Data sources for measuring the socioeconomic gradient of hypertension in rural populations of low- and middle-income countries

Busingye et al. conducted a systematic review and meta-analysis to ‘systematically quantify the association between socioeconomic status (SES) and hypertension in rural populations of low- and middle-income countries (LMICS).’ 1 Given the growing burden of non-communicable diseases (NCDs) and their risk factors in LMICS, the apparent differences in prevalence by level of urbanicity in these settings, and the recent debate on SES gradients of NCDs and their risk factors, 2–5 this is an interesting analysis. However, in order to answer this question comprehensively, all possible data sources should be examined, including those from health information systems that may not be published in the peer-reviewed literature. 6 Specifically, there are unit-level datasets available from nationally representative population surveys that would provide additional information to answer the authors’ question. Two such data sources, the Measure Demographic and Health Surveys (DHS) and the Study on Global Ageing and Adult Health (SAGE), and their potential contribution to the research question are presented. The Measure DHS programme has been in operation since 1984 and has conducted more than 300 surveys in 90 countries. 7,8 The DHS have traditionally focused on maternal and child health, nutrition and fertility. However, recent country surveys have collected an array of biomarkers, including the measurement of blood pressure. 8,9,Table 1 summarizes the 15 surveys from 13 LMICS that included blood pressure measurement (in rural and urban populations) and for which unit-level data are available, as presented on the Measure DHS website. 10 In addition, there are five surveys (from Egypt, Ghana, India, Lesotho and South Africa), recently completed or ongoing, that included blood pressure measurement and will have data available in the near future. 10 Measure DHS and SAGE surveys with blood pressure measurements and unit-level data available Data summarized from the DHS website, SAGE website and country survey reports. DHS, Measure Demographic and Health Surveys; SAGE, Study on Global Ageing and Adult Health. a Response rates not reported. Measure DHS and SAGE surveys with blood pressure measurements and unit-level data available Data summarized from the DHS website, SAGE website and country survey reports. DHS, Measure Demographic and Health Surveys; SAGE, Study on Global Ageing and Adult Health. a Response rates not reported. The World Health Organization’s SAGE is a longitudinal study of adults aged 18 years and over, conducted in nationally representative samples in six countries (China, Ghana, India, Mexico, Russia and South Africa). 11 The first round of SAGE surveys was conducted in 2007-10, and the second round, aiming to follow up the same participants, is currently under way. A further third round is planned for 2016. 11,12 The samples comprise both rural and urban populations, including a large sample of those aged 50 years and over, who were the focus of the study ( Table 1 ). 11,12 Busingye et al . included 36 articles from 19 countries in their systematic review, with a total approximate sample population of 310 000. 1 Only one nationwide survey was included in their analysis. 1 The DHS and SAGE surveys together included 85 432 individuals in rural areas of 18 countries, including 13 countries (and 5 regions) not included in Busingye et al .’s analysis. 1 Furthermore, the DHS and SAGE aim to provide nationally representative samples of their target population ( Tables 1 and 2 ). Busingye et al. assessed the association between hypertension and three measures of SES status for their meta-analysis. 1 There were data available from: 13 countries for education; 7 for occupation; and 7 countries and 1 region (sub-Saharan Africa) for household assets, income and social class ( Table 2 ). 1 By comparison, each DHS and SAGE survey provides all three measures of SES status for 18 countries ( Table 2 ). 10,11 It is important to note some of the limitations of DHS and SAGE data. For example, some surveys covered only women, others did not include older age groups, and the response rates were low in a couple of surveys ( Table 1 ). Countries by type of socioeconomic status measure included in Busingye et al .’s meta-analysis compared with countries covered by Measure DHS and SAGE survey DHS, Measure Demographic and Health Surveys; SAGE, Study on Global Ageing and Adult Health; SES, socioeconomic status. a Seven countries and the region of sub-Saharan Africa. Countries by type of socioeconomic status measure included in Busingye et al .’s meta-analysis compared with countries covered by Measure DHS and SAGE survey DHS, Measure Demographic and Health Surveys; SAGE, Study on Global Ageing and Adult Health; SES, socioeconomic status. a Seven countries and the region of sub-Saharan Africa. An interesting aspect of the SAGE datasets is that they are designed to provide longitudinal data by resampling the same households in future rounds (2014 and 2016). 11 Though the data from subsequent rounds of SAGE are not yet available, they have the potential to improve our understanding of the evolution of the association between SES and hypertension in rural populations, a need highlighted by Busingye et al . 1 Another important feature of DHS and SAGE is that they cover a broad range of health issues, including communicable diseases, which some argue still form the majority of the disease burden in lower SES populations. 2,10 Thus, these data can contribute to further informing the debate around prioritization of health interventions in lower SES populations. 2,3 There are several advantages to using unit-level data for analysis. Busingye et al . noted the heterogeneity of the study methods and analytical models as a major impediment to the meta-analyses performed. 1 The availability of the unit-level data allows researchers to conduct the analysis required to answer the research question, with the appropriate adjustment for confounding. This eliminates many of the issues which occur when attempting to pool data from heterogeneous studies that have applied varying analytical models to answer different research questions. Additionally, the DHS and SAGE programmes use standardized questionnaires and survey-sampling methodologies, with country-specific adjustments, increasing the comparability of data between countries. 8 In fact, the SAGE data have been used recently for a cross-country comparison of the prevalence, awareness and treatment of hypertension. 13 Accessing unit-level data also avoids the exclusion of relevant datasets because they have not been analysed in the way that is required for the meta-analysis, an issue not clearly documented by Busingye et al.1 As the rising burden of non-communicable diseases in LMIC is being recognized as a significant global health challenge, more epidemiological surveys are being conducted measuring the prevalence of NCDs and their risk factors. Some of these may not be published in peer-reviewed journals, but have unit-level data available to researchers. As highlighted in this letter, these data can play an important role in answering some of the contentious questions, including the question of the SES gradient of these conditions and their risk factors.