-from Epidemiological
Bulletin , Vol. 25 No. 4, December 2004-
Measuring health inequalities is essential for analyzing the determinants of health and advancing a theory, which, in turn, is fundamental for action. Nevertheless, how to carry out these measurements is a subject of debate. There are various measurement methods, with differing levels of complexity, and choosing one rather than another depends on the objective of the study. The purpose of this article is to familiarize health professionals and decision-makers with the methodological aspects of measurement and the simple analysis of health inequalities, utilizing basic data that are reported regularly (for example, mortality, morbidity, and resources), aggregated by geopolitical unit (for example, a country or a state). However, the methods presented are applicable to the measurement of various types of inequalities and at different levels of analysis.
TYPES OF INDICATORS
Methodological considerations
Two areas for analysis of inequalities can be identified: health status and health services. Indicators for the measurement of health status basically use morbidity and mortality data; many of the studies published to date have used secondary mortality data or surveys. Measurements of inequalities in the area of health services use mainly survey data and incorporate concepts such as need, access, efficacy, effectiveness, and others that require a somewhat more complex methodology. This article looks only at measurement of inequalities in health status.
Studies that measure inequalities can be classified according to two factors: time and unit of analysis. With regard to the former, they may be cross-sectional or longitudinal, and with regard to the latter, individual or ecological.
In cross-sectional studies all the observations are carried out at a single point in time; although there may be several replications of each observation, they all refer to one moment in time. These studies tend to use vital statistics that contain information on social status, occupation, schooling, and other individual attributes, although they may also use secondary data from surveys conducted for various purposes, such as the Demographic and Health Surveys, which are carried out in 13 countries of the Region.Alternatively, a specific survey may be conducted to study inequalities. In longitudinal studies, on the other hand, the observations are made over a period of time, prospectively or retrospectively.
In individual studies, the unit of observation and analysis is the subject (all variables are recorded as individual attributes), while in ecological studies the unit of analysis is a conglomerate of individuals who are grouped together according to geo-demographic, socioeconomic, or other criteria. These studies are usually based on secondary data aggregated by geopolitical unit.
The principal drawback of analyses of aggregate data is the risk of assuming that the results found in populations (aggregates) are applicable or reproducible equally among individuals (ecological fallacy).1 Nevertheless, their great advantage is that they take into account social, geographic, and community factors of a contextual nature that cannot be analyzed in individual studies and that act as factors that confound or modify the effect of other proxy variables.
When individual data are used, the variables employed establish a ranking, both among and within the groups. This is the case, for example, with social position, schooling, and income level. In the ecological approach, on the other hand, the only ranking possible is among groups, since the attributes that are used (GNP, percentage of poverty, percentage of literacy, unmet basic needs, income ratio, unemployment rate, and others) lack meaning at the individual level.
This article assumes the use of secondary data already existing in countries, aggregated by geopolitical unit, such as those provided by the “Basic Indicators” initiative, which relies on subnational data already available in various countries of the Region. Analyzing health inequalities through these data will be very useful for public health policy-makers.
Because of the regional nature of the analyses carried out by PAHO/WHO, the examples given here use aggregate information by country, but the methods presented can be used for analysis in smaller geopolitical units (state, municipality, community, or neighborhood), depending on the objectives of the study.
The majority of traditional health indicators, such as mortality or morbidity from infrequent diseases, have very large standard errors and are therefore unstable when applied to small populations (under 100 000 inhabitants). Classical statistical techniques, both descriptive and inferential, are not applicable in these cases, and it is necessary to rely on weighting and the use of distributions appropriate to very infrequent events, such as the Poisson distribution.
In the examples given in this document, each geopolitical unit is considered an observation. These units can, in turn, be aggregated in socioeconomic groups according to the number of units studied, the indicator used, and the type of comparison to be drawn. In general, when observations are grouped together, information is lost and biases tend to appear (among them the ecological bias) when effects or associations are estimated. Whether or not to group geopolitical units is a decision made by the investigator.
There are various options for defining socioeconomic groups. One of them 2 consists of using per capita GNP to form groups, such that internal homogeneity is maximized. An example of aggregation of geographic units based on a socioeconomic indicator is the use of quintiles, which is one of the simplest ways of creating groups.
When analyzing social inequalities in the field of health, selection of the socioeconomic indicator is crucial because this variable defines the groups and the ranking within and between groups. A poor selection of indicator or of the categories created can bias the study. Obviously, when a single variable is used to define the socioeconomic status of the geopolitical units, as occurs in the examples in this article, the results cannot be extrapolated to the other factors that define socioeconomic status. The generalizations in some examples presented in this article should not be taken literally as their intent is purely didactic. Selection of the wrong socioeconomic indicator or an inappropriate definition of the categories of this indicator is one of the difficulties with aggregate studies.
Not all health inequalities are of social origin, but this article focuses on the type most frequently found in the literature. Social inequalities in health are health differences between groups of people categorized a priori on the basis of some significant characteristic relating to their socioeconomic position.1
Characteristics of indicators
Several major reviews of methodology for measuring inequality in health status exist. Those of Mackenbach and Kunst 3 and Wagstaff et al.4 have been taken as the basis for this article.
Each indicator has advantages and disadvantages and each serves different purposes. The selection of the indicator should be consistent with the theoretical framework and objectives of the study. Good indicators of inequality should:
1) reflect the socioeconomic dimension of health inequalities,
2) use information on all population groups defined by the indicator, and
3) be sensitive to changes in the distribution and size of the population across socioeconomic groups. 4
Regardless of the type of indicator used, it is extremely important that the information be of good quality and can be validated. Any method used should include a descriptive analysis of variation among groups in the phenomenon studied.
Indicators differ in complexity depending on the objectives of the study. Mackenbach and Kunst 3 recommend that decision-makers use simple methods, but that investigators confirm the results using more complex methods.
Measurements can be expressed as relative differences (e.g., rate ratio) or absolute difference (e.g., rate difference); both are important and tend to have complementary value. Relative measures are more stable and readily understood. In some cases, however, absolute measures are more useful for decision-makers, especially when goals have been set, because they enable a better appraisal of the magnitude of the public health problem. Absolute measures can be derived from relative ones and vice versa.5
Another methodological option is to use measures of the effect or impact of socioeconomic status on health. The essential difference between the two options is that impact measures take into account the actual socioeconomic situation and measure changes in health conditions that can be expected as a result of potential interventions. For this reason, impact measures are especially important for decision-making and the formulation of public policies designed to achieve equity.2
Measures of effect are based on fixed categories of the socioeconomic variable (e.g., primary schooling versus university education). Measures of impact, in contrast, use categories defined by a socioeconomic indicator quantifiable in population terms (e.g., highest income quintile versus lowest income quintile), such that if the distribution of the indicator varies, it also modifies the measurement of inequality.
Among measures of effect, rate ratio and rate difference are two of the most frequently used indicators. Another is regression- based effect index. One of the best-known indicators of total impact in the health field is population attributable risk (PAR), adapted from epidemiology. This indicator can also be estimated through regression analysis. Regression is also used to estimate the slope index of inequality (SII) and the relative index of inequality (RII). The index of dissimilarity is another example of an impact measure.3,4
Indicators from the field of economics are also used to measure inequalities in health, such as the Gini coefficient, with its corresponding Lorenz curve, and variants of both, such as the concentration index and curve, which combine indicators with graphic representations.
Notes:
A) Braveman P. Challenges in monitoring social inequalities in health: examples from a few continents (draft). Rockefeller Foundation Global Health Equity Initiative, 1999.
B) The relationship between the concepts of effect and impact can be compared to the one that exists between the well-known epidemiological concepts of relative risk and attributable risk.
REFERENCES (Part I):
1) Greenland S, Morgenstern H. Ecological bias, confounding and effect modification. Int J Epidemiol 1989;18:269–274.
2) Pan American Health Organization. The health situation. In: Annual Report of the Director. 1996 Edition. Washington, DC: PAHO; 1997. (Official Document No. 283).
3) Mackenbach JP, Kunst AE. Measuring the magnitude of socio-economic inequalities in health: an overview of available measures illustrated with two examples from Europe. Soc Sci Med 1997;44:757–771.
4) Wagstaff A, Paci P, Van Doorslaer E. On the measurement of inequalities in health. Soc Sci Med 1991;33:545–557.
5) Kunst AE, Mackenbach JP. Measuring socioeconomic inequalities in health. WHO Regional Office for Europe, 1994. (Document EUR/ ICP/RPD 416). http://www .who.dk/Document/PAE/Measrpd416.pdf. 12 November 2002.
Source: Originally published with the title “Métodos de medición de las desigualdades de salud” in Pan American Journal of Public Health 12(6), 2002.
Authors: Maria Cristina Schneider, Carlos Castillo-Salgado, Jorge Bacallao, Enrique Loyola, Oscar J. Mujica, Manuel Vidaurre, and Anne Roca.
Return to index
Epidemiological Bulletin , Vol. 25 No. 4, December 2004