India’s Integrated Child Development Services programme; equity and extent of coverage in 2006 and 2016

Abstract Objective To investigate coverage and equity of India’s Integrated Child Development Services programme across the continuum of care from pregnancy to early childhood, before and after the programme was expanded to provide universal access. Methods The programme offers nutrition and health services to pregnant and lactating mothers and young children. We used data from nationally representative surveys in 2005–2006 and 2015–2016, including 36 850 mother–child pairs in 2006 and 190 804 in 2016. We assessed changes in the equity of use of programme services by socioeconomic quintile, caste, education and rural or urban residence. We used regression models to investigate the determinants of programme use. Findings The mean proportion of respondents using programme services increased between 2006 and 2016, from 9.6% to 37.9% for supplementary food, 3.2% to 21.0% for health and nutrition education, 4.5% to 28% for health check-ups and 10.4% to 24.2% for child-specific services (e.g. immunization, growth monitoring). Wealth, maternal education and caste showed the largest positive associations with use of services. However, expansion in service use varied at the sub-national level. Although overall use had improved and reached marginalized groups such as disadvantaged castes and tribes, the poorest quintiles of the population were still left behind, especially in the largest states that carry the highest burden of undernutrition. Conclusion India’s policy reforms have increased coverage of the programme at the national level, including for marginalized groups. With further scaling-up, the programme needs to focus on reaching households from the lowest socioeconomic strata and women with low schooling levels.


Introduction
In 2013, reviews of effective nutrition interventions estimated that scaling-up a set of proven nutrition-specific interventions could reduce stunting globally by 20% and reduce child mortality by 15%. 1 Other modelling exercises have attempted to estimate the potential impact of scaling-up key interventions on progress towards sustainable development goals or on economic growth. 2,3 Few studies, however, have examined how programmes expand to achieve interventions at scale. Fewer still have assessed the extent to which programme expansion reaches the most vulnerable populations. However, it is the juxtaposition of coverage and efficacy that explains progress in reducing malnutrition or its absence.
Nutrition programmes that include targeted food assistance have been implemented at large-scale worldwide for several years, such as the Women, Infants and Children Programme in the United States of America 4 and the Oportunidades programme in Mexico. 5 Studies of targeted programmes have documented respondents' characteristics and enrolment in and coverage of the programme, 5 but less is known about factors affecting uptake of universally offered programme services. India provides a case study to examine the scale-up and uptake of a large-scale, universal, food-assisted maternal and child nutrition programme. 6 India launched its Integrated Child Development Services programme in 1975 7 and expanded it to all states in 2000. However, services were patchy throughout the early 2000s. 8 In 2006, India's Supreme Court ruled that the programme was to be offered universally and, soon after, the government expanded the availability of programme services across India, with a goal of ensuring about 1.4 million programme centres across the country. The programme now serves about 82 million children younger than 6 years and over 19 million pregnant women and lactating mothers. 6,9 Services currently include take-home supplementary food and hot cooked meals, health and nutrition education, and health check-ups delivered at rural child-care centres (called anganwadi) or at home. 7 Immunizations, growth monitoring and pre-school care services are also available for children at the centres. 10 There is mixed evidence on the coverage and effectiveness of India's programme on nutrition and related outcomes. Studies from the 1990s generally found programme placement that was skewed towards the well-off districts and no effects on anthropometric outcomes. 11,12 Studies from the following decade reported positive impacts, but with important caveats. For instance, a study based on 2005-2006 data found that girls who received supplementary food from the services were on average taller. 13 However, only 6% (329) of 5364 girls aged 0 to 2 years and 14% (1113) of 7951 girls aged 3 to 5 years in that study received the supplementary food, despite the fact that over 90% of villages (3522/3849) had an integrated child development centre. This result raised questions about the factors influencing programme uptake. 13 These studies preceded major reforms in the services between 2006 and 2009. 7 The reforms included greater financial outlay from the central government and provision of supplementary food in a rights-based framework. 14 Most research on the delivery of India's Integrated Child Development Services programme in the period after the reforms has focused on performance in implementation of the programme by states, 10 with limited evidence on individual and household uptake or use of the programme. In view of these gaps, we investigated changes in the use of the services over the continuum of care from pregnancy up to early childhood between 2006 and 2016. We assessed equity gaps and factors associated with use of services. Our analysis has policy implications for India, but also offers global lessons to other countries embarking on scalingup integrated programmes to address maternal and child health, nutrition and child development.

Outcome variables
Our primary outcomes were receipt of programme interventions among pregnant and lactating women aged 15 to 49 years and their children aged 0 to 59 months. We restricted analyses to lastborn children to minimize recall bias. We analysed 12 outcomes of the programme services for mothers and children. We grouped them into four types of services (supplementary food, counselling on nutrition, health check-ups and early childhood services), offered over three phases in the care continuum (pregnancy, lactation and early childhood). Detailed definitions of the indicators are provided in the data repository. 17 For the analyses of inequity and determinants of use of services, we constructed composite indicators to represent receiving services during all three phases of the care continuum. Mathematically, for each service k, the composite indicator Y k was defined as: Where Y kx denotes receiving service k in period x, and x can be pregnancy (p), lactation (l) or early childhood (c). We analysed receipt of the following services by mother-child pairs: supplementary food, counselling on nutrition practices and health check-ups. For services that are only provided during early childhood (immunizations, pre-school education and growth monitoring services) the composite indicator is simply an indicator for the child having received the service. We also examine the changes in the frequency of receiving supplementary food in the 12 months preceding the survey. This question is only available for the child and does not clarify CI: confidence interval. a Index of socioeconomic status was constructed by factor analysis using: household access to improved drinking water, improved latrine, clean cooking fuel, electricity, ownership of a house and land, housing materials for floor, roof and wall, and possession of 15 assets (including a mattress, pressure cooker, chair, bed, table, fan, television, sewing machine, phone, computer, refrigerator, watch, bicycle, motorbike and car) and livestock (cow, goat and chicken). b Education categories were grouped as follows for total number of years of education attainted by the individual: no schooling (0 years), primary school (1 to 5 years), secondary school (6 to 10 years) and high school or higher (≥ 11 years). Notes: n is the total number of respondents. We analysed data from India's national family health surveys whether the supplementary food received is take-home rations or daily food at the rural child-care centres, so the full amount of food received is uncertain. Therefore, we report on this question, but did not include it in the composite indicator.

Explanatory variables
Household level variables included household size, socioeconomic status, religion, caste category, urban or rural residence and access to health insurance. The socioeconomic status score was constructed using a factor analysis of multiple variables including water source, toilet type, materials used in dwellings, and ownership of a house, land, livestock and durable assets. 18 Mother-and childspecific variables included the mother's age and education, and the child's sex, age and birth order.

Statistical analysis
We tested changes in outcomes and determinants from 2006 to 2016 using regression models. Equity analyses were conducted for the 12 individual outcome variables by socioeconomic status quintiles at the national level, for residential areas, castes and maternal education for the two survey rounds, adjusting for sample probability weights. We plotted changes in the services received within socioeconomic status quintiles for these categories.
We used multivariate logistic regression models to determine the association between explanatory variables and each of the four composite indicators. To examine changes over time, we conducted the regressions for 2006 and 2016 separately, with clustering of standard errors within states. Dummy variables for the 35 Indian states were included in all specifications.
Thus, for household h in states in time t, we estimated: Where P (Y khst = 1) denotes the probability that the composite indicator Y k took on value 1 for household h in state s in time t. We controlled for J covari-ates at the household level, as listed above under explanatory variables, and for state-fixed effects, γ st . Regressions were run separately for each composite indicator k.
For robustness checks, we conducted the regression analyses individually for all 12 services. Since different services are provided by child's age group, we also examined the summary statistics for service coverage by age (< 6 months, 6 to 35 months and 36 to 59 months). We adjusted our models for statespecific performance using an index of programme performance developed using a survey from 2013-2014. 10 The performance data covered programme infrastructure, knowledge and service provision of workers at rural child-care centres, and annual expenditure per child. 19,20 Finally, we conducted separate regression analyses for two states with the highest burden of undernutrition in India (Uttar Pradesh and Bihar) to investigate if determinants differed compared with national-level estimates. 21

Results
Several individual and household characteristics changed appreciably between the two rounds of national family health surveys (  23.5 percentage points and health and nutrition education increased 17.9 percentage points. These patterns were consistent for the child age subgroups (data repository). 17 The frequency of receiving supplementary food also improved, with a nearly 8 percentage point increase for children who received food monthly ( At the sub-national level there was substantial variation in the expansion of service use across the two survey rounds (Fig. 1). With some exceptions (Tamil Nadu, Odisha, Chhattisgarh and Jharkhand), the coverage of supplementary food during pregnancy and lactation was < 25% in most states in 2006. By 2016, this had improved in almost all states. The state-wise coverage of supplementary food during childhood was higher than in the other two categories, and coverage increased in all states by 2016, with increases to ≥ 50% in many of the central and southern states. Use of the four categories of services among pregnant and lactating women and their children by respondents' characteristics are shown in Table 3. Factors associated with use of services also changed over time (Table 4 and Table 5  In our robustness checks, we found similar findings when using the 12 individual services as outcomes (data repository), 17 and when using the composite coverage indicators by child's age group.
Finally, we saw that trends and determinants of service use in Uttar Pradesh and Bihar did not differ substantially from those for overall national data (data repository). 17 We found higher odds ratios for use of all services among scheduled caste groups compared with other groups in both 2006 and 2016.

Discussion
Using two nationally representative data sets we provide evidence on how the use of India's Integrated Child Development Services programme has changed in the decade after reform of the programme. India appears to be well on its way to scaling-up of nutrition-specific interventions using the integrated services. This large expansion in services is laudable given challenges such as decentralization of implementation to the state level, high numbers and diversity of the population, constraints on funding, and lack of community awareness, among others. 6 Indeed, these challenges are reflected in our findings that expansion in use of services varied considerably at the state level and by sociodemographic characteristics. Even though households in the poorest quintile were better reached by the services in 2015-2016, the wealth inequality in use widened over the decade studied.
The exclusion of the poorest people from services is concerning. Most of the poor who were left behind were from states known to be weak performers on the programme, such as Uttar Pradesh 22 and Bihar, 23 suggesting that overall poor performance in high-poverty states could lead to major exclusions. The exclusion of the poorest quintile could also be due to the challenges of reaching remote or difficult geographical areas, even in better-performing states, despite attempts to close equity gaps district- We analysed whether the service was received at all three phases in the care continuum: during pregnancy, during lactation and in early childhood.
c Immunizations, pre-school education and growth monitoring. d Index of socioeconomic status was constructed by factor analysis using: household access to improved drinking water, improved latrine, clean cooking fuel, electricity, ownership of a house and land, housing materials for floor, roof and wall, and possession of 15 assets (including a mattress, pressure cooker, chair, bed, table, fan, television, sewing machine, phone, computer, refrigerator, watch, bicycle, motorbike and car) and livestock (cow, goat and chicken).
e Education categories were grouped as follows for the total number of years of education attained by the individual: no schooling (0 years), primary school (1 to 5 years), secondary school (6 to 10 years) and high school or higher (≥ 11 years).
Notes: The total number of respondents were 35 570 and 185 101 in 2006 and 2016, respectively. For continuous variables the numbers represent the total number of individuals who received the service and had valid data for the respondent characteristics.
(. . .continued) by-district. 24 There can also be local challenges of exclusion within villages due to caste or location. 25 Further investigations around the potential reasons for exclusions could help in addressing inequity gaps.
The risk of exclusion of the poorest households has also been documented in the coverage of India's safe motherhood cash-transfer programme, the Janani Suraksha Yojana. 26,27 Even in a state such as Odisha, with a well-performing health system, similar inequitable patterns of use were seen for a conditional cash-transfer programme for nutrition and health. 28 Some authors have attributed the exclusion of the poorest households in certain programmes to Table 4.  We analysed whether the service was received at all three phases in the care continuum: during pregnancy, during lactation and in early childhood. b Index of socioeconomic status was constructed by factor analysis using: household access to improved drinking water, improved latrine, clean cooking fuel, electricity, ownership of a house and land, housing materials for floor, roof and wall, and possession of 15 assets (including a mattress, pressure cooker, chair, bed, table, fan, television, sewing machine, phone, computer, refrigerator, watch, bicycle, motorbike and car) and livestock (cow, goat and chicken). c Education categories were grouped as follows for total number of years of education attained by the individual: no schooling (0 years), primary school (1 to 5 years), secondary school (6 to 10 years) and high school or higher (≥ 11 years). Notes: n is the total number of respondents. As some data were missing for the mother's and the child's age we used data only for the set with all complete variables. The following 12 outcomes of integrated child development services were analysed over three phases in the care continuum. During pregnancy: (i) supplementary food; (ii) health and nutrition education; (iii) health check-ups. During lactation: (iv) supplementary food; (v) health and nutrition education; (vi) health check-ups. During early childhood: (vii) supplementary food as take-home rations or hot cooked meal; (viii) health check-ups; (ix) nutrition counselling for the mother after the child was weighed; (x) childhood immunizations; (xi) early childhood care and preschool education; and (xii) growth monitoring. We adjusted estimates are for sampling weights and standard errors clustered at the state level. All specifications include state-fixed effects (N-1 dummy variables for 35 Indian states).

Table 5. Factors associated with use of health check-ups and child-specific services in the Integrated Child Development Services programme among pregnant and lactating women and their children in India, 2006 and 2016
difficulties in complying with the programme conditions. 29 It is also unlikely that the use of services has many barriers that prevent poorer people from using services that are locally available. Except for entry-level barriers, such as pregnancy registration, the unconditional nature and universal scope of the services should make it accessible to all. As noted above, a more plausible explanation is weak performance of service implementation in states or districts with the highest proportions of poorer people. Indeed, in the two states of Uttar Pradesh and Bihar, we found greater exclusion by caste as well, suggesting that exclusions in these states is due to overall poor performance, leading to low coverage for all, rather than targeted exclusions for some. In separate policy-focused descriptive analyses of coverage, 30 we provide district-specific coverage estimates for all programme services that can inform policy-makers in India and elsewhere. Despite lingering caste discrimination in India, 31 it is reassuring that caste and tribe-based exclusion from the programme services has declined. The caste differences appear to favour the traditionally marginalized scheduled castes and scheduled tribes groups compared with the general castes, after controlling for wealth. In Odisha and in Chhattisgarh, where there are large pockets of tribal populations, efforts to strengthen overall programme services with a view to improving equity of access 10 has likely helped close gaps for tribal communities. In Maharashtra, a targeted focus on tribal areas as part of the state nutrition mission 32 has also likely helped to close some gaps.
Despite the hypothesis that girls are discriminated against in health service use in India, 33 we did not find evidence of son preference in families' use of programme services. Instead, we found that being a female child significantly improved the likelihood of receiving the range of services, although the difference was slight. Similarly, there was no indication that children who were higher in birth order had lower use of services.
Health insurance coverage improved 10-fold between 2006 and 2016, likely due to the introduction of a national health insurance programme, the Rashtriya Swasthya Bima Yojana, for households below the poverty line in 2008. This programme is currently operational in 25 states. In our study, those with health insurance were about twice as likely to use the services in 2016. However, since these services are free of charge there is no direct role for insurance in gaining access. This finding could reflect self-selection; households that take up government health and nutrition services are also more likely to be enrolled in the national health insurance. Also, the same states could perform better both on the expansion of the national health insurance programme and on child development services.
Our study had several strengths. It offers an in-depth analysis of individual and household access to a large, universally-offered nutrition programme that targets the first 1000 days of life. The study examines the inequity in the expansion in use of the programme and investigates the complex factors associated with use. By comparing the coefficients of these covariates over time, we have documented how the child development services in India have become more inclusive between 2006 and 2016 and yet, how some groups, such as the poorest wealth quintiles, are still not being adequately served. Finally, our results are robust to several alternative regression specifications, including adjustments for state performance.
Although the survey questions remained the same over time (thus enabling comparisons), they are limited to participation in the last 12 months for children or during pregnancy and lactation. Finer-grained questions could enquire about the regularity and intensity of participation, the actual consumption of foods provided through the programme and the regularity of the use of specific services, such as growth monitoring. Such questions could help planners and policy-makers make more informed assessments about strengthening programme services.
Instead the study focused on changes over time, it is not longitudinal at the individual level. Our ability to analyse the intensity of participation in the programme is limited by the availability of the survey questions included in India's national family health survey questionnaires. Finally, an impact assessment of the programme on maternal and child health is beyond the scope of this paper and is an important area for future research. The post-reform period of the programme coincided with a period that also saw a 10-percentage point reduction in stunting among children aged 0 to 5 years. 34 India's policy reforms have increased coverage of the Integrated Child Development Services programme at a national level and reached marginalized groups. With further scaling-up, the programme needs to focus on effective subnational implementation to reach households from the lowest socioeconomic strata and women with low schooling levels, as these high-need groups are currently more excluded in India. ■