Socioeconomic-related inequities in child immunization: horizontal and vertical dimensions for policy insights

Abstract

Background

The incomplete immunization has potentially exposed vulnerable children, especially from the socioeconomically disadvantage group, to vaccine preventable diseases. The schemes would maximize social benefit only when the immunization is effectively distributed on an equitable principle.

Method

The empirical study is based on unit level data from India’s National Sample Survey: “Social Consumption: Health Survey- NSS 75th Round (2017-18) database. The nationwide survey is designed on the stratified multi-stage sampling method with an objective to make the sample representative. The egalitarian equity principle requires that distribution of vaccine should be based on health needs of children, irrespective of their socioeconomic and regional factors and the principle is broadly based on two aspects - horizontal and vertical equity. The horizontal inequity (HI) is a direct form of injustice, when children with equal needs of routine immunisation are treated differentially due to their socioeconomic status, while vertical inequity (VI) is indirect form of injustice when children with differential health needs and risks exposure do not receive appropriately unequal but equitable immunisation. Using Indirect Standardisation Method and Erreygers’ Corrected Concentration Index, we measure the degree of horizontal and vertical inequities, and then linearly decompose them to identify the major factors contributing towards the respective indices.

Conclusion

Our findings show that incomplete immunization is significantly concentrated among children belonging to poorer households. After controlling for the confounding effects of need factors, the inequity is still significantly pro-poor (i.e., horizontal inequity). The decomposition reveals that lower education, lower consumption and rural habitation are the major factors driving the corresponding inequity. Further, the differential effect of the needs between all and the target groups (at least based on education), is observed, however, is not statistically significant enough to realize inequity (i.e., no vertical inequity). Overall, the inequity is being induced via non-need factors. We further find that community health services (like anganwadi) have contributed towards reducing the inequity in child immunization significantly. The paper highlights the policy recommendation that the child immunisation program should target factors driving HI and need to align their distribution in terms of risks exposures.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13561-024-00566-8.

Introduction

The immunisation is highly cost-effective strategy for combating infant mortality and morbidity burden, especially in the developing countries [6, 36]. A high rate of child morbidity and mortality, due to vaccine preventable diseases, not only affecting vulnerable households with significant economic burden but also threatening their future life prospects, by trapping them in a vicious cycle of poverty. Evidence suggests that around five million under five deaths happen globally every year and roughly one third of this mortality are reported from Nigeria and India alone1. Despite the existence of a longstanding Universal Immunization Program (UIP) that offers free childhood vaccines, India continues to have one of the lowest immunization rates globally [26], and still accounts for more than half million-under-five mortality due to vaccine preventable diseases2. It is more disconcerting that the marginalized groups (Scheduled Caste and Scheduled Tribes) still experiencing disproportionately higher child mortality rate [4].

To prevent such avoidable death, the government had introduced one of the largest public health programmes called Universal Immunization Programme (UIP) in 1985. It has an annual mandate of covering 26.7 million new born and 29 million pregnant women3, against 9 diseases. The programmer included mandatory doses of BCG, OPV, DPT, and TT and subsequently added measles, Vitamin A supplements and Pentavalent for children under five years of age. To strengthen and re-energize the UIP, the government subsequently launched Mission Indradhanush in 2014 and then again introduced Intensified Mission Indradhanush (IMI) in 2017 to ensure at least 90% immunization coverage among underprivileged and difficult to reach children by 2018. The policy initiative was to increase the mandatory vaccination uptake among children and reduce the socioeconomic related inequity in child immunisation. However, as per the latest National Family Health Survey-2019-21 (NFHS-5), the program effectively increased child immunisation coverage from 62.3 to 76.4% for the age group of 12–23 months. Despite a concerted policy intervention, 23.6% Indian children within the same age category are still under incomplete immunization4. Incomplete immunisation is found to be one of the major factors contributing towards child mortality [10]. The first dose of vaccine (BCG) at birth is almost 93% but the pace of prescribed remaining vaccine uptake declines to 79% (measles) within 60 months of a child age (see Fig. 1). The delay in the uptake of successive recommended vaccines may be due to the lack of proper awareness and motivation regarding the need for vaccination, behavioral inertia, social norms, difficulty in accessibility, high opportunity costs etc.5All these factors are highly correlated with the socioeconomic factors.

Fig. 1.

Mandatory Vaccines Uptake across Gender between 13–60 Months. BCG: Bacillus Calmette-Guérin (BCG) vaccination is given to new born babies to protect them from contacting tuberculosis. OPV (1, 2, 3): Oral Polio vaccine is given to new born babies in three consecutive oral dozes for the prevention of poliomyelitis. DPT (1, 2, 3): DPT vaccine (administered in 3 injections) protects the child and adults against diphtheria (D), pertussis (P, also known as whooping cough) and tetanus (T). Source: Social Consumption: Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

In the comparison of Table 1; Fig. 2, two distinct patterns emerge: first, incomplete immunization is disproportionately concentrated among children from socioeconomically disadvantaged backgrounds; second, there is a discernibly lower rate of vaccine uptake within this group when compared to children from households with heads who have attained graduate-level education or belong to the top 20% of per capita consumption expenditure quintiles. For instance, the data reveal that 31.49% of females in households headed by individuals with no formal education remain incompletely immunized, in contrast to 24% of females in households where the head has completed higher secondary education or above. Similarly, among males, 34.54% of those in the poorest 20% CEQ are incompletely immunized, whereas only 15.27% in the richest 20% CEQ experience incomplete immunization. On average, the proportion of incompletely immunized children in the richest households is nearly 10% points lower than that in the poorest households. This unequal vaccine uptake underscores the persistent inequality in child immunization, with incomplete immunization being more pronounced among those situated lower on the socio-economic ladder.

Table 1.

Proportion of children between 13–60 months with incomplete and complete immunisation across socioeconomic status

	% of Children under Illiterate Household Head		% of Children under Graduate or Above Household Head
Variables	Incomplete	Complete	Incomplete	Complete
Male	31.49	68.50	24.39	75.60
Female	29.93	70.06	24.42	75.57
Age 13–23	45.31	54.68	32.44	67.55
Age 24–33	36.58	63.41	32.67	67.32
Age 34–43	28.23	71.76	22.61	77.38
Age 44–60	22.32	77.67	16.69	83.30
Observation	3,017	7,069	884	2,445
	% of Children under lowest 20% CE Quintile		% of Children under Top 20% CE Quintile
Variables	Incomplete	Complete	Incomplete	Complete
Male	34.54	65.45	23.89	76.10
Female	33.22	66.77	24.04	75.95
Age 13–23	48.06	51.93	36.92	63.07
Age 24–33	39.74	60.25	30.63	69.36
Age 34–43	31.80	68.19	22.67	77.32
Age 44–60	25.04	74.95	15.27	84.72
Observation	2,273	4,532	2,230	6,192

Complete: All eight mandatory vaccine uptake; Incomplete: Less than or equal to seven mandatory vaccine uptake; Proportion is calculated at sample weight; Age: is expressed in terms of the months of children; CE: Per-Capita Monthly Consumption Expenditure (categorised into five quintile); Source: Social Consumption - Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Fig. 2.

State-wise percentage of immunised children between 13-60 months, 2017-18. Source: Social Consumption: Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Figure 2 highlights the spatial heterogeneity of incomplete child immunisation uptake among the Indian states, attributed with differential social and economic status. Majority of Indian states still have more than 25% of eligible children with incomplete immunisation. Moreover, the socioeconomic inequity in child immunisation has reduced post IMI intervention [24], but disparity in vaccination coverage still persist. The data reveals significant disparities in child vaccination rates across Indian states that don’t consistently correlate with economic status. While some economically disadvantaged states like Bihar (48.1%) and Uttar Pradesh (54.6%) have moderate vaccination rates, others like Manipur (75.1%) surprisingly lead the pack despite ranking low in GDP per capita. Conversely, some economically advantaged states such as Delhi (47.8%) and Goa (59.7%) show unexpectedly low vaccination rates. North-eastern states exhibit wide variations, from Manipur’s high rate to Nagaland’s strikingly low 12.8%. This inconsistency suggests that factors beyond economic indicators, such as healthcare infrastructure, education, awareness campaigns, and local policies, play crucial roles in determining vaccination uptake. Hence, it is necessary to investigate the socioeconomic inequity in incomplete child immunisation for a targeted policy framework.

The egalitarian equity principle requires that the distribution of vaccine should be based on health needsof children, irrespective of their socioeconomic and regional factors [16, 33]. The equity principle is broadly categorises into two aspects i.e., horizontal and vertical [28]. The recent studies have indicated that the distribution of health care based only on horizontal equity principle is not sufficient, and it is necessary to incorporate vertical equity for a comprehensive assessment [27]. In the context of child immunisation, the horizontal equity principle requires that all the eligible children should be provided with all the prescribed basic immunisation within the prescribed time, irrespective of their socioeconomic and regional status. However, as the children are heterogeneously exposed to risk factors owing to their socio-economic positions, prioritising and treating unequally but appropriately on equity principle is vertically justifiable. The horizontal inequity (HI) is a direct form of injustice when children with equal needs of routine immunisation are treated differentially due to their socioeconomic status, while vertical inequity (VI) is indirect form of injustice when children with differential health needs and risks exposure do not receive appropriately unequal but equitable immunisation. The rationale for incorporating vertical inequity in the analysis is owing to the fact the risks of vaccine preventable diseases are not homogeneously distributed across socioeconomic gradient, and children belonging to the lowest socioeconomic status are highly prone to contiguous risks. Further, there are children who are the victims of omission (due to a variety of inefficiency in healthcare distribution across geography) or commission (households deliberately avoid or delay vaccine for their children, specifically based on gender due to socioeconomic and regional factors) [7, 13]. Hence, providing vaccines to children who are the victim of such consequences, requires policy impetus also based on the vertical equity principle.

Studies on inequity in health care are mostly focused on HI, while little attention has been given towards VI. The two plausible reasons are that health care system across the world is mostly guided by horizontal equity principle, while VI is based on strong value judgment, and therefore, separating the inequity is extremely complicated. Moreover, HI is based on well-established method and therefore has wider applications [1, 22, 28, 34, 35, 39]. Whereas, due to its subjective value judgement, VI analyses vary with underlying assumptions and methodology with respect to the variables of interest [1, 8, 15, 18, 19, 23, 27, 28, 32]. Very few studies have addressed both HI and VI in a comprehensive manner in health economic literature [1, 22, 27–29]. These studies are limited to analysing inequity in treatment intervention (like outpatient visit, inpatient stay, etc.) but have not been explored the inequity in preventive intervention like child immunisation. The existing studies only focused on socioeconomic related inequality in child immunisation (Geweniger & Abbas., 2020; Khan & Saggurti., 2020; Mishra et al., 2020; Masters et al., 2019; Boulton et al., 2018; Asuman et al. [3]) and failed to check the potential inequity in immunisation distribution arising in response to heterogeneous risk factors. So, in this context, we investigate: (i) Whether there is socioeconomic-related inequity (i.e., horizontal and vertical) in incomplete immunisation of the children? (ii) What are the major factors contributing towards such inequities?

The paper is presented as follows: in Sect. 1, we discuss the contemporary literature; in Sect. 2, we elaborate the methodology, data and estimation method; in Sect. 3, we empirically analyse HI and VI; in Sect. 4, we conclude our findings with a discussion on generalisability and policy implications.

Methodology, estimation method and data

Methodology

Given that the child immunisation program is universal (ideally the distribution should be invariant to the heterogeneity), in case, if there is sub-optimal distribution then it is likely to be demand driven. We assume, that the demand for vaccine uptake is a function of risks exposure, demographic profile, socioeconomic and regional status. In consonance to the policy emphasis, the vaccine uptake across socioeconomic status either should be statistically insignificant, i.e., invariant or should be positively associated with respect to risk exposure to reflect the vertical equitable distribution. Therefore, we assume that the appropriate input requirement function of immunisation uptake $(v_k)$ is a function of the level of risk exposure (p) and age (m),

$$v_k=f\left(p_k,,,m_k\right)$$ 1

The risk exposure is measured as a composite index based on the available proxy variables like ‘unprotected latrine’, ‘unprotected garbage disposal’ and ‘unprotected sources of cooking’, thus reflecting the relative ranking of children exposed to risks factors. (see the details in Appendix: Section F Table F2). Since, the proxy variables are in binary, the standard PCA does not work well and therefore becomes redundant. The index is derived using the Non-Linear Principle Component Analysis(NL-PCA) [16, 17]. It allows for non-linearity in variable and, also, is a similar data reduction method where a set of uncorrelated principal components are generated by compressing the observed variables. The advantages of NL-PCA are that it can address and detect the nonlinear associations between variables and is capable of handling both nominal and ordinal variables [37]. Our selected proxy variables account for almost 70% of variation in composite index(see Appendix: Section F Table F 3). To further simplify the interpretation, we have normalised the index between 0 (no risk) and 1 (highest risk)6.

We further included gender (g) as an explanatory need variable to capture the variation in immunisation uptake which is generally not reflected in generic risk index. It has been increasingly acknowledged that gender related constraints are critically responsible for high incomplete immunisation and gender inequality is strongly and positively associated with inequalities in immunization uptake [9, 12, 38, 31]. The rationale for adopting gender as a need variable is further reinforced from the WHO Immunization Agenda 2030 (IA2030) which highlights about prioritising and mainstreaming gender as a strategic instrument for achieving complete immunization [38]. This recognition is embedded in that established linkage.

The other important potential determinants of child immunisation uptake are hosts of socioeconomic and regional factors $(y_k)$. Therefore, the regression equation explaining child immunisation uptake is,

$$v_k=f\left(p_k,,,m_k\right)+\gamma g_k+\sum _{i=1}^n{\delta }_i{y}_{\mathit{ki}}+{ϵ}_k$$ 2

Further, if $y_k$ influences $v_k$ and are correlated with health needs then omission of socioeconomic factor would produce biased estimates of $f(.)$. Therefore, in our study, we have used indirect standardisation method to avoid the bias [14]. To account for the impact of public institutions on inequity we have included government financedanganwadi 7 service as a control to arrive at net effect of socioeconomic related net inequity. Since our outcome variable is dichotomous, we have used GLM model with probit link for estimating the parameter of equation,

$$v_k=P\left(f,\left(p_k,,,m_k\right),+,\gamma ,,g_k,+,\sum ,_{i=1}^n,{\delta }_i,y_{\mathit{ki}},+,{ϵ}_k\right)$$ 3

It is important that the functional relationship of health need is appropriately accounted to capture the inequity in the system, otherwise we would not able to measure the degree of HI. Considering the likelihood of non-linear influence of risks exposure and age, we have incorporated both the determinants up to second order polynomial8. Since, in the study the focus is on incomplete child immunisation, therefore here, $v_k$ is a binary variable and assume 1 if ‘incomplete immunised’ or 0 otherwise. On equity principle, we expect that the distribution of vaccine should be such that the children exposed to higher risks are supposed to have lesser incomplete immunisation, and likewise children with higher age should also have lesser incomplete immunisation. Thus, the regression equation for incomplete immunisation is,

$$v_k=P\left(\alpha ,,+,{\beta }_1,p_k,+,{\beta }_2,p_k^2,+,{\beta }_3,m_k,+,{\beta }_4,m_k^2,+,\gamma ,,g_k,+,\sum ,_{i=1}^n,{\delta }_i,y_{\mathit{ki}},+,{ϵ}_k\right)$$ 4

Here ${\beta }_1,{\beta }_2,{\beta }_3,{\beta }_4,\gamma$ and ${\delta }_i$ are parameters of respective need and non-need variables and ${ϵ}_k$ is an error term

Horizontal inequity

The test on horizontal equity principle, it requires that under the null hypothesis, the parameters $({\delta }_i)$ in Eq. (4) should be statistically zero i.e., $\sum _{i=1}^n{\delta }_i=0$ i.e., non-need factors should be invariant or should have no effect on incomplete immunisation uptake. If ${\delta }_{i}s$ are statistically non-zero, it implies that we fail to prove that immunisation uptake is invariant to socioeconomic and regional factors. To calculate the HI inequity, we follow Vallejo-Torres & Morris., [28] methodology. We first estimate the need-predicted incomplete immunization uptake, after controlling for the effect of non-need factors at their mean (using Eq. (4). The need-predicted incomplete immunization uptake equation predicts how much a household would be left with incomplete immunization if they are treated, on an average, equally in the system9. Thus, the estimated need-predicted incomplete immunization uptake is:

$${\widehat{v}}_k=P\left(\widehat{\alpha },+,{\widehat{\beta }}_1,p_k,+,{\widehat{\beta }}_2,p_k^2,+,{\widehat{\beta }}_3,m_k,+,{\widehat{\beta }}_4,m_k^2,+,\widehat{\gamma },g_k,+,\sum ,_{i=1}^n,{\widehat{\delta }}_i,{\stackrel{-}{y}}_i,+,{ϵ}_k\right)$$ 5

Here, the corresponding coefficients of need and non-need variables are estimated using GLM with probit link.

We then estimate indirect need-standardised of incomplete immunisation uptake (by taking the difference of the actual and need-predicted incomplete immunization uptake after neutralizing the effect non-need variables) using Eq. (6). Following Gravelle [14], the estimated indirect need-standardized incomplete immunization uptake is:

$$v_k^{\mathit{INS}}=v_k-{\widehat{v}}_k+\frac{1}{n}{\sum }_{k=1}^n{\widehat{v}}_k$$ 6

We then calculate the Erreygers’s [11]10Corrected Concentration Index (CCI) of indirect need-standardised of incomplete immunisation uptake using Eq. (7) to account for the degree of HI in the system (see Appendix: Section C for detail). Thus, the HI is calculated as,

$${\mathit{CCI}}_{\mathit{HI}}^{\mathit{INS}}=\frac{8}{n^2\left(b_L,-,a_L\right)}\sum _{k=1}^n{v}_k^{\mathit{INS}}{r}_k$$ 7

The value of ${\mathit{CCI}}_{\mathit{HI}}^{\mathit{INS}}$ is bounded between − 1 and 1. The value of the index measures the degree of unequal vaccine uptake due to socioeconomic status and for a given the health need i.e., HI. The sign would reveal the concentration of incomplete immunisation. If it is statistically positive (negative), it implies that incomplete immunisation is mostly concentrated among the rich (pro-poor). The major determinants contributing to HI is identified using the decomposition method [30] (see the detail in Appendix: Section D).

Vertical inequity

The test of vertical equity principle requires that the estimated effect of need variable in whole sample must be equal to estimated effect of need variable of the target groups11. So, for the target groups, we have selected children in households with head as a graduate or households in top 5th quintile groups (see Table 2). We find that the children in target groups have lower incomplete immunization uptake compared to the system as a whole12.

Table 2.

Percentage of children (12–59 months) with incomplete immunised across the two target groups and overall sample

Target Group	All	Graduate or above	Top 5th CEQ
Incomplete Immunisation	29.12%	22.74%	22.88%
Observation	35, 658	3,330	8,424

Proportion is calculated at sample weight; Source - Social Consumption: Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Further, the selection of target groups is based on the fact that after controlling for socioeconomic factors, children of households belonging to the target groups have lowest probability of incomplete immunization, compared to all [25]. The probability curve of incomplete immunization is strictly decreasing in age in the respective groups, indicating that with increase in children age the likelihood of incomplete immunisation decreases, however again the probability of incomplete immunisation across age among the target group is relatively low (Fig. 3).

Fig. 3.

Predicted probability of incomplete immunization with respect to risk exposure and age. Source: Social Consumption: Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

We subsequently check the odds of incomplete immunization between all and the target groups using Eq. 4. Under the null hypothesis, we assume that$\left(\sum ,_{i=1}^n,{\widehat{\beta }}_i,+,\widehat{\gamma }\right)=\left(\sum ,_{i=1}^n,{\dot{\beta }}_i,+,\dot{\gamma }\right)$, where ${\widehat{\beta }}_i,\widehat{\gamma }$ and ${\dot{\beta }}_i,\dot{\gamma }$ are the corresponding coefficients of need variables for the incomplete immunization of all and the target group respectively. If it is statistically different, we reject the null hypothesis and conclude that incomplete child immunisation uptake failed to be vertically equitable. Following Vallejo-Torres et al., (2012), the VI is measured by calculating the concentration index of indirect target standardised incomplete immunisation uptake. Indirect target standardised incomplete immunisation uptake is difference of need-predicted and targeted predicted incomplete child immunization (where the need variables are assumed to have optimal effect (i.e., minimum incomplete immunization compared to the overall population), after neutralizing the effect of non-need variables [33]. Thus, the estimated target-predicted incomplete immunization uptake is,

$${\dot{v}}_k=P\left(\dot{\alpha },+,{\dot{\beta }}_1,p_k,+,{\dot{\beta }}_2,p_k^2,+,{\dot{\beta }}_3,m_k,+,{\dot{\beta }}_4,m_k^2,+,\dot{\gamma },g_k,+,\sum ,_{i=1}^n,{\dot{\delta }}_i,{\stackrel{-}{y}}_i,+,{ϵ}_k\right)$$ 8

To avoid omitted variable bias, we estimate the target-predicted child immunization by adjusting the non-need variables at their mean [14]. The target-predicted incomplete immunisation estimate is the second-best solution13, that is supposed to hold for general population. Following Gravelle [14], the indirect target-standardised CI is:

$$v_k^{\mathit{ITS}}={\widehat{v}}_k-{\dot{v}}_k+\frac{1}{n}{\sum }_{k=1}^n{\dot{v}}_k$$ 9

Likewise, the corresponding indirect target-standardised CCI (see the Appendix: Section C), which measure the VI, is calculated as:

$${\mathit{CCI}}_{\mathit{VI}}^{\mathit{INS}}=\frac{8}{n^2\left(b_L,-,a_L\right)}\sum _{k=1}^n{v}_k^{\mathit{ITS}}{r}_k$$ 10

The value of index is bounded between − 1 and 1. If index is positive (negative) it implies that there is a pro-rich (pro-poor) VI in child incomplete immunisation. When the vaccine distribution is based on equal need, it might be possible that children exposed to greater need (high risk exposure) are not provided with equitably more vaccine. The VI captures the degree of inequity arises due to difference in health needs. To identify the major determinants contributing to differential effect of need is identified using Oaxaca decomposition method [30] (see the details in Appendix: Section D).

For robustness and sensitivity check; (1) We re-estimate the inequity after dropping poorest and richest 5% of the populations (to neutralise the effect of possible outliers); (2) We assume possible linear and non-linear functional form of health needs for sensitivity check; (3) and alternatively, we also estimate bounded Wagstaff’s (2003) corrected concentration index as complementary evidence for inequity in child immunisation.

Data

We have used unit level data from National Sample Survey: “Social Consumption: Health Survey- NSS 75th Round (2017-18)” for the study. The survey is designed on the stratified multi-stage sampling method with an objective to make the sample representative. The data on child immunisation includes information on households’ socioeconomic background, expenditure incurred, status of child immunisation, and the sources of immunisation. A child is said to be incompletely immunised when at least one of the eight doses of vaccines (BCG, OPV-1, OPV-2, OPV-3, DPT-1, DPT2, DPT-3 and measles) are not administered within the prescribed time period of one year after the birth. We have selected children above the age of 12 months and below or equal to the age of 60 months, and those children are dropped who are identified as a transgender (only 0.03%). Out of 70,258 children, a sample of 35,670 children is selected under the criteria. The box plot graph in Fig. 4 indicates that only 25% of children have followed up with all the vaccine jab, at an age of roughly 28 months, which exceeds the recommended age (i.e., 24 months) but more than 50% of children are found to be delayed their vaccine jab beyond the age of 40 months. So, we cannot follow the age threshold recommended by WHO (i.e., 24 months) because much significant share of full immunisation is delayed. The bivariate Erreygers’s [11] Corrected Concentration Index (CCI) requires that the households must be ranked on some socioeconomic scale to reflect the relative position of households, so we have used information on the uniform monthly household per-capita consumption expenditure 14as a proxy [20].

Fig. 4.

Distribution of full immunisation among children between 13 to 60 months. Proportion is calculated at sample weight; Source: Social Consumption - Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Empirical analysis

The summary table regarding the children left with incomplete immunisation across the socioeconomic and regional determinants is available in the Appendix: Section A Table (A1). The proportion in first column shows the share of population of children, in the second column highlights the share of children with incomplete immunisation and corresponding multivariate t-test for difference in mean (which reveals the heterogeneity in uptake). It reveals that children in better off socioeconomic status and from urban regions are relatively having lesser incomplete immunisation uptake. For example, children in the bottom consumption expenditure quintile have more than 10% incomplete immunisation uptake.

Cumulative distribution of Incomplete Immunisation

The Erreygers’ Corrected Concentration Index (CCI) of incomplete immunisation is − 0.0680, and corresponding concentration curve15 (CC) is concave and strictly dominating 45-degree line (see Fig. 5) (also see the Appendix: Section F Table F4) for dominance test), thus indicating that the incomplete child immunisation is concentrated significantly among the poorer households.

Fig. 5.

Concentration curve of child immunisation. MHCE: Monthly household consumption expenditure; Source: Social Consumption - Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Concerning fact is that the uptake of those vaccines which are administered to children phase-wise, after a gestation period, are significantly lesser among poorer households i.e., DPT3 and Measles (see Table 3). Thus, resulting into a significant cumulative concentration of incompletely immunised children among the poorer households.

Table 3.

Inequity in the distribution of non-uptake of the respective dose of Vaccine

Vaccine	BCG	OPV1	OPV2	OPV3	DPT1	DPT2	DPT3	Measles	Overall Incomplete Immunisation
Erreyger’s CCI	-0.0088 (0.0299)	-0.0010 (0.0289)	-0.0141 (0.0263)	-0.0187 (0.0219)	-0.0149 (0.0286)	-0.0147 (0.0254)	-0.0412*** (0.0206)	-0.0573*** (0.0176)	-0.0681*** (0.0131)

*Significant < 0.10 level; ** significant < 0.05 level; *** significant < 0.01 level. The standard error in parenthesis is calculated using bootstrap technique; Source: Social Consumption - Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Regression analysis

As discussed in the preceding Sect. (2.1.1.) that the unequal distribution of immunisation is inducing the HI to evolve, which is further driven by non-need factors. Table 4 highlights the odds of incomplete immunisation uptake against demographic, socioeconomic and regional determinants based on GLM with probit link function. In Model 1, we regress incomplete immunization over health need variables; In Model 2, we drop need variables and include non-need variables; In Model 3, we include both need and non-need variables. To check for the sensitivity of the result, we have dropped richest and poorest 5%, and then estimated the Model 3 for robustness (see the Appendix: Section F Table F5). We have expected that the odds of incomplete immunisation uptake would be lower with respect to risks and age. The estimates apparently are consistent with our expectation based on theoretical insights. The odd of incomplete immunisation is decreasing with the risks exposure, but it is statistically insignificant. Likewise, the odds of incomplete immunisation is found to be increasing in age, indicating possible delayed vaccination uptake. Given the variation in the odds of incomplete immunisation uptake, horizontal equity principle requires that there should not be such variation with respect to non-need factors. However, with respect to education, children of household heads with illiteracy and lower level of education have higher likelihood of incomplete immunisation. Similarly, children in lowest consumption expenditure quintile have higher chance of remaining with incomplete immunisation compared to children in higher consumption quintile. The intervention of public institutions like anganwadi have reduced the likelihood of incomplete immunisation and is statistically significant. Children residing in urban region have significantly lower probability of incomplete immunisation compared to children in rural region. Given the objective of the immunisation program, the difference in uptake due to socioeconomic factors, specifically due to education, income and regional factors indicate horizontal unequal distribution.

Table 4.

GLM-Probit link model of incomplete immunisation uptake

	All			Target Groups
	Model 1	Model 2	Model 3	Graduate or above	Top 5th CEQ
Dependent Variable (Incomplete Immunisation)	Odds Ratio	Odds Ratio	Odds Ratio	Odds Ratio	Odds Ratio
Need Variable
Risks	1.08 (0.1328)		0.9330 (0.1087)	0.5593 (0.1932)	0.7598 (0.1847)
Risk-Square	1.04 (0.1248)		1.13 (0.1268)	2.93** (1.21)	1.28 (0.3632)
Age	0.9650*** (0.0032)		0.9649*** (0.0032)	0.9598*** (0.0082)	0.9603*** (0.0055)
Age-Square	1.00*** (0.0000)		1.00*** (0.0000)	1.00** (0.0001)	1.00*** (0.0000)
Female	1.01 (0.0157)		1.01 (0.0158)	0.9715 (0.0415)	0.9822 (0.0298)
Non-need Variable
Illiterate		1.12*** (0.0398)	1.11*** (0.0382)
Primary Schooling		1.15*** (0.0354)	1.15*** (0.0352)
Secondary Schooling		1.09*** (0.0243)	1.09*** (0.0241)
Casual Labour		0.9281* (0.0333)	0.9199* (0.0334)
Self-Employed		0.9770 (0.0360)	0.9711 (0.0361)
Other Occupation		0.9435 (0.0413)	0.9308 (0.0420)
Quintile 1		1.14** (0.0489)	1.12** (0.0489)
Quintile 2		1.04 (0.0374)	1.04 (0.0395)
Quintile 3		1.01 (0.0366)	1.01 (0.0377)
Quintile 4		1.01 (0.0246)	1.00 (0.0254)
Scheduled Tribe		1.05 (0.0580)	0.9606 (0.0548)
Scheduled Caste		1.01 (0.0297)	0.9431 (0.0539)
Other Backward Class		0.9953 (0.0300)	0.9513 (0.0528)
Muslim		0.9981 (0.0332)	1.00 (0.0347)
Public - Anganwadi		0.7949*** (0.0293)	0.7878*** (0.0313)
Rural		1.06 * (0.0348)	1.06 (0.0370)
Constant	1.25 (0.1521)	0.4953*** (0.0499)	1.74*** (0.2643)	1.68* (0.3599)	1.26 (0.2195)
State Fixed Effect	Yes	Yes	Yes	Yes	Yes
AIC	1.10	1.13	1.12	1.03	1.03
BIC	-374388.3	-373282.8	-373694.5	-26243.3	-75849.53
Log pseudolikelihood	-21876.89	-22371.66	-22139.31	-1864.25	-4797.76
${\chi }^2$(5)				14.68	6.53
Prob >${\chi }^2$				0.0118	0.2583
Observations	35, 670	35, 670	35, 670	3,322	8,416

*significant < 0.10 level; **significant < 0.05 level; ***significant < 0.01 level. Base Variables (all head of the households): Male, Above Secondary Education, Salaried, Quintile 5th (Top 20% CEQ), Unreserved Social Group, Hindu and others, Non-anganwadi source, Urban; The standard error in parenthesis is calculated using bootstrap technique; Sample weight is used in the estimation; CEQ is abbreviated for monthly per-capita consumption expenditure quartile Source: Social Consumption - Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

It is argued that immunisation distribution should be either invariant or positively associated with risks exposure. It is observed that the association of incomplete immunization with respect to need variables are theoretically consistent, but does the effect is same across the socioeconomic groups? If there is a significant difference in the cumulative effect of health need variables on incomplete immunization across socio-economic gradients, it indicates heterogeneous effect, and thus implies vertically unequitable distribution. It is apparent that the odds of incomplete immunization uptake, although low, is much lower for target groups. Overall, there is a significant differential effect of health need variables between all and the graduate or above but it is not significant between all and to top 5th CEQ (consumption expenditure quartile). We re-examined the associations after dropping the richest and poorest 5% from the sample for a robustness check. We find the results are consistent, that is the odds of incomplete immunization in the sub-sample are higher with lower level of education and consumption expenditure. However, there is no significant differential effect of health need variables between all and the target groups (see the Appendix: Section F Table F5)).

Measuring socio-economic related inequity

Using Eqs. (7, 10 and 18), we identify and measure the degree of actual inequity, HI and VI in incomplete immunisation (see Table 5 and see the Appendix: Section F Table F6) for robustness). For robustness, we re-estimate the same using the reduced linear form in need variable. We find that the actual socioeconomic-related inequity of incomplete immunisation is negative thus implying significant concentration of incomplete vaccine uptake among the poorer households. The need-standardised HI is also found to be significantly negative, implying that the incomplete immunisation is horizontally more concentrated among the poorer households. The decomposition results (see Fig. 6 and Appendix: Section F Table F7) indicates that HI is mostly driven by the effect of non-need factors, specifically low level of education, low consumption expenditure, and rural region. The VI in incomplete immunisation differs based on the estimated targeted effect of those households’ heads having graduate degree or above and those households head in top 5th CEQ. We find with respect to graduate or above as a target group, the VI is statistically insignificant and positive. Similarly, with respect to top 5th CEQ as a target group, we again find statistically insignificant pro-poor VI in incomplete-immunisation (or insignificantly pro-rich in linear form). The difference in the likelihood of immunisation uptake between all and target groups with respect to need variables is weak $P\left({\chi }^2,\left(5\right)\right)=0.0118$ with respect to education and $P\left({\chi }^2,\left(5\right)\right)=0.2583$ with respect to top CEQ, and hence it has not culminated into vertical inequity significantly. After accounting for the combined effect of HI and VI, the total inequity in incomplete immunisation is found to be significantly concentrated among the poor, and HI dominates VI. Overall and low education, low consumption expenditure, and regionality are the major factor driving inequity in incomplete immunisation towards the poor, while availability of public service institutions (like anganwadi) is found to be reducing the inequity. Similarly, after truncating out richest and poorest 5% households the similar pattern of concentration in inequity in incomplete immunisation is observed (i.e., pro-poor) and is majorly driven by the factors associated with horizontal inequity.

Table 5.

Equity estimates of Incomplete Immunisation

	Erreygers’ CCI		Wagstaff’s CCI
Concentration Index	Non-Linear	Linear	Non-Linear	Linear
Actual CCI	-0.0681*** (0.0131)		-0.0824*** (0.0158)
Horizontal Inequity
CCIHI	-0.0541*** (0.0098)	-0.0536*** (0.0098)	-0.0577*** (0.0105)	-0.0592*** (0.0109)
Vertical Inequity
CCIVI (Grad or Above)	0.0014 (0.0069)	0.0009 (0.0073)	0.0014 (0.0070)	0.0009 (0.0076)
CCIVI (Top 5th CE Quintile)	-0.0011 (0.0073)	0.0071 (0.0099)	-0.0012 (0.0080)	0.0072 (0.0100)
Total Inequity
CCITI (Grad or Above)	-0.0517*** (0.0093)	-0.0513*** (0.0094)	-0.0544*** (0.0098)	-0.0558*** (0.0102)
CCITI (Top 5th CE Quintile)	-0.0536*** (0.0098)	-0.0532*** (0.0098)	-0.0565*** (0.0102)	-0.0583*** (0.0107)

*significant < 0.10 level; **significant < 0.05 level; ***significant < 0.01 level. CCI: Corrected concentration index; HI, VI and TI in the suffix are socioeconomic related horizontal inequity, vertical inequity, and total inequity; The standard error in parenthesis is calculated using bootstrap technique; Sample weight is used in the estimation; Source - Social Consumption: Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Fig. 6.

Decomposition of Erreygers’ corrected concentration index. Note: Principle Activities: Casual Labour, Self-employed, salaried head; Source: Social Consumption: Health Survey- NSS 75th Round (2017-18); Authors’ own calculation

Decomposition of horizontal and vertical inequity

The decomposition of CCI reveals the major observable policy relevant determinants, driving the HI and VI in incomplete immunisation. The contribution of each determinant in HI is calculated by taking the multiplication of CCI and average marginal effect of the respective determinants. The contribution of vertical inequity is due to differential effect of CCI and average marginal effect of health need variables between all and the target groups. Figure 6 shows the share of each determinant contributing towards HI, VI, and TI. Under HI and VI, we have concentration index (CCI), average marginal effect (ME) (estimated using GLM with probit link model), and the absolute contribution (A.C) of the respective determinants. Under total inequity, we have total absolute contribution (T.A.C) and the percentage contribution to total inequity of each determinant. The decomposed total inequity constitutes the observable factors driving the HI and VI respectively. The individual contribution of the non-need variables towards the inequity are the source of HI, whereas the differential effect of the need variables under the estimated and target provides the evidence for VI. Our model explains almost 75% with respect to top or 5th CEQ group and it explains 44% of TI with respect to graduate or above as a target group. We find that the major non-need determinants like low level of education (contributing 16%), low per-capita consumption expenditure (contributing 36%), low social status (contributing 2.5%) and regional heterogeneity (contributing 14%) are the factors driving HI in incomplete immunisation uptake, while the activities of anganwadi is found to be reducing HI (by almost 14%) in incomplete immunisation uptake.

If we do not account for VI and infer our judgment on HI alone, then it would lead to bias (overestimate or underestimate depending on the target group) in the estimation of the socioeconomic related total inequity (see Table 5). We have observed differential effect of education and income on incomplete immunisation uptake between all and the target groups (see Sect. 3.2). We check whether the pro-poor VI is emanating due to significant differential effect of need variables between the target group and the whole sample. On decomposition, we find that the differential effect of age between all and the target groups is contributing to VI, however such differential effect have no significant contribution to VI. The total inequity indicates that the differential exposure to risks factors between all and the target groups is contributing to inequity in incomplete immunisation uptake (64% corresponding to education as a target and 13% corresponding to consumption as a target). Hence, it implies that the non-need socioeconomic factors are actually inducing pro-poor total inequity in full immunisation than need factors (see the Appendix: Section F: Table F7) for details).

Conclusion

Ensuring complete immunisation to the children is an uphill policy challenge in a highly fractionalised societies in India. This is particularly critical when more than one-third of global child mortality occur in India due to vaccine preventable diseases [5]. It is evident that there is a significant concentration of incomplete immunisation among the poorer households due to the difference in socioeconomic, regional and spatial gradients. These unequal immunisation uptakes inform us about the inequality in child immunisation and the inequality in disease burden, but does not comprehensively reveal the inequity in the system. Given, the difference in risks exposure and the socioeconomic status, to achieve the policy target of 90% coverage on equity principle, it is necessary to investigate and monitor inequity comprehensively.

The inequity in child immunisation may arise due to horizontal and vertical differences among the children across the socio-economic gradient. The HI is the need adjusted inequity emanating from non-need (socioeconomic and regional) factors, while VI is the inequity emanating due to the differential effect of the need indicators (risks exposure and age) on immunisation uptake. The inequity measurement would be partial and it would lead to different conclusion on nature and extent of the inequity in the system if we do not account the differences arising due to heterogeneous risk exposure. Using bivariate concentration index and its decomposition method, we attempt to measure socioeconomic-related inequity in incomplete child immunisation and determine the major policy relevant factors contributing towards the inequity.

We find that the CCI of actual inequity is negative, indicating the children with incomplete immunisation is significantly concentrated among the poorer households. After adjusting for the need indicators, the HI is also found to be negative, implying pro-poor inequity. The linear CCI decomposition indicates that the non-need factors like low education level, low per-capita consumption expenditure, low social status and regional difference are the major factors driving pro-poor HI, while community Health service like anganwadi is the major factor contracting the pro-poor HI. Further, we do observe weak differential effect of need variables between all and the target groups on incomplete immunisation uptake, but it is only significantly different with respect to graduate or above as a target group. However, the differential effect is not realised into significant VI. Also, depending on the target group the CCI of VI is ambiguous. The decomposition of VI reveals that differential effect of age between all and the target groups is major drivers of VI. Overall, with respect to graduate or above as a target group, the model explains 44% of total socioeconomic related inequity, while with respect to top 5th CEQ the model explains 75% of the total inequity. Based on both the target groups, there is an unanimity about pro-poor concentration of incomplete immunisation. The share of determinants contributing towards total inequity are risk exposures and the factors driving HI. Under child immunisation scheme all the mandatory vaccines are freely available to eligible children, but still the total inequity come out to be pro-poor, indicating the provision is heavily biased against the disadvantaged. Children in poorer households are not being able to obtain effectively appropriate vaccine distribution according to child health care needs. Our finding suggests that child immunisation program should target factors driving HI and need to align their distribution in terms of risks exposures.

The role of anganwadi CHWs at the grassroots levels need to be highlighted as the intervention has proved to be substantially effective in reducing the inequity in immunization uptake, although there are regional variations in their degree of participations. CHWs may have the highest impact in increasing the vaccination coverage through door-to-door canvassing [11, 21] and help in narrowing the gap between the healthcare providers and the community respectively. Our estimated inequity in child immunization would have been magnified in the absence of the CHWs interventions.

The study has some limitations. The recorded immunisation uptake is self-reported. The selected proxy variables which are used to capture the risk exposures are not exhaustive and only reveals a partial correlation. Other risk factors (like hesitancy rate, immunodeficiency, etc.) which causes vaccine preventable diseases, could not be incorporated due to data constraints. The selection of target function for the analysis of VI is again not straight forward but rather based on the value judgment, hence, it suffers from some subjectivity. Further, the target function is based on the appropriate effect of a section of socioeconomic group with least unmet immunisation, and therefore the VI captures the partial inequity emanating due to distributional consequence across the consumption groups. Hence, the estimated inequity in this exercise reveals partial inequity in the system and suffers from under-estimation. At macro level, Indian states differ on demographic profile and as well as on income level, and therefore might reveal Kuznets’ behaviour in the vaccine uptake [24]. So, it is likely the dimension of inequity may differ across states.

Authors’ contributions

Dr. A. Akhtar has contributed in drafting introduction, write up design, conceptualization, formalization of methodology, data curation, preliminary analysis, statistical analysis, and interpretation of data. Dr. I. Roy Chowdhury is responsible for write up design, interpretation of data, critical revision, validation, and supervision. Ms. P. Gogoi and Ms. S.P. Reddy have contributed in drafting of the introduction and literature survey of the manuscript. All the authors have reviewed and approved the final draft of the manuscript

Data availability

Analysis is solely based on the nationally representative secondary database on India. These data sets can be available (on request) from the public domain of the Government of India, Ministry of Statistics and Program Implementation (MOSPI).: Unit Level data & Report on NSS 75th Round for Schedule- 25.0, July 2017 -June 2018, (Social Consumption: Health) Link: https://microdata.gov.in/nada43/index.php/catalog/152.

Declarations

Competing interests

The authors declare no competing interests.