Short communication
Measurement of horizontal inequity in health care utilisation using European panel data

https://doi.org/10.1016/j.jhealeco.2008.09.008Get rights and content

Abstract

Measurement of inequity in health care delivery has focused on the extent to which health care utilisation is or is not distributed according to need, irrespective of income. Studies using cross-sectional data have proposed various ways of measuring and standardizing for need, but inevitably much of the inter-individual variation in needs remains unobserved in cross-sections. This paper exploits panel data methods to improve the measurement by including the time-invariant part of unobserved heterogeneity into the need-standardization procedure. Using latent class hurdle models for GP and specialist visits estimated on 8 annual waves of the European Community Household Panel we compute indices of horizontal equity that partition total income-related variation in use into a need- and a non-need related part, not only for the observed but also for the unobserved but time-invariant component. We also propose and compare a more conservative index of horizontal inequity to the conventional statistic. We find that many of the cross-country comparative results appear fairly robust to the panel data test, although the panel-based methods lead to significantly higher estimates of horizontal inequity for most countries. This confirms that better estimation and control for need often reveals more pro-rich distributions of doctor utilisation.

Introduction

An equitable system of health care delivery appears to remain a core objective in most of the OECD member states with comprehensive and universal coverage and proposed health system reforms usually quote equity preservation or improvement as an important goal (Van Doorslaer et al., 2006). Because in many countries horizontal equity is being interpreted as the principle of equal treatment for equal need, health economists have typically approached the measurement of inequity using inequality measures (Wagstaff and Van Doorslaer, 2000a). In most empirical work, horizontal inequity is measured as the degree to which utilisation is still related to income after differences in needs across the income distribution have been appropriately standardised for (Wagstaff and Van Doorslaer, 2000b). Several cross-country comparisons have adopted variants of these methods to compare across countries in the European Union (Van Doorslaer et al., 2004), in the OECD (Van Doorslaer et al., 2006) and in Asia (Lu et al., 2007).

Invariably, these comparative studies have relied on cross-sectional surveys and have adjusted for needs by comparing actual utilisation distributions (by income) with need-predicted utilisation using some regression-based standardization procedure. This means that adjustment can only be made for need differences that are observed in general, self-reported, health questions which are common across a large number of surveys. Typically, only a small fraction of the inter-individual variation in utilisation measures like doctor visits can be explained by these models. And while an individual's demographic and self-reported health characteristics are known to be very powerful predictors of health care utilisation, nonetheless most of the inter-individual variation remains unexplained.

This paper aims to go beyond the earlier approaches in at least three ways. First, the availability of the full eight waves of the European Community Household Panel (1994–2001), and the development of appropriate models for analysing panel utilisation data (Bago d’Uva, 2006, Bago d’Uva and Jones, 2009) provides an opportunity to further examine the unexplained variation in use. We use estimation results of the preferred specification to model our data in Bago d’Uva and Jones (2009), the latent class hurdle model. This specification captures the time-invariant components of individual unobserved heterogeneity, by modelling the latent class membership probabilities as functions of individual characteristics. Not only does this explain a greater share of the variation, it also allows for the possibility to partition the contribution of individual unobserved heterogeneity into need and non-need factors. In that way, some of the explicitly modelled individual heterogeneity can be included in the computation of the inequity index in much the same way as the observed heterogeneity.

Secondly, by using a multi-year period to assess the degree to which there are any deviations between actual and needed utilisation distributions, we move from a short to a more robust long-run perspective. Jones and López Nicolás (2004) have proposed short-run and long-run measures of income-related health inequality to examine the phenomenon of health-related income mobility. Adapting their method to our analyses of inequity in health care use, we are able to adopt a long-run perspective that accounts for this phenomenon. We find that the upwardly mobile in the income distribution are more likely to use health care, especially specialist services.

Finally, we propose a new measure of horizontal inequity in health care use that differs from the standard measure in the way that the variation left unexplained by the regression models is regarded, and which we label “conservative” index. The “conventional” index of horizontal inequity is defined as a residual and labels as inequity all income-related inequality in use that is not demonstrably related to needs. That means that all the residual income-related variation, not explained by either the need or the non-need variables, is assumed to be inequitable. But some of this residual variation may in fact be due to unobservable need differences. An alternative is to treat only the income-related inequality that is demonstrably related to non-need variables as our index of inequity. The difference between the two indices depends on the degree to which the income-related variation that is not due to included need and non-need variables is pro-rich or pro-poor. Our comparison shows that the “conservative” index tends to give higher estimates.

In what follows, we first explain how we will proceed with measuring inequity using panel data, then we describe the data we have used and the results obtained for the European countries considered. In the final section we discuss what we can and cannot conclude from this study.

Section snippets

Cross-sectional and longitudinal measures of inequality

We measure income-related inequality in the utilisation of health care (GP and specialist visits) in each wave t, using the concentration index CIt (Wagstaff et al., 1991, Kakwani et al., 1997) of the number of doctor visits. The concentration index takes on a positive/negative/zero value when there is pro-rich/pro-poor/no inequality.

Until recently, research on health equity by economists was focused on measures of socioeconomic inequalities in health and health care that were designed for use

Data

The data are taken from the European Community Household Panel User Database (ECHP-UDB). The standardised questionnaire allows for cross-country as well as longitudinal comparisons. We use all 8 waves available for 10 EU member states: Austria, Belgium, Denmark, Finland, Greece, Ireland, Italy, Netherlands, Portugal and Spain. Austria joined the survey in 1995 (wave 2) and Finland only in 1996 (wave 3).5

Results

This section presents an analysis of horizontal inequity in the utilisation of primary (GP) and secondary (specialist) care for 10 EU member states. Predictions of (non-)need-expected health care utilisation were obtained as explained in Section 2.3, from LCH models estimated separately for each country and for GP and specialist visits by Bago d’Uva and Jones (2009). We then use the methods described in Section 2.2 to compute short-run indices (by wave), and their weighted average, and long-run

Conclusion

Achieving equitable access to health care for all citizens, irrespective of their incomes, remains an important public health policy goal in Europe's largely publicly funded health care systems. Key to the monitoring of the extent to which various systems are successful in attaining this goal is the appropriate and reliable measurement of the degree of inequity in the distribution of health care. Over the last two decades, Europe has invested heavily in the collection of comparable data to

Acknowledgements

The European Community Household Panel UsersDatabase, version of December 2003, was supplied by Eurostat. This paper derives from the project “The dynamics of income, health and inequality over the lifecycle” (known as ECuity III Project), which is funded in part by the European Commission's Quality of Life and Management of Living Resources programme (contract QLK6-CT-2002-02297). Teresa Bago d’Uva was funded by Fundação para a Ciência e Tecnologia, under PhD grant SFRH/BD/10551/2002. Teresa

References (23)

  • A. Clark et al.

    Heterogeneity in reported well-being: evidence from twelve European countries

    The Economic Journal

    (2005)
  • Cited by (94)

    • Biomarkers, disability and health care demand

      2020, Economics and Human Biology
    View all citing articles on Scopus
    View full text