Equity in distribution of public subsidy for noncommunicable diseases among the elderly in India: an application of benefit incidence analysis

National Sample Survey (NSS) 71st round unit level data on Social Consumption: Health (2014) has been used for the study. This is the most recent data available in India on morbidity, healthcare utilization and related out-of-pocket expenditure (OOPE). NSS adopted stratified multistage design to collect data. The census villages (in the rural sector) and urban blocks (in the urban sector) were considered as the first stage unit (FSU) and the ultimate stage units (USU) were the households. Considering Census 2011 population, the sample villages were selected by probability proportion to size with replacement (PPSWR) to form the FSUs. On the other hand, in the urban sector, number of households of the urban frame survey (UFS) blocks has been used to form the FSUs following the PPSWR method¹ [25]. The data has been collected from 3,33,104 individuals living in 65,932 households. It is a nationally representative survey covering all states and Union Territories (UTs). Overall, 8.18% of the total sample are in the age group of 60 year and above. The share of elderly is 7.92% in the rural sector and the corresponding figure for the urban sector is 8.52%. Whereas, 17% of the total hospitalized and 28% of the total outpatient visits are recorded by the aged members. NSS collects information on household level as well as individual level characteristics. It records the details of morbidity, hospitalization and corresponding OOPE for doctor’s consultation, diagnostic tests, medicine, transport and other related costs. It has to be noted here that, NSS records the self-reported morbidity and hospitalization information in the survey. The reference period for inpatient care was 365 days and for outpatient care it was 15 days.² Sources of finance for treatment and insurance coverage information is also available in the data. Sample weights are reported for each household and individual in the data. This survey weights have been applied to scale up the estimates at the population level.

Following World Development Report (1993) [26] we have cross classified the ailments into three broad categories – Communicable (CD), Noncommunicable (NCD) and other diseases (OD). The classified data has been analyzed to calculate the utilization of healthcare services for inpatient and outpatient care of the aged people suffering from NCDs. NSS also reports usual³ monthly expenditure of the households. Following Cain et.al., (2010) and Srivastava et.al., (2016) [27, 28], the Organization for Economic Co-operation and Development (OECD) equivalence scale has been applied to construct the monthly per capita expenditure (MPCE) class from the household expenditure [27]. Specifically, we have divided the total monthly expenditure of a household (T) by the square-root of the household size (N) to get the MPCE⁴ (= T/√N). Then the MPCE has been arranged in ascending order and grouped into four quarters – Poorest (P), Lower Middle (LM), Upper Middle (UM) and Richest (R). However, cost of living largely varies across states and the sectors (rural and urban) within the states. To accommodate these differences, MPCE classes have been formed separately for each state⁵ and sector.

The OOPE has been reported under various heads in the NSS data separately for inpatient and outpatient care. This information has been used to calculate the OOPE on medicine and medical care.⁶ OOPE for transport and other non-medical services like food, expenditure on escorts, lodging charges etc. has been added with the medical expenditure to get the total OOPE. In case of inpatient care, OOPE has been reported for each hospitalization episode separately. However, for the outpatient care, NSS reports the total OOPE under various heads for all the outpatient visits together (for multiple visits within the reference period). Therefore, for the present paper we have considered only those aged individuals who have reported noncommunicable diseases related outpatient care utilization in all the visits.

Benefit incidence analysis (BIA) is a method generally applied in the literature to study the extent of equity in any public system. It has both the horizontal⁷ and the vertical equity dimensions [28, 30]. It has to be mentioned here that BIA has been applied in various studies to examine the extent of equity in public subsidy distribution across socio-economic classes. Few such studies are in the Indian context and based on NSS data. A recent study on utilization and benefit both at the national and state level, used NSS and showed that utilization of inpatient and delivery services are pro-poor. It is also revealed that in most of the states out-patient services is also pro-poor. However, users of public healthcare facilities are forced to spend considerable amount as out-of-pocket to supplement the government services [31]. Another study by Mahal et al., (2001) showed the inequity in subsidy distribution across income groups in India for various healthcare services using BIA [32]. Chakraborty (2012), Acharya (2011) and Ngangbam (2015) also applied BIA to study the extent of equity in distribution of public subsidies in the healthcare sector [33‐35]. However, there is dearth of literature with a focus on equity in distribution of public subsidy for NCD related treatment among the elderly in India. In this context, applying BIA, the present study attempts to through some light in this area.

According to the literature, benefit incidence is the net government subsidy weighted by the utilization rate [36, 37]. Mathematically, the benefit incidence could be estimated by the formula – Where,

π_j = Benefit of public subsidy enjoyed by group j (here MPCE class); α_ij = utilization of service i (here inpatient and outpatient care for NCDs) by group j; k = number of individual; α_i = utilization of service i by all groups together; γ_ik = government’s net expenditure on individual k for service i and δ_ij = group j’s share of utilization of service i.

Specifically, α_ij is number of aged people from a particular MPCE class (j) who are suffering from NCDs and utilizing the public healthcare facilities (forms the numerator). Total number of aged people suffering from NCDs and utilizing public facilities for treatment in all the MPCE classes together forms the α_i (the denominator). The ratio of α_ij and α_i is the utilization rate. To calculate the net government subsidy, we have applied the methodologies available from the literature [28, 38‐40]. In short, the net subsidy is the difference between the actual cost of providing a service and the user charges. The user charge for utilizing public facilities is available from the NSS data. However, the major challenge is to get the actual cost of providing the service in the public facilities. Following Bose 2018, Srivastava et.al., 2016; Bose & Dutta 2015; Bose 2014 [28, 38‐40], we have considered the modal OOPE of the private hospitals as the proxy for actual cost of providing the services in the public hospitals. To accommodate the differences in quality of care, severity of illness and cost of providing the services, state, sector, MPCE class and duration of stay in hospital (for inpatients only)/ total duration of ailment (for outpatient only) have been considered during calculation of modal OOPE in the private hospitals. The assumption we have made here that the OOPE for utilization of healthcare services are the actual cost of the government to provide the service in the public hospitals. However, it is indeed true that the cost of providing services in public and private sectors may vary. But, reason behind choosing the proxy could be justified from different angles. Primarily, there are various types of private healthcare facilities available for treatment and the prices largely varies based on the regions of operation, quality of services etc. To accommodate the differences and to normalize the variation in prices, analysis has been made using large unit level dataset. Additionally, we have tried to reduce the variation in prices, if persists, by analyzing the data separately for state, sector, MPCE class and duration of stay in hospital (for inpatients only)/ total duration of ailment (for outpatient only). On the other hand, the primary objective of benefit incidence analysis is to study the relative position of various socio-economic groups in access and benefit of public subsidy distribution [30, 37]. Therefore, some level of under or over estimation of public subsidy benefit would not affect the relative position of the MPCE groups in the subsidy ladder. Finally, private sector is the only option remains after the public facilities for treatment. Therefore, the difference between the expenditure in public and private facilities for treatment would be the best available proxy for the shadow price of subsidy in public facilities. OOPE for each hospitalization episode (or outpatient visit) has been subtracted from the corresponding modal private OOPE to calculate the net subsidy. Then the net subsidy is weighted by the utilization rate to get the subsidy benefit of the individual.⁸ We have added all the individual subsidy of a particular MPCE class to get the MPCE group specific subsidy. Share of each MPCE class in total subsidy benefit is the estimate of benefit incidence.⁹ Following O’Donnell et al., (2008) and O’Donnell et al., (2016) we have also calculated the concentration index and curves [41, 42]. Concentration curve could be used as a measure of inequality. Specifically, concentration curve could be used to assess the public subsidies are targeted towards the poor or not [42]. To derive the concentration curve for health inequality, we need two variables – (a) the health variable (here the public subsidy), the distribution of which are subject of interest and (b) a socio-economic variable (here MPCE class) against which the distribution is to be assessed. Concentration index on the other hand, represents twice the area between the concentration curve and the line of equality (45⁰ line). The index takes the negative value when the curve lies above 45⁰ line and vice versa.

Ailment group wise utilization of healthcare services for inpatient (IP) and outpatient (OP) care by the elderly people has been reported in Table 1. It is observed that 42% of the total hospitalization and 54% of the total OP visits are for NCDs in India among the elderly. Additionally, 46% of the total hospitalization are for NCD-related treatment in the urban sector and the corresponding figure for the rural sector is 40%.

Share of NCDs in total OP visit among the elderly is around 60% in the urban sector. Around 50% of all the OP visits in the rural India are for NCDs. Further bifurcating the IP and OP utilization across MPCE classes, we could observe that there is a positive relationship between MPCE class and utilization of healthcare services for treatment of NCDs among the elderly in India. As we move from the poorest to the richest class utilization of healthcare services increases both for IP and OP care.

Similar trend is observed in both the sectors. The highest utilization of IP care in the rural sector is among the richest (39%) followed by the UM (25%) and LM (20%) class. Corresponding percentages for the urban sector are 31%, 28% and 23% respectively. Lowest utilization of IP care is recorded by the poorest class (17%) of the rural sector followed by the urban-poorest (19%) class. Richest class of the rural and urban sector utilizes the maximum OP care followed by the UM class. Poorest class of both the regions has the lowest access to OP visits in India. Further analyzing the data to check the utilization pattern of public facilities for the treatment of NCDs among the elderly, we could notice that the richest class of the rural sector has the highest utilization of public healthcare facilities for the IP care and the poorest class has the lowest utilization.

On the other hand, in the urban sector UM class has the highest utilization of public IP care followed by the poorest class. Interestingly, utilization of public facilities for the OP care is the maximum for the poorest class in both the sectors in India. The lowest utilization of public facilities is observed among the LM and UM class for the rural and urban sector respectively.

Table 2 reports the MPCE class wise OOPE for public and private healthcare facilities in India. Three sets of estimates have been made separately for IP and OP care – medicine, medical and total OOPE. It is observed that the total OOPE in the private facilities is around 1.5–2 times high in the rural sector and it is around 2–4 times high in the urban sector compared to the public facilities. This trend is uniformly observed in both the IP and OP care. For both IP and OP care, poorest class is spending the lowest amount to purchase medicine during utilization of public institutions and as we move towards the richest class the medicine-OOPE increases. The same pattern is observed in both the sectors. However, the medical and total OOPE for the poorest class is higher than the LM class for public sector hospitalization in the rural sector. It is clear from the table that medicine is an important component in the total medical expenses for both IP and OP care in public facilities.

In case of IP care, the medicine expenditure ranges from 32 to 53% of the total medical expenses. Interestingly, the share of medicine in total medical expenditure is around 80% for public sector OP visits. In both the sectors, the highest medicine share is recorded by the LM class for IP care and UM class for OP visits. Surprisingly, the poorest class is also spending a substantial proportion of their total medical cost to purchase medicine during public sector hospitalization or OP visit.

The above result then raises a question that who is enjoying the benefit of the public subsidy. The NSS data has been analyzed and the MPCE class wise share of subsidy benefit for IP and OP care has been calculated and reported in Table 3. It is observed that the distribution of public subsidy for IP care is pro rich and the maximum subsidy share is recorded by the richest class and the poorest class has the minimum share. It is also observed that the poorest class has the minimum share in total subsidy for inpatient care in the rural sector followed by the UM and the LM class. The richest class of both the regions are benefited the maximum from the public subsidy during their hospitalization. In the urban sector, on the other hand, the lowest subsidy share is recorded by the LM class followed by the poorest class.

Interestingly, there is no specific trend in distribution of public subsidy for OP care. However, the richest class are benefited the maximum from public subsidy followed by the poorest class. The lowest subsidy is observed for the UM class. Sector wise bifurcation of the benefit incidence shows that the poorest class of the rural sector has the maximum benefit share followed by the richest class for OP care. On the other hand, benefit share of the richest class is the maximum followed by the LM class in the urban sector. For better understanding of the distribution of the subsidy, we have derived the concentration curves and presented in Fig. 1.

The distribution of public subsidy is clearly showing a pro-rich trend as the subsidy benefit distribution line lies below the equal distribution line for the IP care in both the sectors. We have also represented the 95% confidence interval for each benefit distribution curve (grey shaded area). Analyzing the distribution of public subsidy for the OP care, we could observe that the benefit distribution line has crossed and merged with the equal distribution line. Therefore, it is very difficult to comment on the distribution pattern of the subsidy benefit from the graphs. To know more about the distribution pattern of the public subsidy we have calculated the concentration indices. The sign of the concentration index indicates the direction of the public subsidy – pro-rich or pro-poor and the magnitude of the index reflects the strength of variability.

The concentration indices are presented in Table 4. It has to be noted here that a positive value of the index denotes a pro-rich distribution of subsidy benefit and a negative value signifies pro-poor distribution. It is observed that the distribution of subsidy benefit is pro-rich for IP care and pro-poor for OP care. Specifically, for both the sectors pro-rich distribution is observed for the IP care and for the OP care the distribution is pro-poor in the rural sector. However, it is observed from the above table that the indices of the urban sector are not statistically significant. Therefore, it is very difficult to comment on the result of the urban sector. On the other hand, the estimates of the rural sectors are highly significant for the IP and OP care. Therefore, significant level of inequity in distribution of public subsidy is evident from the result in case of IP care in the rural region.