An equity dashboard to monitor vaccination coverage

Abstract Equity monitoring is a priority for Gavi, the Vaccine Alliance, and for those implementing The 2030 agenda for sustainable development. For its new phase of operations, Gavi reassessed its approach to monitoring equity in vaccination coverage. To help inform this effort, we made a systematic analysis of inequalities in vaccination coverage across 45 Gavi-supported countries and compared results from different measurement approaches. Based on our findings, we formulated recommendations for Gavi’s equity monitoring approach. The approach involved defining the vulnerable populations, choosing appropriate measures to quantify inequalities, and defining equity benchmarks that reflect the ambitions of the sustainable development agenda. In this article, we explain the rationale for the recommendations and for the development of an improved equity monitoring tool. Gavi’s previous approach to measuring equity was the difference in vaccination coverage between a country’s richest and poorest wealth quintiles. In addition to the wealth index, we recommend monitoring other dimensions of vulnerability (maternal education, place of residence, child sex and the multidimensional poverty index). For dimensions with multiple subgroups, measures of inequality that consider information on all subgroups should be used. We also recommend that both absolute and relative measures of inequality be tracked over time. Finally, we propose that equity benchmarks target complete elimination of inequalities. To facilitate equity monitoring, we recommend the use of a data display tool – the equity dashboard – to support decision-making in the sustainable development period. We highlight its key advantages using data from Côte d’Ivoire and Haiti.


Introduction
The 2030 agenda for sustainable development calls upon the international community to prioritize the needs and rights of the most vulnerable, so that no one is left behind. 1 Some population groups in low-and middle-income countries continue to be systematically missed by lifesaving health interventions, such as childhood immunization. 2 Determining effective ways to measure and monitor progress in addressing social exclusion is a priority for those involved in implementing the sustainable development goals (SDGs), including for Gavi, the Vaccine Alliance.
Gavi's strategy for 2016-2020 includes a focus on equity in vaccination coverage. 3 The 2016-2020 strategy has catalysed a re-examination of Gavi's approach to monitoring equity. In an earlier phase of its operations, Gavi's principal indicator of equity was the difference in coverage between a country's richest and poorest wealth quintiles. 4 This indicator is also used by many other international organizations and has been widely applied in health equity analyses. 5 Several questions have emerged in relation to Gavi's equity monitoring for the 2016-2020 period. Is the earlier approach, based largely on a single dimension and measure, sufficient to monitor progress in addressing inequalities in the SDG period? Is the wealth quintile the most appropriate indicator to assess inequalities in access to vaccines? Does the coverage gap between rich and poor capture the full extent of inequality and allow meaningful international comparisons?
To answer these questions, we conducted a systematic analysis of inequalities in childhood vaccination coverage based on different measurement approaches, using the most recent (as of May 2015) demographic and health surveys (DHS) in 45 Gavi-supported countries. 6 Our findings enabled us to for-mulate recommendations on the equity monitoring approach that would best support policies and practices tailored to the most vulnerable. In particular, we propose the use of a data display tool -the equity dashboard -to monitor inequalities and support decision-making in the SDG period. The purpose of this article is to explain the rationale for the recommendations and for developing the equity dashboard. We illustrate the key advantages of the dashboard using examples from specific countries.

Dimensions of vulnerability
A first consideration in equity monitoring is how to define the vulnerable; that is, to determine which population characteristics describe those who are left behind. For example, SDG17 highlights income, sex, age, race, ethnicity, migratory status, disability and geographical location as important sources of vulnerability. This calls for a disaggregation of data according to these factors and other characteristics relevant to national contexts. 7 For Gavi, the indicators selected must be applicable globally and must allow comparisons across countries. Therefore, although they may be highly relevant in specific settings, certain characteristics such as race, ethnicity or religion will be impractical for the global monitoring of inequalities since they are measured differently across countries and reflect local social structures. 8 We used a framework from the World Health Organization's (WHO) Commission on Social Determinants of Health 9 to review potential determinants of child health inequities. From this framework, we identified five indicators and composite measures: (i) the wealth index; (ii) maternal education; (iii) place of residence; (iv) child sex; and (v) the multidimensional poverty index. These were chosen as they Abstract Equity monitoring is a priority for Gavi, the Vaccine Alliance, and for those implementing The 2030 agenda for sustainable development. For its new phase of operations, Gavi reassessed its approach to monitoring equity in vaccination coverage. To help inform this effort, we made a systematic analysis of inequalities in vaccination coverage across 45 Gavi-supported countries and compared results from different measurement approaches. Based on our findings, we formulated recommendations for Gavi's equity monitoring approach. The approach involved defining the vulnerable populations, choosing appropriate measures to quantify inequalities, and defining equity benchmarks that reflect the ambitions of the sustainable development agenda. In this article, we explain the rationale for the recommendations and for the development of an improved equity monitoring tool. Gavi's previous approach to measuring equity was the difference in vaccination coverage between a country's richest and poorest wealth quintiles. In addition to the wealth index, we recommend monitoring other dimensions of vulnerability (maternal education, place of residence, child sex and the multidimensional poverty index). For dimensions with multiple subgroups, measures of inequality that consider information on all subgroups should be used. We also recommend that both absolute and relative measures of inequality be tracked over time. Finally, we propose that equity benchmarks target complete elimination of inequalities. To facilitate equity monitoring, we recommend the use of a data display tool -the equity dashboard -to support decisionmaking in the sustainable development period. We highlight its key advantages using data from Côte d'Ivoire and Haiti.

Policy & practice
Monitoring equity in vaccination coverage Catherine Arsenault et al.
describe population groups at risk of differential vaccination coverage; are available in household surveys in low-and middle-income countries, such as DHS and multiple indicator cluster surveys (MICS); and are measured in a similar way across countries.
Although no single indicator is ideal, each has a claim to be relevant to equity monitoring. The wealth index is the indicator most commonly used in equity monitoring. It measures socioeconomic position and attempts to capture the material aspects of living conditions. 10 Often called the asset index, it is based on a household's ownership of assets, its access to safe drinking water and the quality of sanitation and housing. It is initially calculated as a continuous index and often split into quintiles that allow for straightforward comparisons between the richest and poorest quintiles within a country. 11 Nonetheless, the wealth index has an important limitation for cross-country comparisons, as it only measures relative socioeconomic position, i.e. specific to a given country. The bottom (poorest) quintile in a middle-income country may be better off than the top (richest) quintile in a low-income country. 12 Parental education, another indicator of socioeconomic position, is related to health and vaccination status through multiple pathways. 9 Educated mothers generally have better knowledge of good medical practices, a higher social status and greater autonomy and decisionmaking power, making them more able to communicate with and access health services. 13 Determining whether mothers with little or no education are able to access vaccination services should be a part of Gavi's equity monitoring strategy. In addition, in contrast to wealth quintiles, education levels can be better translated between low-and higher-income countries. 10 However, education has an added complexity for inequality measurement. Unlike wealth quintiles, which divide population samples into five groups, each containing about 20% of all households, the number of individuals in each category of education can be uneven. For example, there may be few (or no) mothers without any formal education in middle-income countries, and few mothers who completed secondary school in low-income country samples.
Geographical location is another important determinant of child health inequities. 9 To compare geographical in-equalities across countries, an indicator for urban or rural residence can be used. Policy-makers should consider whether vaccination services are reaching rural areas. However, urban and rural subgroups are broad categories that could conceal some of the geographical inequality present. For example, relatively wealthy urban areas and urban slums with mass deprivation may be collapsed into a single category. 8 A further breakdown by smaller geographical units at national levels might offer a more precise picture of inequalities.
Although progress has been made in the past decades towards reducing gender inequalities in access to health resources worldwide, differences between the sexes in child health outcomes remain in some countries. 14 Determining whether girls and boys have equal access to immunization services is crucial from a human rights perspective.
Conceptually, another important source of inequity relates to the concurrence of key health determinants that place some children simultaneously at greater risk of acquiring vaccine-preventable diseases and at lower probability of surmounting them. Unimmunized children are often differentially exposed to further health risks, such as inadequate nutritional intake, poor water and sanitation, indoor air pollution and overcrowding, all of which increase the risk of infectious diseases. 9,15 Unimmunized children are also more vulnerable to poor health outcomes once infected, as their parents often lack health knowledge and access to other preventive interventions and medical care. 15,16 These systematic, overlapping deprivations are particularly important dimensions of vulnerability.
Finally, the multidimensional poverty index, reported annually since 2010 in the United Nations Development Programme's Human development report, is a measure of overlapping deprivations at the household level in developing countries. The index addresses the concept of systematic exposure to multiple burdens, an approach which is neglected by traditional measures of socioeconomic position (such as those that focus only on asset deprivation). The multidimensional poverty index shows not only who is poor but also how they are poor: what simultaneous disadvantages they experience. 17 The index is a continuous indicator calculated by the weighted sum of deprivations in 10 indicators of health, education and living standards that are commonly available in household surveys (adult household members' years of schooling, child school attendance, child mortality, nutrition, cooking fuel, sanitation, water, electricity, house-flooring material and assets). 17 By including indicators for child mortality and nutrition, the index directly identifies children at higher risk of adverse health outcomes and hence at highest priority for vaccination from an equity standpoint. Nonetheless, a child's nutritional status can itself be affected by infection and lack of vaccination. We believe that the index is particularly valuable for descriptive surveillance of vaccination coverage and for orienting policy. However, like all other indicators presented here, its association with vaccination coverage should not be interpreted in causal terms. It has been widely recommended that equity monitoring include several diverse definitions of vulnerability. 5,8,18,19 The WHO Health Equity Monitor database, for example, reports estimates for several maternal and child health outcomes disaggregated by wealth quintiles, maternal education, sex and area of residence (urban or rural), derived from the DHS and MICS. 20 Similarly, the Countdown to 2030 collaboration disaggregates coverage of health interventions by wealth quintiles, maternal education, sex, area of residence and country region. 21 In addition to these indicators, we believe that the multidimensional poverty index, as well as its individual components of health, education and living standards, is particularly relevant to monitoring inequalities in vaccination coverage.
To formulate recommendations based on empirical findings, we measured inequalities in vaccination coverage across 45 Gavi-supported countries and compared results from different measurement approaches (Arsenault C et al., McGill University, unpublished data, 2016). Using the most recent DHS in each country, we measured inequalities in the receipt of the third dose of diphtheria-tetanus-pertussiscontaining vaccine (DTP3) and measlescontaining vaccines according to the five selected indicators. Coverage with DTP3 and measles-containing vaccines are widely accepted as standard indicators of how well a country's immunization system is performing. 22 Although the five indicators were correlated, they differed in their ability to identify vulnerable groups in specific contexts. We found the largest inequalities according to the multidimensional poverty index, maternal education and the wealth index. Inequalities by place of residence and child sex were lower on average, but revealed important inequalities in specific countries. Many equity analyses include only measures of wealth-based inequality. However, the wealth index does not by itself reflect the full complexities of social disadvantage. Our analysis showed that while a country may have equitable coverage according to the wealth index, inequalities could be present in other dimensions that may be equally relevant to policy-making.

Measures of inequality
Equity monitoring entails another important consideration: the choice of appropriate measures to quantify inequalities. The absolute and relative coverage gaps are the measures most commonly used. The absolute coverage gap is typically applied to wealth quintiles and consists of measuring the difference in coverage between the richest (Q5) and poorest (Q1) quintiles (Q5 minus Q1). The relative coverage gap is the ratio of coverage between the richest and poorest quintiles (Q5 divided by Q1). When indicators have only two subgroups, such as child sex or urban versus rural residence, the absolute and relative coverage gaps are the most straightforward way to measure inequality. However, when the whole population is not included in the two subgroups compared, the coverage gaps have limitations. For example, when comparing Q5 and Q1, quintiles 2, 3 and 4 are ignored. This can conceal important heterogeneity and may provide a limited view of the inequalities present across the entire range of social groups. 23,24 This measure is also impractical for multiple education levels or continuous indicators such as the multidimensional poverty index which require arbitrary subgroups to be defined.
Alternative measures of inequality may overcome these limitations. The slope index of inequality and the relative index of inequality have been recommended for multiple subgroups or continuous indicators because they consider information on all subgroups, summarize inequality across the whole distribution and reflect the direction of the gradient in health. [23][24][25][26][27] The slope and relative indices of inequality can be obtained by regressing the child's vaccination status on her or his cumulative rank in the population's socioeconomic distribution. 23,25 In our empirical analysis, we first tested the use of both the coverage gap (absolute and relative) and the slope and relative indices of inequality to assess education, multidimensional poverty and wealth-related inequalities in vaccination coverage. We found that the coverage gaps were more likely to generate imprecise estimates if there were few individuals in the extreme categories of education or of the multidimensional poverty index, this made comparisons of the level of inequality across countries more difficult. In contrast, the slope and relative indices of inequality were less likely to be affected by sampling error and produced more reliable country comparisons. Although they involve additional complexity and assumptions, the slope and relative indices of inequality may be better suited than coverage gap measures for international comparisons of the levels of inequality. 23,28 We also found that comparisons of the level of inequality across countries differed substantially whether absolute or relative measures were used. The fact that absolute and relative measures can lead to different conclusions when comparing inequalities across time and place has been frequently reported. 27,[29][30][31][32] To obtain a complete picture of inequalities and a reliable measurement of progress over time, we designed the equity dashboard to include both absolute and relative measures.

Equity benchmarks
Equity monitoring aims to promote social justice and the realization of human rights, and these concepts can be reflected in the definition of benchmarks and targets. In its 2011-2015 strategy period, Gavi established a minimum equity benchmark requiring that DTP3 coverage in the poorest wealth quintile be no more than 20 percentage points lower than coverage in the richest wealth quintile. 4 This benchmark was set empirically to provide a realistic and attainable target for Gavi-supported countries. However, acceptance of up to a 20% points difference in coverage between the richest and poorest is difficult to justify from an ethical standpoint.
We propose a new target for the SDG period, inspired by the Indian economist and philosopher Amartya Sen's vision of human development as the freedom to achieve well-being. 33 Following Sen, we view health as valuable both in itself, and for its contribution to expanding people's effective freedoms or opportunities to undertake the activities that they find valuable and to live the lives that they find meaningful. 34 Because vaccination is a key determinant of the potential for health over the life course, equality in vaccination is crucial to achieving true equality of opportunity in the capability for health.
We therefore argue that the appropriate target for immunization equity should be equal coverage. We interpret this to mean that there should be no meaningful inequalities across social groups: a judgement that must incorporate both quantitative metrics and knowledge of the characteristics of the country and sample. In practice, this requires testing whether the magnitude of inequality is statistically distinguishable from zero. To avoid confusing moral significance with statistical significance, this benchmark requires adequate sample sizes to achieve precise estimates. Precision benchmarks could also be established by defining acceptable ratios of 95% confidence limits. 35 An equal coverage benchmark is conceptually defensible in that it reflects a commitment to the moral equality of persons and fulfilment of the human right to health. It is also realistic and attainable. Among the 45 Gavi-supported countries we analysed, 38% (17 countries) had already met the benchmark for equal coverage across the wealth index (Arsenault C et al., McGill University, unpublished data, 2016).

Equity dashboard
Based partly on early results from our analysis, Gavi has lowered the equity benchmark to 10 percentage points and added indicators related to geographical equity and maternal education. This added complexity in Gavi's strategy for 2016-2020 raises new challenges, prompting us to propose the use of an equity dashboard for more effective monitoring of equity in vaccination coverage in the SDG period.
To illustrate the advantages of this approach we present the vaccination equity dashboards for two countries that have similar levels of national vaccination coverage but very different equity profiles. The equity dashboards for Côte d'Ivoire and Haiti ( Fig. 1 and  Fig. 2) were based on data from children   36,37 These show the national average coverage for DTP3 and measles-containing vaccines, and the inequalities in children's receipt of these vaccines across the five selected indicators of vulnerability. The slope and relative indices of inequality were used for indicators that were continuous or in multiple ordered subgroups: maternal education (no education, incomplete primary, complete primary, incomplete secondary, complete secondary, higher education attended), the wealth index and the multidimensional poverty index. The absolute and relative coverage gaps were used for indicators with two subgroups: place of residence (urban versus rural) and child sex (male versus female). Equity benchmarks are marked in the dashboard as met if the 95% confidence interval (CI) of the inequality estimate includes the null value, or unmet if the 95% CI excludes the null value.
In Côte d'Ivoire ( Fig. 1), the equity dashboard shows inequalities in DTP3 coverage across all five dimensions, including child sex. DTP3 coverage is 10 percentage points higher among boys than among girls (slope index of inequality: 0.10; 95% CI: 0.01 to 0.19). The inequality estimates revealed by the multidimensional poverty index are also substantially higher than those revealed by wealth and education. National coverage with DTP3 and measles-containing vaccines are 62.9% and 62.3% respectively.
In the case of Haiti (Fig. 2), the dashboard reveals virtually no inequalities by children's sex or place of residence or between the richest and poorest households according to the wealth index. Gavi's initial approach required that the absolute coverage gap between the richest and poorest wealth quintiles be no more than 20 percentage points. As this coverage gap was 10 percentage points (95% CI: -0.05 to 0.25) in the 2012 DHS, 37 Gavi would have considered Haiti as having achieved equitable vaccination coverage. However, the equity dashboard reveals substantial inequalities by maternal education and multidimensional poverty. For example, DTP3 coverage is on average 34 percentage points higher (slope index of inequality: 0.34; 95% CI: 0.20 to 0.48) among children of mothers at the top versus the bottom of the distribution of maternal education, and measles-containing vaccine coverage is 23 percentage points higher (slope index of inequality: 0.23; 95% CI: 0.09 to 0.38). Similarly, the least deprived children are 1.5 times (relative index of inequality: 1.48; 95% CI: 1.14 to1.81) more likely to receive DTP3 than children deprived in most or all of the multidimensional poverty index components. These measures imply systematic associations between low vaccination coverage and, respectively, low education and high multidimensional poverty. More importantly from a policy perspective, relying solely on the wealth index would fail to identify these vulnerable groups. In addition, the dashboard highlights that, although Haiti meets the equity benchmarks across three dimensions (wealth index, place of residence and child sex), national coverage with DTP3 and measles-containing vaccines remain at only 62.2% and 63.3%. In addition to equity benchmarks, national coverage benchmarks could be established to target policy efforts.
Comparing the two dashboards also demonstrates that, although na-  Monitoring equity in vaccination coverage Catherine Arsenault et al.
tional coverage with the two vaccines is almost identical in both countries, the magnitude of inequalities is considerably higher in Côte d'Ivoire compared with Haiti. The equity dashboard offers a single snapshot of a country's overall achievement and inequalities across multiple dimensions of social exclusion and vulnerability. We propose that the dashboard improves equity monitoring and can serve as a decision support tool for policy development. It may also facilitate conversation and be readily useable by those without technical expertise in inequality measurement. By providing a comprehensive and standardized analysis of inequalities, the dashboard also facilitates comparisons across countries (and between different vaccines). As new data become available, the dashboard should also include historical trends of inequalities in each dimension to track changes over time. A third criterion could then be introduced for the equity benchmarks, represented by yellow cells in the dashboard, to indicate whether progress was being made in reducing inequalities.

Conclusion
Given current global efforts to achieve universal health coverage (SDG3), reduce inequalities (SDG10) and increase the availability of disaggregated data (SDG17), countries and development partners must make equity monitoring a priority. 38 Gavi and development partners should develop equity dashboards tailored to their specific policy needs. An equity dashboard may also be a useful decision-support tool for other stakeholders seeking to monitor equity in coverage of other health services. This type of monitoring may help determine where and why inequalities arise and ensure that policies are successful in improving the health of the most vulnerable. 38