Measuring socioeconomic status in multicountry studies: results from the eight-country MAL-ED study

Socioeconomic status (SES) is a theoretical construct encompassing individual, household, and/or community access to resources. It is commonly conceptualized as a combination of economic, social, and work status, measured by income or wealth, education, and occupation, respectively [1, 2]. SES has been linked to a wide range of health-related exposures and outcomes, including child undernutrition, chronic disease, and infection [2, 3]. In a review of risk factors for adverse outcomes in child cognitive development, Walker and colleagues [4] conceptualized poverty as underlying more proximal psychological and biological risk factors, including maternal depression and nutrient deficiencies. More recently, researchers have highlighted connections between childhood SES and lifetime health outcomes, such as heart disease and chronic obstructive pulmonary disease [5].

Literature on SES measurement distinguishes between wealth, or accumulated financial resources, and income, a measure of shorter-term access to capital [2]. Researchers have identified challenges in collecting income data, particularly in low-income settings, due to monthly fluctuations, informal work, and reporting biases [6]. Recent empirical work has drawn attention to the approach of supplementing or replacing information on income with direct measures of wealth, such as household assets [7]. Perhaps the most widespread approach to direct measurement of household wealth is that used by the Demographic and Health Surveys (DHS), implemented in more than 90 countries since 1984 [8]. Using nationally representative data from India, Filmer and Pritchett [7] created an index based on household ownership of assets and housing materials to serve as a proxy for wealth. The resulting index was internally valid and coherent, and robust to the choice of assets. Using additional data sets from Indonesia, Nepal, and Pakistan, they further argued that a composite asset index is as reliable as data on household consumption and is less subject to measurement error [7]. Their statistical approach, using principal components analysis (PCA), has since been adapted to create a household wealth index in each DHS dataset [8]. Concerns about this approach include its over-representation of urban settings, and its failure to distinguish between the poorest of the poor, particularly in rural areas [9]. Furthermore, this approach requires lengthy surveys of household assets. Several studies have found that rapid wealth appraisals requiring as few as four survey questions perform as well as the DHS wealth index in categorizing households [10] and predicting mortality [11].

Multicountry studies pose an added challenge to measuring SES. While approaches focused on asset ownership are often sufficient for homogenous populations, studies in more diverse populations must explore whether variables that measure SES, such as ownership of specific assets, have the same meaning across populations. The DHS wealth index is derived using country-specific data rather than globally pooled data. One result is that a household in the poorest wealth quintile in Egypt might be wealthier than a household in the richest wealth quintile in Ethiopia. Therefore, controlling for SES in pooled analyses using this approach, either by raw score or wealth quintile, is inappropriate. The United Nations Development Programme (UNDP) sought to overcome the challenge of comparing household SES across countries through the Multidimensional Poverty Index (MPI), introduced as an experimental measure in the 2010 Human Development Report. The MPI includes three equally weighted dimensions of household SES: education (years of schooling, school attendance), health (child mortality, nutrition), and standard of living (household attributes, asset ownership) [12].

The Malnutrition and Enteric Infections: Consequences for Child Health and Development (MAL-ED) study seeks to explore relationships between early exposures to malnutrition and enteric infections and their consequences for child growth and cognitive development across eight sites. Geographic, cultural, and socioeconomic differences between these sites present an added challenge to developing a measure of SES that is relevant in all sites. We sought to compare different approaches to measuring SES in resource-limited settings, and provide guidance for measuring SES accurately and simply in epidemiologic studies of diverse populations.

We compared three approaches to measuring household wealth with maternal education and selected the random forests approach due to its consistent association with HAZ and simplicity of use and interpretation of selected assets relative to PCA. Although maternal education was the simplest approach as it only required one variable, it did not reflect theoretical models of household SES, an important consideration in choice of variables [24]. In addition, maternal education is strongly influenced by culture. Our data support previous evidence that, while wealth and education are important components of SES, in many settings they do not measure the same exposures [25].

We then combined our selected wealth measure with access to improved water and sanitation, maternal education, and household income to form a complete SES measure. Krieger and colleagues [26] argue that the term SES “blurs distinctions between two different aspects of socioeconomic position: (a) actual resources, and (b) status, meaning prestige- or rank-related characteristics.” We use the term SES rather than socioeconomic position because the latter is intended to explicitly include prestige-based measures, as linked to social class. Our measure focuses only on the actual resources in a household because we did not collect prestige-based indicators in our study. Hackman and Farah [25] emphasize the importance of a clear conceptual framework driving the measurement of socioeconomic status. In the MAL-ED study, which focuses on the relationships between enteric infections in infancy and subsequent growth and cognitive development, access to improved water and sanitation sources are important risk factors that are closely related to SES. Our final measure requires 12 variables, most of which are easily collected across diverse settings, which is a requirement in order for an SES measure to be applicable in a multicountry study.

Compared to maternal education alone, as well as more complete measures of wealth, the WAMI index demonstrated a significantly stronger association with HAZ across our eight study sites. These results provide evidence of the importance of using a robust measure of household SES, even as an adjustment factor, rather than a measure of wealth or education alone. Based on analyses of a nationally representative sample in the United States, Braveman and colleagues [25] found that, depending on the choice of income, education, or both as a control for SES, the ranking of racial groups by odds of poor health changed dramatically. Many studies also seek to simplify measurement of SES by identifying the variables that are most strongly associated with their outcomes. Daly and colleagues [27] found that wealth and income indicators were more strongly associated with mortality than education and occupation indicators in a United States cohort. In settings where collection of income data is not feasible, an expanded measure of SES including wealth, education, and water and sanitation is still a significant improvement over wealth or maternal education alone (results not shown).

Our study has a number of important strengths. It includes data from households from eight country sites located in South Asia, sub-Saharan Africa, and Latin America to derive a multicountry index. The experienced staff in these eight sites used a standardized protocol with identical questionnaires to collect demographic and SES information. The survey included extensive questions related to SES, due to an a priori interest in identifying an appropriate approach to measuring this construct. Our analyses systematically compared both commonly used and new approaches to measuring wealth and SES. Our final selection of a measure of socioeconomic status balances statistical and theoretical strength with feasibility in a field research setting.

The results of this study should be considered in light of some limitations. Although the MAL-ED study sites are located in eight diverse country settings, they are not nationally representative samples. Therefore, the results cannot be generalized to national comparisons or compared to country-level indices such as the DHS. More broadly, the resulting components of an ideal measure of SES are likely to vary across settings and study objectives. Consistencies across our study sites indicate that it is possible to identify measures of SES that are relevant in diverse settings. However, the actual variables that form an ideal measure will be informed by the study populations, settings, and research questions. Rather than selecting indicators to be used by all studies, our findings are intended to demonstrate the importance of developing a measure of socioeconomic status that is theoretically sound and contextually relevant, especially in multisite or multicountry studies. There is no gold standard measure of socioeconomic status against which to compare our proposed measure, particularly in a multicountry study setting. However, associations between the WAMI index and HAZ demonstrate construct validity of the measure, since we expect these constructs to be associated theoretically. Another limitation of our study is that we do not have a measure of occupation. Much of the work on socioeconomic status that includes occupation as a key component is based on research in high-income settings. In many low-income settings, this concept does not distinguish between households as readily, either due to homogeneity or instability of income sources. Alternatives include caste or religious group. While we collected data on these groupings, we did not feel that they were sufficiently comparable across the eight study sites. Similarly, indicators of prestige or rank-related characteristics have been included in measures of socioeconomic position [26], but these are unavailable in our dataset. With the exception of data on access to electricity, our available SES indicators were measured at the household level, rather than the community level. Previous research has shown that the addition of community-level variables to SES measures can be helpful in exploring trends and inequalities in health outcomes [28].

In summary, novel classification approaches such as random forests provide an alternative to the more widely used PCA for the measurement of household wealth. However, assets alone are not sufficient to capture the full domain of socioeconomic status. We developed a simplified socioeconomic index that combines measures of improved water and sanitation, assets, maternal education, and household income that may be applicable to a multicountry setting. We believe that this measure is an improvement over the commonly used PCA-based wealth measurement approach for several reasons: 1) It is a robust measure that more fully reflects a theoretical understanding of SES; 2) it reduces the data collection burden by highlighting a priority set of indicators for measurement, in contrast to commonly used PCA approaches, which require collecting data on a full set of indicators, even if some are irrelevant; and 3) it is computationally simple to apply, once the priority assets have been selected using the random forests technique. With further validation, this simplified WAMI index may provide a standardized approach for adjustment across diverse study populations.