Subsidized health insurance coverage of people in the informal sector and vulnerable population groups: trends in institutional design in Asia

The eligibility rules are an expression of a country’s perception of which population groups are considered “vulnerable” to be entitled for subsidy, thus determining the number of subsidized people. In general, “vulnerable groups” can be defined as those groups that due to economic, demographic, geographic or other characteristics have an increased susceptibility to not having adequate access to health services with financial protection [39]. However, the subsidized need to be distinguished from the insured family dependents when they are covered via their principal affiliate through the principle of cross-subsidization by contributors. Family dependents usually include children up to a defined age, non-working spouses (most often women), or in some cases elderly parents.

As Table 3 reveals, the most frequent population groups being subsidized are those below the (national) poverty line or the 20 % poorest (five countries), followed by older persons and children (three countries) and those outside formal employment (three schemes). Women form a considerable part of these eligible population groups. A special case is India’s Yeshasvini, where eligibility is restricted to members of rural cooperatives and their family. Vietnam is also unusual with 25 eligible groups defined, including various socio-demographic and socio-economic groups such as children, students, informal sector workers, ethnic minorities in geographically defined areas with high poverty rate as well as veterans, so-called persons of merit and their dependants, state serving individuals and other non-Ministry state serving individuals (e.g. members of the National Assembly and People’s councils) [15].

Most countries employ some form of indirect or direct targeting, and often both, to identify and select individuals eligible for subsidization, but with considerable variation found (see Table 3). Direct targeting is based on (proxy) means testing by determining household income and/or assets) [55]. Indirect targeting is applied to include persons on the basis of socio-demographic (e.g., age, gender), socio-economic (occupation, or employment status) or socio-geographic characteristics. These characteristics are easily observable and identifiable as well as correlated with vulnerability [56].

A few countries employ only indirect targeting, such as India’s Yeshasvini, and China’s NRCMS (with geographic targeting criteria) and URBMI (with socio-economic targeting criteria), whereas their complementary Medical Financial Assistance (MFA) program applies a direct targeting approach. In Mongolia, the main targeting approach is indirect using socio-demographic criteria, but some specific groups are targeted directly. Likewise, the Vietnam Social Security (VSS) scheme applies both indirect and direct targeting for about equal numbers of subsidized people. On the other hand, the Philippines, Indonesia, Cambodia and all Indian schemes (except Yeshasvini) use a direct targeting approach only.

All countries that apply direct targeting use means or proxy-means testing to identify the poor. Several subsidization schemes draw on pre-existing databases built up to grant other forms of social assistance, such as the ration cards system in India [12] or the Equity Access Card, also known as the Poor Card in Cambodia [57]. However, the thresholds that separate eligible and ineligible individuals vary significantly across countries: India’s Rashtriya Swasthya Bima Yojana (RSBY) only includes individuals below the country’s poverty line, whereas Vietnam, Indonesia and the Philippines also attempt to target the near-poor. In these three countries as well as in the complementary MFA in China, the local communities play an important role in the identification process [15, 44, 58, 59].

In a universalist approach, all individuals fulfilling a very broad criterion (e.g. permanent residency in the country, lack of coverage in another SHI scheme) are eligible. Thailand is indeed the only country in this group with no targeting and hence a universalist approach: all permanent legal residents who are not covered by the formal sector health insurance schemes are entitled to the state budget financed Universal Coverage Scheme (UCS). In China, URBMI’s eligibility criteria try to capture all urban legal residents that are not part of the formal sector schemes, thus also coming close to a universalist approach.

Another decisive aspect is the actual level of subsidization and the remaining share of co-contribution. In 10 out of the 14 subsidization schemes, the eligible beneficiaries are fully subsidized from government revenues and do not co-contribute (see Table 5). There is an annual registration fee of 30 Rs. in India’s RSBY, which comes to 4–7 % of the amount that the government provides as a subsidy, varying across states [24].

In the remaining four schemes (China’s NRCMS and URBMI, Vietnam, India’s Yeshavini), some or all groups, as listed in Table 5, receive a partial subsidy only and have to partially contribute. In China, the subsidy level is above 80 % in most regions (up to 85 %), but as low as 23 % elsewhere. The complementary MFA pays the remaining co-contribution for a small proportion of beneficiaries (for 6 % of NRCMS and 7 % of URBMI enrolees in 2011) [31]. In Vietnam, the subsidy level ranges from 30 to 70 % of the set contribution depending on the groups, but the poorest individuals as well as other population groups receive a full subsidy. The subsidy level is about 40 % in India’s Yeshavini, whereby the subsidy level varies from year to year depending on available resources and willingness-to-pay findings [75].

As to the type of transfer mechanism, all countries follow the same logic with the transfer being an implicit or explicit per-capita amount per beneficiary, as per their regulatory or legal provisions. The alternative would be to provide a lump sum that is unrelated to the precise number of number of eligible beneficiaries. There are three approaches found to set this per capita subsidy amount which determines the average amount available for health care expenditure of each subsidized member (see Table 5). One logic is to base the subsidy amount on estimations of future health care expenditure per subsidized member. This method is applied in China as well as in Thailand, although in a more implicit way. In the two Chinese schemes, the level of central government allocations also depends on the regions’ level of economic development as well as affordability considerations [78]. Estimated health care expenditure was also the starting point in Indonesia, and allocations are now adjusted based on expenditure of the previous year. A second logic is to base the subsidy amount upon the contribution rate of contributing members that is applied to a reference salary, such as in the Philippines and Vietnam, or on a set contribution amount (that of non-salaried people), like in Mongolia. Third, in all Indian schemes other than Yeshasvini the actual amount of government revenue transfers is determined through a tender process where insurance companies bid for a “premium” at which they would provide insurance coverage for a family in a pre-defined population group and benefit package [12].

These approaches serve to make government revenue transfers more predictable and needs-adapted by prospectively fixing the per capita subsidy amount, rather than depending on the country’s economic situation or budget allocations and negotiations. However, this has only been partially successful. Independent of the type of approach to set the transfer amount, the actually transferred funds were found to differ from the amount established as per the calculation formula or negotiations. This was the case for example in the Philippines [76, 79, 80], Mongolia [13] and Indonesia [81, 82]. Similarly, in Thailand fiscal constraints resulted in an approved per capita amount that has been lower than the proposed per capita expenditure estimations between 2002 and 2010 [36, 77]. Overall, for the actual subsidy allocation, the regulatory provisions and the related calculation principles appear to have been of limited relevance in some countries, and may be explained by budget constraints, missing data on the number of the eligible or subsidized members, or other political priorities.

In all subsidization arrangements assessed here, the financing source for subsidies are primarily general government revenues. These come from central government revenues in Vietnam, Mongolia, Thailand, Indonesia and the Philippines (since 2011). Sub-national (i.e. state) government revenues are the source used for India’s state-based subsidization schemes. Finally, the remaining schemes—China and India’s RSBY—are financed by both central and sub-national government revenues, whereby the main share comes from provinces/counties and states respectively. A unique feature is found in Cambodia, where subsidies to HEFs are also financed from donor funds. Moreover, in Vietnam, another key source are social security revenues for social security beneficiaries, such as the registered unemployed, pensioners and beneficiaries of social assistance allowances [15]. Notably, the Philippines is the only country which makes use of sin taxes on tobacco and alcohol to subsidize the poorest and most vulnerable groups [35, 83]. After the introduction of this tax in 2012, the government revenues allocated for subsidizing the poor increased significantly [84].

Both single fund systems as well as multiple pool systems are present across Asia, with Table 6 providing country details. In three out of the eight countries studied (Vietnam, the Philippines, Mongolia) the subsidized population groups are part of the same pool as the contributors in a single fund system. In contrast thereto, China, Cambodia, Thailand, Indonesia and India have separate, but non-competing schemes for the subsidized population groups, resulting in fragmentation. In these five countries, a formal sector health insurance scheme or non-competing multiple schemes existed already prior to the introduction of the subsidization arrangement. Importantly, however, as of January 2014, the health insurance schemes of Indonesia are being integrated into a single national health insurance agency [85].

China’s separate subsidization schemes are further fragmented into sub-national pools. Risk pooling by the NRCMS schemes occurs at county level, with an average pool size of 500,000 individuals, whereas risk pooling by URBMI schemes is at municipal level with an average pool size of 2 million members [14, 89]. There are no risk equalization mechanisms in place across these sub-pools, resulting in unequal per capita average expenditures across counties and municipalities respectively [31].

Among the integrated systems, fragmentation along sub-national lines is equally a concern in Vietnam and the Philippines. Differences in revenue generation capacity that determines resource allocation to provincial sub-pools (in Vietnam) and local governments (in the Philippines), and, in the absence of risk adjusted allocations or risk equalization, result in unequal average per capita expenditure across the country. However, this is the case for all insurance members, not only the subsidized ones.

Revenue sources of the subsidization arrangements include government revenue transfers, co-contributions by beneficiaries where applicable, and in integrated schemes these comprise also contributions from contributors. The last column of Table 5 presents data on the total revenue amount as a share of average expenditure, i.e. the revenue-expenditure ratio, based on aggregate data for all subsidized members. This provides insights into the sustainability of the scheme as well as the level of cross-subsidization between contributors and the subsidized in integrated schemes.

While data, especially recent one, is scarce, this ratio strongly varies across countries and even within countries over time. Some schemes have a balance or a ratio above 100 % and others have a ratio below 100 %. In the Thai UCS, total revenues are more or less equal to expenditure (ratio of 95–105 %), as well as in the Philippines and Indonesia prior to 2009. Interestingly, more frequent is the situation where revenues are higher than expenditure (with a ratio above 105 %), such as China’s NRCMS and URBMI, India’s Yeshasvini and Kalaignar (in 2009), Indonesia (in 2010). This ratio even amounts to around 200 % in Vietnam and above in India’s RSBY. Moreover, in Indonesia, there was a high ratio of 140 % in 2009 when the local government-based Jamkesda scheme came into place, which resulted in double coverage of many people. Another reason for Jamkesmas’ relatively lower expenditure was that not all of the enrolees had a health card and hence had no access to benefits. In 2010 this ratio decreased to 112 %, not because of increased utilization rates, but because the per capita transfer amount was reduced. Finally, some subsidization schemes experience a "deficit": the ratio is below 95 % in the Philippines (2010), India’s Rajiv Aarogyasri (2009-10) as well as Mongolia, which reached only 30 % in 2009.

Ratios above or below 100 % indicate that the revenues (subsidy amounts) and expenditure may need to be better linked, particularly in separate schemes. This is because a persisting deficit endangers the sustainability of the system and results in coverage erosion, under-provision and rationing of benefits by providers. A large surplus could also create efficiency losses since resources could have been used for other purposes.

Integrated schemes for both the subsidized and the contributors are financed by contributions from contributors as well as by general government revenue transfers, with the possibility of cross-subsidization between different insured groups depending on the level of revenue coming from each source and per capita average expenditure for different groups. Comparing revenue-expenditure ratios for the subsidized with that of the contributors reveals the degree and direction of cross-subsidization (cf. Table 5). In fact, in Mongolia and the Philippines (for 2009 and 2010), this ratio of subsidized members is lower than that of non-subsidized groups, suggesting that the latter cross-subsidized the former. In contrast, the revenue-expenditure ratio of subsidized members in Vietnam was higher than that of contributors, such that cross-subsidization occurred from subsidized members to contributors (in 2009 and 2010). Specifically, formal sector employees, but also the subsidized groups of the poor, children and students (who are only partially subsidized) cross-subsidized pensioners, the voluntarily insured as well as the partially subsidized group of the near-poor (for 2009 only). Moreover, the revenue-expenditure ratio of the poor and children is considerably above 100 % due to low expenditure levels of these groups. As a result, the government budget transfers also indirectly subsidize other groups [15].

Purchasing consists of the allocation of pooled funds to the health care service providers on behalf of the population for a defined benefit package. The definition of the range of services covered in the benefit package is decisive for financial protection and access to care, as are the cost-sharing mechanisms. The benefit packages are being set by the government in all countries. Moreover, all schemes assessed here provide in-kind-benefits (i.e. health services are directly covered and paid for by the scheme as a third party payer, with no bill settlement of beneficiaries other than paying co-payments). The exceptions are China and the Philippines, where the insured have to advance payments and are later reimbursed by the insurance. Regarding the scope of the defined benefit package (see Table 7), two patterns emerge: The first is a relatively comprehensive package, and the second pattern is a more restricted package primarily covering inpatient care only. The more comprehensive package includes outpatient and inpatient care at all levels and is available in five countries (Vietnam, Mongolia, Indonesia, Philippines and Thailand). These schemes also reimburse pharmaceuticals that are included in the country’s respective essential medicines list. Vietnam’s package is broadest on paper, as all available services are covered in principle, but restrictions come in through a ceiling on the total amount per person per year. The second pattern of a more restricted package is found in all Indian arrangements. China’s NRCMS and URBMI are somewhat in-between the two patterns, with more than half of NRCMS and URBMI counties having some outpatient care in the package [72].

Comparison of the benefit entitlements between the subsidized and contributors reveal crucial differences across these two groups in some countries. In fact, benefits are largely the same or similar in two countries: Vietnam and Mongolia. In contrast, the benefit package for the subsidized is more generous in the Philippines (with a larger outpatient benefit package), Indonesia (with more treatment options) and Thailand (compared to the social security scheme) [36, 60, 95]. On the other hand, they are smaller for those of the Thai UCS (when compared to the civil servant scheme) and for the subsidized in India and China, with a much more restrictive coverage of outpatient services. However, outpatient consultations have been introduced in the benefit package of India’s RSBY in 2014 [91].

The cost-sharing mechanisms and rates vary considerably among countries (see Additional file 1). Thailand and Indonesia are the only countries without any formal cost-sharing, and in Mongolia the subsidized members are exempted from nearly all co-payments. In all other schemes, cost-sharing, e.g. in the form of co-insurance and co-payments and benefit ceilings such as in Vietnam, Mongolia and China apply also to the subsidized. The cost-sharing rates range from 5 % of total health insurance expenses for the poor in Vietnam [15] to 50 % in the Philippines [96] and amount to about 50–60 % in both Chinese schemes [14]. But importantly MFA beneficiaries in China receive cash assistance for some part of outpatient and inpatient cost-sharing [31], thus cushioning the financial burden.

In comparison to the contributors, the cost-sharing rates for subsidized members are the same or similar in Thailand and Indonesia. They are higher in China only, but lower in Vietnam, Mongolia and the Philippines. For India, comparative information is not available, however there are no annual benefit ceilings applied in the contributory Central Government Health Scheme (CGHS) in contrast to the subsidized arrangements [12]. Nonetheless, with or without formal cost-sharing mechanisms in place, people also face informal payments in several countries [24, 31, 37, 49] with potentially negative impacts on access, equity as well as impoverishment.

Here, population coverage is defined as the percentage of the population that is enrolled to a health insurance scheme. Total population enrolment rate vary substantially across countries (see Table 8). They are very high in Mongolia (100 % in 2014), China (99 % in 2012) and Thailand (99 % in 2015). The lowest rates are found in Cambodia (27 % in 2013) and India (25 % in 2014).

Another dimension is to capture the share of the actually subsidized insured members as compared to those who are theoretically entitled or targeted as per entitlement. This gives insights into the targeting effectiveness. The Thai UCS with a universalist approach is the only one with a 0 % exclusion error [77]. In schemes with indirect targeting, the coverage rates of the effectively subsidized as a share of the theoretically eligible number of people vary: it is high for China’s URBMI and NRCMS, covering 93 and 95 % respectively as well as for school children and students in Vietnam (80 %). In contrast, India’s Yeshasvini reaches only 25 % of the intended target group.

Among subsidization schemes with direct targeting, India’s Rajiv Aarogyasri, Vietnam for the poor and China’s complementary MFA reach 100 % or close to 100 % of their target number. Coverage of target groups is much lower in India’s RSBY, but also in Cambodia’s HEFs, which is due to the fact that there is no nation-wide coverage of HEFs nor SUBO. For the group of the partially subsidized near-poor in Vietnam, the coverage rate is only 25 % in 2011. On the other hand, having a closer look on targeting effectiveness also reveals that most schemes have suffered from severe inclusion errors in the past (see Table 8).

Another population coverage indicator is the share of the subsidized as of the total population to reveal the magnitude of the subsidization arrangement. In most countries, this share ranges between 15 and 50 %. This ratio is higher in Thailand and China’s rural population, 74 and 62 % respectively, but extremely low in India’s Vajapayee Arogyasri Scheme (0.12 %) and Yeshasvini (0.25 %). Thus, in all countries with the exception of these two Indian schemes an important share of the population relies on subsidization arrangements to obtain health insurance coverage.

It is equally important to keep in mind the share of the remaining uninsured (see Table 8) and reasons thereto. This share is tiny (about 1 %) in Thailand given its universalist approach, whereby migrants and those who recently became unemployed and transition to the UCS belong to the uninsured. In the Chinese schemes, largely relying on indirect targeting, the requirement of residency status makes it difficult for internal migrants to enrol (about 1 %). Coverage rates in Mongolia reduced in the early 2000s, as herders and part-time students were no longer subsidized and had remained uninsured during these years, and the self-employed and unemployed were also among the uninsured [49, 114].

As Table 8 reveals, the highest rate of uninsured people is found in countries with direct targeting: 62 % in the Philippines [115], 31 % in Indonesia, up to 75 % in India [12] and 73 % in Cambodia. In Vietnam, where direct and indirect targeting each account for about half of subsidized members, the share of the uninsured (33 %) is higher than in schemes with indirect targeting, but lower than in schemes with direct targeting [15].

Global evidence has shown that there is a strong correlation between out-of-pocket (OOP) expenditure as a share of total health expenditure and the share of households experiencing financial hardship [23]. Thus, a starting point is to look at trends in OOP expenditure as a share of total health expenditure. Table 9 presents data on changes over time in OOP expenditure and financial protection measures of the total population, the subsidized target groups or of poor population groups. However, the introduction of subsidized coverage may be just one factor among several to explain changes in the level of financial protection.

A clear decrease in OOP, comparing average values in the years before and after the introduction of the subsidization scheme particularly for the poor, was found for Thailand and in Cambodia’s HEFs and SUBO. For beneficiaries of India’s Yeshasvini and Indonesia’s Jamkesmas, there was some reduction. In Indonesia, it was found that some beneficiaries preferred not to use their cards and instead paid out-of-pocket. They wanted to avoid perceived stigmatization from health providers and longer waiting times as a result of having to complete additional administrative forms [33].

Only a very small OOP expenditure reduction was observed for the beneficiaries of India’s Rajiv Aarogyasri and RSBY and for China’s URBMI and NRCMS. In the Philippines no change or even a small increase was reported for the population, but when differentiated along deciles, the poorest 10 % and richest 40 % of all Philippine Health Insurance Corporation (PhilHealth) members (regardless of subsidization status) experienced a reduction of OOP payments. In contrast to this, in Vietnam, between 2002 and 2010, OOP expenditure increased across all income quintiles [126].

However, overall, in most of the subsidization schemes, beneficiaries still incur considerable OOP expenditure, even though cost-sharing rates are lower for the subsidized than for contributors. This remaining level of cost-sharing might be particularly problematic for the subsidized low-income groups, since they are the ones that most likely forego care [127].

As to the financial protection measures, namely catastrophic and impoverishing health expenditure (for the definition see Legend of Table 1), the evidence is scarce and mixed. Regarding, the incidence of catastrophic expenditure, a clear-cut reduction was only observed for UCS beneficiaries in Thailand, the most notable effect being reported among the poorest beneficiaries again. Some small reductions could also be shown for China’s URBMI beneficiaries, as well as in Vietnam and Indonesia. However in Indonesia, this was partly attributable to lower utilization rates. For beneficiaries of India’s RSBY and Rajiv Aarogyasri, no change in the incidence of catastrophic expenditure was found. In China, NRCMS beneficiaries in the Western regions [120] and those in income deciles 3–10 experienced an increase [118, 120], while another study reports a decrease for the poorest 10 % [118]. In sum, financial protection improved somewhat in some countries and for the poorest, but in a few others, it did not.

It is more difficult to measure effective access to needed services in the absence of data on actual needs. A proxy for access to services are utilization rates [128]. Utilization rates of the subsidized for outpatient and/or inpatient care increased since the introduction of the subsidization schemes and were higher compared to the uninsured in six schemes, namely in Vietnam, Indonesia, China, and Thailand as well as India’s Yeshasvini and Cambodia’s HEFs. In Thailand and China’s NRCMS, the poor have experienced the highest increase in utilization. For China’s NRCMS, Vietnam and Indonesia some studies also report there was no impact at all in certain regions and that among poorer beneficiaries the increase was much more moderate (Table 10). Likewise, in Mongolia, utilization rates for family group practice and district hospitals became more pro-poor, whereas for tertiary care, they became even more pro-rich during the period of 2007–2012 [129].

A comparison of utilization rates between subsidized beneficiaries and contributors reveals some inequities in some countries. In fact, utilization rates are still lower for the subsidized members in Thailand, Vietnam, Indonesia (with regards to inpatient care) and considerably lower in the Philippines. In China, differences in the use of outpatient and maternity care services between different regions and lower and higher income groups have notably reduced since NRCMS and URBMI were established [101]. However, the limited scope of the NRCMS benefit package limits improvements in access to outpatient services.