Financial protection analysis in eight countries in the WHO South-East Asia Region

Abstract Objective To document the financial protection status of eight countries of the South-East Asian region and to investigate the main components of out-of-pocket expenditure on health care. Methods We calculated two financial protection indicators using data from living standards surveys or household income and expenditure surveys in Bangladesh, Bhutan, India, Maldives, Nepal, Sri Lanka, Thailand and Timor-Leste. First, we calculated the incidence of catastrophic health expenditure, defined as the proportion of the population spending more than 10% or 25% of their total household expenditure on health. Second, using World Bank poverty lines, we determined the impoverishing effect of health-care spending by households. We also conducted an analysis of the main components of out-of-pocket expenditure. Results Across countries in this study, 242.7 million people experienced catastrophic health expenditure at the 10% threshold, and 56.4 million at the 25% threshold. We calculated that 58.2 million people were pushed below the extreme poverty line of 1.90 United States dollars (US$) and 64.2 million people below US$ 3.10 (per capita per day values in 2011 purchasing power parity), due to out-of-pocket spending on health. Spending on medicines was the main component of out-of-pocket spending in most of the countries. Conclusion A substantial number of people in South-East Asia experienced financial hardship due to out-of-pocket spending on health. Several countries have introduced policies to make medicines more available, but the finding that out-of-pocket expenditure on medicines remains high indicates that further action is needed to support progress towards universal health coverage.


Introduction
The aim of universal health coverage (UHC), as set out in Transforming our world: the 2030 agenda for sustainable development, 1 is to ensure that all people and communities receive the health care they need, without experiencing financial hardship. The World Health Organization (WHO) South-East Asia Region consists of 11 Member States and almost 2 billion people living in low-and lower-middle income countries. Population health has progressively improved in recent decades, although the Region still lags behind many others, except Africa Region and fragile states elsewhere, and inequities remain. 2 Government spending on health ranges from 0.4% to 2.5% of gross domestic product in all countries of the Region except Maldives and Thailand, lower than what has been suggested as necessary for better performance. 3 As a result, the health financing model relies heavily on out-ofpocket expenditure by households, comprising an estimated 47% of current health expenditure on average in the Region, with a huge variation across countries from 10% to 74%. 4 Such a high level of out-of-pocket expenditure implies a heavy financial burden on households. 5,6 Moreover, the poor may be disproportionately affected due to fewer resources at their disposal; international evidence suggests that the costs of treatment could be prohibitively high for them to access needed health care. 7 There are two widely used approaches to conceptualize financial hardship: (i) catastrophic spending and (ii) impoverishment. Catastrophic spending on health care occurs when out-of-pocket expenditure exceeds certain pre-defined thresholds, affecting households' ability to spend on other necessities of life. Impoverishment refers to situations in which household spending on health pushes people into poverty. The two con-cepts capture different aspects of the economic consequence of out-of-pocket expenditure on households. For instance, for those whose per capita spending is just above the poverty line (threshold), a small amount of out-of-pocket expenditure on health care, although not catastrophic by definition, could lead to impoverishment. By contrast, well-off households may have catastrophic out-of-pocket expenditure, but still stay above the poverty line. Analysing both indicators is therefore important to present a fuller picture.
Efforts to develop the concept of catastrophic expenditure on health care date back to 1986. High out-of-pocket expenditure for illness, defined as a fixed amount of family income, was considered an opportunity cost both for households sacrificing consumption of other items and for societies through loss of labour productivity. 8 Similarly, arbitrarily and exogenously defined fixed thresholds were used to define catastrophic expenditure, but instead of income, total household budget was used as the denominator. 9 A second approach is to use capacity to pay as the denominator, which deducts the spending on necessities defined in a variety of ways (e.g. actual food expenditure, 9 subsistence level food expenditure, 10 maximum saturated level of expenditure on necessities, 11 spending on food, rent and utilities 12 and a multiple of international poverty thresholds 13 ). The evolution in methods highlights the need to better differentiate the budget capacity of poor and rich households to measure the real financial impact of out-of-pocket expenditure. Previous research also underscores the difficulty in coming up with a perfect indicator that can be applicable to a wide variety of countries and surveys.
We aimed to document the financial protection status of eight countries of the WHO South-East Asian Region with the latest available data. Two indicators were calculated, the incidence of catastrophic health expenditure (indicator 3.8. of the sustainable development goals) and the impoverishing effect of households' health-care spending, as defined in the joint World Health Organization and World Bank Global Monitoring reports. 14,15 We also aimed to investigate the main components of out-of-pocket spending both at national level and by quintiles of total household expenditure.

Indicators
Out-of-pocket expenditure on health care is defined as payments made at the point of service, after deduction of any reimbursement. When out-of-pocket expenditure exceeds a threshold of total household budget, the household is defined as having catastrophic health spending. Suppose m i is health expenditure per household i, n i is total expenditure, t j is the threshold, with t 1 = 10%, t 2 = 25%, the catastrophic expenditure under threshold j CHE ij is 1 if m i /n i > t j and 0 otherwise. If we define population We used the change in poverty headcount ratio to calculate how many people were impoverished due to outof-pocket expenditure. The change in poverty gap captures both the number of the households impoverished and the severity of the impoverishment. Equation 2 below defines the gross and net poverty headcount ratio and Equation 3 defines gross and net poverty gap, following previous methods. 16 And the difference between the two, G oop = G net −G gross , measures the change in poverty gaps due to out-of-pocket payment on health, expressed as the percentage of poverty lines in this paper.
To determine the main drivers of out-of-pocket expenditure, we decomposed it by categories of spending and analysed their relative size. Suppose is expenditure by household i on component of k 1 , k 2 , … k j , and let k 1 be out-of-pocket expenditure on medicines as that is universally available of all surveys, while, k 2 , … k j might represent different items across countries. Then the average share of out-of-pocket spending on each component can be defined in Equation 4: When j = 1, the above equation measures the average share of out-ofpocket spending on medicines.

Data sources
We included eight countries of the WHO South-East Asia Region in the study: Bangladesh, Bhutan, India, Maldives, Nepal, Sri Lanka, Thailand and Timor-Leste. We did not include Indonesia as their survey instrument was recognized as being unable to separate actual outof-pocket spending from insurance reimbursement. We also excluded Democratic Republic of Korea and Myanmar, because to our knowledge there were no national surveys at the time that met the criteria for our analysis. We used data from the most recently available household surveys in each country, which were either living standards and measurement surveys or household income and expenditure surveys ( Table 1). These are the most appropriate types of survey for such analysis, because they are nationally representative and have a detailed documentation of household consumption, including that of health care. Some of these surveys have also been used to estimate national poverty ratios and many have been used for National Health Accounts for the estimate of out-of-pocket expenditure. 18 The full lists of variables in each data set used for the analysis are listed in Table 2 (available at: http://www.who.int/bulletin/ volumes/96/9/18-209858).

Data analysis
We used recall periods of 30 days for outpatient care to reduce bias of recall and 12 months for inpatient care to reduce bias due to infrequent occurrence. To generate total household expenditure on health, we separated out items which, although asked about under health modules, do not belong to health services. These include rimdo or puja (or religious treatment, in Bhutan) and transport costs. In rare cases when health-care expenditure was asked both in the health and non-food modules of the survey, only the former was counted in out-of-pocket expenditure.
We used the two international recommended thresholds to define large out-of-pocket health expenditure: above 10% and above 25% of total household expenditure or income. 14,15 The defini-

Research
Financial protection analysis, South-East Asia Hui Wang et al.
tion of poverty usually varies across countries, so for comparison we used the two international poverty lines at the time of the study of 1.9 United States dollars (US$) and US$ 3.1 per capita per day (based on 2011 purchasing power parity exchange rates) 19 to define the incidence of poverty due to out-of-pocket expenditure and the poverty gap. We grouped national population into five economic quintiles based on their per capita consumption level. We used Stata 2014 (StataCorp LLC, College Station, United States of America) for all analyses. Table 3 shows the main sociodemographic and health-system characteristics of the countries analysed. There were large variations in economic development and population health across countries, but a common pattern of heavy reliance of out-of-pocket expenditure. Fig. 1 and Fig. 2 summarize the basic characteristics of out-of-pocket expenditure in each country. On average, in most countries, more than 50% of the population had some level of out-of-pocket expenditure spending. The average ranged from 1.1% to 6.1%

Research
Financial protection analysis, South-East Asia Hui Wang et al.
of total household budget, or purchasing power parity US$ 1.1-21.9 per capita per month. As expected, richer populations had more out-of-pocket spending and the spending was higher both in absolute (dollars) and relative (% of household budget) measures. There was no consistent pattern between rural and urban households across countries. Table 4 (available at: http://www.who. int/bulletin/volumes/96/9/18-209858) presents the incidence of catastrophic health expenditure. For both thresholds, Maldives had the highest share of the population experiencing catastrophic health expenditures, followed by India and Bangladesh. Thailand and Timor-Leste had the lowest (at the 10% poverty threshold). Based on the total populations reported in the corresponding survey years (Table 1), we estimated that across the eight countries, 242.7 million people had catastrophic expenditure at the 10% threshold and 56.4 million at the 25% threshold.

Catastrophic health spending
The finding that poorer households had lower incidence of catastrophic health spending is consistent with global studies, and is aligned with the above findings that poorer households spent less on health care, both in absolute and relative terms ( Fig. 1 and Fig. 2). The pattern across rural versus urban areas was less clear, with the incidence of catastrophic spending much higher in rural than urban areas in Bangladesh, Bhutan, India and Maldives.
When capacity-to-pay was used as an alternative denominator, we found very few people made more than 40% of their non-subsistence spending on health, and the richest quintile was still more likely to spend a bigger share of their budget on health (data are available from the corresponding author). Table 5 shows the impoverishing effect of health-care spending expressed as the share of the population being pushed below the poverty line. In total 58.2 million people were pushed below the extreme poverty line of purchasing power parity US$ 1.90 per capita per day and 64.2 million below the poverty line of US$ 3.10. India and Bangladesh had the highest share of the population affected, translating into 52.5 million and 5.2 million people, respectively, being pushed under the US$ 1.90 poverty line. When US$ 3.10 was used as the poverty line, another two countries, Maldives and Nepal, were also affected. In both cases, Thailand had the fewest people impoverished due to out-ofpocket spending.

Impoverishing health spending
It is worth noting that the value of zero in Table 5, mostly observed in the lowest quintiles, represented those who were already classified as poor; as a result, any out-of-pocket expenditure on health care would only further their financial hardship. Given this, the data clearly show that the poorer suffer much more than their richer counterparts. A typical example is Timor-Leste, where the poorest 40% of the population (at the US$ 1.90 poverty line) were vulnerable to further impoverishment by out-of-pocket expenditure on health. Coupled with the very low out-of-pocket expenditure in Timor-Leste ( Fig. 1 and Fig. 2), the results show that the poor have limited capacity to cope with any out-of-pocket expenditure on health.
Analysis of the changes in poverty gaps induced by out-of-pocket expenditure showed that the impact was highest in Nepal (Table 6; available at: http://www.who.int/bulletin/volumes/96/9/18-209858) indicating that the out-of-pocket expenditure pushed people not only below, but also further away from the poverty lines.   (Table 7). Moreover, in all except two countries the share of out-of-pocket expenditure due to medicines exceeded 70%. In the two exceptions, other important out-of-pocket expenditure components included fees paid to private medical practitioners in Sri Lanka (Fig. 3) and outpatient visits in Maldives (Fig. 4). In general, poorer households spent relatively more on medicines than did their richer counterparts. Data on all health expenditures are available from the corresponding author.

Discussion
Using the latest available surveys, our study provides a cross-sectional description of financial protection against out-of-pocket expenditure for eight countries in WHO South-East Asia Region, with two key findings. First, most countries (except Thailand, Sri Lanka and Timor-Leste) performed below the global median rate for at least one of the indicators. 15 Second, the dominant role of out-of-pocket expenditure on medicines has been observed over the past decade in the Region, as corroborated by earlier studies. [21][22][23][24][25] For instance, medicines constituted 72% of total outof-pocket expenditure payments in India as early as 2004, 21 almost unchanged until 2011.
Our findings suggest that more effective health policies are needed to provide better financial protection of households. Several attempts have already been made in the Region. In India, the Rashtriya Swasthya Bima Yojana scheme was launched in 2008 to provide protection for households below the poverty line. Despite a more than twofold increase in enrolment on average from 2011 to 2016, there were still large gaps in coverage in some states and generosity of benefit packages varied across states. 26,27 Furthermore, evidence suggested that even for the insured households, Rashtriya Swasthya Bima Yojana did not affect either the likelihood or the level of out-of-pocket expenditure spending, thus rendering no financial protection for the most vulnerable populations. 27 By contrast, in Thailand, there was quick expansion of the universal coverage scheme in 2001 to cover the informal sector, with comprehensive inpatient and outpatient health care included in the benefit package. The initiative is one of the reasons behind the successful financial protection of the entire population of Thailand that has been consistently observed. [28][29][30] Maldives had one of the highest levels of catastrophic spending and impoverishment in our study. However, the data represented the situation in year 2009, preceding the launch of the national health insurance programme, Aasandha, in 2012. It would be worth reassessing the financial burden of households against out-of-pocket expenditure in Maldives with data from the 2016 household survey, to see if the nationwide insurance scheme has made a difference.
The high financial burden of medicine expenditure found in our study, draws attention to the limitations of current pharmaceutical policies in reducing out-of-pocket expenditure in these countries. All eight countries have defined and regularly updated their essential medicines list and state their intention to provide medicines free-of-charge in public health-care facilities. However, other studies found that, for several reasons, most people in South-East Asian countries purchased medicines from private pharmacies, 31 exposing themselves to higher risk of financial burden. Studies reported that the poor were more likely to be deterred by the perceived high prices. 32,33 This is particularly worrisome as the burden of noncommunicable diseases, which are associated with higher out-of-pocket expenditure and catastrophic health expenditure 22,34 , is increasing fast in the Region.
While the analysis of impoverishment clearly demonstrated the higher financial burden on poorer households, the incidence of catastrophic expenditure was higher among richer people. This finding is consistent with those from earlier studies in the South-East Asia Region. 14,15,[35][36][37][38] The distribution is sensitive to the ways the denominator and household quintiles are constructed. Using the capacityto-pay approach by deducting food expenditure from household budget is more likely to result in a higher incidence among the poor. 38 A recent study in Bangladesh found a pro-poor distribution of catastrophic spending by using household assets instead of consumption to determine quintiles. However, the results varied by area (rural versus urban) and the threshold selected for the analysis. 39 The lack of robustness of the indicator of catastrophic health spending shows its limitation in conveying policy relevant messages (i.e. the poor suffer more and need more targeted policies). This is partly due to the fact that catastrophic health spending is conditional on being able to spend on health care in the first place, omitting people who cannot afford the medical service at all, which is more likely for very poor households.
Both the financial protection indicators we used have other limitations. First, they do not capture indirect costs associated with illness, such as income loss due to disability. Second, they do not differentiate the households who borrow or reduce their savings to compensate for health care. These households may not have been identified as facing financial hardship due to outof-pocket expenditure in the short-term, but will be economically worse off in the medium term. Therefore, more research is needed to refine the approach.
Variations in the designs of their respective household surveys present a challenge in directly comparing financial protection status across countries. Questions about health spending did not follow the same structure, with different levels of detail and different groupings of out-of-pocket spending components.
The former may overestimate health expenditure 40 while the latter creates difficulty in accurately attributing out-ofpocket expenditure to particular items. The household income and expenditure survey of Sri Lanka is a case in point, where the survey design made it impossible to classify out-of-pocket expenditure into typical inpatient and outpatient care. In addition, the Sri Lanka survey captured out-of-pocket expenditure of both the main households and servants living with the families without further details beyond a lump-sum reporting for the latter. For reporting purposes we assumed that the two groups shared exactly the same structures across out-ofpocket expenditure components, which is unlikely to be true, but is reasonable given its marginal magnitude. Similarly, surveys also varied in the questions asked about non-food, non-health expenditures, both in terms of the type and level of detail of data collected. The poor usually have a larger proportion of spending on food than other categories. Therefore an under-or overestimate of  Notes: The number 0 in the poorer quintiles mean that all people were already below poverty lines and could be pushed below them again due to health expenditure. In other words, they are vulnerable regardless of outcome. In contrast, the 0 observed in richer quintiles mean that no one in the group was pushed below poverty lines due to health expenditure and are therefore not vulnerable.
non-food expenditures may impact on the calculation of the denominator, thus affecting how both the financial protection indicators vary across economic quintiles. Finally, while the majority of countries used a mix of recall periods, Nepal's survey followed a recall period of 12 months. The bias introduced by a long recall period is already well documented. 40,41 A standardization of such surveys, such as following the structure of the Classification Of Individual Consumption According to Purpose, 42 would better support cross-country comparisons.
Despite the limitations, our findings revealed the low-ranking financial protection status of countries in South-East Asia Region and the persistent burden on households from pharmaceutical spending. With the expected increase in demand for health care due to epidemiological and demographic changes, both financing and service delivery policies need to adapt for satisfactory progress towards UHC. We also call for further research efforts to refine the indicators for better monitoring of financial protection and a better reflection of the equity dimension. ■ Table 7.

Research
Financial protection analysis, South-East Asia Hui Wang et al.

Research
Financial protection analysis, South-East Asia Hui Wang et al.

Research
Financial protection analysis, South-East Asia Hui Wang et al.