The Determinants of Health Care Utilization in Ethiopia – Evidence from a Household Survey in four Regional States

(1)

University of Göttingen, Germany University of Groningen, The Netherlands

The Determinants of Health Care Utilization in Ethiopia –

Evidence from a Household Survey in four Regional States

Master Thesis

for attainment of the Double Degree in

Master of Arts International Economics

(University of Göttingen)

&

Master of Science International Economics and Business

(University of Groningen)

presented by

Franziska Clara Jaqueline Schünemann

Period of Work: 17 Weeks

Supervisors: Professor Stephan Klasen, Ph.D. (University of Göttingen) Professor Marcel Timmer, (University of Groningen) Handed in: 6th of July 2012

Student Number: 21012597 (Göttingen) S2238306(Groningen)

(2)

List of Figures

(5)

List of Tables

Table 4.1: List of selected Woredas ...46

Table 4.2: Reported illness in two survey sections ...47

Table 4.3: Use of health care in the conditional sample...48

Table 4.4: Provider of treatment in percent...51

Table 4.5: Treatment options...52

Table 4.6: Percentages of Ethnicities in the regional states ...54

Table 5.1: Satisfaction (perceived quality) with treatment in percentage ...58

Table 5.2: Distribution of treatment providers in the facilities in percent ...59

Table 5.3: Mean of user fees in Birr for each type of health care facility...60

Table 5.4: Mean of individual to facilities distance in kilometers ...60

Table 5.5: Reasons for not getting medical treatment ...61

Table 5.6: Reason for facility choice...62

Table 5.7: Health care use by age group and sex ...63

Table 5.8: Health care use by region, religion and official position ...64

Table 5.9: Health care use by type of education and newspaper use ...65

Table 5.10: Health beliefs and knowledge ...66

Table 5.11: Distribution of wealth by region and religion ...67

Table 5.12: Use of health care facilities sorted by wealth quartiles ...68

Table 5.13: Yearly individual consumption expenditure in Birr...69

Table 5.14: Mean of user fees paid...69

Table 5.15: Sources of health care financing ...70

Table 5.16: Estimated choice probabilities for health care facilities...73

Table 5.17: Results for test of IIA assumption...80

Table B.1: Description of the variables ...xi

Table B.2: Description of variables for sample selection bias test...xii

Table B.3: Multinomial logit estimates for health care provide choice ...xiv

Table B.4: Multinomial logit estimates with sample selection control ...xvi

(6)

List of Abbreviations

BPR Business Process Reengineering

CBHI Community-Based Health Insurance

CIA Central Intelligence Agency

CRDA Christian Relief and Development Association CSA Central Statistical Agency of Ethiopia

FAO Food and Agriculture Organization

FMOH Federal Ministry of Health of Ethiopia

GDP Gross Domestic Product

HSDP Health Sector Development Program

HSEP Health Services Extension Program

IIA Independence of Irrelevant Alternatives

NGO Non-Governmental Organization

OLS Ordinary Least Squares

UN United Nations

UNICEF United Nations Children’s Fund

USAID Unites States Agency for International Development

WHO World Health Organization

(7)

1. Introduction

The absence of economic growth in Sub Saharan Africa since the 1970s has been of major interest to development economists for the last 2 decades with varying conclusions about the reasons of this low growth, ranging from the low quality of institutions (cf. Acemoglu et al., 2001) over the high demographic burden (cf. Bloom & Sachs, 1998) to wrong economic policies and Sub Saharan Africa’s geographical disadvantages (cf. Sachs & Warner, 1997). While all of these issues are likely to have been influencing Sub Saharan Africa’s poor economic performance (cf. Collier & Gunning, 1999), one factor is related to many of these possible determinants - that is the disastrous health situation of the African people. Sub Saharan Africa’s tropical environment has led to a high disease burden that decreased labor productivity enormously and furthermore increased the dependency ratio and demographic burden through low life expectancies combined with high birth rates (cf. Bloom & Sachs, 1998, p. 211). Lower labor productivity and life expectancies exert a direct negative influence on economic growth through lower national income, while indirect effects like lower savings and lower investment in children due to income smoothing from health care expenditure inhibit the development of human capital and add to a further decrease in economic growth (cf. WHO, 2001, p. 30). Although there has been a lot of progress since the introduction of the Millennium Development Goals in 2000, the health situation in most African countries remains disastrous. E.g. 90% of all worldwide deaths from malaria still occur in Sub-Saharan Africa and despite considerable efforts from international organizations malaria incidence has been increasing in some African countries (cf. United Nations, 2011, pp. 42/3).

Ethiopia as one of the largest countries in Sub Saharan Africa in terms of population is no exception from this high disease burden and subsequent low economic development. In order to increase the health of the people the Ethiopian government has been reforming the supply side of its health system for the last 2 decades to increase quantity and quality of health care services. Nevertheless, user fees are still maintained in the country to ensure financing of the reforms that led to a potential health services coverage of about 90% in 2008/9, while the health care utilization rate of available services remains at only 30% per annum. (cf. FMOH, 2008/9; WHO, 2011)

(8)

planning to introduce health insurance consisting of a compulsory social insurance program for the formal sector and a community based health insurance (CBHI) for the urban informal sector and rural areas in the entire country, in order to increase utilization of health services by lowering the cost burden of medical treatment and through that to ensure financing of the health sector as well. (cf. Federal Democratic Republic of Ethiopia (a), 2010, pp. 23-25) However, a policy reform like introducing health insurance that aims at reducing the financial cost of health care use depends crucially on the factors determining the health care utilization. Health insurance can only be successful in increasing health care use and the health of the people if financial factors are indeed determinants of health care utilization. Whereas, if demographic and cultural factors determine health care use, health insurance is not likely to have a great effect on the utilization of health care services. Evidence on the determinants of health care utilization in developing countries remains mixed and highly context-specific so far, especially concerning the role of different individual characteristics like demographic and financial factors, while cultural factors like religion have rarely been examined with regard to health care utilization. Thus, it is extremely relevant to explore the determinants of health care use both from a scientific side, to better understand the importance of cultural and financial factors for health care use, and from a policy side, to assess why a policy measure is expedient or why it is likely to fail. Therefore, a household survey that was conducted in rural areas of four Ethiopian regional states prior to the introduction of a community based health insurance pilot project is used to empirically identify the determinants of health care utilization in Ethiopia. The empirical investigation finds that financial factors are indeed decisive for the decision to use health care and that a policy measure like health insurance that influences the financial side of health care use is likely to increase health care utilization in Ethiopia. On the other hand, differences in culture like religion, ethnicity and regional differences are also important determinants of health care use and should be taken into account when designing health insurance for different groups.

(9)

2. Background

In many respects, Ethiopia is a prime example for a Sub Saharan African country with low economic development and a high disease burden, so that successful reforms could be role models for other countries in the region. To better understand Ethiopia’s situation and the importance of a policy measure that increases health care utilization and secures a financial basis for the health sector as well, this chapter gives an overview of the country’s macroeconomic and health situation and about its health sector and policy reforms.

2.1 Overview and facts about Ethiopia

Ethiopia is a country of contrast and variability. Although it is one of the oldest countries in the world Africa’s oldest independent country, Ethiopia ranks on place 209 of 226 countries in the world in the CIA World Factbook’s ranking of GDP per capita (PPP) with an annual GDP per capita of only 1100 US$ in 2011 and is thus one of the poorest countries in the world. Agriculture is the biggest sector amounting to 41% of GDP and 85% of employment, but suffers extremely from a variable climate through frequent draughts and irregular rainfall, leading to very low productivity. Nevertheless, the country’s real GDP is now growing with about 7,5% per annum, though starting from a very low basis and with an inflation rate over 30% for its currency Birr (data from 2011, cf. CIA, The World Factbook, 2011).

(10)

With about 94 million inhabitants, Ethiopia also amounts to the second biggest country in Africa after Nigeria. In 2005, the total fertility rate amounted to 5.4 children per woman, while population growth has increased from 2.6% per annum in 2008 to an estimated growth rate of 3.2% in 2012 (cf. CIA, The World Factbook, 2011). About 84% of the population live in rural areas, while almost 50% of Ethiopia’s population is younger than 15 years old and a median age of about 17 years clearly indicates a high demographic burden as well as a dependency ratio (ratio of population below 15 and above 64 to the working age population between 15 and 64) of 81,17 in 2010 (cf. Index Mundi, 2011).

The country’s population exhibits also a high ethnical diversity with about 70 ethnicities. The largest are the Oromo with 34.5% and the Amhara with 26.9% of the population. Amharic is the official language. About 63% of Ethiopians are of Christian belief (mainly orthodox) and 34% are Muslims (data from the 2007 census, cf. CIA, The World Factbook, 2011). Eventually, educational attainment remains extremely low, with only 62% of males being able to read and only 36% of women in 2009. (cf. Federal Democratic Republic of Ethiopia (a), 2010, p. 3)

2.2 Health situation

Although many health outcomes have been improved in the last 2 decades, the general health situation in Ethiopia remains rather devastating. The latest Human Development Report of 2011 ranks Ethiopia 174th of 187 countries in the human development index (cf. UNDP, 2011). Life expectancy at birth is currently around 56 years compared to about 80 for the European Union (cf. CIA, The World Factbook, 2011). In rural areas, 64% of the population still use unimproved and often dirty water in the form of surface water or unprotected springs and more than 50% of sanitation is still open defecation (cf. data from 2010, WHO & UNICEF JMP for Water Supply and Sanitation data, 2010).

Ethiopia ranks among the top 15 of high infant mortality countries, which are exclusively to be found in Sub Saharan Africa (cf. CIA, The World Factbook, 2011). More than 90% of these deaths are the result of pneumonia, diarrhea, malaria, neonatal problems, malnutrition and HIV/AIDS, or even a combination of these conditions (cf. Federal Democratic Republic of Ethiopia(a), 2010, p. 3). High maternal death aggravate this problem with 676 deaths per 100,000 live births due to low levels of antenatal care (only about 34% of women) and even lower rates of postnatal check-ups (about 7%) (cf. CSA, 2012, p. 8).

(11)

and a weak referral system at the primary level as well as in demand side hindrances through cultural norms and especially financial barriers (cf. Federal Democratic Republic of Ethiopia (a), 2010, p. 3). These barriers are very likely to impede general health care utilization enormously as well, although about 60-80% of the prevalent diseases would be communicably preventable: As already mentioned above, Ethiopia suffers the fate of a tropical country and exhibits therefore an extremely high malaria prevalence with almost 75% of the land being malarious. About 9 million cases of malaria are being reported annually (average for 2001-5) and the disease was reported to be the major cause of death and morbidity in 2004/5, although the disease can be cured through the intake of drugs. As in many other African countries, tuberculosis is on the rise with an incidence rate of 378 per 100.000 inhabitants in 2009 and likely to increase with the prevalence of HIV/AIDS. (cf. USAID, 2009)

Though, in comparison with many other Sub-Saharan countries, HIV prevalence remains rather low in Ethiopia until now. Only 1.5% of women and men between 15 and 49 have been recorded HIV positive in the demographic and health survey of 2011, while this figure has almost not increased since 2005. Moreover, rural women are even less affected since the HIV prevalence among urban living women is six and a half times higher (5.2%) than among women living in rural areas (0.8%), which also indicates HIV is mainly of problem of the wealthier part of the population. (cf. CSA, 2012, p. 13)

In reality, even if death is not the immediate result of many diseases in Ethiopia, the high disease environment has vastly impeded the human and economic development in Ethiopia as mentioned in the introduction. Therefore, the Ethiopian government with the help of donors and NGOs has been trying to improve the country’s health situation by reforming the Ethiopian health system in order to increase the use of health care. The measures that have been taken as well as their outcomes are examined below.

2.3 The Ethiopian Health system

(12)

(HSDP) in order to increase overall efficiency, increase the focus of the community and the lower level tiers through decentralization and to ensure the access to health for all population segments. Further policy objectives are the development of a health care system with preventive, promotive and curative components with the participation of the private sector and NGOs. Essentially, the community is the priority of the health system and comes always first. Therefore, under the HSDPs, the number of primary health facilities in the community has been increased quite substantially together with the quality of health personnel through training and deployment of new health workforce. Concerning primary health care, the number of health centers has been increased in the course of three HSDPs from 412 in 1996/7 to 2689 in 2010, while it has been possible to enhance the provision of health posts from only 76 in 1996/7 to 14,416 in 2010 and equip most of them with better facilities for emergency and antenatal care. Eventually, the number of public hospitals has been increased from 82 to even 111. Apart from public health care facilities, private-non-profit NGO clinics amount to 277 and private-for-profit clinics number 1788. (cf. Federal Democratic Republic of Ethiopia (a), 2010, pp. 16-18)

Moreover, in 2003, a special program as part of the HSDPs was launched to ameliorate the above-mentioned health situation of children and mothers. The community-based Health Services Extension program (HSEP) assigned health extension workers (mainly females, who were selected by their communities and subsequently trained) to directly attend villages and provide health posts with preventive and promotive health care. The aim was to increase access and utilization of services for children and mothers apart from the health care facilities. Admassie et al. (2009) find that this program has significantly increased child immunization and the use of insecticide treated mosquito nets as a preventive measure against malaria. (cf. Admassie et al., 2009)

However, a big problem is that the number of physician remains extremely low, since the Ethiopian government has been neglecting the education of physicians for a long time. The country exhibits only 0.03 doctors per 1000. However, the number of medical schools has been increased from three in 2003 to fourteen in 2011 and is supposed to solve the problem of too few physicians in the future. (cf. Feysia et al., 2012, pp. 6/11)

(13)

covering 3,000-5,000 people. All three facility types are connected through a referral system. Furthermore, a general hospital encompasses the second tier with a coverage of 1-1.5 million people and a specialized hospital at the third tier covers 3.5.-5 million people. (cf. Federal Democratic Republic of Ethiopia (a), 2010, p. 4)

This development led to an estimated health services coverage of an astounding 90% in 2008/9, while annual utilization per person in the same period stayed at only 30%, making the reform appear rather unsuccessful in terms of health care use (cf. FMOH, 2008/9; WHO, 2011). This low utilization of health care services could be either due to low availability and quality of services, which would imply that the problem lies on the supply side. However, as just illustrated and according to the Ethiopian Health Ministry, both the number and the quality of services have been increased rather substantially. This indicates that the low utilization of services is a problem of the demand side, where low demand could be either due to cultural reasons or monetary reasons or even both, while the latter is closely related to the problem of financing the health care system.

2.3.1 The problem of financing the health sector

The question of financing of these promising reforms remains extremely pressing both from the supply and the demand side. The country’s health sector is not only financed by the government at several administrative levels (about 21%), but also by international donors and NGOs (40%) and to a large degree through out-of-pocket spending of the households (37%) in the form of user fees, while health insurance is basically non existent (cf. Federal Democratic Republic of Ethiopia (b), 2010, p. 3).

The annual per capita health spending of the government increased form 7.1 US$ in 2004/5 to 16.1 US$ in 2007/8, but user fees for health care are still maintained as a means of financing the health care system. Following the Bamako initiative1 in 1987, many countries in Sub Saharan Africa had implemented user fees to finance their ailing health systems and to provide basic health care for all. User fees were not seen as a problem for utilization as health care demand was assumed to be price-inelastic in developing countries (cf. Deolalikar, 1998, p. 93). Nowadays however, these user fees in combination with extreme poverty of the rural population through low and highly seasonal income are blamed for the remarkable under-utilization of health services (cf. Asfaw, 2003, p. 3), leading to low demand in the light of the !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

(14)

rather high service coverage. However, the empirical evidence on the price elasticity of the demand for health care in developing countries is mixed. Akin et al. (1986) for example found no effect of prices and income on health care use and the general finding is a rather inelastic demand for health care in general (cf. e.g. Sauerborn et al., 1994; Akin et al., 1998 & Li, 1996). On the other hand, many studies observe that health care demand elasticities vary for different income groups, while being rather elastic for the poorer quintiles (cf. e.g. Gertler et al., 1987; Sauerborn et al., 1994 & Sahn et al., 2002). For Ethiopia, Asfaw et al. (2004) find a crowding out effect of user fees for especially the poorest, thereby impeding their access to health care enormously.

In order to solve both the financing problem of the health sector and the issue of access to health care the Ethiopian government is now starting to introduce health insurance in their country. This planned Ethiopian health insurance system is a two-part health care financing strategy, which consists of a compulsory social health insurance program for employees of the formal sector and a community based health insurance (CBHI) for employees in the urban informal sector as well as targeted rural areas. The CBHI has started in 2010 with a two-year pilot scheme in 4 regional states (Tigray, Amhara, Oromiya & SNNPR) with the plan to scale the scheme up nationwide. (cf. Federal Democratic Republic of Ethiopia (a), 2010, pp. 23-25)

2.3.2 Community based health insurance

A community-based health insurance is a very promising concept in rural areas, since it pools the risks of its members like every formal insurance, but is based on the village level and not the formal sector. Furthermore, CBHI relies on strong social bonds and mutual trust, which reduces information asymmetries, so that premiums can mostly be set accordingly to the members ability to pay (cf. Creese & Bennet, 1997, pp. 166-168). In this way, community based insurance is able to reach the low-income population in rural areas as well as those working in the informal sector or being self-employed, thereby increasing the access to curative and preventive health care measures through reducing the financial burden of illness for a large part of the population in many developing countries, while many studies find indeed evidence for an increase in health care utilization through CBHI (cf. Jütting, 2003; Gnawali et al., 2009; Saksenaa et al., 2011).

(15)

indicates that there is indeed a scope for CBHI. Moreover, Asfaw (2003) and Asfaw and von Braun (2005) have both tried to assess the prospects of CBHI by looking at the willingness to pay for CBHI in rural areas of Ethiopia and by calculating possible revenues of CBHI for the Ethiopian health system. Both studies find that households in their surveys are generally quite enthusiastic about CBHI and would pay up to 3.5% of their monthly income for health insurance, while universal coverage of CBHI would generate about 75 million US$. This number is remarkable, as the Ethiopian government spent only 15 million US$ on health in 2007/8 as mentioned above.

However, these studies did not examine the prospects of CBHI in terms of increasing use of health care in Ethiopia, which is the main object of the introduction of CBHI in Ethiopia as already mentioned above. Only if individuals really use more health care services can their health status improve and more revenues be collected to support the Ethiopian health system. Essentially, whether health insurance will be successful in increasing access to and utilization of health services depends on the factors, which determine the utilization of services, and whether these factors can be influenced by health insurance. If financial factors like income and user fees determine the use of health care, insurance could solve these problems enormously. On the other hand, if cultural reasons like religion or regional characteristics prevent people from using modern health care, alleviating the financial pressures of health care use might not have a large effect on overall utilization.

(16)

3. Theory and evidence on health care utilization

Although the Ethiopian government has been trying to enhance the utilization of health care services through substantial supply side measures for almost two decades with a potential service coverage of modern health care of almost 90% in 2008/9, utilization did not increase at the same pace and still remains at only 30% of the supply (cf. FMOH, 2008/9). Therefore, it is decisive to answer the question what actually determines health care use and whether these determining factors can be influenced to increase utilization of health services. Due to the increase at the supply side, the problem of low utilization in Ethiopia is likely to be a demand side problem and is mostly attributed to a combination of extreme poverty and the existence of user fees so that the majority of the population is not able to afford modern medical treatment (cf. Asfaw, 2003, p. 3). If this hypothesis is true, then a health insurance that would alleviate the cost burden could increase utilization of health services enormously. However, health care use can be influenced by many different factors and to assess the prospects of a CBHI in Ethiopia in terms of increasing health care use, it is essential to examine whether financial reasons really are determinants in the use of health care utilization, since the CBHI is only able to address these financial problems. To establish the theoretical basis for the empirical analysis, this chapter first defines the term health care utilization and then looks at the theory concerning its determinants. This is followed by an examination of the evidence on the determinants of health care utilization.

3.1 Definition of the term health care utilization

Before looking at the theory and evidence about health care utilization, it is essential to distinguish between three different concepts pertaining to the use of medical services, i.e. need, demand and utilization, since these concepts have been muddled in the health care use literature.

(17)

area without taking into account that demand might be much lower due to travel and opportunity costs for waiting time. (cf. ADB, 2000, p. 27)

Demand itself is an economic concept that defines which quantity or amount of a good, e.g., a

certain kind of medicine, or medical service an individual is willing to buy and consume for a certain price, given a range of other factors like prices of alternatives and substitutes, the individual’s income and its preferences. Essentially, the individual must actually be willing and able to purchase this quantity. Yet, only because an individual has a demand for a service and would like to purchase a certain quantity of it for a certain price does not mean he will be able to do so, since the demand he can realize is dependent on its supply. This means that some individuals will only be able to realize their demand when there is a supply for that health service where they live for a certain price within their willingness and ability to pay. This leads to unfulfilled demand of some individuals due to the lack of accessibility and affordability. Decisively, utilization of health services can be seen as the “realized demand” of those individuals who were able to purchase some service, and denotes actual use of health care services, while it can indeed be different from demand and need. Thus, there can be much more individuals who would like to get, i.e. who demand medical treatment than those who really use medical services. While demand tries to explain how much of what kind of health care an individual would like to use for a certain price if there was a supply for it, utilization assesses how the medical services available to the individual are actually used, especially the question what determines that an individual chooses to use one health care facility over another or even no facility at all. (cf. ADB, 2000, pp. 27/28)

This means that health care utilization is the outcome of the interplay of demand and supply as shown in the Figure 1, which depicts the determinants of health utilization for an industrial country (cf. Layte (ed.), 2009, p. 3). Figure 1 shows that the use of health care is a complex process. Although this process differs slightly for a developing country, the basic drivers are the same.

!

(18)

The assessment of health care utilization and demand is completely distinguished in the economic literature, where studies either look at demand or at utilization of health care, using different models of demand and utilization. This is somewhat misleading, since especially in developing countries it is not possible to measure the actual demand curve empirically due to a lack of supply.

Even if different observations on price and quantity of a service across different locations are available, they cannot be assembled to one large demand curve. This is because the observations are outcomes of the intersection of demand and supply for a specific location at one point in time. Therefore, these observations depend on local and probably temporary demand curves, which are not very likely to equal the actual demand curve. On the other hand, utilization as the “realized demand” can be a good approximation for demand, in particular if many health care options and prices are available, which make the estimation of a local demand curve possible. (cf. Cullis & West, 1979, pp. 97/98)

To really capture demand, it would be necessary to obtain data by asking individuals what kind and how much health services they would use for a range of different prices. Most studies on health care demand however use data from household surveys, where individuals were asked which kind of health care service they used when they were sick, i.e., their realized demand, which is basically nothing else as health care utilization. Following figure 1, health care utilization is the result of an interactive process of demand and supply, which are both the result of interactions of other factors. The general problem is that health care is also in developing countries a public good, since it is at least to some extend offered by the state and not only private providers, which implies that there is no market in its economic sense where supply and demand determine the quantity and market price (cf. Williams, 1977, pp. 308-316). Thus the supply side can basically be seen as exogenous. Even if there are private clinics, which probably have been established due to a kind of market or demand for their service, the majority of health care services in the Ethiopian case is not the result of a market interaction, but of the provision by the government in its effort to satisfy need (cf. Asfaw, 2003, p. 29).

(19)

utilization cannot be distinguished analytically from each other, as individuals pass through the same decision process when supply is given exogenously in both cases. Therefore, the theory on the determinants of health care use illustrated in the next section applies to both concepts of utilization and demand for health care.

3.2 Theory on health care utilization

The theory on health care use tries to explore theoretically why sick people use a certain kind of medical care or even no medical care at all. Different sciences have used different frameworks and models and two broad categories can be distinguished (cf. Kroeger, 1983, pp. 147/8): The anthropological kind of models are pathway models which look at different stages in the behavior of sick individuals and their process of decision-making, however by using predominantly qualitative data. The idea is that health seeking is part of a process of usually five steps of decision-making, through which individuals pass during their illness starting with the recognition of symptoms (1), acceptance of the sick role (2), choosing consultation or medical care (3), choosing treatment or acceptance of the patient role (4) and finally adherence to the treatment or recovery (5). The last 3 steps of the process differ slightly according to the different models, but generally all denote the way in which individuals make a decision which kind of medical care to seek and to adhere to. Every stage is seen as a transition point and the individual decides about his own behavior, which is determined by social and cultural norms and the past experiences with medical treatment. (cf. Suchman, 1965; Chrisman, 1980)

(20)

3.2.1 A behavioral model of health care use

(21)

thus a measure of perceived health status should be preferred. The more needy a person considers himself, the more he should use health care. (cf. Andersen, 1995, pp. 1-3)

With these three types of determinants of health care utilization, Andersen’s behavioral model provides several measures of access to health care: While the enabling resources provide a good measure for potential access, it is the actual utilization of health care determined by all three explanatory variable types that can measure the effectively realized access. Whether there is indeed equitable access to health care must be assessed through looking at the relative importance of the explanatory variables concerning the variance in realized access. Equitable access is most likely if demographic factors like age and need proxies like the health status have the biggest effect on actual use and account for the biggest part in its variability, while a dominance of the social structure variables like belonging to a certain ethnic group and of enabling resources like income and assets would indicate a rather inequitable access. In the case of inequitable access, which is most likely in many developing countries, it is necessary to recognize which determinants of health care use are mutable in order to be changed through appropriate policy measures. The mutability of predisposing factors is considered rather low. Obviously, it is not possible to change demographic variables like age or gender, while it will probably take a long time or is even impossible for policy to influence the social structure variables and health beliefs concerning education or ethnicity respectively. Decisively, most enabling factors can indeed be changed in the short run through the right policies, e.g. income, the amount of assets, the existence of insurance and the information an individual receives, which are so called policy variables. (cf. Andersen & Newman, 1973, pp. 20-24)

The ideas of inequitable access and mutability of some determinants of health care utilization are essential for the hypothesis of this thesis: Only if the enabling factors, that is namely the financial variables like income and assets are indeed determinants of variation in health care utilization and hence show inequitable access to health care can the introduction of a community based health insurance as a policy measure influence the levels of health care utilization and enhance equitable access.

(22)

to health care like the waiting time and structure of the system once the individual has entered, e.g., the referral system. (cf. Andersen & Newman, 1973, pp. 6-8)

The model was further complemented by including a measure of health outcome to assess the effectiveness of health care use (cf. Andersen, 1995, p. 8). However especially in the case of developing countries, it is extremely costly to monitor individuals over their illness and recovery so that health outcome data is usually not existent. Altogether Andersen’s behavioral model was developed for both health care utilization in industrial and developing countries, e.g., Buor (2002) and Lopez-Cevallos and Chi (2010) both apply Andersen’s model to the use of health care in Ghana and Ecuador respectively. Yet, a slightly modified model concerning the utilization in developing countries following Kroeger (1983) is more specific concerning health care use in rural societies of developing countries, which was used by Elo (1992), Develay et al. (1996) and Amaghionyeodiwe (2008) to analyze health care utilization in Peru, Burkina Faso and Nigeria respectively, and is described in the following.

3.2.2 Determinants model for developing countries

Most developing countries are in some form of transition, so that the variables that are potential determinants of health care use must be seen on a somewhat wider range. This is because many factors in these countries are continuously changing, i.e. the socio-economic factors change through economic growth and urbanization, which raise income and economic opportunities on the one hand and destroy traditions and social ties on the other. Although increasing contact with Western culture can evoke changes in the perception of illness and health seeking behavior, they can also simultaneously evoke fear and rather rejection of modern medicine, leading to the coexistence of modern and traditional health care and thus a wider range of health services than in industrial countries. Kroeger (1983) follows the International Collaborative Study of Medical Care Utilization2 (1976, pp. 12-21) in classifying three categories of determinants of health care use for developing countries, which are heavily based on Andersen but appear more plausible in the case of rather different health care facilities as in developing countries. The model is illustrated in Figure 2 below.

Here the health system alone is not seen as a determinants category of health care utilization behavior but rather the characteristics of the different health care services (1), which consist of all variables related to the chosen health care facility including both the health system factors as well as most of the enabling factors, and amount to cost of treatment, the !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

(23)

individual’s accessibility and the quality of treatment. The second category of determinants comprises the characteristics of the individual’s illness (2), whether the disease is chronic or acute on the one hand and whether it is severe or rather trivial like a minor injury on the other hand. Eventually, the third category encompasses all individual characteristics (3) apart from those related to the chosen health care facility characteristics in (1). Kroeger summarizes them all under the term predisposing factors, while this encompasses apart from the above-described demographic factors, social structure and health beliefs also one of Andersen’s enabling factors, i.e. income. Importantly, special individual characteristics of health care users of developing countries are added to the model. First of all agriculture is the most important sector in most developing countries with the rural population usually outnumbering the urban population and many families living from subsistence economy, so that one cannot speak of a regular measurable income. Hence, consumption expenditure or assets like land and livestock are likely to shed a better or at least additional light on the socio-economic situation of individuals than some kind of cash income. Simultaneously, rural societies exhibit different kinds of social interaction than in industrial countries and rely much more on family and tribal ties and often consult family and friends in case of illness, while the roles of the different household members are much more pronounced, e.g., the household head decides over the other members. (cf. Kroeger, 1983)

The influence of these social interactions could be measured by the level of community participation, e.g. the membership in a communal organization or the relationship of the individual towards the household head.

(24)

curative measures for common diseases on the communal level3 (cf. Declaration of Alma-Ata, 1978). After consultation on the primary level usually follows referral to more specialized practitioners for more severe diseases at the secondary, e.g. a hospital, and tertiary health care level, e.g., a specialized clinic, where also inpatient treatment is possible (cf. John Hopkins medicine, 2012).

!

Figure 2: Determinants model for developing countries (source: own composition)

!

This distinction between types and purposes or levels of health care is essential in determining why individuals use one type of health care facility over the other. The determinants for health care utilization are likely to vary enormously between the different types, as secondary and tertiary health care might only be used in very severe cases or private facilities are more expensive and thus used only by richer individuals or traditional health care is preferred by older people. Therefore, it is necessary to distinguish soundly between different facility types when analyzing the determinants of their use. Hence, this examination of the model on the determinants of health care utilization has shown some of the caveats that empirical research !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

(25)

and policy has to take into account when looking at health care use in developing countries. How theory was applied in the literature and what evidence on health care utilization in developing countries was found is examined in the next section.

3.3 Literature on health care utilization in developing countries

(26)

This review looks therefore at the evidence from studies that assessed determinants of health care use with the help of household data and quantitative methods. The starting point here is the methodology concerning the dependent utilization variable, which is examined in the next section.

3.3.1 Measurement of health care utilization

In the empirical literature the measurement of health care utilization is not a straightforward concept. Utilization can for example be defined as the number of visits to a medical facility or the individual’s expenditure for medical services and both types have been most common (cf. Aday, 1972; Diehr et al., 1999). In this case utilization is simply measured as a continuous variable with health care expenditure as a proxy for utilization as the dependent variable. However, this way of measurement applies basically only to the case of industrial countries where there is no question or choice of using medical care at all, but only of how much, since medical treatment is a part of western life, and it is thus also possible to measure quantitative variables of utilization. Moreover, even if expenditure levels are available in developing countries, these rarely mirror market prices since health care is usually subsidized or even for free (cf. Havemann & van der Berg, 2003, p. 4). Furthermore, mandatory referral systems in industrial countries make the choice of health care provider obsolete, since one usually has to attend some form of primary care and will then receive a referral to a higher care facility (e.g. for the EU: cf. European Commission, 2010, p. 115). Concerning health care utilization in most developing countries however, neither universal health systems nor regular health care utilization is encountered due to missing social security systems. Thus, individuals in developing countries face on the one hand the choice between using medical treatment or not in case they are sick and on the other hand the decision of which kind of medical facility to use. In this case, the determinants of some quantities of health care cannot be assessed, but rather the question, which variables influence in which way the probability that a specific health care facility is chosen, leading to the necessity of discrete choice models concerning the individual’s behavior and making health care utilization a qualitative and categorical choice variable.

Continuous variable

(27)

by Buor (2004), who looks exclusively at health care use of women in Ghana. In addition, a longitudinal study of health care utilization in Zaire applies ordinary least squares (OLS) regression to a utilization coefficient for each of 21 health centers, which is calculated monthly by dividing new treatment cases by population served by the center (cf. Haddad & Fournier, 1995). In a similar fashion, Baker and Liu (2006) compute an attendance index for each of three health care facilities over a period of one year in Honduras. Basically all of these studies estimate models with one regression for each health care provider respectively and thus cannot assess why an individual chooses to use one provider over another, since they model only the health care behavior concerning on health care facility at a time.

Qualitative choice

Besides examining the use of a particular health care facility, other studies analyze the decision to use medical treatment or not with a qualitative choice model, e.g. a binary probit or logit model, as it is done by Mekonnen & Mekonnen (2002) for a study in Ethiopia, by Navaneetham & Dharmalingam (2002) in Southern India and by Elo (1992) in Peru. All of these studies look at the use of maternal health care services, where it is of interest whether women used any kind of birth related health service or not. Thus, the advantage of binary choice models is that the determinants that make the choice of taking any kind of medical service over staying at home more likely can be easily assessed. Develay et al. (1996) for example use a binary logit model to examine the determinants of modern health care use in an urban area of Burkina Faso regardless from different facilities. The disadvantage of binary choice models is however the same as of the linear models, namely that they cannot model the decision for one health care facility over another. In the Ethiopian survey, individuals are faced with more than the two alternatives of medical treatment or staying at home, so that information on the determinants of choosing one particular health care facility would be lost by using a binary choice model.

(28)

issues are further discussed in chapter 4 below, when the methodology for the empirical analysis is assessed. The next section continues with examining the actual findings of different studies on determinants of health care use, while it is important to remain cautious when comparing the results of multivariate studies analyzing the use of alternative health care facilities. Types of health care facilities are not only likely to differ across continents and countries, but also across regions in quality, prices they charge and services they offer. Even if it is possible to control for some features of facilities through the number of technical personnel or prices, it is almost impossible to monitor which individual went to which special facility. Thus, findings about certain types of health care facilities are likely to be only rough averages as the facilities individuals really used could differ enormously even within one type of facility. Keeping these caveats in mind, one can better analyze the findings of different studies.

3.3.2 The determinants of health care use in the literature

The preceding section has illustrated the problem of simply comparing the findings of different studies on health care use. Apart from using different measurements of utilization, the explanatory factors included in the models and their measurements vary as well, depending on the data available. Therefore, findings can often only be seen as country- or context-specific. Nevertheless, general conclusions are drawn if possible and if it is not mentioned otherwise, only statistically significant result are considered. The review examines determinants following the model from figure 3.2 and starts with individual characteristics.

3.3.2.1 Individual Characteristics

Age

(29)

these results were estimated against the background of not choosing any kind of health care at all and thus cannot assess whether young people are more likely to use modern than traditional health care. Similarly, Chawla and Ellis (2000) find that older people are less likely to choose formal health care in comparison to informal health care.

Gertler et al. (1987) detect for Peru that higher age has a significant positive effect on the likelihood of using hospitals and private doctors, but a negative effect on clinic use. Amaghionyeodiwe (2008) and Lawson (2004) observe for Nigeria in a similar fashion that older people are more likely to attend public and private hospitals than clinics. In both papers, clinics provide some form of primary health care indicating that specialized services needed by older people are offered rather by hospitals and more specialized health care facilities. Thus age increases not only the likelihood to seek any kind of health care, but older people are also more likely to seek more specialized health care.

Sex & Relationship to household head

At first, a gender bias seems likely in many developing and rural settings, as women face usually more time constraints than men through working, raising children and looking after the household. On the other hand, women are likely to seek more health care than men on average due to pregnancy and maternal issues. Brown and Theoharides (2009) however do not find any significant effect of gender on health care use in China. This finding goes in line with Kroeger (1983, p. 148), who suggests that gender is only likely to affect health-seeking behavior in societies where the roles of gender are strictly defined, in particular in Islamic countries. Lawson (2004) estimates different regressions for men and women using health care in Uganda. He detects similarly to Mwabu et al. (1993) that women are more constrained by user fees and by income in using health care than men. This indicates that women’s autonomy indeed plays a role in using health care.

(30)

sex of the household head. Amaghionyeodiwe (2008) explores a positive effect for attending public facilities more than private facilities, both if the household head is male as well as female. Eventually, Sahn et al. (2002) notice that men are less likely than women to use any kind of health care facility, both private and public, while being married had a negative effect of seeking private health care for both women and men.

In sum, the assumption of a gender bias in health care use is not such a clear picture. Women are certainly in more need of health care due to pregnancies, but might not seek treatment for example out of cultural reasons. Women’s autonomy seems to be quite important in the health seeking behavior of women, but probably also in every other part of daily life.

Culture: Ethnicity & Religion, Social Interactions & Health Beliefs

Cultural variables are likely to play an enormously role in health care seeking behavior, as they shape beliefs, social relationship and social position. However, many of these cultural variables are extremely difficult to measure quantitatively and can only be assessed qualitatively. Moreover, it is also difficult to draw conclusions from one culture to another. Unfortunately, the fewest empirical studies on health care provider choice have looked at cultural variables in one way or other, although some cultural variables like ethnicity are easily controlled for.

Lopez-Cevallos and Chi (2010) for example explore the effect of ethnicity on health care use and find that indigenous individuals in Ecuador use significantly less of curative health care than descendants from Spanish immigrants. Moreover, rural individuals were less likely to be hospitalized. Lawson (2004) on the other hand detects significant regional (central-south-north) differences concerning the use of different health care providers in Uganda, but no urban-rural differences. De Allegri et al. (2011) report that women in Burkina Faso are more likely to deliver in a health care facility if they belong to a certain ethnicity, among them the largest ethnic group. Brown and Theoharides (2009) examine whether health care use of emigrants is different from local Chinese, but cannot claim a significant effect of being emigrant.

(31)

husbands were away. The author also reports that for some types of malaria, it is common to consult traditional healers and take herbs, as these are believed to cure the disease. On the other hand, the cost argument was overall the most important in the choice of medical care in that study. Cocks and Moeller (2002) explore the use of traditional medicine in South Africa and find that traditional medicine is rather seen as augmenting than replacing modern drugs. Individuals across all religions, income and education statuses purchased traditional medicine in order to avert evil spirits and invite good fortune. The WHO also reports, that traditional medicine is used in developing countries both due to its cheap price and its embeddedness in wider belief systems (cf. WHO, 2002).

These findings suggest that cultural beliefs play an important role and it is therefore necessary to further explore this issue in quantitative and qualitative research, especially concerning the choice of different health care providers.

Education

Education is one of the determinants of health care use that has received a lot of attention. People with better education can take better care of their health, as they have better knowledge about diseases and about staying healthy (cf. Folland et al., 1997, pp. 96/97). Moreover, low education fosters ignorance and superstition, leading to more use of traditional health care, while educated people are more likely to use modern health care (cf. Amaghionyeodiwe, 2008, p. 222). However, studies have found mixed results concerning the effect of education. Ichoku and Leibbrandt (2003) report a significant positive effect of average years of formal education in the household on health care use in Nigeria, which is contradicted by the study of Gertler and Van der Gaag (1990) in Cote d’Ivoire and by the study Sauerborn et al. (1994) for Burkina Faso that both do not find a significant effect of education. Ichoku and Leibbrandt (2003) argue that the insignificance of education in the Cote d’Ivoire case is due to the low level of education in their study area.

Some studies also use the education of the household head as a proxy for knowledge on health as an explanatory variable. Brown and Theoharides (2009) do not find a significant effect of the head’s education on health care use in China. Decisively, Frederickx (1998) detects a positive effect of household head education on the use of hospitals in Tanzania only for the richer individuals. She argues that the returns to education in areas with traditional agriculture are very low and thus investments in education might not have positive cross-effects on health there.

(32)

education in Nigeria makes the use of public and private hospitals more likely and the use of primary care less likely, which the author thinks could be due to the fact that the perceived quality of care is higher at hospitals. Moreover, Lawson (2004) notes for Uganda, that men and women with higher education are significantly less likely to use public health care, while Mwabu et al. (1993) report the opposite.

More findings concerning education can be listed in a similar fashion, since probably all studies on some kind of health care used added education or schooling as an explanatory variable in their empirical analysis, however with contradicting results. This could be linked to the fact that studies use different measures of education, sometimes for the individual, sometimes for the household head or mother. On the other hand, the effect of education on health care use appears to be rather context-specific and probably dependent on the general returns to education.

Income & consumption expenditure

The decisive role of income and assets in general in the use of health care in developing countries has been acknowledged by numerous studies. Not only is higher income likely to lead to better health through a better diet and way of living, but it severely influences the health care seeking behavior of individuals when they are ill. Usually health care is seen as a normal good, so that with increasing income, the demand for health care increases as well (cf. Gertler et al., 1987, p. 70 & Mwabu et al., 1993, p. 849). In many studies, income is proxied by household consumption expenditure, which is easier to collect in rural regions with mostly subsistence economy where some kind of cash income cannot be obtained. Chernichovsky & Meesook (1986) report that a 10% increase of monthly household expenditure would significantly increase monthly health care expenditure by 7% in Indonesia. Likewise, Ichoku and Leibbrandt (2003) detect a significant positive effect of higher total consumption expenditure on the likelihood to use health care in Nigeria. This means the poorer households are severely constrained in using health care, when their members are sick.

Concerning the choice of health care provider, Amaghionyeodiwe (2008) observes a significant effect of income on private care facilities, indicating that the richer individuals use more private than public health care as they consider private facilities of higher quality. Mwabu et al. (1993) also find that higher income increases the probability of seeking any kind of health care as opposed to self-care in Kenya enormously, while it is highly significant for private facilities. Tembon (1996) confirms these results for Cameroon.

(33)

many studies have found the demand for health care to be rather inelastic in general when examined over all income groups (e.g. Gertler & van der Gaag, 1990; Sauerborn et al., 1994; Akin et al., 1998). On the other hand basically all studies that analyze demand elasticities for different income groups observe that the poor income groups have a much more elastic demand for health care than the richer ones (cf. Gertler et al., 1987; Sahn et al., 2002; Asfaw et al., 2004). This indicates that those who are disadvantaged in any way are most affected by policies such as user fees and less likely to seek and receive treatment even if it is for free, as they have to pay transport costs for example.

These findings imply on the one hand, that poverty is one of the most important hindrances in the use of health care, and that policies that alleviate poverty and make health care use cheaper for the poorest can be decisive to increase health care utilization. On the other hand, some studies found income to be less important for health care use and it is therefore essential to subject the role of income to careful scrutiny.

3.3.2.2 Characteristics of the illness

In an equitable health care system, need and the type of illness from which an individual suffers should be the only factor in his decision to seek medical treatment and choice of medical provider. Unfortunately in many developing countries, poverty is likely to be as important for health care using behavior as need. Since it is probably not feasible to introduce dummy variables for every disease in a regression equation, unless the sample is very large with many individuals having suffered from the same diseases, the characteristics of the disease are usually added in terms of its severity and whether it is chronic or acute. The more severe the disease, the higher would be the expected likelihood for using health care. On the other hand, in the case of chronic illnesses self-treatment is more likely as a physician cannot do much to cure the disease. Information on illness characteristics evaluated by professionals is practically always missing in household surveys, as this would be much too costly to obtain. Another option is to ask individuals about their perceived illness characteristics, which is probably a very good way to obtain this information. Individuals are very likely to know whether their illness was chronic or acute, i.e., an emergency, and how many days they could not work or were sick to evaluate severity. Unfortunately, this was not done in many household surveys, so that the decisive illness variable is not included in many studies. Nevertheless some studies report quite the expected results.

(34)

treatment in Nigeria. Kroeger (1983, p. 150) notes that individuals with chronic diseases mostly attend traditional healers, since they might have already attended modern health care facilities without being cured. On the other hand he finds, in the case of severe diseases, modern health care facilities are more likely to be used. Develay et al. (1996) confirm this view: Individuals in urban Burkina Faso were more likely to use modern medicine, the more severe their illness was. Asfaw et al. (2004) detect a significant effect of severity of illness on using both hospitals and traditional healers in Ethiopia, which stands against Kroeger’s findings concerning traditional healers. It could well be that traditional healers are the closest and it is therefore the fastest way to get initial medical treatment in the case of emergencies. Gertler et al. (1987) observe that in Peru individuals with an acute illness (emergency) are more likely to attend hospitals and primary care than a private doctor, while respiratory illnesses decrease the likelihood of seeking any kind of treatment. Respiratory illnesses are likely to be of chronical nature, which explains the higher likelihood of self-treatment. The higher likelihood to attend public facilities in case of emergencies could be due to the fact that private doctors might be more specialized and only attended for certain diseases. Moreover, Amaghionyeodiwe (2008) finds for Nigeria an increase in the likelihood of using public care (both hospitals and clinics) and private clinics in an emergency, while individuals were less likely to use private hospitals, which could be due to the same explanation as in the Peruvian case.

As expected, basically all studies that used characteristics of illness as an explanatory variable found a significant effect on health care use. However, this does not mean that the countries have equitable health care systems, as need variables are usually not the only determinants of health care use, so that it is necessary to explore the role of all determinants for health care use.

3.3.2.3 Characteristics of the Health care facilities

(35)

difficult. However, Mwabu et al. (1993, p. 852) believe that treating supply as exogenous might lead to endogeneity problems, so that one should look at the results more cautious. Since provider choice is usually assessed through a discrete choice model, the measurement of these variables poses great problems. In the random utility maximization models of discrete choice, all individuals are presumed to know about all characteristics of every provider available to them, so that it is necessary to have information on every provider with regard to every individual in the sample. However, for each individual the only information about access variables usually available is about the health care facility that he actually used. To overcome this problem, different procedures have been used. Some studies were able to fill the missing information with data from surveys on health care providers (cf. Akin et al., 1986). Others have estimated hedonic pricing equations (cf. Gertler et al., 1987; Gertler & van der Gaag, 1990). This approach regresses the access variables of each health care facility on individual characteristics of the users. With the help of the resulting estimated coefficients, values for individuals with similar characteristics are predicted (cf. Asfaw, 2003, p. 115). The third approach is to estimate village means for each alternative from the information available and assign them to every individual without controlling for individual characteristics (cf. Li, 1996; Sahn et al., 2002; Asfaw, 2003). Asfaw at al. (2004) have further developed this last method by adjusting user fees for quality and different diseases. All three methods have their advantages and disadvantages. The first method is likely to fail in many cases due to a lack of data. The second method might fail in some circumstances, e.g., Asfaw (2003) observed unreasonable imputed prices after using the hedonic method. Estimating village means is probably the most feasible method and makes much sense when trying to find out the average distance to some health care provider as it is indeed the same for all villagers, but might be rather vague in the case of users fees. Measuring quality is a problem with every method, since it is very unlikely that all hospitals or all health posts are of the same quality, so that assigning some quality value to each provider type does not capture the quality differences within the different provider types. Unless it is recorded which particular facility of which type an individual used, it is not possible to adjust for this problem. Despite of these problems, access variables have been found to be extremely important for the decision on health care use and are discussed in the following.

Distance

(36)

transportation but also time. Distance can be measured as travel costs and might then be seen as part of user fees. However, distance is mostly measured as travel time in minutes or in kilometers. Gertler et al. (1987) find a highly significant negative effect of travel time on the use of all types of medical treatment (public hospital and primary care, private doctor) in Peru. Lawson (2004) can confirm this finding for Uganda: If modern health care facilities were within three kilometer reach, individuals were more likely to use them. Other studies that find a negative effect of distance on health care use are Frederickx (1998) for Tanzania, Mwabu et al. (1993) for Kenya, Chawla and Ellis (2000) for Niger and Amaghionyeodiwe (2008) for Nigeria. Similarly, Chernichovsky and Meesook (1986) observe that women are more likely to use modern services if they were available in their own village.

Tembon (1996) is one of the few who detects a positive effect of distance on health care use in Cameroon. The author ascribes this result to the phenomenon that in the analyzed district, individuals often used health care facilities that were further away due to other socio-cultural factors such as relatives living nearby a certain facility.

Quality

Although quality is probably one of the most important determinants of provider choice, it is also one of the most problematic in terms of empirical assessment and cannot really be measured if not obtained through health facility surveys or as perceived quality from household surveys (cf. Lavy & Germain, 1994). Usually, higher quality is also associated or correlated with higher user fees and therefore, many studies treat it as more or less unobservable. Gertler et al. (1987) for example assume that quality of health care facilities varies by region in Peru and therefore introduce regional dummies into their regression, for which no overall picture can be observed. However, simply introducing regional dummies as proxies for quality seems rather arbitrary and other studies have specifically examined facilities or obtained information from surveys.

(37)

Sahn et al. (2002) are able to use data on perceived doctor quality for all their health care facilities and on perceived drug and environment quality for their public non-hospital alternative in Tanzania. Perceived quality had a significant positive effect on the use of all facilities except for private hospitals (being insignificant), which indicates that perceived quality does play at least an important role as evaluated quality. Just like distance, quality offers a rather clear picture in its effect on the choice of health care facility.

User fees

The Determinants of Health Care Utilization in Ethiopia – Evidence from a Household Survey in four Regional States