Validation of the EQ-5D in Patients With a History of Acute Coronary Syndrome*

By Ellis, Jeffrey J; Eagle, Kim A; Kline-Rogers, Eva M; Erickson, Steven R

Key words: Acute coronary syndrome * EQ-5D * Health-related quality of life * Health status *Psychometrics

ABSTRACT

Objective: To analyze the construct validity of the EQ-5D in patients with acute coronary syndromes (ACS).

Methods: All ACS-diagnosed patients discharged from a university- affiliated hospital during a 3-year period were mailed a questionnaire that included the EQ-5D and the SF-8. The EQ-5D includes a visual analogue scale (EQ VAS) to measure self-reported current healthstatus (0-100) and a five-item descriptive system measuring mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Also included were disease severity measures [Duke Activity Status Index (DASI), cardiac symptom count (SC), patient-perceived cardiac disease severity], comorbidity measures (Charlson comorbidity index, total medication count), and other demographic and disease-related items.

Results: Of 1217 patients, 490 (40.3%) responded. Patients averaged 65.2 (SD 11.3) years of age; 71.0% male; 91.9% Caucasian; 64.3% history of MI. Only 0.2%-0.4% of EQ-5D items and 8% of the EQ VAS were left unanswered by respondents. The nine most common health states were identified based on the five EQ-5D item scores. Levels of responses to EQ-5D items and the EQ VAS score were significantly better for patients with very mild/mild perceived disease severity compared to severe/very severe, for patients with lower comorbidity, for patients with lower symptom responses, and for patients with a higher cardiac-related functioning. EQ VAS score and SF-8 subscale score correlation coefficients ranged from 0.527 to 0.798 (all p < 0.0001). Significant differences were observed between the response level of individual EQ-5D items and scores of comparable SF-8 subscales.

Conclusions: This study demonstrated the construct validity of the EQ-5D in a populationbased sample of patients with a history of ACS.

Introduction

Following a myocardial infarction or an event of unstable angina (acute coronary syndrome or ACS), patients receive pharmaceutical, surgical, and/or intracoronary revascularization interventions with the goals of improving coronary blood flow, maximizing cardiac function, maintaining or improving functional status, and reducing mortality. The measurement of interventional outcomes from a patient perspective is conducted by using surveys which quantify such concepts as health status, work performance, symptom burden, and satisfaction. Health-status is measured by general and/or disease specific instruments that are often generalized as measuring health- related quality of life (HRQL). Several disease-specific instruments exist for cardiovascular disease1-5. The value of disease-specific instruments is their ability to focus the respondent on functional assessment related to a specific illness or related treatment, thus leading to a more sensitive measure of HRQL. For population studies however, it is often desirable to include both general and diseasespecific instruments to maximize both sensitivity and generalizability of the results.

A general measure of health status allows the researcher to compare the HRQL of patients with the illness of interest to the HRQL of patients with other illness. Not only do these general instruments assess patients’ perceptions of health status, but several also provide preference-based weights that can be used for costutility comparisons. Preference based measures of health status have become an important set of instruments for estimating the health state values used to calculate quality adjusted life years (QALYs). They are being used more often to evaluate the economic impact of interventions in clinical trials and population-based studies. One such general health state measure is the EQ-5D6-8. The EQ-5D is a brief, generic instrument used to describe and value health status or HRQL and for which population-based preferences for various health states defined by the EQ-5D descriptive system are available.

The purpose of this study was to determine the validity of the EQ- 5D in patients with a history of ACS. Validity, which relates to the extent that an instrument measures what it is intended to measure, can be defined and assessed in several ways. This study assessed the construct validity of the EQ-5D, which refers to the agreement between an instrument and a theoretical concept related to the disease state in question. For this study, the association between the domains of the EQ-5D and cardiac function and symptom burden were used to assess construct validity. It was hypothesized that as disease severity worsens, the EQ-5D domain scores will also worsen. Additional construct validity checks were conducted by examining the degree of association between EQ-5D domains and analogous domains of another general HRQL instrument, the SF-8.

Methods

Patients and methods

This cross-sectional study utilized a mailed survey completed by patients prospectively chronicled in the Acute Coronary Syndrome/ Myocardial Infarction Patient Registry maintained at the University of Michigan. Patients eligible for this registry include those admitted to and discharged from the University of Michigan Hospital for treatment related to acute coronary syndrome (unstable angina, ST-segment elevation or non-ST-segment elevation myocardial infarction) who were 18 years of age or older at the time of presentation. Registry patients obtained from July 1999 to November 2002 and alive at the time of this survey (December 2002 to mid- January 2003) constituted the study sample. In total, 1217 eligible patients were mailed the survey. The human investigations committee of the affiliated medical school approved this study and written informed consent was obtained from those participating in this study.

The initial survey contained a cover sheet, the questionnaire, informed consent document, and preaddressed and postage-paid envelopes. A reminder postcard was sent to all subjects a week later. Subjects who did not respond by 4 weeks after the initial mailing were sent a duplicate packet of questionnaires.

Data and data sources/questionnaires

Patient and disease characteristic data were obtained through chart review and patient self-report. Age, gender, and type of ACS (unstable angina and myocardial infarction) diagnosed were obtained from the hospital medical record. Information regarding race, education, number of medications currently taken, and household income were acquired by self-report using items on the survey.

Comorbidities, a potential mediator of functional status and performance, were assessed through selfadministration of the questionnaire version of the Charlson Comorbidity Index9’10 included in the survey. The Symptom Distress Checklist (SDC), included in our survey, is a questionnaire developed for use in clinical trials to identify the existence of cardiovascularrelated symptoms and treatment-related adverse effects and the impact they have on the patient11-12. Symptoms contained in the SDC include those with high relevance to patients who have ischemic heart disease (chest pain, numbness in extremities, slow or irregular heartbeat) but also symptoms not only common in heart disease but that may be common to other illnesses or treatments. Subjects indicated whether or not the symptom occurred in the previous 4 weeks. The sum of symptoms experienced served as the symptom measure for this study. The Duke Activity Status Index (DASI) was utilized as part of the survey in order to assess cardiac functional status at the time of questionnaire administration13. The DASI is a 12-item questionnaire designed to assess functional capacity based on approximations of metabolic expenditure necessary to complete an assortment of personal, household, occupational, and recreational activities (e.g. ‘Can you run a short distance?’). The DASI selfadministered questionnaire has been shown to provide a valid measure of functional capacity and correlates significantly with peak oxygen uptake (VO^sub 2^)14,15. The DASI score may range from O to 58.2. Threshold DASI levels indicative of disease severity have not been established. As such, the study population median score served to dichotomize the population into patients with less severe and more severe cardiac disease.

In addition to the DASI, patient-perceived cardiac disease severity was assessed through response to the survey question ‘How severe do you think your heart disease is?’ Five response options were available ranging from ‘very mild’ to ‘very severe’. For purposes of this study, this ordinal measure was trichotomized to describe perceived severity as ‘high’ (‘very severe’ and ‘severe’), ‘moderate’, and ‘low’ (‘mild’ and Very mild’).

The SF-8 Health Survey was used to assess physical and mental health status16. The SF-8 measures the health concepts of physical functioning, role limitations due to physical health problems, bodily pain, general health, vitality, social function, role limitations due to emotional problems, and mental health. The 8 items of the SF-8, originating from the SF-36, were used to calculate physical (PCS) and mental (MCS) summary scores. Final scores were determined by the normativescoring based method, with a range from O to 100, with 50 being the mean score for the US population.

The EQ-5D descriptive systemcomprises five dimensions of health: mobility, self-care, usual activities, pain/discomfort, and anxiety/ depression6″8. Each dimension comprises three possible ordinal levels of severity: 1 = no problems; 2 = moderate problems; and 3 = extreme problems. Responses generate a five-dimensional descriptive system of health status, signified by a fivedigit number, with a total of 243 (i.e. 33) theoretically possible self-classified health states, ranging from 11111 to 33333. US adult population-based preferences for each health state have been derived using time tradeoff (TTO) utility analysis exercises to estimate values for each of the 243 self-classified EQ-5D health states17. Once a health state has been derived for a patient, it can be coupled with the population preference score for that health state. This number is called the EQ-5D Health Index score (EQ Health Index score). EQ Health Index scores are normalized so that a score of 1 (maximum score) represents perfect health and a score of O represents death. Certain severe health states have a score less than O and are thus considered worse than death by the general public6,7,17. The preference-based score can then be used for cost-utility analyses. The EQ-5D also includes a General Health Visual Analogue Scale (EQ VAS). Utilizing a vertical 20-cm graduated line, patients reported their current level of health state by placing a mark on the line. The scale is anchored with 0 equal to the worst imaginable health state and 100 equaling the best imaginable health state.

Statistical analyses

Univariate analyses were conducted to characterize patient, disease, treatment, and health state descriptive variables using mean and standard deviation or frequency and percentage.

To assess construct validity of the EQ-5D, several hypotheses were developed. It was hypothesized that the EQ VAS score and the EQ Health Index score are related to the disease-specific measures such as the DASI, perceived severity, number of symptoms, and comorbidity. It was also hypothesized that as disease severity worsened, the number of patients in the EQ-5D domain worse functional response categories would increase. Therefore, construct validity was determined by examining the frequencies and percentage of respondents categorized into each functional level of the EQ-5D compared with the level of severity of illness determined by the DASI (median split at < 18 and ≥ 18), Perceived Cardiac- Disease Severity (3 categories of very mild/mild, moderate, severe/ very severe), and Number of Symptoms (median split at ≤ 7 or > 7). Chi-squared tests were used to compare the number of patients in each functional category by disease severity category while either the Kruskal-Wallis H test or Mann-Whitney U test was used for the EQ VAS and the EQ Health Index score.

An additional check of construct validity was to hypothesize that the domain scores of like constructs of the EQ-5D and the SF-8 would be similar and related. For this analysis it was hypothesized that the EQ-5D and SF-8 domain-matches included: mobility and physical functioning; self-care and physical functioning; usual activities and role-physical; pain/discomfort and bodily pain; and anxiety/ depression and mental health, respectively. This analysis was done by examining the median SF-8 scale score of the scale hypothesized to be related to the EQ-5D domain using the KruskalWallis H test. Spearman Rank Correlation was used to determine the relationship between each SF-8 domain and clinical measures with the EQ Health Index score and the EQ VAS scores. Significance levels of p < 0.05 were considered statistically significant.

Results

Of the 1217 potential patients eligible to receive the survey, 490 (40.3%) responded. There were no significant differences between respondents and nonrespondents for variables available for comparison (age, gender, and type of ACS event). Table 1 provides a description of the respondents. The mean age was 65.2 (SD 11.3) years, 71% were male, and 91.9% were Caucasian. Over 60% attended college. There was a fairly even distribution in annual household income. Most respondents experienced a myocardial infarction versus unstable angina. Also, the time since the most recent cardiovascular event (MI, unstable angina, CABG, or PCI) was fairly evenly distributed from 6 months to 3 years. Over half of the respondents reported their perceived cardiac illness as either mild or moderate in severity. The mean DASI score of 20.4 (SD 8.2) indicates moderate impairment. Patients reported a mean of 8.0 (SD 5.1) symptoms and, on average, had existing comorbid conditions.

Table 1. Demographic characteristics of survey respondents (n = 490)

Table 2, Distribution of responses to EQ-5D self classifier

Table 3. Most frequently reported EQ-5D self-classified health states

The percentage of missing responses to the individual EQ-5D items ranged from 0.2% to 0.4%. Eight percent of respondents failed to answer the EQ VAS. Table 2 provides the overall results of the EQ- 5D scores for each domain. Over half of respondents indicated no problems in any domain except for pain/discomfort, where most indicated moderate problems. Very few respondents chose extreme problems for any domain. The EQ VAS mean score was 72.6 (SD 20.7), median 80.0, and a range from O to 100. The EQ Health Index mean score was 0.81 (SD 0.18), median 0.83, with a range from 0.12 to 1.0. A ceiling effect was observed for the EQ Health Index score, with 30.0% of respondents reporting a health state that had an accompanied EQ Health Index score of 1.0, the maximum score. Table 3 lists the nine most common health state descriptions. The highest functional health state (11111) was indicated by 30% of the respondents. The nine most common health states covered 78.1% of respondents.

Table 4. Construct validity of the EQ-5D

EQ-5D results were compared to measures of disease including the DASI, perceived cardiac illness severity, symptoms, and comorbidity (Table 4). Levels of responses to EQ-5D domains, EQ VAS, and EQ Health Index scores were significantly better for patients with a DASI ≥ 18 (less severe cardiac disease) compared to a DASI < 18 (more severe cardiac disease). The same pattern was seen when comparing EQ-5D results based on a condensed patient perceived severity of cardiac illness, as shown in Table 4. As severity of illness worsens, more patients responded that they have moderate or severe problems in each domain. The least amount of change between severity categories is found in the SelfCare domain. The mean and median scores for EQ VAS and EQ Health Index scores decline significantly with worsening perceived cardiac illness severity as well. The same patterns were observed for EQ-5D responses when comparisons are based on the number of symptoms and comorbidities when dichotomized at median value. Of note, significant differences in the EQ Health Index scores and EQ VAS between dichotomized symptom burden groups were identified. Based on the median split of the number of reported symptoms, the group of patients with seven or fewer symptoms had a mean EQ Health Index score of 0.86 (SD 0.17) while the group with eight or more symptoms had a mean EQ Health Index score of 0.65 (SD 0.26). The same relationship was seen for the EQ VAS, with the group of patients reporting seven or fewer symptoms with a mean EQ VAS of 82.6 (SD 15.6) compared to a mean of 63.9 (SD 20.7) for the patient group reporting eight or more symptoms.

Construct validity was also assessed by comparing the EQ-5D results to conceptually similar domains of the coadministered SF-8. Significantly different within-domain SF-8 scale scores were observed when categorized based on the reported level of disability from the conceptually similar EQ-5D domain. These results are found in Table 5. For example, it was hypothesized that the SF-8 domain of physical functioning would be analogous to the mobility domain of the EQ-5D. As shown in Table 5, the SF-8 physical functioning scale score was incrementally lower for progressively lower functioning categories for the EQ-5D mobility domain. Similar significant differences in the within-scale SF-8 domain scores were observed between the response levels of individual EQ-5 D items when comparing scores of the other analogous SF-8 scales (Table 5). Correlations between all individual SF-8 scale scores and the measure of overall general health status, the EQ VAS, were highly significant (Table 6). The physical summary measure (PCS) had a higher correlation than the mental summary measure (MCS). A similar pattern was observed with the correlation between the SF-8 domains and the preference-based score, the EQ Health Index score. All correlations between the clinical variables and EQ VAS and EQ Health Index score were also significant.

Table 5. Construct validity: EQ-5D and comparable SF-8 scale scores*

Discussion

This analysis sought to determine the feasibility of assessing health-related quality of life through selfadministration as well as the construct validity of the EQ-5D in patients with a relatively recent history of acute coronary syndrome. The EQ-5D is short and simple enough that many subjects can complete the form without assistance. There were few missing data on this survey. The EQ-5D’s short length and preference-based weighting for the EQ-5D Health Index score make it an attractive instrument for use in clinical studies as well as for patient care. This study supports the use of the EQ-5D in cross-sectional population-based studies as a general instrument to measure the HRQL of patients who have experienced unstable angina or myocardial infarction. Additional work must be conducted to determine its usefulness in longitudinal studies, where reliability and sensitivity of the instrument must be further established.

Table 6. Construct validity: Spearman rank correlation coefficient (r) between EQ V\AS score, EQ Health Index score and SF- 8 subscale scores and clinical measures

The construct validity of an instrument refers to the extent to which it correlates with criteria from an established measure. We used bivariate and correlation analyses to test the hypothesis that as the burden of illness increased using measures of cardiac- related functioning, symptoms, and perceived severity of illness, health status as measured by the EQ-5D would diminish. Indeed, there was deterioration in the EQ VAS and EQ Health Index scores and an increase in disability as assessed by the EQ-5D domains as the burden of illness increased. Likewise, the distribution of SF-8 scores by level of impairment indicated on the EQ-5D and the correlation analysis demonstrated construct validity of the instrument. This relationship demonstrates the construct validity of the EQ-5D for cross-sectional studies of patients with a history of ACS.

A study by Johnson and Coons examined the psychometric properties if the EQ-5 D in a small sample of respondents in the United States18. Of patients who noted they had heart disease in this sample, the distribution of EQ-5 D domain responses was similar to that in our study. A more recent study by Nowels et al. also demonstrated validity of the EQ-5D in a postmyocardial infarction population19. The EQ-5D has been tested in other illnesses and conditions with evidence of its validity as a measure of health status reported20-23.

Of the individual dimensions of the EQ-5D, anxiety/ depression, pain/discomfort, mobility, and usual activities have the greater spread between no problem and some or extreme problem. Few respondents indicated problems with self-care. The two overall measures, the EQ VAS and the EQ Health Index scores both showed significant distribution between levels of impairment. The single item EQ VAS appears to be a good, easily understood measure of general health perception in this population. However, 8% of respondents failed to complete this item, the most of any on the survey.

As noted by other researchers, a ceiling effect is present with the EQ-5D. Few patients responded to the worst level of functioning in most domains. This may be a potential concern for some researchers who wish to use a brief general measure of health state. However, the distribution of the EQ VAS as well as the health state preference score of the EQ Health Index increases the usability of this instrument. If researchers desire a more detailed examination of health status relevant to a particular illness or treatment, a disease-specific instrument can be used alongside the EQ-5D. Cardiovascular disease-specific measures of HRQL exist1-5.

It may be reasonable to administer a disease-specific instrument at the same time for specific circumstances when a more refined, focused examination of HRQL related to a target illness is desired. However, in many population studies where many illnesses or comorbid conditions exist, a general measure may be required. For example, the Medication Expenditure Panel Survey contains the EQ-5D and the SF-12 to measure respondent health status. Based on results of this and other studies, researchers should feel confident in assessing health status of patients with particular illnesses in these datasets using the EQ-5D.

One of the limitations of this study was that it was crosssectional and not longitudinal. Without longitudinal data we were unable to test for two important psychometric properties of survey instruments:test-retest reliability and sensitivity to change (responsiveness). These properties are essential for an instrument to be useful in day to day clinical practice. Another limitation is the lack of an objective measure of cardiac function. However, when measuring the health of a population, such measures cannot be obtained. We instead used the DASI, a clinically proven instrument that correlates cardiac function with physical functioning of the cardiac patient. Another limitation of this study was the inability to truly determine the acceptance and understanding of the EQ-5D in this population sample. It was true that very few respondents failed to answer the EQ-5D items and 8% did not respond to the EQ VAS question. Further research may be warranted using cognitive debriefing techniques to truly understand how patients answer the questionnaire.

If the main focus of a study is measurement of global health status, the EQ-5D provides a brief, valid instrument that patients with a history of acute coronary syndrome can complete. Additionally, the EQ-5D can also be used to estimate quality weights to construct QALYs, since it provides a preference-based score, the EQ Health Index score. This is a benefit that can be used in cost- effectiveness analyses.

Conclusion

This study demonstrated the validity of the EQ-5D in patients with a history of ACS. The demonstrated validity will allow for its use in ACS-related populationbased studies where brief measures of HRQL are desirable. When used in concert with the newly published general United States population preferencebased valuations of EQ- 5D health states, this general HRQL instrument can provide useful information for health policy decision making.

Acknowledgement

Funding for this project was provided by the Clinical Research Resources Committee of the University of Michigan College of Pharmacy Clinical Sciences Department.

* Presented at the International Society for Pharmacoeconomics and Outcomes Research meeting, 17-19 May 2004, Washington DC

References

1. Spertus JA, Winder JA, Dewhurst TA, et al. Development and evaluation of the Seattle Angina Questionnaire: a new functional status measure for coronary artery disease. J Am Coll Cardiol 1995;25:333-41

2. Rumsfeld JS, Magid DJ, Plomondon ME, et al. Predictors of quality of life following acute coronary syndromes. Am J Cardiol 2001;88:781-4

3. Spertus JA, McDonell M, Fan V, Fihn SD. Health status predicts long-term outcome in outpatients with coronary disease. Circulation 2002; 106:43-9

4. Dixon T, Lim LLY, Oldridge NB. The MacNew heart disease health- related quality of life instrument: reference data for users. Qual Life Res 2002;! 1:173-83

5. Valenti L, Lim L, Heller RF, Knapp J. An improved questionnaire for assessing quality of life after acute myocardial infarction. Qual Life Res 1996;5:151-61

6. EuroQol Group. EuroQol: a new facility for the measurement of health-related quality of life. Health Policy 1990; 16:199-208

7. Brooks RG. EuroQol: the current state of play. Health Policy 1996;37:53-72

8. Kind P. The EuroQol instrument: an index of health-related quality of life. In: Spilker B, editor. Quality of life and pharmacoeconomics in clinical trials, 2nd ed. Philadelphia (PA): Lippincott-Raven Press; 1996. p. 191-201

9. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987;40:373-83

10. Katz JN, Chang LC, Sangha O, Fossel AH, Bates DW. Can comorbidity be measured by questionnaire rather than medical record review? Med Care 1996;34:73-84

11. Anderson RB, Testa MA. Symptom distress checklists as a component of quality of life measurement: comparing prompted reports by patient and physician with concurrent adverse event reports via the physician. Drug Inf J 1994;28:89-114

12. Anderson RB, Nackley JF, Testa MA. Symptom distress check lists as a component of quality of life measurement; comparing symptom reports with responses to multiple choice questionnaires. Drug Inf J 1995;29:1689S-1707S

13. Hlatky MA, Boineau RE, Higginbotham MB, et al. A brief selfadministered questionnaire to determine functional capacity (the Duke Activity Status Index). Am J Cardiol 1989;64:651-4

14. Nelson CL, Herndon JE, Mark DB, et al. Relation of clinical and angiographie factors to functional capacity as measured by the Duke Activity Status Index. Am J Cardiol 1991;68:973-5

15. Jaeger AA, Hlatky MA, Paul SM, Gortner SR. Functional capacity after cardiac surgery in elderly patients. J Am Coll Cardiol 1994;24:104-8

16. Ware JE, Kosinski M, Dewey JE, et al. How to score and interpret single-item health status measures: a manual for users of the SF-8″ Health Survey. Lincoln (RI): Quality Metric Incorporated; 2001

17. Shaw JW, Johnson JA, Coons SJ. US valuation of the EQ-5D health states. Development and testing of the Dl valuation model. Med Care 2005;43:203-20

18. Johnson JA, Coons SJ. Comparison of the EQ-5D and SF-12 in an adult US sample. Qual Life Res 1998;7:155-66

19. Nowels D, McGloin J, Westfall JM, Holcomb S. Validation of the EQ-5D quality of life instrument in patients after myocardial infarction. Qual Life Res 2005;14:95-105

20. Cleemput I, Kesteloot K, Moons P, et al. The construct and concurrent validity of the EQ-5D in a renal transplant population. Value Health 2004;7:499-509

21. Honig HH, Ulshofer A, Gregor M, et al. Validation of the EuroQol questionnaire in patients with inflammatory bowel disease. Eur J Gastroenterol Hepatol 2002;14:1205-15

22. Prieto L, Sacristan JA, Hormaechea JA, Casado A, Badia X, Gomez JC. Psychometric validation of a generic health-related quality of life measure (EQ-5D) in a sample of schizophrenic patients. Curr Med Res Opin 2004;20:827-35

23. Wu AW, Jacobson DL, Frick KD, et al. Validity and responsiveness of the EuroQol as a measure of health-related quality of life in people enrolled in an AIDS clinical trial. Qual Life Res 2002; 11:273-82

CrossRef links are available in the online published version of this paper: http://www.cmrojournal.com

Paper CMRO-3017_3, Accepted for publication: 06 June 2005

Published Online: 01 July 2005

doi: 10.1185/030079905X56349

Jeffrey J. Ellis(a), Kim A. Eagleb, Eva M. Kline-Rogers(b) and Steven R. Erickson(c)

a Lincoln Surgical Hospital, Department of Pharmacy, Lincoln, NE, USA

b Department of Internal Medicine, Division of Cardiology, University of Michigan Medic\al Center, Ann Arbor, MI, USA

c College of Pharmacy, University of Michigan, Ann Arbor, MI, USA

Address for correspondence: Steven R. Erickson, PharmD, Associate Professor, University of Michigan, College of Pharmacy, 428 Church Street, Ann Arbor, Michigan 48109-1065, USA. Tel.: +1 734 763 4989; Fax: +1 734 763 2022; email: serick@umich.edu

Copyright Librapharm Aug 2005