Long‐term trend in socioeconomic inequalities and geographic variation in the utilization of antenatal care service in India between 1998 and 2015

Abstract

Objective

To investigate the temporal trend of socioeconomic and rural‐urban disparities and geographical variation in the utilization of antenatal care (ANC) services in India before and throughout the Millennium Development Goals era.

Data Sources/Study Setting

For this temporal analysis, secondary data from the Indian National Family Health Surveys between 1998 and 2015 (Waves 2, 3, and 4) were used.

Study Design

We analyzed the trend in inequality for at least one and four ANC visits to a health care professional (ANC1+ and ANC4+, respectively) by education, wealth, and residence type. Multilevel logistic regression models were used to assess the temporal trend and the relative contribution of communities and states to the overall variation in ANC1+ and ANC4+.

Data Collection/Extraction Methods

Data on utilization of ANC services for the last birth of women aged 15‐49 years during the three or five years preceding the survey (depending on the survey year) were used.

Principal Findings

Educational and wealth inequality in ANC1+ and ANC4+ worsened between 1998 and 2005 and improved between 2005 and 2015 (for ANC4+, OR [95% CI] = 0.22 [0.19‐0.25] in Wave 2; OR [95% CI] = 0.19 [0.17‐0.22] in Wave 3; and OR [95% CI] = 0.38 [0.36‐0.40] in Wave 4 for the poorest). Rural‐urban inequality showed a consistent decline (for ANC4+, OR [95% CI] = 0.59 [0.54‐0.64] in Wave 2; OR [95% CI] = 0.63 [0.59‐0.68] in Wave 3; and OR [95% CI] = 0.82 [0.79‐0.85] in Wave 4 for rural area). The relative contribution of the community to the total geographic variation in the utilization of ANC services increased more than four times during the study period.

Conclusions

The use of ANC services remains disproportionately lower among women with low socioeconomic status. Efforts to directly target these women are necessary to tackle inequality in ANC utilization in India.

What This Study Adds.

• We identified only two relevant studies on the changes in economic inequality in utilization of antenatal care (ANC) across time. However, these two studies are now outdated, used indicator for unadjusted inequality, or were not geographically inclusive. In addition, no studies examined the trend in geographic variation in ANC utilization.

• We found that socioeconomic inequality was widened during 1998‐2005, but narrowed between 2005 and 2015, which might be partly due to decrease in utilization of ANC of women from the richest and the most educated group.

• We also found the difference in trend of inequality between at least one ANC visit and at least four ANC visits, as well as difference in trend between educational inequality and economic inequality.

• This study presents the first evidence on the importance of community in contributing to the geographic variation in ANC utilization over time.

1. INTRODUCTION

Reducing maternal and childhood mortalities has remained a significant interest in international society for decades.1 Millennium Development Goals (MDGs) initiative yielded 43.9 and 50 percent overall reduction in maternal mortality ratio (MMR) and neonatal mortality rate (NMR), respectively, between 1990 and 2015.1, 2 However, several developing countries continue to report high figures.3 India, despite the substantial decline (approximately 70 percent) in both maternal and infant mortality during 1990‐2015, has failed to achieve the MDG target, ranking the 128th in MMR among 183 countries and the 147th in infant mortality among 195 countries in 2015,4, 5 and continues to contribute almost 20 percent of global infant deaths.2 What is worse is the persistent inequalities within the country by socioeconomic status (SES).6, 7

Approximately 80 percent of maternal deaths in developing countries are attributed to problems preventable by a few maternal interventions such as proper antenatal care (ANC) services or delivery by skilled birth attendance.8 Proper ANC services have been demonstrated to have positive effect on maternal and neonatal mortality and morbidity, directly through diagnosis, management, and prevention of pregnancy‐related complication or concurrent disease,9 and indirectly through identification of women or girls at higher risk of developing complications during delivery, and thus referring to a higher level of facilities.10, 11 Health education, health promotion, and social support provided by antenatal consultation are other indirect benefits of ANC services.12 Evidence suggests that ANC visit is associated with a higher likelihood of childhood vaccination and skilled birth attendance.13, 14

Factors influencing utilization of health care service, including ANC, may come from either demand side or supply side, or most likely both. The demand‐side barrier includes a lack of knowledge and information on health care choices/providers, attitude and norms, and financial difficulty to pay due to low income. Indirect costs such as distance cost and opportunity costs are also important.15 The supply‐side barriers include availability and quality of services (technology, staffs, equipment., etc).15 All these factors are likely to be closely related to socioeconomic factors.

In India, ANC initiatives have a long history, dating back to the 1950s. “Five Year Plan,” a nationwide economic development program, which started in 1951, had a maternal and child health service component. Although this program was originally geared toward family planning service provision, there was a modest achievement of an increase in access to ANC among rural women during the Third Five Year Plan (1961‐1966) by hiring auxiliary nurse midwives (ANMs) and health assistants.

Major initiatives to promote maternal and child health, especially among the poorer populations, were undertaken as important components of the Family Welfare Programme of the Government of India in the 1990s, namely the Child Survival and Safe Motherhood Programme (CSSM) and the Reproductive and Child Health (RCH) program (phase I). Subsequently, the National Rural Health Mission (NRHM) was launched in 2005, which is still ongoing and covers 18 states. The NRHM is more focused on the poor compared to prior programs operated during the 1990s in terms of improving availability, accessibility, and quality of health care for the underprivileged populations.16 Importantly, all of these initiatives encompassed components to promote the utilization of ANC among pregnant mothers in India.

While the prevalence of the utilization of ANC service has been repeatedly investigated to monitor the effectiveness of these programs, there are two important gaps in the current literature. First, evidence of how equally Indian women improved in utilization of ANC service over time is limited. Most of the studies on the socioeconomic gradient in the utilization of ANC in India are cross‐sectional, cover limited geographic area, or are outdated.7, 17, 18, 19 Second, the changes in geographic variation in the ANC services utilization by multiple contextual levels have not been explicitly assessed. Given that factors at state, district, and community levels are likely to simultaneously shape the distribution of ANC services utilization, it is important to partition the total variation in ANC services utilization to identify the relative contribution of each unit and their changes over time. Such analyses may have important implications as to which contextual level needs to be targeted more to reduce inequality of ANC services utilization in India.

In the new Sustainable Development Goal (SDG) era, there is an urgent need to provide a comprehensive assessment of the long‐term trend in socioeconomic inequality and geographic variation in the utilization of ANC services given the economic liberalization and a number of health care initiatives aimed at the poor that were implemented during the MDG period. Multilevel modeling using three rounds of the National Family Health Surveys (NFHSs) from 1998 to 2016 provides a unique opportunity to achieve the following objectives: (a) assess temporal trend in socioeconomic and rural‐urban disparities in utilization of ANC services and (b) assess temporal trend in the relative contribution of important geographic levels to the total variation in utilization of ANC services.

2. METHODS

2.1. Data

The data used in the present study were taken from the three waves of the NFHS 2 (1998‐1999), NFHS 3 (2005‐2006), and NFHS 4 (2015‐2016).20, 21, 22 The NFHS is an Indian version of the Demographic Health Survey (DHS), which is a nationally representative survey that provides information on fertility, mortality, morbidity, family planning, utilization of maternal and child health services, and other indicators related to maternal and child health.20, 21, 22 NFHS 4, unlike previous waves, was designed to provide information for all 29 states and 7 union territories in India.22

NFHS 2 and NFHS 3 adopted a two‐stage sample design in rural areas and a three‐stage sample design in urban areas, while the NFHS 4 sample was selected by a stratified two‐stage sampling for both urban and rural areas. For NFHS 2 and NFHS 3, the Primary Sampling Units (PSUs) in rural areas, corresponding to villages, were selected with probability proportional to population size (PPS) in the first stage, followed by random selection of households within each PSU at the second stage. In urban areas, wards were selected with PPS sampling in the first stage, followed by the second stage of random selection of one census enumeration block (CEB) from each sample ward. Finally, households were randomly selected within each selected CEB. For NFHS 4, village and CEB served as PSU in rural and urban areas, respectively. For a sub‐sample for state module in NFHS 4, about one‐third of the main sample PSUs were randomly selected and 22 households were selected per each PSUs.

2.2. Outcome

The present study focuses on two indicators of the utilization of ANC service. The first outcome is having had at least one visit to a health care provider for ANC during the last pregnancy (hereafter, ANC1+). This is the definition of ANC used in the Composite Coverage Index (CCI) developed to measure inequalities in coverage of maternal and child interventions.23 The second outcome is having had at least four ANC visits to a health care provider during the last pregnancy (hereafter, ANC4+). ANC4+ has been the World Health Organization (WHO) recommendation since 2001.24, 25 Health care providers include doctors, nurses, or midwives. NFHS 2 (1998‐1999) collected information on utilization of ANC service for the last two births during the three years preceding the survey, whereas NFHS 3 and NFHS 4 asked about the utilization of ANC service only for the most recent birth in the five years preceding the survey. We limited the study sample to the last birth for all three waves to ensure comparability. These two outcome variables (ANC1+ and ANC4+) were dichotomized to yes vs. no.

2.3. Inequality measures

In the assessment of long‐term trends in inequality in the utilization of ANC service by SES, maternal education and household wealth were the key variables of interest. Maternal education was categorized into four groups: No education, primary graduate or less, secondary graduate or less, and college or above. In the NFHS, household wealth index was estimated from a set of economic proxies including 33 durable asset ownership, access to utilities, and infrastructures and housing characteristics.26 Each household asset was assigned a weight generated through the principal component analysis (PCA) to construct the composite wealth index. Then, percentile distribution of wealth score was estimated from the composite wealth index and the distribution was divided into five equal quintiles.22 The urban‐rural disparity in the utilization of ANC service was assessed by the respondent's type of residence.

2.4. Other covariates

In order to better assess the temporal trends in inequality in ANC service utilization by SES and place of residence, we adjusted for the following covariates that were selected a priori based on our literature review. The association between these covariates and ANC service utilization is not interpreted in detail as this was beyond the primary objectives of this paper. The covariates included maternal age at pregnancy (<20, 20‐29, 30‐39, and ≧40 in yrs),7 experience of terminated pregnancy (yes vs. no),7, 27 marital status (currently married or living with partners vs. widowed or divorced or separated or never married),27 religion (Hindu, Muslim, Christian, and others),7 family size (3 or less, 4‐6, and 7 or above),28 birth order (first, second, third, and fourth or below),27 and caste groups (scheduled caste, scheduled tribe, other backward class, and others).7 Three proxy variables for women's autonomy were additionally considered in a sensitivity analysis because the relevant questionnaire was asked for only a small subset of the study sample. Women's autonomy is known to be highly correlated with both our inequality measures (SES and type of residence) as well as the outcome of interest (utilization of ANC services).29, 30 We operationalized women's autonomy following prior studies' conceptualization as women's freedom to exercise their judgment in order to act for their own interest in three dimensions: decision making autonomy, freedom of movement, and financial autonomy.30, 31, 32, 33 More specifically, decision making autonomy was measured based on responses to “who decides on 1) obtaining health care, 2) purchasing jewelry or other major household items, and 3) visits to family or relatives.” For each question, a binary variable was created, with “1” representing decisions made by a respondent alone or jointly with her husband or others and “0” representing decisions made by husband alone or other household members. These values were summed up, resulting in a score ranging from 0 to 3. To measure freedom of movement, two items were used for NFHS 2 (whether women were allowed to 1) go to the market and 2) visit relatives or friends), and three items were used for NFHS 3 and NFHS 4 (whether women were allowed to go to 1) the market, 2) the health facility, and 3) the place outside the village/community). Each of these questions was answered as “alone,” “with someone else only,” or “not allowed at all.” These were coded as 2, 1, and 0, respectively, and then summed resulting in a score from 0 to 4 for NFHS 2 and from 0 to 6 for NFHS 3 and NFHS 4. Financial autonomy was measured with a question about whether a respondent was allowed to have money for her own use, which was answered with “yes” or “no.”

2.5. Statistical analysis

First, we performed descriptive statistics to show the distribution of our study sample and the crude prevalence of ANC services utilization by the selected variables across all three surveys. Then, multilevel logistic regression models with random intercepts for individual women i (level 1), community j (level 2: village for rural and CEB for urban), and state k (level 3) were performed to investigate the differences in utilization of ANC services by SES and rural/urban, while accounting for the clustering of observations at the community and state levels and adjusting for other salient covariates.

The multilevel model is useful to address a few gaps that single‐level analysis cannot address. First, single‐level studies implicitly assume all observations to be independently and identically distributed. However, this assumption is not plausible, especially in large countries like India. Multilevel model considers the clustering of observations across multiple levels of geographic units. Second, the multilevel model allows examination of the relative importance of multiple relevant levels while the single‐level models cannot.

$$\log \text{it}\left({\pi }_{\mathit{ijk}}\right)={\beta }_0+\mathrm{B}{X}_{\mathit{ijk}}^{\prime }+\left(v_{0k}+u_{0jk}\right)$$

$$[v_{0k}]\sim N\left(0,{\sigma }_{v0}^2\right)$$

$$\left[u_{0jk}\right]\sim N\left(0,{\sigma }_{u0}^2\right)$$

This model estimates the log odds of π_ijk (ANC1+ or ANC4+) while adjusting for a vector ($X_{\mathit{ijk}}^{\prime }$) of the aforementioned independent variables measured at the individual level. Parts inside the bracket are random effects which are residual differential for each state k(v _{0 k}) and community j (u _{0 jk}). This model was performed separately for each wave. The changes in the magnitude of inequality in ANC1+ and ANC4+ were assessed by the differences in odds ratios (ORs) associated with education, wealth levels, and residential area over the three waves.

Calculation of the proportion of variation in the log odds of ANC1+ or ANC4+ attributable to each level, known as variance partitioning coefficient (VPC), was based on variance estimates of random effects. For instance, the proportion of variation attributable to states was estimated by dividing between‐state variance by the total variance, specified as ${\text{VPC}}_{\text{state}}={\sigma }_{\text{state}}^2/\left({\sigma }_{\text{state}}^2+{\sigma }_{\text{community}}^2+{\sigma }_{\text{indivudual}}^2\right)$. Since the variation at the lowest level cannot be directly obtained in multilevel logistic model, it was approximated as 3.29, treating between individual variation as having a variance of a standard logistic distribution.34

Two sensitivity analyses were performed. First, we repeated main analysis for NFHS 4 after excluding seven union territories, which were covered only in NFHS 4 but not in prior surveys. Second, we additionally adjusted for women's autonomy variables. Women's autonomy variables were asked only in the 15 percent sub‐sample of NFHS 4. Therefore, we presented the results from the model additionally adjusting for women's autonomy variables as a sensitivity analysis in supplementary material. MLwiN 3.02 software was used for all estimates.

3. RESULTS

3.1. Study population

The number of eligible women for this analysis was 28 978, 36 850, and 190 898 in NFHS 2 (1998), NFHS 3 (2005), and NFHS 4 (2015), respectively. After excluding women with missing information on the variables, the final analytic sample comprised of 25 261, 34 775, and 180 023 in NFHS 2, NFHS 3, and NFHS 4, respectively. The derivation process and hierarchical structure of the analytic sample were provided in Figure S1.

Across the survey years, 63.0‐70.5 percent of the women in our analytic sample were found to be in their 20s. The average education level has improved over the study period. While less than 34.8 percent of the women had attained secondary or higher education in 1998, 57.0 percent achieved this level of education in 2015. The share of women who have no education decreased from 48.9 percent in 1998 to 29.0 percent in 2015 (Table 1).

Table 1.

Descriptive statistics of women and the prevalence of ANC1+ and ANC4+ in NFHS Waves 2, 3, and 4

Variable	NFHS 2 (1998)			NFHS 3 (2005)			NFHS 4 (2015)
	N (%)	Prevalence (%)		N (%)	Prevalence (%)		N (%)	Prevalence (%)
		ANC1+	ANC4+		ANC1+	ANC4+		ANC1+	ANC4+
Age at pregnancy
10s	5873 (23.2)	64.5	29.4	6165 (17.7)	78.3	40.2	24 009 (13.3)	78.6	45.1
20s	15 910 (63.0)	67.1	35.6	22 872 (65.8)	79.3	46.9	127 008 (70.5)	78.9	46.2
30s	3273 (13.0)	55.6	27.7	5386 (15.4)	68.8	38.3	27 013 (15.0)	71.6	41.0
40s	205 (0.8)	42.4	12.7	370 (1.7)	43.2	17.3	1993 (1.1)	53.8	25.0
Terminated pregnancy
No	20 707 (82.0)	64.3	32.4	28 699 (82.5)	76.6	43.6	151 720 (84.3)	76.6	44.4
Yes	4554 (18.0)	67.3	35.3	6076 (17.5)	79.8	46.4	28 303 (15.7)	81.9	48.4
Maternal education
No education	12 349 (48.9)	45.1	13.5	13 328 (38.3)	58.7	18.8	52 131 (29.0)	60.0	24.1
Primary graduate or less	4121 (16.3)	71.7	30.9	4938 (14.2)	77.5	37.0	25 245 (14.0)	75.2	38.0
Secondary graduate or less	6398 (25.3)	86.6	53.9	13 405 (38.5)	90.4	61.8	83 399 (46.3)	85.3	54.6
Collage or above	2393 (9.5)	96.4	80.8	3104 (8.9)	98.6	87.9	19 248 (10.7)	93.3	69.9
Marital statusa
Married	24 916 (98.6)	64.9	33.0	34 125 (98.1)	77.2	44.2	176 912 (98.3)	77.5	45.1
Non‐married	345 (1.4)	58.8	25.8	650 (1.9)	72.5	38.2	3111 (1.7)	74.7	44.2
Religion
Hindu	18 839 (74.6)	63.1	31.3	25 183 (72.4)	79.1	45.5	134 400 (74.7)	78.1	45.1
Muslim	3579 (14.2)	66.7	35.2	4822 (13.9)	73.7	40.6	23 061 (12.8)	75.9	43.2
Christian	1690 (6.7)	75.3	42.2	3248 (9.3)	68.1	35.9	14 770 (8.2)	71.7	41.8
Others	1153 (4.6)	71.0	40.0	1522 (4.4)	75.4	49.8	7792 (4.3)	82.0	56.0
Type or residency
Urban	6717 (26.6)	87.2	59.5	13 754 (39.6)	89.3	64.2	45 214 (25.1)	88.0	61.2
Rural	18 544 (73.4)	56.7	23.3	21 021 (60.4)	69.2	31.0	134 809 (74.9)	73.9	39.7
Family size
≦3	12 996 (51.4)	61.4	30.4	14 212 (40.1)	73.8	30.6	68 828 (38.2)	75.8	43.4
4‐6	10 669 (42.2)	67.4	36.0	17 201 (49.5)	78.5	47.4	94 476 (52.5)	78.1	48.3
≧7	1606 (6.4)	75.3	47.0	3362 (9.7)	84.3	55.8	16 719 (9.3)	80.3	51.6
Birth order
First	7046 (27.9)	77.6	47.8	9794 (28.2)	87.9	60.0	57 988 (32.2)	85.4	55.8
Second	6 727 (26.6)	72.9	40.6	10 317 (29.7)	85.4	54.3	58 882 (32.7)	81.3	49.9
Third	4519 (17.9)	62.3	27.5	5 936 (17.1)	74.4	37.5	31 247 (17.4)	73.1	37.6
More than four	6969 (27.6)	45.7	14.0	8728 (25.1)	57.3	18.7	31 906 (17.7)	60.1	24.0
Caste
Scheduled caste	4639 (18.4)	59.9	24.9	6243 (18.0)	75.5	37.4	34 910 (19.4)	75.8	42.1
Scheduled tribe	3732 (14.8)	52.7	19.6	5654 (16.3)	63.5	27.3	37 296 (20.7)	70.9	40.0
Other backward class	7236 (28.6)	62.3	33.1	11 769 (33.8)	76.7	42.6	73 458 (40.8)	77.7	43.8
Others	9654 (38.2)	73.8	41.9	11 109 (31.9)	85.4	58.1	34 359 (19.1)	85.6	56.3
Wealth level
Poorest	4643 (18.4)	36.4	8.7	5897 (17.0)	52.0	11.8	44 745 (24.9)	57.3	21.6
Poorer	4799 (19.0)	48.0	14.0	6095 (17.5)	63.7	21.7	40779 (22.7)	73.9	37.3
Middle	5211 (20.6)	63.3	25.1	6979 (20.1)	76.1	37.6	35 916 (20.0)	83.7	50.6
Richer	5745 (22.7)	79.5	43.5	7661 (22.0)	87.8	55.0	31 182 (17.3)	89.3	60.4
Richest	4863 (19.3)	92.8	70.6	8143 (23.4)	96.3	79.7	27 401 (15.2)	93.8	70.2
Total	25 261	64..8	32.9	34 775	77.1	44.1	180 023	77.4	45.1

Abbreviations: ANC1+, at least one visit to a health care provider for ANC for the last pregnancy; ANC4+, at least four visit to a health care provider for ANC for the last pregnancy; NFHS, National Family Health Survey.

^a“Married” includes currently married/living with partner while “non‐married” includes Widowed/divorced/separated/never married.

3.2. Trends in prevalence of utilization ANC service

Findings from the descriptive statistics revealed a few similarities and differences between ANC1+ and ANC4+. The overall prevalence of both ANC1+ and ANC4+ constantly increased over time (by 12.6 and 12.2 percent, respectively), with a greater margin of increase between 1998 and 2005 than between 2005 and 2015 (Table 1). The overall prevalence for ANC1+ was much higher than ANC4+ throughout the entire period (64.8, 77.1, and 77.4 percent for ANC1+ vs. 32.9, 44.1, and 45.1 percent for ANC4+ in 1998, 2005, and 2015, respectively).

A large gap in the prevalence of ANC1+ and ANC4+ by SES and residence area persisted throughout the study period with the prevalence being higher in more educated, wealthier, and urban‐living women than their counterparts. While gaps in the prevalence of ANC1+ utilization between the two extreme strata of SES have continuously decreased over the study period, gap in the prevalence of ANC4+ utilization widened between 1998 and 2005 and narrowed down again afterward (Figure 1A–C for ANC1+, Figure 1E–F for ANC4+).

Figure 1.

Trends in the prevalence of ANC1+ and ANC4+ by educational level, wealth level, and residence type in 1998, 2005, and 2015 (X axis: year/Y axis: prevalence(%)/ANC1+: at least one visit to a health care provider for ANC for the last pregnancy/ANC4+: at least four visit to a health care provider for ANC for the last pregnancy) [Colour figure can be viewed at http://wileyonlinelibrary.com]

The trend in the utilization of ANC1+ and ANC4+ over the study period varied according to SES strata and residence area. For instance, while the prevalence of ANC4+ constantly escalated over time for women in the bottom two educational groups (from 13.5 and 30.9 percent in 1998 to 24.1 and 38.0 percent in 2015 for no education and education of primary graduate or less groups, respectively), it increased between 1998 and 2005 but declined afterward for the top two educational groups (from 53.9 and 80.8 percent in 1998 rising to 61.8 and 87.9 percent in 2005, but declining to 54.6 and 69.9 percent in 2015 for secondary graduate or less and college or above, respectively) (Table 1, Figure 1D). The same trend was observed according to the wealth level and residence area. For those who belonged to the richest wealth quintile or were living in urban area, the trajectory in utilization of ANC4+ was upward between 1998 and 2005 but turned downward from 2005, while those belonging to lower wealth quintiles or residing in rural area showed a constant increase in the utilization of ANC4+ (Figure 1E,F).

3.3. Inequalities in the utilization of ANC service by SES and residence area

The main analyses using multilevel logistic regression demonstrated persistently strong disparities in the use of ANC1+ and ANC4+ across the education, wealth level, and residence area during the whole study period even after controlling for other covariates (Table 2).

Table 2.

Adjusted ORs of ANC1+ and ANC4+ by SES and residence type from multilevel logistic regression from the NFHS Waves 2, 3, and 4

	NFHS 2 (1998)		NFHS 3 (2005)		NFHS 4 (2015)
	OR	95% CI	OR	95% CI	OR	95% CI
ANC1+
Maternal education (ref: ≧ college)
No education	0.21***	0.17‐0.27	0.17***	0.12‐0.22	0.46***	0.42‐0.49
Primary graduate or less	0.32***	0.25‐0.40	0.24***	0.18‐0.32	0.60***	0.55‐0.64
Secondary graduate or less	0.46***	0.36‐0.58	0.36***	0.27‐0.49	0.73***	0.68‐0.78
Type or residency(ref = urban)
Rural	0.57***	0.51‐0.64	0.67***	0.61‐0.74	0.84***	0.79‐0.88
Wealth level(ref = richest)
Poorest	0.24***	0.20‐0.28	0.19***	0.16‐0.22	0.29***	0.27‐0.32
Poorer	0.29***	0.25‐0.35	0.25***	0.21‐0.29	0.40***	0.37‐0.43
Middle	0.39***	0.33‐0.45	0.32***	0.28‐0.37	0.54***	0.51‐0.58
Richer	0.56***	0.48‐0.65	0.50***	0.43‐0.58	0.71***	0.66‐0.76
ANC4+
Maternal education (ref: ≧ college)
No education	0.26***	0.23‐0.30	0.28***	0.25‐0.32	0.58***	0.55‐0.61
Primary graduate or less	0.37***	0.32‐0.42	0.37***	0.32‐0.43	0.67***	0.64‐0.70
Secondary graduate or less	0.53***	0.47‐0.60	0.53***	0.47‐0.60	0.79***	0.76‐0.82
Type or residency(ref = urban)
Rural	0.59***	0.54‐0.64	0.63***	0.59‐0.68	0.82***	0.79‐0.85
Wealth level(ref = richest)
Poorest	0.22***	0.19‐0.25	0.19***	0.17‐0.22	0.38***	0.36‐0.40
Poorer	0.28***	0.25‐0.32	0.27***	0.25‐0.30	0.49***	0.47‐0.52
Middle	0.38***	0.34‐0.42	0.37***	0.34‐0.41	0.61***	0.59‐0.64
Richer	0.56***	0.51‐0.62	0.53***	0.49‐0.58	0.74***	0.71‐0.78

Abbreviations: ANC1+, at least one visit to a health care provider for ANC for the last pregnancy; ANC4+: at least four visit to a health care provider for ANC for the last pregnancy; NFHS, National Family Health Survey.

^* P < .05; **P < .01; ***P < .001.

The educational gradient of ANC1+ and ANC4+ was statistically significant across all three time periods (P‐values < .001), but the trend in inequality slightly differed for the two outcomes. Specifically, educational inequality in ANC1+ widened between 1998 and 2005 (ORs for the least educated were 0.21 in 1998 and 0.17 in 2005, P values < .001) while inequality of ANC4+ narrowed during the same period (ORs for the least educated were 0.26 in 1998 and 0.28 in 2005, P values < .001). Educational inequality narrowed down between 2005 and 2015 for both ANC1+ and ANC4+ (Figure 2A for ANC1+ and Figure 2D for ANC4+). Wealth gradient worsened between 1998 and 2005 for both ANC1+ and ANCS 4, followed by improvement between 2005 and 2015 (Figure 2B for ANC1+ and Figure 2E for ANC4+). For instance, the ORs of receiving ANC1+ for the poorest women versus the wealthiest women were 0.24 (P < .001) in 1998, 0.19 (P < .001) in 2005, and 0.29 in 2015 (Table 2) (Figure 2D). The ORs of receiving ANC4+ for the poorest women compared to the wealthiest women were 0.22 (P < .001) in 1998, 0.19 (P < .001) in 2005, and 0.38 (P < .001) in 2015 (Table 2) (Figure 2E). Disparities in both ANC1+ and ANC4+ between urban and rural areas declined between 1998 and 2015 even after controlling for other covariates (Figure 2C for ANC1 and Figure 2F for ANC4).

Figure 2.

Trends in inequality in ANC1+ and ANC4+ by educational level, wealth level, and residency type in 1998, 2005, and 2015 (X axis: year/Y axis: odds ratio/ANC1+: at least one visit to a health care provider for ANC for the last pregnancy/ANC4+: at least four visit to a health care provider for ANC for the last pregnancy)

When stratified by the residence area, we observed a similar pattern as the main results for SES inequalities in ANC1+ and ANC4+, with a widening gap between the highest and the lowest group in terms of education and wealth level during 1998‐2005, followed by a sharp reduction during 2005‐2015 (Table S1).

Results on the association of other covariates with the utilization of ANC were presented in Table S2. Birth order and three variables representing women's autonomy were significantly associated with the utilization of ANC across the three waves.

3.4. Geographical variations in the utilization of ANC1+ and ANC4+

Table 3 provides the variance estimates (in Logit scale) with standard errors at community and state level for all India as well as the proportion of variation in ANC1+ and ANC4+ attributable to each level. After adjusting for all covariates, the proportion of variation attributable to states constantly declined from 15.0 percent in 1998 to 12.3 percent in 2015 for ANC1+ and from 19.7 percent in 1998 to 14.4 percent in 2015 for ANC4+. By contrast, the proportion of variation attributable to communities increased from 6.6 percent in 1998 to 22.6 percent in 2015 for ANC1+ and from 1.9 percent in 1998 to 9.9 percent in 2015 for ANC4+. Results from all 36 states and 29 states were comparable (Table 3).

Table 3.

Variance estimates in logit scale (SE) and proportion of total variation in ANC1+ and ANC4+ attributable to community and state level in null and fully adjusted models from the NFHS 2, NFHS 3, and NFHS 4

	NFHS 2 (1998)				NFHS 3 (2005)				NFHS 2 (2015)
	Variance estimate(SE)		% variation		Variance estimate(SE)		% variation		Variance estimate(SE)		% variation
	Null	Full	Null	Full	Null	Full	Null	Full	Null	Full	Null	Full
ANC1+
State	0.867 (0.243)	0.631 (0.179)	17.9	15.0	0.7 (0.188)	0.547 (0.147)	14.0	12.6	0.993 (0.236)	0.623 (0.151)	16.2	12.3
Community (Village/CEB)	0.693 (0.034)	0.277 (0.024)	14.3	6.6	0.997 (0.04)	0.501 (0.03)	20.0	11.5	1.862 (0.027)	1.142 (0.021)	30.3	22.6
state(N)	26				29				36
community(N)	3182				3805				27 808
individual(N)	25 261				34 775				180 023
ANC4+
State	0.954 (0.268)	0.826 (0.232)	20.6	19.7	0.757 (0.201)	0.612 (0.163)	16.2	15.0	0.824 (0.195)	0.625 (0.149)	17.4	14.4
Community (Village/CEB)	0.391 (0.023)	0.08 (0.019)	8.4	1.9	0.63 (0.026)	0.185 (0.017)	13.4	4.5	0.601 (0.011)	0.432 (0.01)	12.7	9.9
state(N)	26				29				36
community(N)	3182				3805				27 808
individual(N)	25 261				34 775				180 023

Abbreviations: ANC1+, at least one visit to a health care provider for ANC for the last pregnancy; ANC4+, at least four visit to a health care provider for ANC for the last pregnancy; NFHS, National Family Health Survey.

3.5. Sensitivity analyses

Descriptive statistics of the sub‐sample of NFHS 4 after excluding seven territories were presented in Table S3. There was no difference in the characteristics of the study population and the prevalence of ANC1+ and ANC4+ utilization when seven union territories were excluded from the NFHS 4.

The results on the SES inequality and geographic variation were almost same when 7 union territories were excluded from the analysis for NFHS 4 (Tables S4 and S5).

Including the three variables constructed for women's autonomy—each representing women's decision making, freedom of movement, and financial autonomy—reduced the number of study samples to 25 184, 34 040 and 30 882 in NFHS 2, NFHS 3, and NFHS 4, respectively.

Higher scores on the autonomy for decision making and freedom of movement were associated with higher utilization of ANC4+ (ORs for the autonomy of decision making are 1.04, 1.08, and 1.08 and ORs for the freedom of movement are 1.07, 1.06, and 1.02 in NFHS 2, NFHS 3, and NFHS 4, respectively. All P‐values < .01). Women with financial autonomy were more likely to complete more than 4 times of antenatal visits than their counterparts (OR = 1.13, 1.15 and 1.10 in NFHS 2, NFHS 3, and NFHS 4, respectively. All P‐values < .01) (Table S2).

Attenuation in ORs of the main independent variables was trivial when women's autonomy variables were additionally adjusted for. ORs for the use of antenatal care services by education, residence area, and wealth level have changed only within a range of ±0.2, and statistical significance remained unchanged (Table S6).

4. DISCUSSION

We present three major findings from our examination of the trend in prevalence and socioeconomic and geographic disparities in the utilization of ANC services in India from 1998 to 2015. First and foremost, the improvement of inequality in ANC1+ and ANC4+ between the highest and the lowest SES groups occurred between 2005 and 2015. During the early stage of the study period (between 1998 and 2005), only educational inequality for ANC4+ slightly decreased while wealth inequality for ANC1+ and ANC4+ and educational inequality for ANC1+ worsened.

Our findings on the worsening inequality from 1998 to 2005 are consistent with the result from Pathak et al7, where the gap in probabilities of uptake of ANC between the poor and the non‐poor was bigger in 2005 than in 1998. When interpreted in the context of the rise in the overall prevalence of ANC, decrease in poverty, and improvement in the average level of education in India during 1998‐2005,35 our finding suggests that the growth in ANC coverage occurred unequally and was not inclusive enough to embrace socioeconomically disadvantaged mothers.

India launched CSSM in 1992 to provide essential obstetric care for all, enable early detection of complication, and deliver emergency obstetric care in order to prevent maternal deaths. Accordingly, efforts to encourage antenatal visits were integrated into the program.36 However, assessment of the CSSM program conducted in 1997 presented somewhat disappointing results indicating that emergency obstetric care was not functioning well and that people still did not consider pregnancy as an event requiring any special medical attention.36, 37 Afterward, CSSM and other programs for family planning were merged to create the RCH in 1997 (phase I ended in 2004) to improve the performance of its Family Welfare Program (FWP) in reducing infant and child and maternal mortality. Report on the program's progress in 2005 appraised that even if there had been improvements in a few indicators, for example, rise in percentage of women receiving full ANC (more than three visits) from 39.6 to 47.5 percent, the objectives were only partially achieved because the aim of increasing access in disadvantaged area was not met.38 Our result of a substantial rise in the crude rate of ANC but widening educational and wealth inequality between 1998 and 2005 might be explained in line with this.

Another possible driving force behind the widening educational inequality shown in ANC1+ during the early stage is an improper implementation of focused antenatal care (FANC).25, 39 FANC is a new model promoted by WHO, whereby the earlier recommendation on monthly ANC visits was replaced by a minimum of four visits at a specific time point in the course of pregnancy if women were deemed uncomplicated and of low risk. The visit schedule was supposed to be made by trained health providers based on the expected date of delivery, presumed risk, and health status of each individual.25, 39 However, evaluation studies on the effect of new model reported inadequate training of health providers,27 which might have led to the delivery of a false message to pregnant women that pregnancy is a healthy physical state that does not require professional care when there is no perceived threat to the pregnancy.40 Therefore, a risk‐focused new model of ANC might have misguided less educated women to not visit at all.

It was in RCH phase II that addressing the poor was placed as a core principle of the program, and therefore, it was aimed that resources are used where health status was the worst and need the greatest.41 RCH phase II was rolled out in 2005 as a component of NRHM which was planned to strengthen the public health service delivery infrastructure, particularly at the village, primary, and secondary level, and decentralize a village‐ and district‐level health planning and management. Janani Suraksha Yojana (JSY), a cash‐transfer program, also launched in 2005, for women from lower SES to give birth in health facilities, was one of the important initiatives of NRHM along with an “Accredited Social Health Activists (ASHA)” to ensure wide outreach.42 However, JSY was not successful from the outset. In the early stage of the program, it only focused on uptake of institutional delivery, neglecting other components such as ANC, which led to stagnancy in uptake of ANC.19 This concern finally led to changes to JSY guideline in several states around 2010 to ensure that provision of ANC was one of the eligibility conditions for payment of incentive to ASHAs.43 Vellakkal et al's19 evaluation on the effect of NRHM depicted this situation. Inequality in the ANC declined to a much greater extent in the later period of the NRHM (2011‐2012) than in the early period of the NRHM (2007‐2008).19 This aligns with our findings on the reduction in inequality between 2005 and 2015 although our data do not allow us to look into the progress separately between early and late NRHM periods.

On the other hand, improved inequality in utilization of ANC services between 2005 and 2015 should be interpreted with caution because the result might be driven by not only an increase in the utilization of ANC1+/ANC4+ among the poorer and less educated groups but also a decrease in utilization among the richest and most educated groups. A decline in the utilization of ANC among women from high SES group deserves serious attention. They might have been neglected while the more disadvantaged group has been a target from global society.44

Second, the different patterns between ANC1+ and ANC4+ also merit attention. The gap in the overall rate between ANC1+ and ANC4+ was substantial. The coverage expansion in the lowest level of education and wealth was greater between 1998 and 2005 for ANC1+ than the period between 2005 and 2015, whereas greater expansion happened between 2005 and 2015 for ANC4+. Initiating the first ANC visit might be accomplished with relative ease by intervening on the demand‐side factors such as changing their perception or emphasizing the importance of ANC through communication and education, all of which can be carried out with relatively low resources and within a short time. However, ensuring women to complete full four visits requires intervening on a more complicated array of factors that go beyond user's willingness and knowledge. For example, a variety of direct or indirect financial barriers may hamper women from keeping visits. Specifically, a fee for a check‐up, which is largely free in public, is not free in a private clinic. Transport expenses and loss of income from absence in the workplace is also burdensome.45, 46 Low quality of service would also be a discouraging factor for completing four visits. Since people tend to comply with the utilization of medical service when they perceive the benefits of service to outweigh the cost,47 pregnant women are likely to be reluctant to continue antenatal visits if they feel that the quality of ANC service is not good enough. Removing financial barriers and improving service quality require more sophisticated and comprehensive interventions that cannot be obtained within a short time, which might explain our findings on the delayed increase in coverage of ANC4+ for the lowest SES group.

Third, the proportion of variation attributable to smaller geographic units—villages for rural area and CEB for the urban area—has grown while the relative contribution of state level decreased. Previous studies have also demonstrated the greater importance of the micro‐level environment for poverty and catastrophic health spending in India.48, 49 The increase in the contribution of villages to the geographic variation in the ANC utilization may be partially explained by the growth of ASHAs which is female volunteering community health worker selected and deployed by the unit of the community and act as health educators and promoters in their communities. Heterogeneity in ASHA's capacity, attitude, and level of efforts across villages might have shaped the community variation.

This study has a few limitations. First, while district might be a meaningful geographic entity that may affect the distribution of ANC service utilization, we were not able to include district as a level for our multilevel analysis because only NFHS 4 provided district identifier. Second, we did not intend to test hypotheses or propose a new conceptualization of factors related to the utilization of ANC services. Instead, we focused on investigating the trend of SES inequality and geographic variation in the utilization of ANC services over time. Therefore, we do not provide a profound interpretation of the association between ANC service utilization and other covariates (such as women's autonomy). Third, women's employment status could not be included in our main analyses due to a large number of missing values in NFHS 4 despite the possibility that it may be associated with women's health service utilization behaviors by affecting the availability of time and money. Therefore, adjustment for women's autonomy was restricted to sensitivity analysis. Fourth, the ANC data we used are subject to recall bias depending on the women's ability to accurately recall their ANC utilization during their pregnancy. However, we tried to minimize this by restricting the sample to the mother's most recent delivery within the last three (NFHS 2) and five (NFHS 3 and NFHS 4) years.

5. CONCLUSION

This is the first study to examine the long‐term trends in inequality in the utilization of ANC services by SES and residence type as well as its geographic variation. Although the prevalence and inequality in the utilization of ANC services have improved during the last two decades, India still lags far behind comparable countries and there remains deeply ingrained inequality. The fluctuation in inequality over time indicates potential reversibility in the recent improvements seen between 2005 and 2015 if programs and policies fail to monitor target groups with greater needs.

Ensuring equal opportunities to ANC for all women is important not only for maternal health, but also for the long‐term health benefits for the newborns because proper ANC helps babies not only be born healthy but also stay healthy after birth by detecting disease risk or encouraging child vaccination, etc, and thus, sets them in a good trajectory of social mobility to attain higher SES in their later life.

The findings from this study indicate the need to continue investments in reducing inequality as well as improving overall coverage of ANC utilization. Efforts to remove barriers for the lower SES population and understanding of community‐level factors need to be expanded and strengthened. At the same time, efforts should be inclusive to ensure that the prevalence in ANC utilization does not drop for women of higher SES.

CONFLICT OF INTEREST

The authors declare that they have no conflict of interest.

Supporting information

 

 

Lee H‐Y, Oh J, Kim R, Subramanian SV. Long‐term trend in socioeconomic inequalities and geographic variation in the utilization of antenatal care service in India between 1998 and 2015. Health Serv Res. 2020;55:419–431. 10.1111/1475-6773.13277

DATA AVAILABILITY STATEMENT

This study used existing datasets from the Demographic and Health Survey (DHS) Program. Data are accessible free of charge on registration with the DHS Program from the following DHS website: https://dhsprogram.com/data/available-datasets.cfm.