Inequalities in full immunization coverage: trends in low- and middle-income countries

Abstract Objective To investigate disparities in full immunization coverage across and within 86 low- and middle-income countries. Methods In May 2015, using data from the most recent Demographic and Health Surveys and Multiple Indicator Cluster Surveys, we investigated inequalities in full immunization coverage – i.e. one dose of bacille Calmette-Guérin vaccine, one dose of measles vaccine, three doses of vaccine against diphtheria, pertussis and tetanus and three doses of polio vaccine – in 86 low- or middle-income countries. We then investigated temporal trends in the level and inequality of such coverage in eight of the countries. Findings In each of the World Health Organization’s regions, it appeared that about 56–69% of eligible children in the low- and middle-income countries had received full immunization. However, within each region, the mean recorded level of such coverage varied greatly. In the African Region, for example, it varied from 11.4% in Chad to 90.3% in Rwanda. We detected pro-rich inequality in such coverage in 45 of the 83 countries for which the relevant data were available and pro-urban inequality in 35 of the 86 study countries. Among the countries in which we investigated coverage trends, Madagascar and Mozambique appeared to have made the greatest progress in improving levels of full immunization coverage over the last two decades, particularly among the poorest quintiles of their populations. Conclusion Most low- and middle-income countries are affected by pro-rich and pro-urban inequalities in full immunization coverage that are not apparent when only national mean values of such coverage are reported.


Introduction
Despite the improvements made in global immunization coverage for children over the past decade, 1,2 an estimated 21.8 million infants worldwide are still not being reached by routine immunization services. 3 In 2013, most of the World Health Organization's (WHO) regions reached more than 80% of their target populations with three doses of diphtheria, pertussis and tetanus (DTP) vaccine but coverage with such vaccine remained well short of the 2015 goal of 90%, particularly in the African (75%) and South-East Asia regions (77%). 2,3 Many barriers exist to achieving good vaccination coverage, including a lack of parental education, low income, poor access to health facilities and traditional beliefs. [4][5][6][7][8][9][10][11][12][13] As progress in this field is commonly expressed in terms of national or regional mean values, many of the underlying disparities among and within countries go unobserved or, at least, unreported. If routine immunization is to be made fast and equitable, we need multi-country studies that use the same types of stratification to document and understand the inequalities in vaccination coverage at both national and regional level. 1,3,14,15 We also need to know the percentages of children who receive the full set of standard vaccines recommended by WHO. In India, for example, national immunization coverage has been increasing since the early 1990s but the proportion of children who, in 2006, had received all of the immunizations recommended for their age group as part of WHO's Expanded Programme on Immunization was still under 50%. 16 Failures or delays in the vaccination of children in high-risk groups can limit the impact of vaccine programmes on the burden of disease. 17 The main objectives of the present analyses were: (i) to assess the proportions of children in low-or middle-income countries who receive a basic set of routine vaccinationsthat is one dose of bacille Calmette-Guérin vaccine, one dose of measles vaccine, three doses of vaccine against DTP and three doses of polio vaccine -at the appropriate ages; (ii) to document between-country and within-country inequalities in such coverage -in terms of socioeconomic status and other characteristics commonly recorded in national surveys; and (iii) to assess temporal trends in such coverage and in the associated inequalities.

Methods
In May 2015, we accessed publicly available data sets collected during the most recent Demographic and Health Survey 18 and/ or Multiple Indicator Cluster Surveys 19 in each of the 86 lowor middle-income countries in which at least one such survey had been conducted since the year 2000 (Table 1). Our study outcome was full immunization coverage, which we defined as the proportion of children who, at any age, had received one dose of bacille Calmette-Guérin vaccine, one dose of measles vaccine, three doses of -trivalent, tetravalent or pentavalent -vaccine against DTP and three doses of polio vaccine. For the 20 study countries where measles vaccine was routinely administered at the age of 18 months, full immunization coverage was measured among children aged 18-29 months. Similarly, for the three study countries where measles vaccine was routinely administered at the age of 15 months, full immunization coverage was measured among children aged [15][16][17][18][19][20][21][22][23][24][25][26] Objective To investigate disparities in full immunization coverage across and within 86 low-and middle-income countries. Methods In May 2015, using data from the most recent Demographic and Health Surveys and Multiple Indicator Cluster Surveys, we investigated inequalities in full immunization coverage -i.e. one dose of bacille Calmette-Guérin vaccine, one dose of measles vaccine, three doses of vaccine against diphtheria, pertussis and tetanus and three doses of polio vaccine -in 86 low-or middle-income countries. We then investigated temporal trends in the level and inequality of such coverage in eight of the countries. Findings In each of the World Health Organization's regions, it appeared that about 56-69% of eligible children in the low-and middleincome countries had received full immunization. However, within each region, the mean recorded level of such coverage varied greatly. In the African Region, for example, it varied from 11.4% in Chad to 90.3% in Rwanda. We detected pro-rich inequality in such coverage in 45 of the 83 countries for which the relevant data were available and pro-urban inequality in 35 of the 86 study countries. Among the countries in which we investigated coverage trends, Madagascar and Mozambique appeared to have made the greatest progress in improving levels of full immunization coverage over the last two decades, particularly among the poorest quintiles of their populations. Conclusion Most low-and middle-income countries are affected by pro-rich and pro-urban inequalities in full immunization coverage that are not apparent when only national mean values of such coverage are reported.

Inequalities in coverage
We investigated inequalities in full immunization coverage that related to three characteristics recorded in all or most of the surveys that had provided the data that we used: socioeconomic status, urban/rural residence and sex of the child. The data for three study countries -Cuba, Djibouti and Jamaica -could not be used to estimate the wealth index that we used as a measure of socioeconomic status. We calculated a mean full immunization coverage for the study countries in each WHO region.
In each of the surveys we used as a data source, urban or rural residence had been defined by the local census bureaux and the study households had been categorized into five asset-based wealth quintiles. The quintiles had been derived, using principal component analyses, from variables representing household goods, materials used for housing construction and available infrastructure such as types of water access and sanitation facilities. 20 To summarize any wealth-related inequalities in full immunization coverage, we calculated four indicators. 21 Two of these were based on simple comparisons of the coverage recorded for the lowest wealth quintile and that recorded for the highest wealth quintile: (i) the difference, in percentage points, between the two values; and (ii) the ratio between the two values. We also calculated two indicators of inequality that take the whole distribution of wealth into account: (i) the slope index of inequality -which uses a logistic regression model to express the absolute difference in coverage, in percentage points, between the extremes of the wealth distribution; 22 and (ii) a concentration index 23 that is similar in concept to the Gini index for income distribution. The concentration index was expressed on a scale from −100 to +100, with full equality indicated by a value of zero. Both summary indicators tended to be positive, indicating that full immunization coverage was higher for the rich than for the poor. We calculated standard errors for each summary indicator and corresponding P-values for the probability that there was no inequality. Our absolute measures of inequality -i.e. the difference in coverage between the lowest and highest wealth quintiles and the slope index of inequality -give an idea of the effort that will be needed to close the gap. Our two relative measures -i.e. the ratio between the coverage for the lowest and highest wealth quintiles and the concentration index -give an idea of the degree of disparity.
We made similar comparisons of the coverage recorded for urban children and that recorded for rural children -again, the difference, in percentage points, between the two values and the ratio between the two values. We evaluated the statistical significance of the difference, assuming a binomial distribution. In a similar fashion, we investigated inequalities in coverage according to the sex of the child.
Standard errors, expressed in percentage points, were calculated for all of the coverage estimates. Like the tests for statistical significance, these took into account the sample weights and clustering. When the unweighted number of children in a specific subgroup -e.g. a wealth quintile -was less than 25, we ignored the results for that subgroup.

Temporal trends in inequalities
Time-trend analyses were conducted for a subset of eight low-income study coun-tries that had previously been identified as having the greatest within-country disparity in vaccine coverage: Central African Republic, Chad, India, Madagascar, Mozambique, Nigeria, Pakistan and Viet Nam. 24 We investigated the temporal trends in routine immunization coverage , for children aged no more than 23 months, by both wealth quintile and urban/rural residence. Whenever possible, pre-2000, 2000 -or close to 2000 -and post-2000 records were included (available from the corresponding author) to cover periods before Gavi was launched, when Gavi was launched and when Gavi's main strategies had been implemented, respectively.
Variance-weighted least squares regression was used, with survey as the independent variable, to test the statistical significance of the observed temporal trends, taking into account the clustered nature of the survey samples.

Detected inequalities
Wealth Table 1 shows national levels of full immunization coverage for all 86 countries that we assessed. For the 83 study countries for which the relevant data were available, it also summarizes the full immunization coverage for the lowest and highest wealth quintiles and the corresponding summary indicators of inequality according to wealth. Five countries -Afghanistan, Central African Republic, Chad, Georgia and Somalia -showed national levels of full immunization coverage that were below 20%. Nine countries -Albania, Armenia, Belarus, Costa Rica, Egypt, Jordan, Maldives, Rwanda and The former Yugoslav Republic of Macedonia -showed corresponding coverage above 90%. Lesotho and Senegal, each with a full immunization coverage of 63%, represented the 50th percentile of the country ranking. In most of the study countries, the increase in coverage with wealth was monotonic, that is, coverage in the lowest wealth quintile was lower than that in the second-lowest, coverage in the second-lowest quintile was lower than that in the third-lowest and so on.
Of the 83 study countries for which the relevant data on wealth were available, 65 each gave a positive slope index of inequality that indicated the existence of a pro-rich inequality in coverage ( Table 1). For 45 of the countries with a positive slope index of inequality, that index was significantly different from zero. Although 18 countries had negative slope indices, indicating a pro-poor inequality in coverage, only five of the 18 negative slope indices were significantly different from zero. The corresponding results for the concentration index were very similar: we recorded 64 positive and 19 negative concentration indices, of which 45 and four, respectively, were significantly different from zero.
In terms of the slope indices of inequality, Nigeria showed the greatest pro-rich inequality in full immunization coverage, followed by Pakistan, India, Turkey, Madagascar, Yemen, Cameroon and Liberia. The corresponding patterns for the concentration indices were similar. Seven countries -in descending order of pro-rich inequality, Chad, Central African Republic, Nigeria, Somalia, Yemen, Ethiopia and India -gave concentration indices above 20 (Table 1). In terms of one or both of our summary indicators, only four of our study countries showed distinctively pro-poor inequalities in their full immunization coverage: Gabon, Gambia, Mauritania and Uzbekistan. However, Gabon and Mauritania had relatively low national levels of coverage.
Countries that appeared similar in terms of their national values for full immunization coverage could show very different degrees of inequality. For example, Côte d'Ivoire and Mali had national values of about 50% but very different slope indices of inequality -of 8 and 33 percentage points, respectivelyand very different concentration indices -of 3 and 11, respectively. Likewise, the Plurinational State of Bolivia and Philippines had national values of about 79% but very different slope indices -of 3 and 30 percentage points, respectively -and very different concentration indices -of 0.7 and 6, respectively. Table 2 shows the mean values for full immunization coverage in the low-and middle-income countries we investigated in each WHO Region, which varied from 55.5% in the Eastern Mediterranean Region to 68.9% in the Region of the Americas. Globally, according to the most recent survey data available in May 2015, just over 60% of all eligible children in low-and middle-income countries had received full immunization. There was wide variation in the level of full immunization coverage within a given Region (Fig. 1). For example, in the African Region, the mean level of full immunization coverage varied from just 11.4% in Chad to 90.3% in Rwanda. Fig. 2 shows that on average, the wealth inequalities in full immunization coverage were less marked in the low-and middle-income countries in the Region of the Americas and the European Region than in such countries in other regions. The Eastern Mediterranean Region not only presented the highest absolute and relative wealth-related inequalities in such coverage but also the lowest mean level of such coverage. Table 3 summarizes mean levels of full immunization coverage, split according to urban/rural residence, in our 86 study countries. Although 28 countries had higher levels of coverage in their rural areas than in their urban areas, the differences were generally very small and only those for Mauritania, Swaziland and Uzbekistan were statistically significant. The remaining 58 countries had higher levels of coverage in their urban areas than in their rural areas and the differences for 35 of these countries achieved statistical significance. In Ethiopia, which showed the greatest absolute pro-urban inequality, the mean level of full immunization coverage in urban areas was 28 percentage points higher than that in rural areas. Côte d'Ivoire, Madagascar, Nigeria, Turkey and Yemen also showed pro-urban differences of at least 20 percentage points. In contrast, Mauritania, Swaziland and Uzbekistan showed pro-rural differences of at least 10 percentage points. In terms of full immunization coverage, no other countries showed a statistically significant advantage for rural children.

Residence
In terms of the effects of urban/ rural residence on full immunization coverage ( Table 2 and Fig. 3), the Region of the Americas and the European Region appeared more equitable than other Regions. The Eastern Mediterranean Region showed the largest pro-urban inequalities, where mean levels of full immunization coverage were about 60% higher among urban children than among their rural counterparts.

Sex of child
Although most of our study countries showed higher levels of full immunization coverage among boys than girls (Table 3), such sex-related differences were of less than three percentage points in each of 59 countries and only achieved statistical significance in Azerbaijan, Belize, India, Mali and Somalia. The absolute levels of sex-related inequality were relatively high in the South-East Asia and Western Pacific Regions. In the Western Pacific Region, the mean level of full immunization coverage was 10% lower among boys than among girls (Fig. 4).

Temporal trends
In all eight countries included in our investigation of temporal trends, there   (. . .continued) in the national level of full immunization coverage over our study period. The level of absolute pro-rich inequality declined -since the absolute reduction in coverage was most marked among the richest quintile -but the level of relative pro-rich inequality increased. In Chad, however, the national level of coverage remained low and stable over our study period and wealth-related and urban/ rural inequalities remained largely unchanged. In India, the national level of coverage increased but, as in Chad, wealth-related and urban/rural inequalities remained fairly stable. The results for Madagascar and Mozambique, which both showed increasing national levels of coverage over time, were more encouraging. The relative pro-rich inequality observed in Madagascar also decreased over time, although the absolute pro-rich inequality and all measures of urban/ rural inequality did not decline. In Mozambique, much of the increase seen in the national level of coverage was linked to increasing coverage in the poorest quintile of the population. Although the country's pro-rich and pro-urban inequalities decreased over time, the decrease observed in the pro-rich inequality was partly attributable to a decrease in coverage among the children from the richest quintile. In Nigeria, increases in the level of full immunization coverage in some areas had little impact on the overall national level, which remained relatively low. When we compared the most recent data we investigated for each of the eight countries, Nigeria showed the largest absolute pro-rich inequality as well as a large level of pro-urban inequality. Although Nigeria's relative pro-rich inequality appeared to have decreased over our study period, this was only the result of mean coverage in the poorest quintile increasing from nearly zero to just over 10%. In Pakistan, the level of coverage in all quintiles except the poorest showed improvement over time -with a consequent increase in the level of pro-rich inequality. In Viet Nam, coverage in the poorest quintile remained fairly stable while that in each of the other wealth quintiles -like the levels of pro-rich and pro-urban inequalities -initially increased but then declined.

Discussion
Our findings indicate that, despite progress, much remains to be done if the ben-   ‫التغطية.‬ efit of routine childhood immunization is to be maximized. Reports of regional levels of vaccination coverage may mask local challenges, inequalities and variation. We observed pro-rich inequalities in full immunization coverage in most lowor middle-income countries, although they were, in general, relatively small in the Region of the Americas and European Region -and relatively large in the Eastern Mediterranean and Western Pacific Regions. Pro-urban inequalities were also common. They were generally very small in the low-or middle-income countries in the Region of the Americas and largest in such countries in the Eastern Mediterranean region, where coverage was about 60% higher among urban children than among rural children. Low-or middleincome countries in the South-East Asia Region showed the largest absolute promale inequalities. We observed that, whether related to sex, wealth or urban/rural residence, inequalities in full immunization coverage varied substantially between and within our study countries. Inequalities related to wealth and urban/rural residence appeared to be ubiquitous and persistent and to be larger, in general, than the corresponding sex-related inequalities. Although some countries have made substantial progress in reducing such inequalities, some other countries have seen such disparities increase.
Among the eight countries included in our investigation of temporal trends in coverage and coverage inequality, Madagascar and Mozambique appeared to have made the most progress in improving national levels of coverage -in both cases by achieving particularly rapid increases in coverage in the poorest quintile. In general, the factors that are believed to have contributed to global improvements in immunization coverage include national multi-year planning, district-level planning and monitoring and the establishment of national budget lines funded via domestic and external resources for the strengthening of immunization services. 25 Our observation of markedly different temporal trends in coverage and coverage inequality in eight countries needs to be followed up with case studies aimed at documenting the factors -within and beyond the health sector -that might explain such variation.
It is important to note that the data on coverage being reported here are solely based on survey information. As international agencies estimate vaccination coverage using a combination of data from surveys and data from health information systems, the coverage levels reported here will not necessarily be consistent with the estimates given in official documents produced by national governments and the United Nations. However, such estimates cannot be disaggregated by wealth quintile or place of residence and can rarely be used to determine the level of the full immunization coverage that we wished to investigate. To allow consistent and meaningful comparisons, we confined our investigation to vaccines that are available in almost all countries of the world. We ignored several new vaccines that are included in the national immunization programmes of a few of our study countries. Another limitation of the present study is that, when vaccination cards are not available, the information collected on child immunization in national surveys has to be based on the recall of mothers or other caregivers.
Our results indicate that the ultimate goal of the Global vaccine action plan 2011-2020 -i.e. universal access to immunization 3 -will only be achieved if the relevant health workers, policymakers and stakeholders can: (i) develop and implement strategies for reaching those who are difficult to reach and for promoting the need for full immunization among those who have contact with health services for other interventions; (ii) expand vaccination programmes to include underserved groups; (iii) improve the quality of the monitoring of immunization coverage; (iv) use monitoring data to ameliorate programme performance; and (v) explore additional cross-sectoral strategies -particularly in those low-or middle-income countries with the worst inequalities in coverage. The improvements in coverage and equitable access to routine immunizations achieved by some Latin American countries may serve as useful examples. 26