Area-based units of analysis for strengthening health inequality monitoring

Inequalities in health persist worldwide and one of the starting points for remedial action is collecting data that reveal patterns of inequality. Current discussions about the best ways of monitoring health inequalities emphasize disaggregating data by variables such as socioeconomic status, geographical area or sex. The sustainable development goals (SDGs) adopted in 2015 include a call for countries to increase the availability of disaggregated data as part of the aim to strengthen data monitoring and accountability (SDG target 17.18).1 Yet countries have varying capacities for monitoring health inequality. This is due in part to data-related issues such as weaknesses in the health information systems, especially in many lowand middle-income countries; lack of availability or poor quality of health data; and a limited ability to disaggregate data across all health topics within countries.2 Overcoming these challenges in the long term requires substantial investments in the health information infrastructure.3,4 In the short-term, countries need innovative approaches to best harness the potential of their existing data to improve monitoring efforts. Current approaches to health inequality monitoring tend to focus on data collected through household health surveys. These provide two streams of data – about health indicators and about the dimensions of inequality – at the individual or household level. This makes such surveys the main source of data for within-country monitoring of health inequality especially in lowand middleincome countries. However, household health surveys have certain limitations. In many lowand middle-income countries they tend to cover only a narrow set of topics, such as reproductive, maternal, newborn and child health. Other health topics, such as infectious diseases or road traffic injuries, are rarely the focus of household surveys. Household health surveys and their consequent reporting tend to be done outside the regular activities of the health information system, and are resource intensive. Furthermore, data from household surveys may not be representative of small subpopulations of interest, and so cannot be used for certain purposes, such as assessing cross-district inequality, due to too small sample size at that administrative level. By increasing the use of area-based units of analysis, including greater integration of data from other reliable data sources – including vital registration systems, censuses and administrative data – the possibilities for health inequality monitoring may be strengthened and expanded across health topics. In this article we make the case for stratifying data at the level of subnational geographical regions, such as provinces, states or districts. The wider use of an area-based unit of analysis as a complementary way to analyse data at the individual or household level has certain practical advantages that are relevant to lowand middle-income countries as well as high-income countries. First, this approach opens up new possibilities concerning the data that can be used for within-country monitoring, in terms of both health data and data about dimensions of inequality. In some cases, individual or household data on both health and inequality dimensions may be unavailable in one data source; if these data were available from different data sources (e.g. those that collect data at the level of subnational regions), alternative ways of capturing area-level estimates may provide an insight into the extent of inequality. For instance, whereas data about economic status, race, ethnicity, migratory status or disability may not always be collected alongside health data at an individual or household level, they may be available by region. Subnational regions are often aligned with administrative districts, which facilitates the use of administrative-level data. For example, the distribution of health system inputs and outputs (e.g. service delivery) can be compared to health determinants (e.g. district-level poverty, education or employment). Second, since interventions to reduce inequities are likely to be implemented at the local administrative level, regional monitoring of health inequalities may be a useful tool for benchmarking, with implications for resource allocation, planning and evaluation. This is particularly true when a country’s health system administration is decentralized because substantial differences may exist across geographical areas.5 Third, area-based measures may provide a more intuitive understanding of health inequalities and may help to identify possible points for intervention. Geographically defined subpopulations are by nature easy to identify and locate, and health interventions may thus be effectively targeted to disadvantaged regions.6 For example, measuring health inequality on the basis of household wealth using asset-based indices may pose limitations in terms of identifying and reaching disadvantaged subpopulations, as the poorest segment of the population may be located throughout different regions of a country.7,8 Alongside these advantages, some caution is needed when adopting an area-based unit of analysis. There is the risk of committing a so-called ecological fallacy (i.e. making assumptions about individuals based on population-level patterns, or in this case, erroneously drawing conclusions about the health of individuals using area-based data). For instance, if richer districts were found to have a higher prevalence of road traffic injuries it could not be assumed that road traffic injuries are more prevalent among richer individuals. Also, ethical Area-based units of analysis for strengthening health inequality monitoring Ahmad Reza Hosseinpoor & Nicole Bergen