Patterns and determinants of out-of-pocket health care expenditure in Sri Lanka: evidence from household surveys

Abstract

This article examines patterns and determinants of the likelihood and financial burden of encountering out-of-pocket healthcare expenses in Sri Lankan households as, on average, more than 60% of households incur such costs. This percentage varies substantially across household categories in demographic properties, sectors and ability-to-pay. Households comprising more than one elderly person, pre-school children, members with chronic illnesses, and literate household heads are at significant risk of incurring out-of-pocket payments and bearing a higher financial burden. Rural and estate sector households are more likely to bear a higher burden. The marginal effects of household income show that the burden of private healthcare is less sensitive towards changes in household income and that households’ burden in private healthcare was regressive in 2006/2007. Hence results imply that low-income households need to be protected. Analysis of supply side factors shows that availability of closer government hospitals, bed numbers and dentists in government hospitals reduce the burden of out-of-pocket expenses. However, more government doctors lead to higher likelihood and burden of incurring such healthcare expenses and create a government-doctor-induced cost. Therefore, the results show a convincing need for the expansion of healthcare infrastructure by government and a policy framework for its doctors that will lessen the financial burden in Sri Lankan households, particularly the poor.

Introduction

Academics and policymakers have examined the impact on households due to illness and rising out-of-pocket expenditure on healthcare (McIntyre et al. 2006; Garcia-Diaz and Sosa-Rub 2011; Sekyi and Domanban 2012; Shahrawat and Rao 2012). There is now a paradigm shift away from efficiency considerations of healthcare provision to equity: the notion of aligning healthcare costs to household income levels is taking root (McIntyre and Gilson 2002; World Health Organization 2010a). This is more relevant for low- and middle-income countries, where empirical findings show that fees for public and private healthcare impose considerable financial burdens on households (O’Donnell et al. 2008; Garcia-Diaz and Sosa-Rub 2011; Shahrawat and Rao 2012). Hence a ‘critical re-evaluation of the health sector reforms that prevailed in low- and middle-income countries over the past two decades is highly warranted’ (McIntyre et al. 2006, p. 858).

Health reforms in the low- to middle-income strata of Sri Lanka are highly debated (Jayasinghe 2010/2011; Bandara 2011; UNDP Sri Lanka 2012). Although Sri Lanka’s status as a welfare state has allowed it to maintain a satisfactory level of health indicators, regardless of any economic decline over time (Rannan-Eliya and Sikurajapathy 2008; World Health Organization 2010b; World Bank 2014), demand for healthcare is increasing rapidly (Jayasinghe 2010/2011; UNDP Sri Lanka 2012). This trend will continue because of a number of economic, demographic and social changes. Bandara (2011) argues that Sri Lanka’s middle-income status may result in increasing demand for healthcare, because economic expansion causes people’s lifestyles to change (e.g. consuming more ‘fast food’), to work more and to experience higher levels of occupational stress. An aging population is also a major driver of increased healthcare expenditure (Jayasinghe 2010/2011; UNDP Sri Lanka 2012). The elderly population (aged 60 or more) was 11.8% of the total in 2010 and projected to increase to 25.7% by 2050 (UN 2012). With this change come added health requirements of elderly people with limited social security or extended family support. Sri Lanka faces the problem of looking after people affected by the three-decade civil war, which ended in May 2009 but left many disabled and suffering from severe traumatic impacts: of those affected, 8–12% currently suffer from Posttraumatic Stress Disorder and require psychological treatment (Jayatunge 2010).

Private financing accounts for 52% of total healthcare expenditure, of which 95% is out-of-pocket (Amarasinghe et al. 2013). Macroeconomic data from the World Health Organization (2010b) for the period 1995–2011 show the percentage out-of-pocket component in total healthcare expenses ranged from 42% to 48%, with an average value of 45%. The lower standard deviation of those percentages (1.81%) shows a lesser variability of the household out-of-pocket healthcare component in their total healthcare expenses.

Government financing accounts for 48% of total healthcare expenditure (Central Bank of Sri Lanka 2014). The Ministry of Health and provincial councils manage the public healthcare system through annual allocations. The majority of in-patient care and the total volume of preventive and public health services are administered by the central government (World Bank 2013). Though people initially contact public hospitals, they end up with the private sector, paying out-of-pocket for laboratory tests, drugs and medical devices (Berendes et al. 2011; Institute of Policy Studies 2012; Rannan-Eliya et al. 2015). Those unable to afford payments must wait for months until facilities become available in public hospitals (The Economist 2014).

Even where public provision is subsidized, patients who can afford the cost often seek private care because of perceptions of better amenities and quality. There is ‘a steep income gradient in use of private hospitals, with the richest quintile of people accounting for 45% of all private sector admissions, and public sector admissions being equal to pro-poor in their distribution’ (Rannan-Eliya et al. 2015, p. i47). Rannan-Eliya and Sikurajapathy (2008) argue that public sector health services reach the poor more than does the private sector and does ‘more with less’.

Government allocation for healthcare remains stagnant at 1.7–2.0% of GDP (World Health Organization 2010b). This results in government hospitals facing a shortage of medicine and other resources needed for patient care. Limited access to specialist treatment and inconsistent service standards further exacerbate problems in public hospitals (The Economist 2014). There is now an increasing trend in out-of-pocket direct payments for private healthcare (Withanachchi and Uchida 2006; Institute of Policy Studies 2012). This is further aggravated by lack of social insurance aimed at healthcare; private health insurance and employer contributions play a negligible role in financing healthcare. Health insurance is provided by firms as a product in their general insurance portfolio, and as a rider cover to life insurance (Govindaraj et al. 2014). Due to the limited role played by healthcare insurance and social security systems in South Asian countries, households predominantly depend on informal sources, such as extended families and traditional community support (Schieber and Maeda 1999; Samaratunge and Nyland 2007).

Private healthcare in Sri Lanka is currently handled by four companies, Nawaloka, Asiri Hospital Holdings, Lanka Hospital and Durdens, mainly locally owned. Public-private partnership in ownership is minimal. These companies, based primarily in Colombo, provide a large volume of private healthcare (Govindaraj et al. 2014). The main barrier to new entrants in the sector is high fixed costs; the four main companies have decided to expand outside Colombo, but foreign investors remain nervous about ownership in the health sector and have limited their partnership to < 3% (The Economist 2014).

The Government may facilitate private investors, local and foreign, in the sector to reduce initial fixed costs through fiscal policy. It should try to avoid the oligopolistic nature of the private healthcare market, promote competition and thereby reduce the cost of private medical practice, which is the largest portion of out-of-pocket healthcare expenses in households.

Wagstaff and Doorsaler (2000) argue that substantial dependence on such spending creates issues for equity, because out-of-pocket payments are the most inequitable source of healthcare financing. Hence a need arises to analyse such factors in Sri Lanka, because there is a huge pressure on government to replace that expenditure with more equitable financing. Only through careful analysis of patterns and determinants of households’ out-of-pocket expenditure on healthcare, policy reforms aimed at its financing can be designed. Additionally, the likelihood of spending for private healthcare and its financial burden vary across households, depending on factors at household and individual level (Malik and Syed 2012).

Accordingly, this study analyses such expenditure to reveal its patterns and determinants. It investigates the nature of the impact of an array of demand-side and supply side factors and estimates relevant elasticities of healthcare expenses. Income elasticity of expenditure allows us to know how much households spend out-of-pocket for healthcare as income fluctuates, and this can be informative for government, health-related NGOs and the private sector in allocation decisions. It also specifically considers post-conflict Sri Lanka, where the Government is in major reconciliation and reconstruction mode. This is ideal timing for understanding patterns and determinants of out-of-pocket healthcare expenses, because this will enable adjustment of public policies to better fit equity needs.

The next section reviews relevant literature; the third elaborates the conceptual framework, the methodology and data employed. Empirical results and main findings are then discussed, followed by conclusions and policy implications.

Out-of-pocket healthcare expenditure in other countries

A number of determinants influence household out-of-pocket healthcare expenditure. These vary substantially according to the developed or developing status of a country. Growth of public sector health spending is relatively higher in developed than in developing countries due to higher levels of government quality, stability and efficiency, and therefore the household burden of out-of-pocket expenditure does not vary considerably because of supply side factors (OECD 2010; Liang and Mirelman 2014). Also, as Xu et al. (2011) argue, the literature on determinants of spending on healthcare services in OECD countries is not entirely applicable in developing countries.

US studies show that determinants include household size, composition, financial constraints and the level of health insurance. Fan et al. (2000), using a US consumer survey, found that out-of-pocket expenditure is notably associated with household size, composition and financial constraints. They considered a variety of variables to control demand-side and supply side factors that impact this expenditure. Hwang et al. (2001), using the US 1996 medical expenditure panel survey, concluded that out-of-pocket expenses increase as the number of chronic conditions for elderly and non-elderly increases, the highest out-of-pocket expenses are observed among uninsured individuals and they vary by age.

Mondal et al. (2014) study for India showed that the main determinants involve number of illnesses, prevalence of chronic conditions among household members and inpatient care and childbirth. They provided compelling evidence to show that the size of such expenses is significantly affected by household size and people’s living sector.

Studies have also thoroughly investigated elderly and non-elderly persons’ impact on out-of-pocket expenses, because elderly people have different health requirements. According to Mohanty et al. (2014), monthly out-of-pocket healthcare expenditure in elderly households in India is considerably larger than in non-elderly households. Brinda et al. (2014) investigated the relationship between expenditure and antecedents using a Tanzanian national household survey and found that increased age, female gender and obesity and functional inability result in increased expenditure for adult individuals (aged 18–59). For elderly Tanzanians (age > 60), out-of-pocket expenses are higher among those who suffer from functional inability and use traditional healer services.

A number of studies have estimated the relationship between household income and size of out-of-pocket expenditure for healthcare. However, there is no clear-cut relationship between these, because the relationship varies from country to country. Bock et al. (2014) found a positive relationship in Germany. They reached the same conclusion for Turkey, with convincing evidence from a Satori sample-selection approach applied to Turkish household budget surveys from 2003 to 2008. Since their study allowed for possible selection biases, their conclusion, that the likelihood of seeking healthcare privately is relatively lower for poor Turkish households, is very convincing. Atanasova et al. (2012) found no significant association between household income and out-of-pocket expenditure in Bulgaria. They argued that the burden of such expenditure is approximately the same across poor and rich households. Some studies concluded that out-of-pocket expenses and income are negatively related: according to Sanmartin et al. (2014), though expenditure increased for households in all income quintiles in Canada, households in lower-income quintiles tended to spend more than those in richer-income quintiles.

In order to analyse the association between the burden of such expenditure and household income, some studies have used macroeconomic data. Using household expenditure and socio-economic surveys, O’Donnell et al. (2008) investigated equity issues of healthcare financing in 13 Asian countries. The data show that out-of-pocket healthcare expenses in high-income Asian countries are proportional or regressive. In low-income Asian economies, such as Sri Lanka, richer households spend relatively larger out-of-pocket amounts for healthcare and enjoy a larger scope of services. Kwon (2011) found that healthcare in many Asian countries relies on very limited financial sources, except in countries like Japan, Korea, Taiwan and Thailand. This imposes a higher financial burden on poor households in those countries due to out-of-pocket expenses. This warrants further research in individual countries by considering household income as one variable of interest. Therefore, we are particularly interested in income elasticity of out-of-pocket healthcare burden in Sri Lanka.

Conceptual framework

We consider household composition in terms of the number of elderly, young dependants and members with chronic illnesses as potential determinants. The characteristics of household head, household income and geographic location of households are also included. Although the literature indicates the influence of health insurance on out-of-pocket expenses (Parker and Wong 1997), it cannot be tested here for Sri Lanka due to unavailability of relevant data in the surveys used. However, from the experiences of other countries, we can predict for Sri Lanka that the availability of health insurance leads to a decline in households’ out-of-pocket healthcare burden: as Parker and Wong (1997) reported, uninsured Mexican households are more sensitive to fluctuations in household income levels than insured households when spending privately for healthcare.

Figure 1 shows the conceptual framework for the current study.

The study tests the influence of supply side factors that influence the burden of out-of-pocket healthcare expenses. Most studies have not tested variables from both sides. This study fills that gap by providing evidence from Sri Lanka.

Method and the data

We employ the standard Probit regression model to analyse the relationship between households’ likelihood of out-of-pocket expenses on healthcare and various covariates concerned and the standard Tobit model to analyse how households’ burden of out-of-pocket healthcare expenditure is determined by and is responsive to each covariate. In order to ensure the robustness of findings, we calculate alternative estimates by using Adult-Equivalent Scaled (AES) per capita expenditure along with the out-of-pocket healthcare burden alternatively defined as the percentage of out-of-pocket healthcare expenses in households’ total food and non-food expenditure (Supplementary A1 of Appendix 01).

This study uses the 2006/2007 and 2009/2010 Sri Lanka Household Income and Expenditure Surveys (HIES 2006/2007, 2009/2010) from the Department of Census and Statistics. The data on demography, education, health and income are collected and recorded at the individual level. However, data on expenditure are available at the household level and therefore the unit of analysis of this study is the household. The field data collection of both HIES 2006/2007 and HIES 2009/2010 was in 12 monthly rounds, from July 2006 to June 2007 and from July 2009 to June 2010, respectively.

The data on healthcare supply side factors, such as availability of hospital beds, doctors and dentists, are from the District Statistical Handbook compiled by the Department of Census and Statistics of Sri Lanka. These district data are combined with HIES 2006/2007 and HIES 2009/2010 to generate a comprehensive dataset for the main analysis (Supplementary A2 of Appendix 01).

Results and discussion

Patterns

This study considers relevant expenditure as the sum of payments for the main categories of healthcare, including private medical practices, medical and pharmacy products, private hospitals and nursing homes, medical laboratory services, and other facilities. Figure 2 depicts nominal monthly expenditure per household by main categories.

The largest portion of expenses has been for fees to private practices, accounting for 50% of household total out-of-pocket healthcare expenses on average. Payments for medical and pharmacy products and private hospitals have been among the major categories of total expenditure of households. Recently (Figure 2), payments for private hospitals and nursing homes, accounting for ∼20% of expenses, have become more important than payments for medical and pharmacy products. Payments for medical laboratory services have become more important over the period considered. All expenditure categories exhibit an increasing trend in monthly nominal value per household during the period 2001/2002 to 2009/2010.

Figures 3 and 4 provide more information on the utilization of private healthcare across household expenditure quintiles. The curve of utilization across expenditure quintiles takes an ‘inverted-U’ shape for both HIES 2006/2007 and HIES 2009/2010 data (Figure 3). There is no substantial difference between households in poorer quintiles and those in richer quintiles in burden as a percentage of out-of-pocket healthcare expenses in non-food expenditure. However, this explains that poorer households allocate relatively lower out-of-pocket amounts for healthcare than their richer counterparts in terms of absolute monetary value.

Figure 4 extends Figure 3 by combining items of healthcare. No substantial difference between poorer and richer households in terms of burden for private medical practices and pharmacy products is evident. Utilization of private services is relatively lower for poorer households when considering the items of consultation of specialists, private hospitals and laboratory services: for those, richer households allocate larger absolute monetary amounts by bearing a higher financial burden.

O’Donnell et al. (2008) confirmed that utilization of public services is more pro-poor in Sri Lanka than in any other emerging-market Asian country. They calculated concentration indices for outpatient services, inpatient services and public dispensaries. All were negative values (−0.1183, −0.2138 and −0.004, respectively), indicating that utilization of public sector facilities is higher in poorer households—that is, government healthcare in Sri Lanka performs better in equity.

We provide descriptive statistics relating to the percentage of households with positive expenditure across demographic and other profiles for HIES 2006/2007 and HIES 2009/2010 (Table 1) and summary statistics for households considered in this study (Table 2).

Table 1 shows that 69.1% of households incurred out-of-pocket healthcare expenses in 2006/2007, against 60.5% in 2009/2010. On average, 35% do not spend private money for healthcare. However, this percentage varies substantially across household categories, depending on household demography, geography and ability-to-pay. In households with elderly members, a relatively larger percentage had out-of-pocket expenditure compared with the households with no elderly members. As Table 1 shows, of households with no elderly members, 66.8% spent out-of-pocket for healthcare in 2006/2007. This increases to 74.8% for those with only one elderly member and to 81.8% for more than one elderly member. A similar pattern can be observed in 2009/2010, indicating that households with more elderly members have a relatively higher tendency to spend out-of-pocket expenses.

The percentage of households with positive out-of-pocket expenditure varies greatly with the number of household pre-school children. The percentage with positive out-of-pocket expenditure gradually increases when moving across categories of ‘households with no pre-school children’, ‘only one pre-school child’ and ‘more than one pre-school child’ for both surveys (Table 1), exhibiting a relatively higher tendency for households with more pre-school children to have out-of-pocket expenses.

The percentage of households with positive out-of-pocket expenditure varies across categories according to characteristics of the household head. Of households headed by educated members, a relatively larger portion had out-of-pocket healthcare payments (Table 1). Only 60.5% headed by members with no schooling had out-of-pocket spending according to HIES 2006/2007. This percentage increased to 66.5% for households headed by members with only primary education and to 71.0% for households headed by members with secondary or tertiary education. The same pattern is observable in HIES 2009/2010, suggesting that incurring out-of-pocket healthcare expenses is more observable among households headed by more literate members. Out-of-pocket expenditure is relatively more observable among households headed by members with no employment: on average, 72.7% had out-of-pocket payments in HIES 2006/2007, against 64.0% in HIES 2009/2010.

The percentage of households with positive out-of-pocket expenditure varies substantially depending on chronic illnesses at household level. Incurring expenditure is relatively less common among households headed by members with no chronic illnesses: 64.6% had expenses in HIES 2006/2007 and 55.3% in HIES 2009/2010, both below the overall averages, whereas households headed by members with chronic illnesses showed 81.7% in HIES 2006/2007 and 76.2% in HIES 2009/2010, well above the respective overall averages. Out-of-pocket expenses are more observable among households with multiple members with chronic illnesses. Among those with no members with chronic illnesses, only 59.4% had expenses in HIES 2006/2007 (below the overall average). This increases to 79.5% for those with only one member with chronic illnesses and to 86.8% for those with more than one such member. A similar trend was observable in HIES 2009/2010, indicating that healthcare spending out-of-pocket is more observable among those with more members with chronic illnesses.

A relatively larger percentage of households in richer income quintiles spend out-of-pocket. The percentage of households with positive payments gradually rises from 55.6% to 76.0% from the poorest income quintile to the richest in HIES 2006/2007. The percentage changes gradually from 45.7% to 72.6% in HIES 2009/2010. Spending out-of-pocket is more observable in urban than in estate and rural households. However, the tendency towards out-of-pocket payments does not vary greatly across urban, rural and estate sectors in HIES 2009/2010. Poverty in the estate sector is the highest of all three sectors but the highest number of poor persons is recorded from the rural sector (Department of Census and Statistics 2009).

Table 2 provides summary statistics of variables for out-of-pocket expenses. A household spends LKR 235.78 monthly out-of-pocket for healthcare on average per person, about a 21% increase from HIES 2006/2007. This represents 2.4% of average per capita total household expenditure or 5.5% of its per capita monthly non-food expenditure. Considerably larger standard deviations and wider ranges show that these figures are extremely volatile across households.

The presence of larger variability of household burden of out-of-pocket spending further encourages researchers to investigate what makes some households bear larger burdens, whereas others bear relatively smaller burdens.

Determinants

Table 3 shows the marginal effects on the burden of out-of-pocket spending derived from the Probit and Tobit models. The multicollinearity of the explanatory variables is explored by estimating the Variance Inflation Factor (VIF). All VIF values remain <2.5, indicating that multicollinearity is not an issue.

Our Tobit model, as a whole, fits significantly, because the P value of the F-test is approximately zero for both surveys. The P value of the Lagrange Multiplier (LM) test confirms that we do not reject the normality hypothesis. Moreover, we report standard errors of all the estimations as robust to heteroskedasticity. Therefore, in an environment where there is no considerable issue of multicollinearity, non-normality and heteroskedasticity, the Tobit model generates consistent estimators.

Table 4 provides the results based on alternative specifications for the burden of expenses and AES per capita household expenditure.

Household composition

As shown by the conditional and unconditional marginal effects of the Tobit model, out-of-pocket spending imposes a considerably higher burden on households with more than one elderly member compared to those with none. This is robust across the two surveys and alternative estimates which account for the impact of household food expenditure and household composition. Generally, elderly people require more frequent and expensive medical services (Minh et al. 2013), yet only one elderly member living in a household does not impose a significant impact either on the probability of incurring out-of-pocket expenditure or on the related burden according to HIES 2009/2010.

The probability of spending privately for healthcare and the related burden are more sensitive to the number of pre-school children than to the number of elderly people. Those with one pre-school child are more likely to spend out-of-pocket for healthcare and bear a higher burden compared with those with no pre-school children. Being robust across two surveys and alternative specifications, the marginal effects are even larger for households with more than one pre-school child, showing that the number of pre-school children is a stronger predictor of out-of-pocket payments than the number of elderly. Pre-school children might need more preventive healthcare and often experience early-age illnesses (Brown et al. 2014). Households with one or more school-age children are more likely to incur out-of-pocket expenditure than those with no school-age children. Therefore, the number of school-age children may predict the probability of incurring spending, but not steadily predict the burden of such spending. These findings raise an important social question as to whether parents spend more on children than on elderly relatives, which is not the focus of this article.

We have controlled for the impact of chronic illnesses of household members on the likelihood and the burden of healthcare. As predicted, results strongly confirm that the likelihood of incurring payments and the burden are higher when households consist of more members with chronic illnesses. As explained by the Ministry of Healthcare and Nutrition Sri Lanka (2009a), the main reasons for chronic illnesses, particularly for non-communicable disease, are unhealthy diet and physical inactivity. Therefore, national and local policies are essential to promote healthy eating habits and physical exercise.

Characteristics of household head

The majority of variables relating to household heads were insignificant in determining the probability and burden of expenditure. In both Probit and Tobit models, the dependent variables do not significantly vary according to gender, employment and condition of chronic illnesses of household head; even household head’s age has a negligible positive impact on burden.

However, Tables 3 and 4 demonstrate that literate household heads largely influence spending while bearing a higher burden. Heads with secondary education are positively associated with a higher probability of seeking healthcare privately by bearing a higher burden compared with those with no education. Relatively broader understanding of consequences of health hazards and higher tendency towards promoting good health practices among household members by literate heads might be reasons for this association.

Living sector

The living sector of people is a significant determinant of spending (Tables 3 and 4). We find that estate and rural households bear relatively a higher burden from spending compared with urban households. UNDP Sri Lanka (2012) elaborates that basic knowledge of health practices of rural and estate sector people is significantly lower than that of their urban counterparts. Hence, the former seek services when their illnesses develop to a more critical stage that primary and secondary care cannot treat. Ultimately, they have to depend on an urban-based public hospital or a private hospital for tertiary healthcare, including highly technical and advanced diagnosis and treatment, which require higher expenses. As pointed out by UNDP Sri Lanka (2012), usage of unsafe drinking water and unhygienic toilets is more common among rural and estate households, leading to a higher likelihood of illness. Therefore, enhancing the access and quality of related infrastructure will play a role in reducing the burden of out-of-pocket healthcare payments.

Utilization of Ayurveda¹ and traditional treatment is relatively more observable in rural Sri Lanka. Approximately 5% of rural households frequently use traditional and Ayurveda treatments. However, the total cost of a visit is higher compared to Western healthcare due to the high cost of drugs and other supplies (Weerasinghe and Fernando 2009). The main concern with Ayurveda and traditional healthcare is that there are many traditional and spiritual healers without formal qualifications due to the lack of a formal system to register practitioners (Weerasinghe and Fernando 2009): traditional treatments may make illnesses more critical, requiring higher expenses subsequently.

Household income

Household per capita income is a significant predictor of the likelihood of incurring out-of-pocket expenditure (Tables 3 and 4). Across alternative estimations, the conditional and unconditional marginal effects derived from the Tobit model based on HIES 2006/2007 are negative and significant, indicating that households with relatively lower per capita income are likely to bear a higher burden. However, the situation changed in 2009/2010, because the Tobit marginal effects have become positive and significant, indicating that higher per capita income households are more likely to bear a higher burden of expenses. However, these marginal effects are income elasticities of the out-of-pocket burden. Tables 3 and 4 also show that the conditional and unconditional income elasticities of out-of-pocket healthcare burden are less than unity, indicating that there is low flexibility in expenses in relation to income fluctuations, making private healthcare a ‘necessity’. This creates a relatively larger burden on poor households, which lack ways to finance health services.

This inelastic demand for expenses with respect to income implies a need for universal health insurance to protect poor and vulnerable households. Brazil has successfully expanded health insurance coverage and access for poor and vulnerable households through its family health strategy, reducing household burden (Couttolenc and Dmytraczenco 2013). Thailand has expanded health insurance coverage to its total population through three main public schemes, the universal coverage scheme, the social security scheme and the civil servant medical benefit scheme (Hanvoravongchai 2013). These can be considered case studies of universal health insurance adaptable for Sri Lanka, reducing out-of-pocket portion to a larger extent.

Supply side factors

The study discusses how supply side factors affect the likelihood of incurring private expenses and the associated burden. Tables 3 and 4 show that basic resources provided by government are strong predictors of out-of-pocket expenditure. Our results show that a shorter distance to the nearest government hospital leads to reduced likelihood and burden, while a shorter distance to the nearest private health centre leads to increased-likelihood and -burden. Our results further show that the number of inpatient beds in government hospitals, accounting for ∼93% of available bed capacity (Central Bank of Sri Lanka 2014), is a key determinant of the burden of expenses. This implies a need to expand bed capacity to lead to reduced-burden.

Similarly, dentists employed in government hospitals lessen the burden of expenses. Sri Lanka’s population desperately needs widespread dental facilities, as the majority have tooth decay and gum diseases (Silva and Gamage 2011). Approximately 10% visited government hospitals in 2007 and ‘51% of the 1.8 million dental visits per year were for dental extractions’ (Brailsford and De Silva 2015, p. 1568). Sri Lankans largely depend on dentists in government hospitals. But there are many unqualified dentists in the market: Silva and Gamage (2011) estimated the number of unqualified dentists as increasing by 9% between 2007 and 2012. Currently, there are 1942 dentists in government hospitals. Our results imply the need for increasing this number to reduce the burden on households. This further confirms that redistribution of basic government healthcare resources (hospitals, beds and dentists) will significantly lower the burden of out-of-pocket healthcare expenses.

Our results indicate that a larger number of doctors, including specialists employed in public hospitals, leads to a higher probability of incurring expenses and a larger burden. According to Govindaraj et al. (2014), the vast majority of private practitioners are doctors in the public sector. Approximately 30% of such officers work part-time in private hospitals (Central Bank of Sri Lanka 2011). Since 1977, the Government has allowed public sector doctors to undertake private practice outside their official hours. However, as the Ministry of Healthcare and Nutrition Sri Lanka (2009b) reports, the ‘dual practices’ of doctors being regulated and monitored loosely means that doctors tend to divert patients from the public to the private sector by manipulating service quality, which reduces access and efficiency of public health services (Hipgrave and Hort 2014). The health sector has been criticized for public medical doctors’ prioritizing their private channelling services (Silva 2012, cited in Thresia 2013).

Conclusion, limitations and future research agenda

Our findings show that the number of elderly members and pre-school children in a household are strong predictors of the probability and financial burden of encountering out-of-pocket healthcare expenditure. Expenditure is more sensitive to the number of pre-school children than to the number of elderly. Hence more attention needs to be paid to households with pre-school children and with more than one elderly member when formulating healthcare policies.

The study reveals that the burden of expenses does not vary substantially according to variation in income. Hence, such expenses can create considerable hardship and financial impoverishment, particularly in poor households. Poor households tend to obtain loans or to sell assets to cover out-of-pocket expenses (Leive and Xu 2008), implying more public sector involvement to better redistribute resources. Subsidized or free tertiary healthcare services targeting households with low per-capita income will enable them to utilize these with a lower burden. Including a ‘healthcare component’ in the current poverty alleviation programmes like ‘Samurdhi’ (a Government-funded social development programme) might reduce the burden borne by poor households. However, future research should empirically test the potential of such a proposal.

Policy intervention is required to protect households in rural and estate sectors from a relatively higher burden. Sri Lanka’s rural development policies should be flexible enough to integrate health concerns into other policy elements. This implies more policy intervention to ensure equitable distribution of resources, regardless of geographic location.

As the study robustly finds, more government hospitals, beds and dentists will significantly reduce the burden of household expenditure. Future research should take into account tax implications of further allocations for public health and equity concerns. However, provision of public hospitals and hospital beds through public-private-partnership programmes may reduce the burden of expenses without creating many tax implications. More government intervention is also required to reduce the burden induced by public doctors. Here regulatory policies ensuring private sector affordability and increasing public salary and incentives might play a role.

In the absence of panel data, this study has focused on patterns and determinants of the burden of expenses in households. Therefore, it has not accounted for panel properties, including household fixed and random effects. Also, the results do not represent the situation of areas affected by the civil conflict, because of missing data on important supply side variables. The calculation of marginal effects for the Probit and Tobit models is done only at the mean of each covariate.

Our study primarily elaborates associations between household financial burden of out-of-pocket healthcare expenses and demand-side and supply side factors. More empirical evidence is required to answer the question of why certain demand-side and supply side factors cause higher levels of burden for households. This study is more about explaining association than identifying causal mechanisms. A future study may combine qualitative evidence from social experiments and field surveys with quantitative evidence to establish such causal mechanisms.

The Government established the National Medicines Regulatory Authority Act in March, 2015, to ensure quality and affordable medicines. This benefits households through reduced out-of-pocket healthcare payments, because a considerable portion is spent for medical devices and drugs. Being the central regulator for these, the authority needs to evaluate the potential and comparative advantage of producing medicines and medical devices locally.

Since there is a widespread network of public hospitals, the rate of utilization of healthcare is satisfactorily high even among poor households. On average, a household can find a public hospital within a radius of 15 km. Also, the rate of utilization of public clinics is high. As explained by Saksena et al. (2010), in developing countries public healthcare facilities are more dominant in inpatient services and their proportion out of total volume exceeds that of private facilities. In Sri Lanka, the share of inpatient visits at public sector facilities has been >80%, of which the majority has been from households in lower-income quintiles. However, treatments by public hospitals and clinics include an out-of-pocket component which is generally regressive (Institute for Policy Studies 2012). From the public policy point of view, it is more valuable to analyse the interplay between utilization and financing mechanisms by considering out-of-pocket payments as a percentage of total healthcare expenditure.

Although Sri Lanka HIESs record out-of-pocket payments for healthcare comprehensively, they do not record complete information on household utilization of services. A future study should focus on how the incidence of payments depends on both healthcare utilization and financing. Also, public-private partnership is minimal in health and there is an urgent need for fostering a dialogue on opportunities for increased collaboration between these two sectors. This warrants fresh thinking on how to finance and deliver health services.

Acknowledgements

The authors wish to acknowledge that this article was made possible through the support and guidance provided by the ‘Australia Awards Fellowship Programme for Sri Lanka–2014’, funded by the Department of Foreign Affairs and Trade, Australia, and co-hosted by Monash University, Australia and the Sri Lanka Institute of Development Administration. Also, we would like to thank both reviewers of the journal for their insightful and helpful comments, which greatly enriched our article.

Conflict of interest statement.None declared.

References