39: 379-392
      1. Bryan S., Jowett S., Hardyman W., Bentham P. Does the EQ-5D anxiety/depression item measure anxiety, depression, both or neither? 21st Plenary Meeting of the EuroQol Group. Chicago, US; 16-18 September 2004.

      2. Macran S., Kind P. EQ-5D valuations from a British national postal survey. In: 17th Plenary Meeting of the EuroQoL Group, Pamplona, Spain; September 18-29, 2000.

        • Engel L.
        • Haagsma J.A.
        • Janssen B.
        • Whitehurst D.G.T.
        • Mulhern B.
        An exploratory analysis of the pain/discomfort dimension of the EQ-5D-5L in people living with physical and mental health conditions.
        in: 37th Plenary Meeting of the EuroQol Group (Virtual). 16-17 September 2020
        • Spronk I.
        • Bonsel G.J.
        • Polinder S.
        • van Baar M.E.
        • Janssen M.F.
        • Haagsma J.A.
        Exploring the relation between the EQ-5D-5L pain/discomfort and pain and itching in a sample of burn patients.
        Health Qual Life Outcomes. 2020; 18: 144
        • Feng Y.
        • Herdman M.
        • van Nooten F.
        • et al.
        An exploration of differences between Japan and 2 European countries in the self-reporting and valuation of pain and discomfort on the EQ-5D.
        Qual Life Res. 2017; 26: 2067-2078
        • Kennedy-Martin M.
        • Slaap B.
        • Herdman M.
        • et al.
        Which multi-attribute utility instruments are recommended for use in cost-utility analysis? A review of national health technology assessment (HTA) guidelines.
        Eur J Health Econ. 2020; 21: 1245-1257
        • Rencz F.
        • Gulácsi L.
        • Drummond M.
        • et al.
        EQ-5D in Central and Eastern Europe: 2000-2015.
        Qual Life Res. 2016; 25: 2693-2710
        • Brooks R.
        EuroQol: the current state of play.
        Health Policy. 1996; 37: 53-72
        • Peasgood T.
        • Mukuria C.
        • Carlton J.
        • Connell J.
        • Brazier J.
        Criteria for item selection for a preference-based measure for use in economic evaluation.
        Qual Life Res. 2021; 30: 1425-1432
        • Menold N.
        Double barreled questions: an analysis of the similarity of elements and effects on measurement quality.
        J Off Stat. 2020; 36: 855-886
        • Grant Levy S.
        Deconstructing a double-barreled alternative: evolution and creationism.
        Psychol Rep. 2019; 122: 1995-2004
        • Bassili J.N.
        • Scott B.S.
        Response latency as a signal to question problems in survey research.
        Public Opin Q. 1996; 60: 390-399
        • Keeley T.
        • Al-Janabi H.
        • Lorgelly P.
        • Coast J.
        A qualitative assessment of the content validity of the ICECAP-A and EQ-5D-5L and their appropriateness for use in health research.
        PLoS One. 2013; 8e85287
        • Whalley D.
        • Globe G.
        • Crawford R.
        • et al.
        Is the EQ-5D fit for purpose in asthma? Acceptability and content validity from the patient perspective.
        Health Qual Life Outcomes. 2018; 16: 160
        • Matza L.S.
        • Boye K.S.
        • Stewart K.D.
        • Curtis B.H.
        • Reaney M.
        • Landrian A.S.
        A qualitative examination of the content validity of the EQ-5D-5L in patients with type 2 diabetes.
        Health Qual Life Outcomes. 2015; 13: 192
        • Rencz F.
        • Mukuria C.
        • Bató A.
        • Poór A.K.
        • Finch A.P.
        A qualitative investigation of the relevance of skin irritation and self-confidence bolt-ons and their conceptual overlap with the EQ-5D in patients with psoriasis [published online ahead of print, 2022 Apr 26].
        Qual Life Res. 2022; https://doi.org/10.1007/s11136-022-03141-y
        • Gudex C.
        The descriptive system of the EuroQol instrument.
        in: Kind P. Brooks R. Rabin R. EQ-5D Concepts and Methods: A Developmental History. Springer, Dordrecht, The Netherlands2005: 19-27
        • Yang F.
        • Jiang S.
        • He X.-N.
        • et al.
        Do rural residents in China understand EQ-5D-5L as intended? Evidence from a qualitative study.
        Pharmacoecon Open. 2021; 5: 101-109
        • Ashkenazy S.
        • DeKeyser Ganz F.
        The differentiation between pain and discomfort: a concept analysis of discomfort.
        Pain Manag Nurs. 2019; 20: 556-562
        • Tighe P.J.
        • Sannapaneni B.
        • Fillingim R.B.
        • et al.
        Forty-two million ways to describe pain: topic modeling of 200,000 PubMed pain-related abstracts using natural language processing and deep learning-based text generation.
        Pain Med. 2020; 21: 3133-3160
        • Fetters M.D.
        • Curry L.A.
        • Creswell J.W.
        Achieving integration in mixed methods designs-principles and practices.
        Health Serv Res. 2013; 48: 2134-2156
        • Herdman M.
        • Gudex C.
        • Lloyd A.
        • et al.
        Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L).
        Qual Life Res. 2011; 20: 1727-1736
        • Devlin N.J.
        • Shah K.K.
        • Mulhern B.J.
        • Pantiri K.
        • van Hout B.
        A new method for valuing health: directly eliciting personal utility functions.
        Eur J Health Econ. 2019; 20: 257-270
        • Evans J.D.
        Straightforward Statistics for the Behavioral Sciences.
        Brooks/Cole Publishing Co, Belmont, CA1996
        • Elo S.
        • Kyngäs H.
        The qualitative content analysis process.
        J Adv Nurs. 2008; 62: 107-115
        • Rencz F.
        • Brodszky V.
        • Gulácsi L.
        • et al.
        Parallel valuation of the EQ-5D-3L and EQ-5D-5L by time trade-off in Hungary.
        Value Health. 2020; 23: 1235-1245
        • Knott R.J.
        • Black N.
        • Hollingsworth B.
        • Lorgelly P.K.
        Response-scale heterogeneity in the EQ-5D.
        Health Econ. 2017; 26: 387-394
        • Smith A.B.
        • Cocks K.
        • Parry D.
        • Taylor M.
        A differential item functioning analysis of the EQ-5D in Cancer.
        Value Health. 2016; 19: 1063-1067
        • Whynes D.K.
        • Sprigg N.
        • Selby J.
        • Berge E.
        • Bath PM; ENOS Investigators
        Testing for differential item functioning within the EQ-5D.
        Med Decis Making. 2013; 33: 252-260
        • Spronk I.
        • Bonsel G.J.
        • Polinder S.
        • van Baar M.E.
        • Janssen M.F.
        • Haagsma J.A.
        The added value of extending the EQ-5D-5L with an itching item for the assessment of health-related quality of life of burn patients: an explorative study.
        Burns. 2021; 47: 873-879
        • Swinburn P.
        • Lloyd A.
        • Boye K.S.
        • Edson-Heredia E.
        • Bowman L.
        • Janssen B.
        Development of a disease-specific version of the EQ-5D-5L for use in patients suffering from psoriasis: lessons learned from a feasibility study in the UK.
        Value Health. 2013; 16: 1156-1162
        • van Dongen-Leunis A.
        • Redekop W.K.
        • Uyl-de Groot C.A.
        Which questionnaire should be used to measure quality-of-life utilities in patients with acute leukemia? An evaluation of the validity and interpretability of the EQ-5D-5L and preference-based questionnaires derived from the EORTC QLQ-C30.
        Value Health. 2016; 19: 834-843
        • Hoogendoorn M.
        • Oppe M.
        • Boland M.R.S.
        • Goossens L.M.A.
        • Stolk E.A.
        • Rutten-van Mölken M.
        Exploring the impact of adding a respiratory dimension to the EQ-5D-5L.
        Med Decis Making. 2019; 39: 393-404
        • Geraerds A.J.L.M.
        • Bonsel G.J.
        • Janssen M.F.
        • Finch A.P.
        • Polinder S.
        • Haagsma J.A.
        Methods used to identify, test, and assess impact on preferences of bolt-ons: a systematic review.
        Value Health. 2021; 24: 901-916
        • Brazier J.E.
        • Peasgood T.
        • Mukuria C.
        • et al.
        The EQ-HWB: overview of the development of a measure of health and well-being and key results.
        Value Health. 2022; 25: 482-491
        • Feng Y.-S.
        • Kohlmann T.
        • Janssen M.F.
        • Buchholz I.
        Psychometric properties of the EQ-5D-5L: a systematic review of the literature.
        Qual Life Res. 2021; 30: 647-673
        • Buchholz I.
        • Janssen M.F.
        • Kohlmann T.
        • Feng Y.S.
        A systematic review of studies comparing the measurement properties of the three-level and five-level versions of the EQ-5D.
        Pharmacoeconomics. 2018; 36: 645-661
        • Bansback N.
        • Hole A.R.
        • Mulhern B.
        • Tsuchiya A.
        Testing a discrete choice experiment including duration to value health states for large descriptive systems: addressing design and sampling issues.
        Soc Sci Med. 2014; 114: 38-48