An Empirical Examination of the Inequality of Forgone Care in India

ABSTRACT Understanding how well a health system is meeting the needs of the population is critical to achieving the policy aspirations of universal health coverage. This study focuses on assessing the inequity of forgone care for priority maternal and child health services across India. We utilize data from the 4th round of the Indian National Family Health Survey (NFHS-4) to examine inequality of forgone care. Our outcomes include forgone institutional delivery, antenatal care, medical care for a child with fever or cough, and medical care for a child with diarrhea. Wagstaff’s standardized concentration indices (CIs) are computed at the national level, over urban and rural sub-populations, and by state. Regression decomposition is performed to determine the influence of specific drivers on overall inequality. There was significant variation in the national-level prevalence and CIs for forgone antenatal care (17.8%, CI: −0.423), forgone medical care for a child with fever or cough (32.4%, CI: −0.199), forgone medical care for a child with diarrhea (33.8%, CI: −0.172), and forgone institutional delivery (24.5%, CI: −0.436). For all outcomes, forgone care is disproportionately concentrated among the poor, particularly in rural areas. There is also significant heterogeneity in state-level inequalities. Decomposition analyses show that socioeconomic status, maternal education, rural status, and state-level per capita health spending are the leading drivers of observed inequalities in forgone care. Results suggest attending to both the operation and financing of India’s health care system as well as the social determinants that make poor women more likely to forgo maternal health care.


Introduction
In India, as in many countries around the world, government is focusing attention on achieving universal health coverage (UHC). India's recent national health policy seeks to promote people's health and wellbeing through preventive and promotive health care and "universal access to good quality health care services without anyone having to face financial hardship." 1 The policy builds on its experience with the National Health Mission to address health needs in under-served areas of India.
A critical aspect to achieving UHC involves trying to identify and address barriers to health care according to people's needs. Whereas most studies focus on questions of health insurance coverage or health care access, utilization, or quality, less attention has been paid to understanding questions of unmet health need. Forgone care is an important component of unmet health needs (in addition to delayed and low-quality care), and involves an individual not using health care despite the need for care. However, health systems have difficulty in evaluating whether they are meeting health needs, either because they don't examine the question directly, or because of the wide variation in how "health needs" are defined and measured. 2 A common concern is identifying who is most affected by forgone care-typically marginalized and impoverished groups-and what can be done to overcome barriers to specific health services in particular contexts.
This study examines data in India that identifies forgone care in specific maternal and child health (MCH) services that have been the focus of attention for the National Health Mission. In particular, we assess the role of inequities in wealth and how maternal education, sex of the child, geographic location, state-level health spending, maternal age, caste, and health insurance status, affect forgone care for antenatal care, institutional delivery, medical care for childhood fever and cough, and medical care for diarrhea across Indian states. Such information can provide insights on who is forgoing MCH services, and to provide a nationally-representative empirical baseline for future research to assess the impact of programs under Ayushman Bharat, including a scheme to provide primary health care through Health and Wellness Centers, and a National Health Protection Mission to provide free health hospitalization to the poorest 500 million people, on forgone care.

Data
We utilize data from the most recent 4th round of the National Family Health Survey (NFHS-4), a population-level survey containing a representative sample of over 200,000 households across India from 2015 to 2016. 3 Within the NFHS-4, we utilize questions pertaining to care-seeking from the adult women's and child surveys to derive 4 different measures of forgone care: forgone institutional delivery, forgone antenatal care, forgone medical care for a child with fever or cough, and forgone medical care for a child with diarrhea. Each measure is computed as a binary indicator taking the value of 1 if care was forgone when needed and 0, otherwise. For antenatal care, forgone care was defined as no antenatal care visits prior to delivery. These specific care-seeking indicators were chosen because of the ability to link forgone care conditional on the reported display of symptoms or eligibility (e.g., pregnancy) to unmet need. Other variables in the dataset did not permit the association of care seeking with display of symptoms, signifying needed care (e.g., care for cancer). We also compute a second set of forgone care indicators for the child outcomes of diarrhea and fever, which capture whether any care was forgone, or whether either medical care (allopathic or "western" medicine) or non-allopathic care (e.g. Indian systems of medicine) was sought for the corresponding condition.
In addition to the indicators of forgone care, we also create a binary indicator for lacking any type of health insurance coverage, and examine the relationship between self-reported insurance coverage and each of the primary outcomes. For institutional delivery, the reasons for why care was forgone were also reported. We assess these explanations through a series of binary indicators for each stated reason. Socio-economic status is captured through the NFHS-4 wealth index weights and quintiles of the corresponding wealth index are utilized for the analysis of inequity. Finally, data on potential social, political, and geographic determinants of forgone care are also taken from the NFHS-4. These determinants include maternal education level, the age of the mother, sex of the child, caste designation, rural designation, state of residence, per capita health spending in the state of residence, and marital status of respondent.

Statistical Analyses
To examine the inequality in forgone care by socioeconomic status, we compute Wagstaff normalized concentration indices (CI) for each of the binary indicators of forgone care. 4,5 Since all of the care-seeking variables are binary, the Wagstaff normalization procedure ensures that resulting CIs are bounded between −1 and 1. We produce CIs at the national-level for the socio-economic distribution of forgone care, utilizing nationally representative socio-economic quintiles for the ranking variable as well as separately for rural and urban utilizing urban-specific and rural-specific socioeconomic quintiles from the NFHS-4. We also compute CIs for each type of forgone care at the state-level, utilizing state-specific socio-economic quintiles as the ranking variable. We report results separately for all states and Union Territories that have populations greater than 7 million people, representing 98% of the national population.
In addition to presenting the concentration indices, we also conduct a Wagstaff-style regression-based decomposition analysis to examine the drivers of the observed inequalities. 6 The factors considered for drivers include state, rural designation, socioeconomic status, per capita state-level health expenditure, age of the mother, maternal education level, marital status, insurance status, caste, and, for child outcomes, the sex of the child. All analyses were conducted using Stata version 15.1. 7

Results
The total NFHS-4 sample utilized contains data on 259,627 individuals. Table 1 presents summary prevalence statistics and totals for each of our forgone care variables at the national-level, urban and rural levels, and within each state. The totals presented correspond to the number reporting forgone care within each geographic sub-unit.
In general, there are high rates of forgone medical care for childhood fever (32.4%) and diarrhea (33.8%). These prevalence rates of forgone care are lower when considering the combined use of either medical or non-allopathic care for fever (24.5%) and diarrhea (24.2%). For all indicators, with the exception of lacking insurance coverage, the prevalence of forgone care is higher within rural areas than within urban areas. Examining the prevalence of forgone care by state, we see large heterogeneity in both the prevalence and the absolute number reporting forgone care. Across the indicators, states including Assam, Gujarat, and Madhya Pradesh routinely exhibit some of the highest prevalence in forgone child care, whereas Bihar, Jharkhand, and Uttar Pradesh exhibit some of the highest prevalence in forgone maternal care. Prevalence estimates are presented as % with standard errors presented in parenthesis. Total refers to the total number of observations in the NFHS-4 survey exhibiting forgone care in each outcome by geographic unit. *** indicates that the prevalence within urban areas is statistically significantly different from the prevalence in rural areas with a p-value <0.001. Table 2 presents the national-level prevalence of any reported health insurance coverage and also for each potential rationale for why care was forgone, amongst those who exhibited each type of forgone care.
Most of the NFHS-4 sample reports a lack of health insurance coverage (86.1%). However, those exhibiting forgone antenatal care, forgone medical care for fever or cough, forgone medical care for diarrhea, and forgone institutional delivery, exhibit lower rates of reported health insurance coverage. For example, 88.3% of individuals with forgone care for diarrhea did not have health insurance. For forgone institutional delivery, it is possible to examine the specific reasons why people did not seek care. The leading self-reported reasons for forgone institutional delivery were that institutional delivery was not viewed as necessary (27.1%), the facility was too far away (15.1%), and the facility was too expensive (10.9%).
The national-level Wagstaff normalized concentration indices for each metric of forgone care are presented in Table 3. Results are presented for the total national sample, within rural and urban subsamples only, and amongst those reporting any insurance coverage vs. lacking any insurance coverage.
All CIs for every indicator of forgone care are negative indicating that the poor have disproportionately higher levels of forgone care. Lorenz curves for the national-level indicators are presented in Figure 1.
Although the burden of unmet medical need consistently affects the poor disproportionately, there is variation by type of health service. Forgone antenatal care and institutional delivery exhibit the highest degree of inequality across socio-economic status with concentration indices at the national level of −0.423 and −0.436, respectively. For nearly all types of forgone care, with the exception of forgone institutional delivery, the disproportionate burden on the poor of forgone care is less unequal amongst those who have health insurance, compared with those who do not. However, the poor-concentrated distribution of forgone care persists even among those covered with health insurance. Inequality across socio-  economic status for lacking health insurance (CI: −0.079), however, is less unequal than all of the indicators of forgone care. Finally, the degree of inequality in forgone care is greater in rural areas than in urban areas for most indicators, with the exception of forgone institutional deliveries, which were more unequally distributed in urban areas.
We also examine inequalities in the self-reported rationales for forgone care in Table 4.
As with forgone care itself, most of the highlighted self-reported rationales for forgone institutional delivery (98.7% response rate) were disproportionately more prevalent among the poorest than the richest. The explanations with the most unequal distribution for forgone institutional delivery were that care was too expensive (CI: −0.451), the facility was too far (CI: −0.433), and that the husband did not allow it (CI: −0.376).
As with other studies published on the inequalities of healthcare access and healthcare financing, the nationallevel estimates of inequality in forgone care mask a considerable amount of state-level variation. [8][9][10][11][12] A complete table of CIs by state, utilizing state-specific wealth quintiles as the ranking variable, along with standard errors appear in Table A1. Figures 2 and 3 present states ranked by the degree of inequality for   Concentration indices are presented with standard errors presented in parenthesis. * = p < 0.05, ** = p < 0.01, *** = p < 0.001.  Table 5. Socioeconomic status, state-level per capita health spending, age of the mother, and insurance status are significantly associated with forgone care across all outcomes. Sex of the child is significantly associated with forgone care for the child outcomes with female children more likely to experience forgone care than male children. Caste and maternal education level are only strongly associated with forgone care for maternal outcomes including institutional delivery and forgone antenatal care with higher levels of education strongly associated with a lower probability of forgone care. Rural designation is significantly associated with an increased probability of forgone care for all outcomes with the exception of care for diarrhea.
Through the regression decomposition, nationallevel concentration indices are broken down into the percent contribution of each significant driver of overall inequity. These percent contributions are presented in Table 6.
Overall, socioeconomic status is the driving determinant of inequity, contributing to over 60% of the concentration index for each outcome. For institutional delivery and antenatal care, maternal education level is the second most influential determinant, accounting for approximately 25% of observed inequality. For forgone  fever care rural status and per capita state health spending are the second and third most influential determinants, accounting for 10% and 5% of observed inequalities, respectively. For diarrhea, maternal education and per capita state health spending are also leading contributors to observed inequalities, accounting for 6% and 5% of observed inequalities, respectively.

Discussion
The importance of health in human capital and economic development makes investment in health critical. Countries aspiring to UHC aim for all members of their population to be able to obtain the health services they need without experiencing financial hardship. At its core UHC means nondiscrimination; policies that exclude certain individuals or groups are inconsistent with the goals of UHC. 13 In practice, as our study reports, there are high levels of forgone medical and non-allopathic care in India, and forgone care is concentrated among poorer socioeconomic groups and in rural areas. There are several reasons why poorer socioeconomic groups and areas in India experience higher levels of forgone care compared to others. This is partly due to deficiencies in the supply and implementation of health system programs and services, patterns of social and economic inequality, and the inability of politically weak and vulnerable groups to ensure that their rights and entitlements are honored by governments and other stakeholders. 14 In particular, our data suggest that nearly one in five women had forgone antenatal care and one in four women had forgone institutional delivery. We have some idea of why forgone maternity care is concentrated among poorer groups, and these include health system, economic, as well as socio-cultural issues. For example, in the case of forgone institutional deliveries, some of the reasons for forgoing care that were concentrated among the poor included health services availability and acceptability issues ("health facility too far", "health facility not open" "no female provider" "distrust facility"), affordability ("too expensive"), and other demandside barriers ("husband/family did not allow", "[institutional delivery] not necessary", "[institutional delivery] not customary"). Rural designation and lower levels of maternal education were also highly associated with forgone institutional delivery. To reduce the excess forgone care in institutional deliveries and medical care more broadly among poorer socioeconomic groups requires particular attention to the health systems such as locating service delivery points closer to communities, ensuring a greater supply and presence of women providers in the workforce, making health insurance more accessible, and ensuring that patients have access to good quality publicly funded health services. Note that the national Health and Wellness Center program, by bringing health services closer to rural populations, can potentially reduce forgone care.
More challenging are the underlying socio-cultural issues because they are often beyond the traditional remit of health departments. Nevertheless, their presence further highlights the need for strengthening the focus on social determinants of health, especially maternal education. It is also notable that while there is less of a burden of forgone care among the poor when considering both medical and non-allopathic care-seeking, nearly one in five individuals forgoes even nonallopathic care for childhood fever and diarrhea.
As expected, there is considerable variation across states in India in forgone medical care and the concentration of forgone medical care among the poor (Figure 4). For example, Assam (52%), Jharkand (44%), and Bihar (44%) have among the highest levels of forgone medical care for fever/cough and diarrhea. In contrast, states like Kerala (24%), Tamil Nadu (27%), and West Bengal (25%) have much lower levels of forgone care. In general, these patterns indicate that better resourced states have lower average levels of forgone care. In fact, per capita state health spending appears to be a leading driver of inequity in forgone care across all observed types of forgone care examined in this study.  State performance on the socioeconomic distribution of forgone medical care exhibits somewhat different patterns ( Figure 5). As an example, in the case of forgone medical care for diarrhea, some states that had low average levels like Kerala and Tamil Nadu are among states with high levels of inequalities in forgone diarrhea medical care. Conversely, states like Bihar and Rajasthan are more equitable, though they had high average levels of forgone medical care for diarrhea. Further, several other states, particularly those in north-eastern India, continue to have both high average and inequitable levels of forgone medical care. Overall these patterns suggest that inequities in forgone care remain an important concern even in states that might have low overall forgone care levels. As our decomposition analysis hints at, the persistence of inequality in low-prevalence states is likely tied to the socioeconomic and demographic profile of those who are missed by available health services. As such, paying attention to both average levels and the distribution of forgone care is important from a policy perspective.
Patterns in the level and distribution of forgone deliveries across states largely follow those of forgone medical care. For example, in the case of institutional deliveries, states like Bihar (34%), Jharkand (38%), and Uttar Pradesh (32%) have high forgone institutional deliveries, while low forgone care is apparent in states like Kerala (0.1%), Punjab (8%), and Andhra Pradesh (9%). In general, these patterns indicate that better-resourced states have lower levels of forgone care. Notably, several of the northeastern states feature prominently among states with high levels of forgone medical and traditional medicine care. It is also observed that maternal care has higher degree of inequality than curative care for children. This could be related to both the supply-side financing and availability of curative services as opposed to preventative services as demand-side factors including beliefs regarding the importance of curative relative to preventative care.
There are several limitations to our study. The first is that the cross-sectional data utilized provide only a snapshot of inequalities existing at one point in time. As further rounds of the NFHS are conducted utilizing similar sampling techniques and question structure, we recommend conducting more in-depth analyses of trends in inequity both nationally and within states. In particular, it will be important to use future surveys to assess the influence of the new Ayushman Bharat health insurance and Health and Wellness Centers programs, to see whether they can reduce forgone care, particularly among the poor for whom they are targeted. The results presented in this study may serve as a baseline for such future analyses. We were also limited in our ability to examine forgone care by the questions asked in the NFHS-4, which enabled a linkage between care-seeking behavior and reported need for only a few select conditions. Currently, most of the questions in the NFHS-4 for chronic conditions including heart disease, diabetes, and cancer do not permit the assessment of whether the condition is managed or needs active care. Therefore, we were unable to separate unneeded care from forgone care for these conditions. Additionally, the focus of the NFHS-4 on healthcare utilization and care-seeking behavior only permits the examination of demand-side determinants of inequality. Future research should examine the potential for supply-side drivers of inequity including skilled labor supply, stock-outs, physician availability, wait times, and perceived quality of formal care as potential barriers to accessing care. Finally, even in instances where questions were asked regarding the rationales for forgoing care, they were often insufficiently detailed and not consistent across types of care. For example, while a myriad of potential rationales for forgoing institutional delivery were collected, for forgone medical care for children, specific rationales were not available in the data. Instead, reporting of any insurance coverage was used as an imperfect proxy for a potential rationale. Finally, much of the data is self-reported from a 2-week recall so there may be recall bias present in results, which may have linkages to education-level and socio-economic status.
One of the biggest challenges on India's path to achieving UHC is to find ways to reach vulnerable populations-those that are at risk of poor health and health care disparities-with limited health resources. Vulnerable groups have adverse health outcomes compared to others because they live in hard-to-reach places, are excluded from services because of gender, age, ethnicity or other characteristics and may not participate in health programs because they lack of awareness of their entitlements or because of their own beliefs and education or due to financial constraints or the legality of their status. How well vulnerable populations do on health and accessing health services is an important bellwether for the success of UHC policies.

Conclusions
India has high rates of forgone health care. Forgone care is concentrated among poorer socioeconomic groups, in rural areas, and is greater for women's health services than for curative care for children. There is substantial inter-state variation in both the level and distribution of forgone care. Importantly, better average state performance on forgone care does not necessarily correlate with low inequities. Reducing both the level and inequitable distribution of forgone care requires attention to health systems (e.g., locating service delivery points closer to communities, better quality of care), health care financing to make health services affordable, and on social determinants of health, and particularly to the intersection of poverty and gender that makes poor women more likely to forgo maternal health care.

Disclosure of Potential Conflicts of Interest
No potential conflict of interest was reported by the authors.