*3.2. Individual Vaccines (BCG, DTP, Polio, and MCV)*

Focusing on BCG, the eight countries reporting the greatest difference between the multivariate and wealth-based concentration indices are the Maldives (0.178), Afghanistan (0.172), Chad (0.170), Senegal (0.133), Yemen (0.121), Guatemala (0.110), and Madagascar (0.117). However, absolute differences in the AEG vary widely from 1 to 42 percentage points. In contrast, countries with the lowest differences between multivariate and wealthbased concentration indices also had the lowest absolute differences between AEGs. For 44 of the 56 countries in this analysis, the multivariate concentration index is statistically significantly greater than that of wealth-only. For the remaining 12 countries, which include the Kyrgyz Republic, Republic of Congo, Mozambique, Comoros, Benin, India, The Gambia, Sierra Leone, Lesotho, Peru, Malawi, and Ghana, there is no statistical difference between multivariate and wealth-based concentration indices. When looking at total country averages for individual vaccines, BCG has the lowest difference between multivariate and wealth-based estimates with a concentration index difference of 0.046 and an AEG difference of 8.8 percentage points, suggesting that wealth accounts for a significant proportion of the total inequity in this birth-dose vaccine (see Table 1).

If we consider MCV1, the greatest differences in concentration index values are attributed to Guinea (0.230), Afghanistan (0.215), Madagascar (0.166), Angola (0.152), Nigeria (0.145), Ethiopia (0.145), and the Maldives (0.144). Again, we observe a wide range in the differences in AEG values between approaches, ranging from 4 to 39 percentage points. By evaluating inequity with a multivariate approach, it is revealed that the use of a wealth-only ranking metric results in a significant underestimation of inequity for 51 of the 56 countries considered. Countries for which the multivariate concentration index is not statistically different from the wealth-only concentration index include the Kyrgyz Republic, Mozambique, Republic of Congo, Comoros, and Lesotho. Using national averages, the difference between concentration indices as measured by each approach for MCV1 was 0.068 with an AEG difference between approaches of 10.1 percentage points.

For the three-dose vaccines DTP and Polio, the absolute difference between concentration indices generally increases for subsequent doses, though the same trend does not apply to differences in the AEG, suggesting that much of the inequity present after receiving the first dose occurs in the middle of the distribution rather than the tails of the distribution. The greatest difference in DTP concentration index values when comparing the multivariate and wealth-only methodologies are exhibited by Chad (DTP1: 0.192, DTP2: 0.216, and DTP3: 0.268) and Afghanistan (0.186, 0.204, and 0.224). Of all the vaccines included in this study, DTP3 has the highest national average absolute difference between concentration index types at 0.084 and experiences an AEG difference between approaches of 12.9 percentage points, on average. The concentration index differences for DTP1 and DTP2 are 0.053 and 0.066, respectively, with AEG differences between approaches of 9.9 and 11.7, respectively.

The greatest differences between multivariate and wealth-based concentration indices for Polio occur in the Maldives (0.160), Afghanistan (0.130), and Senegal (0.130) for dose 1; Gabon (0.150), Afghanistan (0.142), and Madagascar (0.135) for dose 2; and Angola (0.184), Chad (0.179), and Guinea (0.177) for dose 3. The average differences in concentration index over all countries for Polio doses 1, 2, and 3 are 0.046, 0.059, and 0.080, respectively, with differences in AEG between approaches estimated to be 10.5, 12.4, and 14.5, respectively.

#### **4. Discussion**

This case-study application of the VERSE toolkit to 56 countries demonstrates that using multivariate procedures for measuring vaccine coverage equity results in significantly larger values compared with traditional methods in most settings. The findings indicate that metrics which only utilize socioeconomic status as a basis for measuring inequity, in order to track whether or not access is pro-poor, will miss a significant amount of the variation in the overall equity in vaccination status that is directly correlated with observable characteristics such as education, sex, and geographic location [23,24].

In countries such as Chad, Afghanistan, or Guinea, if inequities in fully immunized status were only captured through the traditional wealth-based concentration indices or absolute equity gaps, the measures would show that there was no systematic inequity in vaccine coverage within the country (concentration indices between −0.006 and 0.020); however, the multivariate concentration index demonstrates otherwise.

Several recent studies on equity also support the empirical findings of this study. A 2022 systematic review by Ali et al. found that besides wealth, maternal education, sex, and geographic access can also systematically and independently affect vaccination coverage [25]. Additionally, a 2020 study by Acharya et al. comparing the inequalities in full vaccination coverage based on maternal education and wealth quintiles also found that in four of the six studied countries, the absolute inequalities arising from a metric using maternal education level were significantly larger than those measured using wealth quintile [26]. These studies further emphasize the importance of utilizing multivariate metrics to holistically measure and work toward reducing systemic inequality.

Multivariate indicators integrating these multiple socio-demographic parameters effectively quantify differences in coverage even in countries with more modest inequity, such as Uganda. Uganda achieved large increases in overall vaccination coverage during the 2000s with its immunization program through the implementation of Family Health Days and other regular health outreach initiatives, which made the coverage distribution significantly pro-poor. However, when considering the other factors included in the VERSE toolkit's approach, we can estimate a residual inequity driven by both supply- and demandside factors such as the district of residence and maternal education [27]. Such an approach revealed aspects of access to vaccines, such as sufficient health literacy and adequate and timely supply across districts, which can help the country consider new approaches to continue to improve coverage equity [28,29].

While the VERSE approach and toolkit can yield a stable metric to track equity over time or between settings, it is also subject to several practical limitations common to

all measures of equity and inequality [15]. The first is the inability to objectively state what a "good" or "bad" level of inequity is using the concentration index alone. Like all concentration indices, the results of the VERSE methods lend themselves more toward assessing relative performance than to categorizing objective performance. Although values closer to 0 are objectively preferred, whether a value of 0.1 is bad or good depends upon the circumstances of a specific setting, the mean level of coverage obtained in the setting overall, and the specific benchmarks associated with the rollout and distribution of each vaccine. For this reason, all equity metrics should be put into the context of the outcome or intervention they are evaluating. To assist with this contextualization, the VERSE toolkit produces an absolute equity gap alongside the concentration index to assist with interpretation. While the AEG is a measure of absolute inequity, and the concentration index measures relative inequity, they are both based on the same ranking procedure. They can therefore complement one another, with the AEG providing important coverage-level context to the concentration index.

Another limitation is the data used to populate the tool. While DHS surveys are designed to be nationally representative, evidence shows that settings like urban slums, conflict areas, and refugee settlements are significantly under-sampled, in addition to being more likely to be under-immunized [30]. As a result, estimates of vaccination coverage generated using the DHS are likely to be systematic overestimates of true immunization coverage, and estimates of coverage inequity are likely to be systematic underestimates of true coverage inequities.
