Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

2.1 International EQ-5D Archive of Population Surveys

The international EQ-5D database archive consists of 27 EQ-5D population surveys collected in 24 countries. Countries with 1 or more population surveys include: Argentina, Armenia, Belgium, Canada, China, Denmark, England, Finland, France, Germany, Greece, Hungary, Italy, Japan, Korea, the Netherlands, New Zealand, Slovenia, Spain, Sweden, Thailand, United Kingdom, United States, and Zimbabwe. The datasets are structured in a standardized format to facilitate comparative research, although each survey also has its own characteristics and variables specific to the individual research context in which they were conducted. In addition, three datasets from Argentina, China, and Sweden (Stockholm area) were analyzed locally and results were added to the book, as the dataset transfer to the central archive was not possible from these countries. The datasets captured by this book currently include observations on 216,703 individuals. For a more detailed account of the data, see Table 2.1.

Table 2.1 National and Regional EQ-5D Population Surveys
Full size table

All of the surveys used a standardized version of EQ-5D-3L. The Dutch, Swedish and Finnish versions were translated in 1987 according to a ‘simultaneous’ process while the remaining versions were translated according to the EuroQol Group’s translation protocol – based on international guidelines. However, some differences between sampling and data collection methods should be noted.

Most importantly, while the majority of surveys were national representative surveys covering the whole of the country, some surveys covered a specific part (such as prefectures, regions or even city areas). Therefore, care should be exercised in generalizing data outside the geographic location captured by the data collection. Results in this book are reported separately for the national and the regional surveys.

Surveys also differed in sample sizes and in the method of data collection. The Argentinean dataset had the largest sample with over 41,000 respondents, while the Greek and the Swedish national surveys had the smallest sample of around 500 respondents. Some of the surveys were postal while others were performed as part of a face-to-face interview or administered by telephone. Since the questions asked in EQ-5D are very simple to answer, there is no reason to believe that there would be a significant impact on results other than differences in response rates.

While only the most recent national surveys were included in this book from each country, the date of data collection varied considerably across countries. Data collection for the majority of surveys took place during or after 2000, however some surveys were older with the United Kingdom and Swedish national datasets being the earliest from 1993 to 1994, respectively. These differences should be considered when interpreting results, given that health-related quality of life in general and specifically EQ-5D ratings and values could have changed over time.

Standardized variables across all datasets included reported problems by the five dimensions, self-reported EQ VAS ratings, and the EQ-5D index values. In addition, all analyses of EQ-5D data presented in this book focused on three main characteristics of the population: age, gender, and education level. Age in most surveys was measured as a continuous variable (life years), while gender was recorded as a categorical variable. Education level in each country was recoded to a three-level scale, distinguishing low (i.e. primary), medium (i.e. secondary), and high (i.e. university degree) education level.

All data analyses were performed using SPSS version 19 and Stata version 12 statistical software packages. All codes were checked and analyses were reproduced by a second analyst. The exact methodologies are described in the remainder of this chapter.

2.2 Methods of Describing EQ-5D Population Norms

The EuroQol Group is frequently asked to provide EQ-5D population reference data (sometimes called normative data) for a specific country or international region. Such data can be used as reference data to compare profiles for patients with specific conditions with data for the average person in the general population in a similar age and/or gender group. This comparison helps to identify the burden of disease in a particular patient population.

Descriptive statistics are provided for EQ VAS, the five dimensions, and EQ-5D-3L index values for the total population and by gender and the following age groups: 18–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75+ years.

EQ-5D index value calculations are provided using the following value sets (Szende et al. 2007):

  • European VAS value set for all countries. Note that the European VAS value set was constructed using data from 11 valuation studies in 6 countries: Finland (1), Germany (3), The Netherlands (1), Spain (3), Sweden (1) and the UK (2). This survey included enough data from different European regions to make the European VAS dataset moderately representative for Europe. (Greiner et al. 2003; Weijnen et al. 2003).

  • Country-specific time trade-off (TTO) value set if available. Note that the time trade-off (TTO) method has played an important role in generating value sets for the EQ-5D as one of the most widely accepted preference elicitation methods for health states (Torrance 1986) for economic evaluation and the method of choice in the first large-scale EQ-5D valuation study (Dolan 1997). Table 2.2 summarizes those 13 countries that have their own TTO value sets and describes the value sets.

    Table 2.2 Coefficients for the estimation of the EQ-5D index values based on TTO valuation studies
    Full size table

  • Country-specific VAS value set if available. Note that the Visual Analogue Scale (VAS) has become the other widely used method to elicit preferences for the EQ-5D, including nine countries. Table 2.3 summarizes countries that have their own VAS based value sets and describes the value sets, including the European value set.

    Table 2.3 Coefficients for the estimation of the EQ-5D index values based on VAS valuation studies
    Full size table

This means that for countries with no available value set from their own general population, only the European VAS value set based EQ-5D index values are summarised. However, for countries with available TTO and/or VAS value sets, additional population norms of EQ-5D index values are calculated.

To summarize key results on reported problems, EQ VAS, and EQ-5D index values, countries are tabulated in alphabetic order and are not ranked. Detailed country-by-country results are provided in the appendices. Because the population norms data are presented by age and gender, there is no need for the sample to have the same age distribution as the general population in each country. Therefore the data that are presented in the tables have not been standardized for age or gender. This means that international comparisons across several age groups should be made with caution as the demographic build-up by age and gender varies between countries, and that the samples of the general population used to create the tables do not necessarily follow that same distribution. However, international comparisons of data contained in a single cell (i.e. 1 age and gender group) are valid. The following section describes the methodology used to analyse cross-country differences in EQ-5D population data.

2.3 Methods of Cross-Country Analysis of EQ-5D Data

Cross-country summary data for reported problems by five dimensions and EQ VAS were estimated using a standardized population structure for all countries with national EQ-5D surveys. Countries were tabulated in alphabetic order. Standardization for age was performed to avoid bias due to the fact that some populations have a relatively higher proportion of elderly people. Age standardization of reported problems by dimension and EQ VAS was based on the European population structure using Eurostat data from 2010 (Table 2.4).

Table 2.4 European population age structure
Full size table

To explore reasons for cross-country differences in EQ-5D data, correlations between country-specific EQ-5D data (EQ VAS and five dimensions) and country-specific economic and health system macro indicators were calculated.

Living standards were estimated by means of Gross Domestic Product (GDP) per capita and unemployment rate. Indicators for health care system performance were health expenditure per capita and health expenditure as a % of GDP, number of hospital beds per 1,000 people and number of physicians per 1,000 people. The indicators were selected on the basis of a presumed or possible relationship with self-reported health. Data were obtained from the World Health Organization Statistical Information System and the World Bank. The data were from 2010 or the closest year with available data (Table 2.5). An alternative set of macro data was also used to see how results might change when using macro data from the same year as the EQ-5D data collection, including variables on gross national income on purchasing power parity, unemployment rate, and health expenditure data.

Table 2.5 Country-specific economic and health system macro indicators
Full size table

For all correlation analyses, non-parametric Spearman rank correlations were calculated. For this calculation, countries were ranked based on mean self-assessed health results, and their living standards and health care system performance characteristics. A high rank correlation means that the ranking of countries on one variable (e.g. prevalence of self-reported health problems) is similar to the ranking of another variable (e.g. GDP per capita).

2.4 Methods of Sociodemographic Analysis of EQ-5D Data

Two main approaches were used to derive socio-demographic indicators based on EQ-5D, based on odds ratios and concentration indices.

Logistic regression age-adjusted odds ratios for reporting problems on each EQ-5D dimension were calculated by age groups, gender, and education. An odds ratio higher than 1 indicates that the examined group reported more health problems than the reference group. The reference group was males, 18–24 years, with medium/high education.

Secondly, the analysis used the concentration index method, which is a single index measure of relative inequalities (Wagstaff et al. 1991; Kakwani et al. 1997). The overall health concentration index measures the mean difference in health between individuals as a proportion of the average health of the total population. This index can also be interpreted as a measure of how unequal the distribution of health is in the population. Health inequality is measured on a scale between 0 (meaning complete equality in health) and 1 (meaning complete inequality in health). Researchers also showed that the concentration index value also corresponds to 75 % of the Schutz index, and as such, it can also be interpreted as the proportion of health that should be redistributed from those above the average level to those below the average in order to equalize the distribution of health. (Koolman and Doorslaer 2004).

The overall concentration index can be decomposed to identify the impact of various factors, such as socio-demographic or quality of life characteristics, in order to determine how much each factor contributes to inequalities (Wagstaff and Doorslaer 2004; Clarke et al. 2010). In the current analysis, overall self-reported health was measured by the EQ VAS. Decomposition analysis was performed to determine inequalities by socio-demographic factors and by the EQ-5D dimensions, as well as in a combined model in which both socio-demographic and EQ-5D dimension variables were included.

The health concentration index for overall self-reported health, as measured by the EQ VAS, was computed by the convenient regression model as proposed by Kakwani et al. (1997):

$$ \frac{2{\sigma}_R^2}{\overline{ EQVAS}} EQVA{S}_i={\alpha}_i+{\gamma}_k{R}_i+{\varepsilon}_i $$

where R i is the relative fractional rank of the ith individual (ranked by the individual’s EQ VAS health), and γk is the estimated concentration index.

For the purposes of the decomposition analysis, the same estimation is used for all explanatory variables (by replacing EQ VAS with the explanatory variable in the equation and also using this variable for ranking purposes).

The total health concentration index can be written as the weighted sum of the concentration indices of the explanatory variables and the generalized concentration index of ε:

$$ \widehat{C}={\displaystyle \sum_k{\widehat{\eta}}_k}{\widehat{C}}_{xk}+G{\widehat{C}}_{\varepsilon } $$

where the weights are equal to the elasticities of EQ VAS score with respect to each explanatory variable in the model:

$$ {\widehat{\eta}}_k={\widehat{\gamma}}_k{\overline{x}}_k/\overline{ EQVAS} $$

where \( {\overline{x}}_k \) (the mean of xk explanatory variables: age, gender, education, EQ-5D problems) is multiplied by the coefficients for each explanatory variable that are taken from the linear regression model to explain EQ VAS:

$$ EQ- VAS=\alpha +{\displaystyle \sum_k{\gamma}_k}{\chi}_{ik}+{\varepsilon}_i. $$