Correlates of Out-of-Pocket Spending in Nepal: Implications for Policy. Indrani Gupta and Samik Chowdhury

Transcription

1 Correlates of Out-of-Pocket Spending in Nepal: Implications for Policy Indrani Gupta and Samik Chowdhury 1. Introduction Health equity has been conceptualized and defined in several ways, as its principles are derived from the fields of philosophy, ethics, economics, medicine, public health, and others. The WHO defines inequity as differences [in health status], which are unnecessary and avoidable, but in addition, are considered unfair and unjust [WHO 2000]. Literature around the Social Determinants of Health (SDH) brings into the health inequity discussion a wide range of issues that result in social gradients in health (Marmot et al 2008, WHO 2008). The focus on avoidable health inequalities between groups of people within countries and between countries that emanate from social and economic conditions have an important corollary: inequities can be addressed to a great extent if one tackles, in the first instance, those that are accentuated because of existing anomalies in a variety of policies. The Commission on SDH notes that access to and utilization of health care is vital to good and equitable health. The health-care system is itself a social determinant of health, influenced by and influencing the effect of other social determinants. Gender, education, occupation, income, ethnicity, and place of residence are all closely linked to people s access to, experiences of, and benefits from health care (WHO 2008). A key factor in inequities in access and utilization is the need and ability of households to pay for health care. The World Health Report (WHO 2010) on health systems financing brings together a large body of evidence that highlight the various barriers to access that enhances the vulnerability to health shocks for a large section of the population in developing countries. Paying for health care from own resources remain an important source of burden on households, and together with other social determinants of health, accentuate existing inequities. Moving towards a system of universal health coverage (UHC) is certainly an important way of bringing down out-of-pocket spending (OOPS) and addressing a significant source of inequity in financing and, therefore, in health outcomes.

2 Nepal is a typical South Asian society with much of its social inequalities coming from economic inequities compounded by caste and ethnicity issues (Gellner 2007). It is among the poorest countries in the world and currently ranks 157th out of 187 countries on the Human Development Index. The regional imbalances also play an important role in Nepal. The latest Demographic and Health Survey (Nepal DHS 2011) indicate that wealth inequality is higher in rural than urban areas. Among the three ecological zones mountain, hill and terai, inequality of wealth is highest in the hill region. Among the five development zone (eastern, central, western, mid-western and far-western), the Central and Mid-western development regions, and midwestern hill sub-region have the highest inequality of wealth. Finally, as in other South Asian countries, Nepal also has inequalities between the gender evidenced in differences in, for example, education and health, as also significant presence of gender-based violence 1. However, it has been able to reduce poverty significantly and inequality has also improved; the overall GINI coefficient (based on expenditures) declined from 41 to 35 between 2003/2004 and 2010/2011 (World Bank 2 ). This is despite Nepal s political instability and internal conflicts, which has resulted in constrained progress and growth of the country. This paper examines the level, variation and trend in out-of-pocket spending in Nepal by using the Nepal Living Standard Survey (NLSS) over two periods. Since regional imbalances are important in the country, we look at the changes in inequalities over time, across ecological zones or belts, development regions and rural-urban residence. This is in addition to consumption quintiles and gender of the head, the other two variables used to analyze inequality in household OOP expenditure on health. We set the context by discussing the health financing system in Nepal and policies that attempted to increase health coverage. 2. Health financing system in Nepal The 2007 interim constitution of Nepal enshrined basic health care as a fundamental right of its citizens. The Government of Nepal, however, had been making incremental efforts at increasing

3 access to health care services even prior to this landmark event, through a comprehensive framework of health policies, strategies and plans. The National Health Policy (1991), the Ninth Five Year Plan ( ), the Nepal Health Sector Programme Implementation Plan (NHSP- IP) ( ) etc. were some of these measures aimed at improving health access and outcomes (WHO, 2007). In spite of these well-intentioned measures, the presence of private sector in health had been steadily growing in Nepal, with a concomitant increase in financial burden at the household level. The first policy intervention aimed at providing financial protection arrived in 2006, when emergency and inpatient care was made free for vulnerable groups like the poor, destitute, elderly etc. at district hospital and primary health care centre (PHCC) levels. In 2007, free health care at the health posts and PHCC level was universalized. Since 2009, all citizens were made eligible for free outpatient, emergency, in-patient services, as well as drugs at District Hospitals (DH), Primary Health Care Centres (PHCC) Health Posts (HP) and Sub-Health Posts (SHP) 3 (HSSP-GTZ 2009, Witter et. al 2011). In 2009, the Government introduced the Aama Suraksha Karyakram 4, a scheme targeting the demand side of safe motherhood. These measures led to creditable improvements in health access and outcomes in Nepal, relative to its income levels 5, with the rate of progress being better than some of its neighbouring countries. However, two important issues still afflict the current health care system and is growing: geographic and income related inequalities in population health outcomes and high household out-of-pocket payments for health care that largely takes place in the yet unregulated private sector (World Bank 2011, WHO 2013). 3 The public health system of Nepal comprises of health clinics, sub health posts (SHP), health posts (HP) and primary health care centers (PHCC) at the primary level, district and zonal hospitals at the secondary level and regional and central hospitals at the tertiary level. Private health care institutions in the form of hospitals, medical colleges and nursing homes have also emerged in the recent past. 4 The Aama Programme includes both consumer led demand side payments and provider payments. Consumers receive 1,500 Nepalese Rupees (NRs) in high mountain districts, NRs.1,000 in hill districts, and NRs.500 in Terai districts, to cover transportation and other access costs. Health staff receive NRs.200 per home delivery assisted, health facilities of up to 25 beds receive NRs.1,000 per delivery and facilities with more than 25 beds receive NRs.1,500 for normal deliveries (WHO 2011) 5 According to UNSTATS, Nepal is a less developed country within a developing region (Asia)

4 The National Health Accounts (NHA) (2006/ /2009) gives a fairly recent picture of the state of health financing in Nepal. The NHA shows that total, as well as per capita health expenditure grew by a substantial amount over the years. But the chief (60%) source of health expenditure has been the private sector, followed by the general government and the international community. The disquieting feature of the system is that about 90 percent of private health expenditure and 55 percent of total health expenditure was spent out-of-pocket by households. The NHA for Nepal collects information on OOP expenditure from health service providers unlike many countries where household surveys are the sole basis for estimation of such expenditures. While the efficacy and robustness of such a procedure can be debated, it allows one to classify OOP expenditure by agents, functions and providers. In Nepal, retail sales and supplies of medical goods, private hospitals, clinic and labs were the main recipients of OOP payments. In terms of functions, outpatient and curative care services together received around 78 percent of the total OOP payments (NHA Shrestha et. al 2012). There have been very few studies examining the extent and variation of the financial burden of OOP health expenses on Nepalese households. One inexplicable reason has been the reluctance to use health expenditure related information from the National Living Standards Survey (NLSS) 6, a nation-wide household survey consisting of multiple aspects related to household welfare. This is in spite of the fact that poverty rates at the national and subnational levels have been computed periodically from the same data source. The NLSS (NLSS-III) was last conducted during , and was made available in November Findings from one study (Hotchkiss et. al 1998) that uses the NLSS-I for household OOP expenses on health indicate that households spend around 5.5 percent of their total consumption expenditure on health care. Rural households spend more on health care compared to urban ones, after controlling for economic status. Another finding from the study is that households spend high amounts on health care when their initial consultation is with a public practitioner, even though such consultations are supposed to be priced nominally. This, according to the 6 Central Bureau of Statistics (CBS) carried out the first Nepal Living Standards Survey (NLSS) in 1995/96 followed by one in 2003/04. The Nepal Living Standards Survey 2010/11 is the latest available round. All the three surveys followed the Living Standards Measurement Survey (LSMS) methodology developed and promoted by the World Bank (Government of Nepal 2011).

5 authors, is a probable indication of private providers complementing public providers of health services. Yet another study (Rous et. al 2003) investigates the determinants of household OOP expenses, controlling for sickness and provider choice, using NLSS-I again. The authors found income to have a direct as well as an indirect effect on health expenditures. While the direct effect was measured in terms of income elasticity of OOP expenditures, the indirect effect was manifested in terms of likelihood of illness and provider choice. Housing and sanitary conditions emerged as significant determinants of illness and therefore OOP expenses. This study also found a higher average health care expenditure among the rural sample compared to the urban one. Another landmark study on Nepal s health financing is a Benefit Incidence Analysis (Silva- Leander 2012) of health subsidies on population subgroups categorised by region, caste, gender, dwelling area, income, poverty and multidimensional poverty. The study uses NLSS-III data for demand side variables and Public Expenditure Review (PER) of the health sector for the supply side. Key findings suggest that the largest per capita gross subsidy goes to the western region of Nepal, whereas the largest net (net of direct cost recoveries e.g. consultation fees) subsidies accrue to the Mid-Western and Far-Western regions. Women receive slightly higher gross subsidies than men. Finally, the fourth income quintile captures the highest public subsidy on health while bottom quintile captures the lowest. Nepal has been seriously considering ways to implement universal health coverage (UHC), though the exact process of implementing it remains undecided as of now. Augmentation and reallocation of public resources within the health system along with regulation of the private sector would remain, however, the core of any UHC but the design, number and implementation level (national or sub-national) remains to be decided. 3. Data and methods The Nepal Living Standards Surveys (NLSS) follow the Living Standard Measurement Survey (LSMS) designed by the World Bank and applied widely in other developing countries as well. Till now there have been three rounds ( , and ) of the NLSS, each reasonably comparable to the other in terms of the range of information and the structure of the

6 schedule. There have been certain amendments though in terms of the recall period. The NLSS- III consists of a cross-sectional sample of 5988 households and a panel sample of 1032 households, 1160 of whom were interviewed in NLSS-I as well. The survey collects information on different aspects of household welfare, including demography, housing, access to facilities, consumer expenditure, education and health to mention a few. The survey covers the entire country, both rural and urban areas. The 75 districts of the country are grouped into three ecological belts and five development regions. The NLSS-I surveyed 3373 households (CBS 2011). This study uses data from the cross-sectional sample of NLSS-I and NLSS-III. The section on health of the NLSS-III schedule consists of four parts chronic illnesses, illnesses or injuries, HIV/AIDS knowledge and immunizations. The first part on chronic illnesses contains among others, information on the type of illness and the expenditure incurred on its treatment in the past twelve months. The second part, which might be interpreted as acute illnesses, also contains information on the same variables but, for a reference period of 30 days. We arrive at the total annual health expenditure of a household by deriving the annual value of the health expenditure on acute illnesses from the monthly figures, and simply adding it to the health expenditure on chronic illness, which already is available on an annual basis. For total consumption expenditure we add up food expenditure, frequent non-food expenditure, infrequent non-food expenditure and value of inventories purchased within a year from the date of survey. Since health spending on the two items mentioned under frequent non-food consumption were quite different from the total health spending obtained from the detailed health section, we adjusted consumption expenditure in the following way: from the total consumption expenditure we deducted the total of those two items, and added the total health expenditure obtained from the health section. Thus, the total consumption expenditure estimated by us is different from the reported total expenditure in other studies that have used NLSS (CBS 2011, Hotchkiss 1998). 4. Out-of-pocket spending on health in Nepal The average per capita OOP health spending in Nepal currently stands at NPR During this 15 year ( to ) period, this has increased seven-fold, in nominal terms. If we make a plausible assumption that prices do not vary substantially by regions and household socio-economic attributes

7 Figure1: Ratio of Per Capita OOP Expenditure to Rural Urban Eastern Central Western Mid Western Far Western Poorest Second Middle Fourth Richest Sector Region Expenditure Quintiles ALL within a country, we confront a scenario, which has serious implications regarding equity. The factor by which nominal OOP health expenditure per capita increases between the two periods, shows huge variation across select socio-economic characteristics. While for urban Nepal the increase was four fold, for rural Nepal it increased by more than seven times. The Eastern and Far-western region experienced a six-fold increase in per capita OOP health spending while the Central region saw an eight fold jump. Perhaps the most significant of all observations emerge when we examine the increase factor across consumption expenditure quintiles. The richest quintile experienced a five-fold increase while the poorest two quintiles currently spent more than ten times the amount that they did in Figure 2 presents the share of OOP in total consumption expenditure (henceforth OOP share) of households between the 2 rounds. Clearly, for Nepal, this share has increased between NLSS I and NLSS III from 3.4 percent to 4.5 percent. This increase has been quite uniform across all the parameters used, except for the Western region. The other point to note is that OOP share is higher for rural areas compared to urban, and among the 3 belts, highest for the Terai belt, where the increase between the two rounds has also been quite sharp. The increase has also been sharp for Eastern, Central and Mid-Western regions.

8 Figure 2: Share of OOP health expenditure in total consumption expenditure of households Rural Urban Mountain Hill Terai Eastern Central Western Mid Western Far Western Poorest Second Middle Fourth Richest Male Female Sector Belt Region Expenditure Quintiles Gender Head ALL (NLSS I) (NLSS III) In terms of consumption quintiles and gender of the household head, figure 2 indicates that the increase between the two rounds has been sharpest for the first two quintiles. In fact the richest quintile has seen a fall in OOP share. There does not seem to be much of a difference in OOP share for male and female- results headed households, but both have seen a rise between the two periods. Overall, however, the indicate that the distribution of OOP was rather progressive in the earlier period and has worsened slightly.. To understand which of the variables discussed above might influence health spending and health share, we ran a simple regression on the pooled data from the two surveys. Two models were tested, the dependent variable being per capita health spending and the health share respectively. The explanatory variables were belt, rural-urban, region, consumption category, gender of the head of household and a created variable indicating the time-period. The results are reported in Table 1. We estimated the equation using the Generalized Liner Model (GLM). As is commonly the practice in modeling health expenditures using GLM, we use the log-link with a gamma error (Manning and Mullahy 2001, Manning et al 2005) for the GLM estimation. Comparing between the two sets of estimates the most robust results seem to be the positive association of OOP spending with income, the lower association for hills and mountains compared to the Terai belt, and higher expenditure in all the regions compared to the far-

9 western (omitted category). Also OOP is significantly lower for the earlier period, compared to the latter period. In case of the health share model, too, the co-efficients are roughly of the same sign and strength. Table 1: Determinants of OOP health spending GLM (Dependent variable: Independent variables GLM (Dependent variable: per capita health spending) share of OOP in total consumption expenditure) Coefficient z Coefficient z Rural * Female head Per capita consumption *** NLSS round (NLSS 1=1) *** *** Belt Hill *** *** Mountain ** *** Region Eastern *** ** Central *** *** Western *** *** Midwestern ** Constant - - Log Pseudoliklihood Observations A consistent negative significant coefficient corresponding to the time (NLSS) variable indicate that OOP health expenses as well as its share have increased between the two rounds. Since from Figure 1 and Figure 2 we see that this increase has not been uniform within groups, it is important to objectively identify inequality within the goups and its behaviour over time. One easy way of verifying this is through the coefficient of variation (CV). We analyse the change in CV in share of health in household consumption across the regional and gender variables. Figure 3 indicates that there has been a significant decline in inequality of OOP spending across the belts and regions, but more so for the regions. Inequality in health share between rural and urban areas has increased slightly but not changed for female and male-headed households.

10 Figure 3: Coefficient of Variation in Share of Health in Total Household Expenditure Belt Regions Rural/Urban Gender of Head (NLSS I) (NLSS III) The conclusion is thatt while per capita OOP spending and its share in total household consumption has increased over the years, the within group inequality in certain select cases, has declined. In other words, while all households are spending a higher amount out-of-pocket, their share does not vary too much within the selected categories. The explanationn is straightforward: from Figure 2 which shows that the share did not increase much for those who were already spending more on health, rather it increased at a faster rate for those with a relatively lower share during Thus, what is driving the inequality results is an increase in OOP at the lower end of the distribution, and therefore, this result is deceptive. This is clear from the finding that in the earlier year the ratio of health share of lowest to highest income quintile was only 0.38, which subsequently increased to 0.92, indicating that the poor are spending as much as the rich on health as a share of total consumption. Further analysiss is needed to understand where exactly the burden if higher OOP falls the most. 5. Distribution of OOP burden In this section, we attempt to look at the possible burden of catastrophic payments across the various categories in Nepal, to understand the distribution of such burden across regions and economic categories. There are many alternative ways researchers have defined catastrophic payment (Pradhan and Prescott 2002; Ranson 2002; Wagstaff and Van Doorslaer 2001; Xu et al ); we define health spending as catastrophic if it is 10 percent or more of total household expenditure. Figure 3 presents the results on catastrophic expenditure i.e., the proportion of households in each category that spends more than 10 percent of their total consumption expenditure on health care. The first point to note is that on an average a higher percentage of households are incurring such expenditures in the latter period (increased from 10.7% to 13% %). The other results are that there has been

11 an increase in proportion in rural areas and a decrease in urban areas, though the difference is small for the latter. Figure 4: Households spending more than 10% of total consumption expenditure on healthh (%) Urban Rural Mountain Hill Terai Eastern Central Western Mid west Far west Poorest Second Third Fourth Richest Ml Male Female Sector Belt Region Expendituree Quintiles Gender Head ALL Among the three belts, Terai has seen the most increase in proportion of householdss with catastrophic expenditure (from 12.9 to 16.25). However, the other two belts have also seen some increase. Among regions, Eastern has seen the most increase, followed by Central and Mid-west. Western region has actually seen a fall. There has been no change in Far-West. As for income quintiles (Figure 5), the biggest increase in proportion of householdss with catastrophic expenditure is from the poorest quintile, followed by the second quintile. The remaining three have remained more or less the same, with slight fall for the third quintile. An important observation is that the catastrophic headcount had a roughly positive income (consumption) gradient in , which is no longer the case, in the recent period. Figure 4 clearly shows that in more households in the lower quintiles had to face the catastrophic impact of OOP healthh expenses, compared to those in the higher quintiles. For gender of head, both types of households have seen an increase in catastrophic burden over the two periods.

12 The most pertinent question for policymakers is the composition of households experiencing catastrophic expenditures. The question can be posed thus: of all households experiencing catastrophic expenditure, what is the distribution across the belts for example? Figure 5: Distribution of households facing catastrophic OOP health expenses Urban Rural Mountain Hill Terai Eastern Central Western Mid-west Poorest Third Richest Second Fourth Sector Belt Region Quintile In Figure 6, we present the distribution over these categories for each year. The most important result that stands out is that the bulk of the households struck with catastrophic payments are from rural areas in both the periods. While it was 90 percent in the earlier period, this has reduced to about 80 percent but still it is an important finding that rural households take the most hit in terms of catastrophic expenditure. The results do not change much for the three belts between the two periods, but Terai region contribute more than 56 percent of the households that experience catastrophic expenditure. Among regions,

13 Eastern and Central now contribute more, with the situation improving considerably for the Western region. The Mid-west and Far-west regions are doing relatively better and contribute much less to the total burden of worse-hit households. As for consumption categories, 45 percent of the worse-hit households come from the two lowest quintiles. However it should be noted that the upper three quintilesalso contribute to the total burden significantly, though relatively less than the two lower quintiles. Finally, there has been an increase in the proportion of female-headed households experiencing catastrophic expenditure, though the majority of such households remain male-headed. 6. Summary and Conclusion OOP spending in Nepal has increased between and from 3.4 percent to 4.5 percent, an increase of over 30 percent. Compared to other South Asian countries like India (OOP share is more than 8 percent in India), this share is still relatively lower. The increase in the average share can be attributed to the increase in health spending by households at the lower end of the economic scale. Thus, there is less progressivity in the latter period, compared to the earlier one. There has been an increase in the incidence of catastrophic expenditure, mostly occurring in rural areas, Eastern and Central regions, the Terai and among the poorer sections of the population. There are important policy implications: firstly, OOP spending can contribute to poverty if not addressed due to its impact on catastrophic expenditure. The vulnerable sections need special attention in any UHC, especially if it is being rolled out in a phased fashion. The most effective use of scarce resources for UHC would be to target the priority areas brought out by the preceding analyses, where most of the vulnerable households reside.

14 References: Hachette F (2009), Free Health Care Services in Nepal: Rapid Assessment of the Implementation and Per Patient Expenditure, Report for GTZ/GFA Consulting Group GmbH, Health Sector Support Programme, Department of Health Services, Kathmandu, Nepal. Summary available at: Witter Sophie, Sunil Khadka, Hom Nath and Suresh Tiwari (2011), The national free delivery policy in Nepal: Early evidence of its effects on health facilities, Health Policy and Planning 2011; 26: ii84 ii91 WHO (2007), Health System in Nepal: Challenges and Strategic Options, Country Office for Nepal, WHO, November Available at World Bank (2011) Assessing Fiscal Space for Health in Nepal. Health Nutrition and Population, South Asia Region, The World Bank, Adhikari Shiva Raj, Devi P Prasai, Sharad K Sharma (2011), A Review of Demand Side Financing Schemes in the Health Sector in Nepal, NHSSP, Government of Nepal, June Shrestha BR, Gauchan Y, Gautam GS, Baral P (2012). Nepal National Health Accounts, 2006/ /09, Health Economics and Financing Unit, Ministry of Health and Population, Government of Nepal, Kathmandu Central Bureau of Statistics (2011). Nepal Living Standards Survey : Statistical Report, Volume 1 and 2, National Planning Commission Secretariat, Government of Nepal, November Hotchkiss David R, Jeffrey J Rous, Keshav Karmacharya, Prem Sangraula (1998) Household Health Expenditure in Nepal: Implications for Health Care Financing Reform, Health Policy and Planning, 13(4): , Manning, W.G and J. Mullahy (2001). Estimating log models: to transform or not to transform? Journal of Health Economics, 20:461-94

15 Rous David, David R Hotchkiss (2003), Estimation of the determinants of household health care expenditures in Nepal with controls for endogenous illness and provider choice, Health Economics 12: , Silva-Leander Sebastian (2012). Benefit Incidence Analysis in Health, Nepal Health Sector Support Programme, Government of Nepal, October Available at al.pdf Wagstaff, A., & van Doorslaer, E (2001). Paying for health care: Quantifying fairness, catastrophe and Impoverishment with applications to Vietnam , World Bank Policy Research Working Paper No Washington DC, World Bank. Ranson, M.K. (2002) Reduction of catastrophic health care expenditures by a community based health insurance scheme in Gujarat, India: Current experiences and challenges, Bulletin of the World Health Organization,80 (8), Pradhan, M. and N. Prescott (2002). Social risk management options for medical care in Indonesia, Health Economics 11: Xu K, Evans D B, Kawabata K, Zeramdini R, Klavus J, and Murray C J L. (2003) Household Catastrophic Health Expenditure: A Multicountry Analysis, The Lancet. 362, July 12. Marmot Michael, Sharon Friel, Ruth Bell, Tanja AJ Houweling, Sebastian Taylor (2008). Closing the gap in a generation: health equity through action on the social determinants of health, The Lancet, Volume 372, Issue 9650, 8 14 November 2008, Pages Gellner David N (2007). Caste, Ethnicity and Inequality in Nepal Economic and Political Weekly. Page , 2007.Manning, W.G., A. Basu and J. Mullahy (2005). Generalized modeling approaches to risk adjustment of skewed outcome data. Journal of Health Economics, 24:

16 .