Uwe Konerding

The validity of the EQ-5D-3L items: an investigation with type 2 diabetes patients from six European countries

BackgroundMost previous studies concerning the validity of the EQ-5D-3L items refer to applications of only a single language version of the EQ-5D-3L in only one country. Therefore, there is little information concerning the extent to which the results can be generalised across different language versions and/or different countries. Here the validity of the EQ-5D-3L items is investigated for six different language versions in six different countries.MethodsData came from 1341 type 2 diabetes patients (England: 289; Finland: 177; Germany: 255; Greece: 165; the Netherlands: 354; Spain: 101). The relationships of the five EQ-5D-3L items with seven different test variables (age, gender, education, previous stroke, problems with heart, problems with lower extremities, problems with eyes), were analysed for each combination of item and test variable. For each combination two logistic regression models with the dichotomised EQ-5D-3L item as dependent variable were computed. The first model contained the test variable and dummy coded countries as independent variables, the second model additionally the terms for the interaction between country and test variable. Statistically significant better fit of the second model was taken as evidence for country specific differences regarding the relationship. When such differences could be attributed mainly to one country the analyses were repeated without the data from this country. Validity was investigated with the remaining data using results of the first models.ResultsDue to lack of variation in the Spanish data only 31 of the originally intended 35 interaction tests could be performed. Only three of these yielded a significant result. In all three cases the Spanish data deviated most. Without the Spanish data only 1 of the 35 interaction tests yielded a significant result. With 3 exceptions, the tendency of reporting problems increased with age, female gender, lower education, previous stroke, heart problems, problems with lower extremities and problems with eyes for all EQ-5D-3L items.ConclusionThe results concerning the European Spanish version are ambiguous. However, the items of the English, Finnish, German, Greek and Dutch versions of the EQ-5D-3L relate in substantially the same way to the test variables. Mostly, these relationships indicate the items’ validity.

What does Cronbach's alpha tell us about the EQ-5D? A methodological commentary to "Psychometric properties of the EuroQol Five-Dimensional Questionnaire (EQ-5D-3L) in caregivers of autistic children".

In a paper just accepted for publication in Quality of Life Research, Khanna et al. [1] investigate the psychometric properties of the EQ-5D with three levels [2] using data from a sample of 316 US—American caregivers of autistic children. The authors examine convergent and discriminant validity of the five EQ-5D dimensions by analyzing the relation of these dimensions with the dimensions of the SF-12v2 [3], and they examine known group validity by analyzing how the severity of the children’s autism relates to the EQ-5D index and the VAS scale. With a few exceptions, the results of these examinations accord with the authors’ expectations. The authors interpret this as evidence for the EQ-5D’s validity. The authors also examine the reliability of the EQ-5D. For this purpose, they compute Cronbach’s alpha and item-total correlations for the five dimensions of the EQ-5D. Cronbach’s alpha is only 0.63, whereas the item-total correlations range from 0.25 for anxiety/depression to 0.58 for usual activities. Following the recommendations of Nunnally [4], the authors judge the result for Cronbach’s alpha as unsatisfactory and the results for the item-total correlations as acceptable. Whilst the authors’ investigations concerning validity are perfectly reasonable, those concerning reliability warrant some critical remarks. The method they apply, that is, the estimation of reliability via Cronbach’s alpha and itemtotal correlation, is not adequate for the kind of measurement instrument investigated, that is, for an index measurement instrument like the EQ-5D. The approach of estimating the reliability of a measurement instrument via Cronbach’s alpha and item-total correlations has clear limitations. It is not adequate for all kinds of measurement instruments, but only for homogenous tests. In a perfectly homogenous test, all items reflect the same dimension, and therefore, deviations from perfect inter-item correlations reflect nothing else but measurement error. For such a test, the deviation of Cronbach’s alpha from the upper bound 1 is a direct function of measurement error, and consequently, Cronbach’s alpha itself is a valid estimation of the test’s reliability. The opposite of homogenous tests are heterogeneous tests. A test is heterogeneous when it is composed of items which address qualitatively different aspects or even qualitatively different dimensions of the object of investigation. In a heterogeneous test, the different items cannot correlate perfectly even if there is no measurement error. Consequently, when Cronbach’s alpha is applied to heterogeneous tests, the true reliability is underestimated. The more heterogeneous a test is, the more the reliability will be underestimated. Cronbach himself already explicitly pointed out this feature of Cronbach’s alpha when he first presented this coefficient [5]. Item-total correlations have similar features like Cronbach’s alpha. The more heterogeneous the test is, the lower the item-total correlations will be. A good index measurement instrument requires a set of questions which cover all relevant different aspects of the object of investigation. Ideally, the different questions should reflect different analytically independent dimensions, because in this case, the informative value provided by each single question would be maximal. Selecting items with regard to this ideal necessarily produces a very heterogeneous test. Consequently, neither Cronbach’s alpha nor item-total correlations for such a test can be particularly high. It would even be extremely unwise to strive for a high Cronbach’s alpha and high item-total correlations, because selecting items with regard to maximizing these U. Konerding (&) Bamberg, Germany e-mail: uwe.konerding@uni-bamberg.de

Which kind of psychometrics is adequate for patient satisfaction questionnaires

The construction and psychometric analysis of patient satisfaction questionnaires are discussed. The discussion is based upon the classification of multi-item questionnaires into scales or indices. Scales consist of items that describe the effects of the latent psychological variable to be measured, and indices consist of items that describe the causes of this variable. Whether patient satisfaction questionnaires should be constructed and analyzed as scales or as indices depends upon the purpose for which these questionnaires are required. If the final aim is improving care with regard to patients’ preferences, then these questionnaires should be constructed and analyzed as indices. This implies two requirements: 1) items for patient satisfaction questionnaires should be selected in such a way that the universe of possible causes of patient satisfaction is covered optimally and 2) Cronbach’s alpha, principal component analysis, exploratory factor analysis, confirmatory factor analysis, and analyses with models from item response theory, such as the Rasch Model, should not be applied for psychometric analyses. Instead, multivariate regression analyses with a direct rating of patient satisfaction as the dependent variable and the individual questionnaire items as independent variables should be performed. The coefficients produced by such an analysis can be applied for selecting the best items and for weighting the selected items when a sum score is determined. The lower boundaries of the validity of the unweighted and the weighted sum scores can be estimated by their correlations with the direct satisfaction rating. While the first requirement is fulfilled in the majority of the previous patient satisfaction questionnaires, the second one deviates from previous practice. Hence, if patient satisfaction is actually measured with the final aim of improving care with regard to patients’ preferences, then future practice should be changed so that the second requirement is also fulfilled.

Investigating burden of informal caregivers in England, Finland and Greece: an analysis with the short form of the Burden Scale for Family Caregivers (BSFC-s)

ABSTRACT Objectives: The burden of informal caregivers might show itself in different ways in different cultures. Understanding these differences is important for developing culture-specific measures aimed at alleviating caregiver burden. Hitherto, no findings regarding such cultural differences between different European countries were available. In this paper, differences between English, Finnish and Greek informal caregivers of people with dementia are investigated. Methods: A secondary analysis was performed with data from 36 English, 42 Finnish and 46 Greek caregivers obtained with the short form of the Burden Scale for Family Caregivers (BSFC-s). The probabilities of endorsing the BSFC-s items were investigated by computing a logit model with items and countries as categorical factors. Statistically significant deviation of data from this model was taken as evidence for country-specific response patterns. Results: The two-factorial logit model explains the responses to the items quite well (McFaddens pseudo-R-square: 0.77). There are, however, also statistically significant deviations (p < 0.05). English caregivers have a stronger tendency to endorse items addressing impairments in individual well-being; Finnish caregivers have a stronger tendency to endorse items addressing the conflict between the demands resulting from care and demands resulting from the remaining social life and Greek caregivers have a stronger tendency to endorse items addressing impairments in physical health. Conclusion: Caregiver burden shows itself differently in English, Finnish and Greek caregivers. Accordingly, measures for alleviating caregiver burden in these three countries should address different aspects of the caregivers’ lives.

Preference‐based index measurement of health‐related quality of life: when does it reflect only arbitrary settings of the researcher?

In health economics, health-related quality of life (HRQoL) is often assessed by means of preference-based index measurement instruments (e.g. EQ-5D, SF-6D, HUI). Each instrument of this kind consists of (1) a multi-attribute classification system for distinguishing health states and (2) a scoring function which assigns a valuation to each health state distinguished within the classification system. Scoring functions are often produced according to the so-called statistical approach which consists of two steps: (1) the valuations of some of the health states of the classification system are empirically determined and (2) the scoring function values for all health states are estimated from the empirically determined valuations using a theoretical model, i.e. an index model. This approach can run into problems because the empirically determined valuations necessarily contain arbitrary settings. This article is concerned with how these arbitrary settings together with the index model affect the final scoring function values. It is shown that for many conceivable index models the final scoring function values have no empirical meaning. Only additive models with a free additive constant are appropriate for representing the information contained in the empirically determined valuation. Only these models should be used within the statistical approach.

The impact of travel distance, travel time and waiting time on health-related quality of life of diabetes patients: An investigation in six European countries

AIMS The effects of travel distance and travel time to the primary diabetes care provider and waiting time in the practice on health-related quality of life (HRQoL) of patients with type 2 diabetes are investigated. RESEARCH DESIGN AND METHODS Survey data of 1313 persons with type 2 diabetes from six regions in England (274), Finland (163), Germany (254), Greece (165), the Netherlands (354), and Spain (103) were analyzed. Various multiple linear regression analyses with four different EQ-5D-3L indices (English, German, Dutch and Spanish index) as target variables, with travel distance, travel time, and waiting time in the practice as focal predictors and with control for study region, patients gender, patients age, patients education, time since diagnosis, thoroughness of provider-patient communication were computed. Interactions of regions with the remaining five control variables and the three focal predictors were also tested. RESULTS There are no interactions of regions with control variables or focal predictors. The indices decrease with increasing travel time to the provider and increasing waiting time in the providers practice. CONCLUSIONS HRQoL of patients with type 2 diabetes might be improved by decreasing travel time to the provider and waiting time in the providers practice.

Some problems with determining the reliability of the EQ‐5D‐3L: commentary to “Value of EQ‐5D in Mexican city older population with and without dementia (SADEM study)”

In a paper published in the International Journal of Geriatric Psychiatry (Sanchez-Arenas et al. 2014), Sanchez-Arenas et al. concern themselves with the reliability of the EQ-5D-3L, which they refer to solely as the EQ-5D. They investigate both the reliability of the five single dimensions and the reliability of the whole instrument. Sanchez-Arenas et al. report that they have determined the reliability for the five single dimensions using Cronbach’s alpha. This approach is mathematically impossible. Cronbach’s alpha can only be computed for scales that consist of at least two items. However, each dimension of the EQ-5D-3L is measured with exactly one item. So, with the description given by the authors, it is completely unclear how the authors have actually produced the numbers they report. Therefore, in further research, these numbers should not be used as estimates for the reliability of the single dimensions of the EQ-5D-3L. Sanchez-Arenas et al. do not report how they have determined the reliability for the whole measurement instrument. However, considering that there is obviously only one measurement point, they cannot have applied the test–retest approach. Then, the only approach that they might have applied is computing Cronbach’s alpha for the five items of the instrument. As elaborated elsewhere (Konerding, 2013), this approach is not appropriate for measurement instruments such as the EQ-5D-3L. Cronbach’s alpha only provides an estimate of reliability if the measurement instrument is one-dimensional. Otherwise, the estimation of the reliability is an underestimation. Each of the five items of the EQ-5D-3L, however, refers to a different dimension. So, for this instrument, Cronbach’s alpha will be lower than the actual reliability. Therefore, in further research, the number reported by Sanchez-Arenas et al. should not be used as an estimate for the reliability of the EQ-5D-3L.