Impact of offering health insurance on chronic illness related
expenditures in Kwara State, Nigeria
MSc thesis Nienke van den Bergh Student number: 5732069
October 2013
University of Amsterdam
Faculty of Economics and Business
MSc Economics, specialization Development Economics Supervisor: dhr. prof. dr. M.P. (Menno) Pradhan
Acknowledgements
This thesis functions as the final challenge to pass the Master’s degree in Development Economics. My first supervisor, Menno Pradhan, has been of great value to the development of my thesis and I like to thank him for that. In favor of me studying a very interesting topic for my thesis, he connected me with the Amsterdam Institute for International Development. During the whole process of writing he criticized the thesis in a very professional way and was always accurately responding to the situation. I am very grateful to the Amsterdam Institute for International Development and the Health Insurance Fund for giving access to the relevant data. Also, I would like to thank Alexander Boers, Daniëlla Brals and Zlata Tanovic for their useful contributions. Finally, I appreciate the critical comments of Jacques van der Gaag on the topic of the thesis in an early stage and I like to thank him for being the second supervisor.
Index
1. Introduction………...4
2. Chronic illness in developing countries………..….6
2.1 Communicable and noncommunicable chronic diseases………6
2.2 Prevalence of chronic diseases……….……….………7
2.3 Economic consequences………..…..8
2.3.1 Microeconomic level.………9
2.3.2 Macroeconomic level.……….10
3. Health insurance in developing countries……….…12
3.1 Health insurance……….………….12
3.2 Adverse selection……….…………...13
4. Case study: Health insurance program in Kwara State, Nigeria……….….14
4.1 The Nigerian context………..….14
4.1.1 Socioeconomic features ……….……..14
4.1.2 Chronic diseases ……….……….15
4.1.3 Healthcare system……….16
4.2 Supported by the Health Insurance Fund: a health insurance program…...16
4.3 Hypotheses………17
4.4 Data description……….……...17
4.5 The burden due to chronic illness………..…...…..20
4.5.1 Health status………..…….……..21
4.5.2 Utilization……….….23
4.5.3 Financial burden………....24
4.6 Impact evaluation………....……….26
4.6.1 The impact of the health insurance program on health expenditures and utilization……….………...26
4.6.2 The impact of the health insurance program on health status…....….28
4.6.3 Adverse selection in context...30
4.7 Discussion………..………….33
5. Conclusion………...35
Bibliography………...37
Appendix 1: Health Insurance Fund Study Concept Sheet & Analysis Plan………....39
Appendix 2: The Nigerian context – extension……….………...41
1. Introduction
Diseases are not a health problem only; they form a development challenge as well. Illnesses can cause individuals and their families to end up into poverty due to several costs that are related to being ill. This thesis focuses on chronic diseases in order to give attention to these ‘underestimated’ diseases. Chronic diseases are known among the public as a rich-world problem, but that is an outdated view. It is estimated that developing countries bear more than 80 percent of the deaths due to chronic diseases (World Health Organization, 2005). Chronic diseases are per definition long lasting. Chronic illnesses progress often very slowly, they are in most cases hardly visible and underdiagnosed. Even though the burden due to chronic diseases is high in developing countries, chronic diseases do not get the attention they ‘deserve’ among policy makers, economists or in academic literature.
This thesis gives a unique insight into the impact of offering health insurance on chronic illness related expenditures in Nigerian context by including a case study. The Health Insurance Fund (HIF) has set up a health insurance program in Kwara State in Nigeria, which provides low-cost health insurance to rural farming communities. HIF is dedicated ‘to protecting low-income families in Sub-Saharan Africa from health-related risks by providing access to affordable quality health care through the introduction of innovative financing mechanisms (including health insurance), and improvement of quality of health services.’ HIF is part of the PharmAcces Group that aims to improve access to quality health care for people in sub-Saharan Africa. The integrated approach consists of complementary initiatives that aim to increase resources, efficiency and effectiveness within the healthcare system. By combining standards for quality improvement, loans for healthcare providers, health plans and in-depth impact research the demand for and supply of healthcare services are simultaneously stimulated. The extensive household surveys, which were made available to the author by HIF in cooperation with the Amsterdam Institute for International Development (AIID), allow to evaluate the impact of offering this program to Nigerian people on chronic illness expenditures.
Formal health insurance is often lacking in developing countries. In contrast with many developed countries, poor countries do not have the resources to implement a social insurance scheme in which participation is compulsory. In practice this means that up-take of health insurance is low. These considerations altogether lead to the following research question: what is the impact of offering health insurance on chronic illness related
expenditures in Kwara State, Nigeria?
To understand the various costs that individuals and households experience due to chronic illnesses and to evaluate the impact of offering health insurance on these expenditures, this thesis contains an literature study with the following content: an analysis of the recent trends of chronic diseases in low- and middle-income countries; the economic consequences of chronic diseases in developing countries; and an analysis of the impact of health insurance in developing countries. The case study consists of an outline of the health insurance program and the sample, an estimation of the burden due to chronic illness at baseline and an impact evaluation of offering the program to the public on chronic illness related expenditures. To broaden the analysis of the impact of offering health insurance the case study will contain a descriptive analysis of the determinants of enrolment in the program afterwards. This analysis can make us aware of possible adverse selection that could influence the impact of private health insurance programs as insurance providers take this into account when setting their insurance structure and premiums. Also, the analysis can give information about the potential this program has as it consequently will have maximum impact when insurance up-take is as high as possible. The exploration of adverse selection on certain characteristics leads therefore to additional information with regard to the impact of offering health insurance on chronic illness related expenditures.
More and profound knowledge about health insurance programs in developing countries contributes to an important ambition: a well designed and implemented health insurance plan has the potential to protect people from health risks and to avoid people falling into poverty because of these risks.
The first chapters provide a thorough literature study and the second part of the thesis is concerned with the empirical findings of the Nigerian case study, ending with a discussion and the conclusion.
2. Chronic illness in developing countries
This chapter first makes the distinction between communicable and noncommunicable chronic diseases and then analyses the prevalence of chronic diseases in low- and middle-income countries. The last section describes the economic consequences of chronic diseases.
2.1 Communicable and noncommunicable chronic diseases
Chronic diseases are defined by the fact that the illness is long lasting, that is more than 6 months. In general these diseases progress slowly. In the group of chronic diseases the noncommunicable diseases (NCDs) can be separated from communicable diseases because of their non-infectious causes.
Long lasting communicable diseases such as HIV/AIDS or tuberculosis (TB) can also be characterized as chronic diseases. In some studies (Nugent, 2008; Suhrcke et al., 2006) ‘chronic disease’ is a synonym for NCDs, thus ignoring the communicable chronic diseases. However, it is because of the (recently) improved medication that for example HIV/AIDS has become a severer chronic disease, as people do not have to die (young) from it anymore. People live longer because of the medication, which implies they have to deal more years with the chronic illness. An infection with HIV/AIDS results in a worsening deterioration of the immune system, breaking down the body's ability to fend off some infections and other diseases. Ongoing research to medicines has already resulted in therapy that prevents the HIV virus from multiplying in the body and in medication that prevents the mother-to-child transmission of the virus well. The force back of high-impact diseases such as HIV/AIDS and TB is included in the Millennium Development Goals (United Nations). However, access to preventive interventions remains limited in most low- and middle-income countries.
Noncommunicable diseases are often characterized as ‘diseases of affluence’ or as problems affecting only the elderly retired population. This is not the case anymore. ‘To the extent that those notions have been common, they may have been responsible for a lack of recognition among economic policymakers of chronic disease as an issue of potential public-policy relevance’ (Suhrcke et al., 2006). NCDs are often hardly visible, develop slowly and underdiagnosed. NCDs such as diabetes, cancers and few of the CVDs are often labeled as ‘lifestyle diseases’ implying that people adopt unhealthy habits from personal preference. Whether these choices are complete voluntarily is a complex issue and still being investigated (Nugent, 2008). Unhealthy and risky behavior remains high worldwide and is increasing in
the majority of low- and middle-income countries (World Health Organization, 2013). Many NCDs can be prevented through more healthy behavior such as no tobacco use, a healthy diet, physical daily activity and limited consumption of alcohol. Healthy behavior decreases also the chance to develop communicable diseases. For example, risk factors such as an unhealthy diet and smoking increase the chance of developing active TB and falling ill. Diseases can provoke other illnesses as well. HIV, for instance, is the strongest risk factor for developing active TB disease. In 2011 approximately 25 percent of all deaths from HIV occurred among people suffering from tuberculosis as well. This confirms Nugent’s statement (2008) that poor healthcare in itself is an enormous risk factor.
2.2 Prevalence of chronic diseases
Noncommunicable diseases are by far the leading cause of death in the world, representing 63 percent of all annual deaths. NCDs kill more than 36 million people each year of which the great majority (80 percent) of deaths due to any NCD occurs in low- and middle-income countries. The cardiovascular diseases (CVDs), which are a group of disorders of the heart and blood vessels, are the number one cause of death globally. In 2008 approximately 17 million people died from CVDs. Low- and middle-income countries are disproportionally affected: more than 80 percent of deaths due to CVDs occur in developing countries (World Health Organization, 2013). From all people in the world with diabetes, it is 80 percent too that lives in low- and middle-income countries (World Health Organization, 2013). In developed countries most people with diabetes are above the age of retirement, whereas in developing countries those most frequently affected are aged between 35 and 64. Remarkably, from the main two types of diabetes, one can easily be prevented by daily moderate 30 minutes of physical activity and a healthy diet (World Health Organization, 2013).
HIV/AIDS and TB belong to the most prevalent chronic communicable diseases. The vast majority of people infected with HIV/AIDS and TB are in low- and middle-income countries. Almost 2.5 million people were newly infected with the HIV-virus in 2011 (World Health Organization, 2013). The figure on the following page shows categories of diseases causing deaths and how it is projected to differ for low-, middle- and high-income countries. From the graph it becomes clear that chronic diseases are not a problem in rich countries only, in fact the number of deaths is much larger in developing countries.
Source: World Health Organization, (2005). Preventing chronic diseases: a vital investment: WHO global report. Note: Chronic diseases include among other diseases cardiovascular diseases, cancers, chronic respiratory disorders and diabetes. HIV/AIDS is not grouped as a chronic disease in this graph, but is included in the group of communicable diseases.
The rising average age of the population and the changing epidemiologic profile of the population are two trends that cause populations to suffer more from NCDs (Nugent, 2008). Nugent (2008) points out that there is a difference in the burden imposed by chronic diseases in rich and poor countries. More morbidity and mortality due to chronic illness occurs before the age of 60 in low- and middle-income countries compared with high-income countries. Next to that, it are mostly the poorest people that have the highest risk of developing NCDs and they are least able to cope with the resulting financial consequences (Suhrcke et al., 2006). According to the World Health Organization (2011) the proportion of premature NCD deaths under 60 years rose to 28 percent in lower-middle-income countries, more than double the proportion in high-income countries.
2.3 Economic consequences
The goal of the World Development Report (1993) was to create awareness for the relationship between health and macroeconomic outcomes. Chronic diseases are a problem to the public health, but these form a challenge to the economic development of a country too.
The costs on the microeconomic level will be clarified in the first section, after which the economic consequences on the macroeconomic level will be discussed.
2.3.1 Microeconomic level
The costs-of-illness on the individual and household level can be separated into three categories. The first category consists of the direct costs or out-of-pocket expenditures, which are mainly costs for treatment, medicines and travel costs. High levels of out-of-pocket spending often contribute to financial difficulties for families, including bankruptcy. Such costs can also induce people to delay or even forego entirely needed medical care. High out-of-pocket expenses are retaining for individuals with any chronic disease. By definition, chronic diseases are long-term conditions and they often require frequent monitoring and ongoing treatment. Research shows that out-of-pocket spending for mainly health services over a two-year period is highest for adults with chronic conditions (Cunningham, 2009). The second category consists of indirect cost, namely the foregone income. Because of illness people are not able to work or less productive and therefore earn less and thus acquire less income. Also, people can lose their job if they are often absent. If it is the wage earner of the household that becomes ill or even dies it can bring the whole household into poverty. Indirect costs are also present when other household members need to take care of someone in the family; time that they could have spent differently (Nugent, 2008). The effects of losing an adult can even persist into the next generation, as children need to help at home and cannot go to school (World Bank 2003). Van der Gaag and Tan (1998) emphasize the importance of qualitative early-childhood development: ‘The benefits range from directly reducing the number of children who suffer from ill health, to enabling the children to enjoy more productive lives as adults, to improving society by, for example, reducing crime rates’. This means that when one or both parents die or are ill, children lose the care and guidance from their parents and that harms their development. The third class of costs contains the intangible costs such as pain, suffering and anxiety. These costs are intangible and hard to measure, but always present when suffering from a chronic disease (Suhrcke et al., 2006).
Wagstaff and Pradhan (2005) discuss the further implications of out-of-pocket expenditures, namely that the exposure to risk of high health out-of-pocket expenditures has an effect on savings and consumption. Precautionary savings can cause households to consume less. Risk can even alter labor supply decisions so that labor supply is well diversified in the household when exposed to this kind of risk. This is confirmed in another
study: ‘Health expenditures can crowd out consumption of essential goods for the household’ (Suhrcke et al., 2006).
Quantitative evidence is missing with respect to chronic illness causing people to live below the poverty line. Even though there is a lack of studies that prove the causal relationship between chronic illness and poverty, the numerous effects that costs of chronic illness have and the smaller pieces of evidence are pointing into that direction (Suhrcke et al., 2006). ‘Chronic diseases impose a significant economic burden on patients, households, communities, employers, health care systems and on government budgets. Typical costs-of-illness studies often underestimate this burden’(World Bank, 2007).
2.3.2 Macroeconomic level
The Report of the Commission on Macroeconomics and Health (2001) distinguishes three channels on the macroeconomic level through which diseases influence wellbeing and economic development. The first channel is the reduced number of years of healthy life expectancy, which means that individuals at the top of their productive working lives are forced to quit. The cost of replacing skilled workers is expected to be substantial (World Bank, 1993). The second channel is the negative effect on investments in children by their parents. Societies with high rates of infant mortality (deaths under 1 year of age) and child mortality (deaths under 5 years of age) have higher rates of fertility. This effect exists in order to compensate for the frequent deaths of children. However, with many children, it is not possible to invest that much in all children as would have happened with fewer children. This channel might not be too relevant with respect to chronic illness as these diseases are not the most prevalent among children. The third channel is the effect of disease on the returns to business and infrastructure investment, besides the effects on worker productivity. Whole industries can be undermined by a high prevalence of disease. In addition, epidemic and endemic diseases can also undermine social capital and even political and macroeconomic stability.
When a population does not grow old it has more widespread implications. Longer-lived persons will invest a higher fraction of their incomes in financial savings, as they will have more time to enjoy the benefits of their investments. For a young population this could imply that credit markets do not function optimal. Another effect is that the governmental budget shift towards more health services, which leaves a smaller budget available for other investments (The Report of the Commission on Macroeconomics and Health, 2001).
The World Development Report of 1993 emphasizes the impact AIDS has on the economy: ‘The macroeconomic impact of AIDS comes partly from the high costs of treatment, which divert resources from productive investments’. ‘As a result of the AIDS pandemic alone, aggregate economic growth slows several percentage points per year in Africa, as individuals in the prime of their working lives are struck down’ (World Health Organization, 2001). That AIDS has such a direct negative impact on economic circumstances is however difficult to prove. ‘HIV infection itself is not directly related to severe negative outcomes; however, weight loss (a known correlate with a more advanced stage of AIDS) is’ (Gustafsson-Wright et al., 2011). The authors suggest that in the long term the HIV epidemic may hit the studied households more severely. The study by Abegunde et al. (2007) estimated the lost economic output of 23 high-burden developing countries to be $84 billion. These costs only include the costs of a reduced productive labor force due to death. This means that direct costs and intangible costs are excluded totally, while indirect costs are included partially.
The causal relationship is difficult to identify, as poverty and chronic illness are interacting in both ways. The following was signaled in Nigeria: ‘Poverty, as an outcome of the collapse of the rural agrarian economy in the Delta, is believed to trigger rural-to-urban migration, fostering sex work and other sexual risk behaviors that make people vulnerable to HIV and other diseases’ (Udoh et al., 2009). In many sub-Saharan African societies high HIV/AIDS prevalence is hypothesized as an outcome of as well as a contributor to poverty. When a lack of money makes it impossible that the illness is being treated or induces risky behavior, the infection gets more easily spread and the prevalence is due to poverty. On the other hand diseases can impose enormous costs on a household or society, which then can be seen as a cause of poverty.
The costs due to chronic illness can have widespread consequences: ‘In poor countries, the immediate costs of chronic ill health rests with the ill individual and his family, but longer-term health and economic consequences may increasingly affect health systems and other units in society, such as workplaces’ (Nugent, 2008).
3. Health insurance in developing countries
The first section of this chapter explores studies in which the impact of other health insurance programs – predominantly on the financial protection - has been estimated. The last section investigates briefly the phenomenon adverse selection in comparable studies.
3.1 Health insurance
In developing countries formal insurance is scarce. In order to smooth consumption, families make use of informal ways of insurance (Suhrcke et al., 2006). It is assumed that all kinds of informal insurance are less efficient than formal health insurance (Wagstaff and Pradhan, 2005).
Formal health insurance has three aims: to increase access and use; to improve health status; and to lower the financial consequences and risk (Escobar et al. 2010). This can be realized by making the health services affordable to everyone and by spreading the risk among a group of people. Health insurance promises to have several advantages in comparison with other interventions. First, health insurance empowers users because the money is aimed at the consumers (the demand side) and not directly to the providers (the supply side), which means that the money paid is based on actual services delivered. Second, whether the services are provided privately or publicly, the health insurance works with all kind of providers. Third, price discrimination is easy to implement, which can be at advantage of the poor. The final advantage of health insurance is that individual health spending has less space to be crowded out by other investments in the health sector when the money is fixed through periodical fee payments. HIF faced the following related problem in Africa: ‘With a lack of quality delivery (supply), the willingness to prepay for care (demand) is low. Without pre-payment of services the revenues for healthcare providers will remain unpredictable. This means that the investment risk remains high and the healthcare delivery sector is unable to attract the necessary financing to improve quality and capacity.’ The implementation of a health insurance scheme is therefore a good remedy.
Health insurance studies are difficult to compare since the insurance schemes are often different from each other and the methodologies used to evaluate the impact as well. Giedon and Díaz (2008) do give an overview of existing studies in developing countries and they find that health insurance improves access to and utilization of health services. The insurance package influences the utilization heavily. In most cases out-of-pocket expenditures decreased, with some exceptions depending on the coverage. For the poor, however, often an
increase of out-of-pocket expenditures was signalled. Whether health really has improved, is difficult to state. Research in Georgia shows that due to the health insurance out-of-pocket expenditures decreased for some groups, that risk reduced but that there was no impact on the utilization of healthcare. Neither did it have influence on people’s management of chronic illness, which comes down to the treatment of the illness and adapting the lifestyle to the illness (Bauhoff et al., 2010). Insurance coverage in Costa Rica did result in better management of diabetes, for other chronic diseases the estimates were not significant. Insured people were less likely to end up in an inpatient bed or the emergency room than the uninsured (Escobar et al., 2010). A study in Vietnam did find proof that health insurance decreased of out-of-pocket expenditures, which resulted in an increase of non-health consumption (Wagstaff and Pradhan, 2005). Escobar et al. (2010) provide an analysis of existing health insurance impact studies too, but experience difficulties with measuring the financial protection: ‘Evaluating the impact of health insurance on out-of-pocket payments and the incidence of catastrophic payments is challenging because it is the result of sequential decisions (whether to use care, what type of care to consume, how much care, and finally the price to pay for care based on the former sequence)’.
As mentioned before, there is not much evidence yet on the relationship between health insurance and health. Neither are there many studies that investigate the influence of the insurance scheme itself on the outcome variables such as health, utilization and financial protection, nor about the relationship between the supply side and these outcomes. It is obvious that health insurance is not a homogeneous type of intervention and that there is a call for more studies that evaluate the impact of health insurance.
3.2 Adverse selection
In general, in the market for insurances there is always the danger of moral hazard and adverse selection. Moral hazard is assumed to be not a problem in developing countries: underuse of health services is a far bigger problem than overuse (Escobar et al. 2010). Adverse selection, however, is often encountered in the health insurance market in developing countries. The information asymmetry in voluntary insurance contracts can lead to many high-risk people that enrol. This leads to an inefficient insurance system as it is exactly the risk pooling of low- and high-risk people that creates the insurance market to operate optimally. Chronic ill people are observed to be more likely to enrol in a health insurance scheme than those who are not (Giedon and Díaz, 2008).
4. Case study: Health insurance program in Kwara State, Nigeria
This chapter contains the Nigerian case study of the health insurance program, which is supported by the HIF. First, the Nigerian context will be sketched and then the health insurance program will be introduced. The next section provides the hypotheses with regard to the burden of chronic diseases and the impact of the program, followed by a description of the dataset. Then, an analysis of the burden that people endure due to chronic illness at baseline is included. In the next section the impact of offering health insurance on these chronic illness related costs will be estimated as well as there will be attention for the underreporting of chronic analysis and possible adverse selection. In the final section the findings of this chapter will be discussed.
4.1 The Nigerian context
The first section describes some socioeconomic features of Nigeria, the second section discusses the prevalence of chronic illnesses and the last section is about the healthcare system.
4.1.1 Socioeconomic features
Nigeria is the most densely populated African country with a population of 162.5 million people (World Bank, 2011). This Sub-Saharan country belongs with a GDP of 224.0 billion US dollars and a GNI per capita (PPP, international dollar) of 2,290$ to the group of lower middle-income countries. From 2003 until 2011, Nigeria experienced an average annual GDP growth of approximately 7 percent, yet the population increased quickly as well in that period: an increase of more than 30 million people. In 2010 62,6 percent of the population was living below the national poverty line.
Life expectancy at birth is 52; only 3 percent of the population reaches an age above 65. The primary school completion rate is 74 percent in 2010, while the gross rate of enrolment in secondary school is 44 percent (World Bank).1
1For an extension of the socioeconomic circumstances in Nigeria see appendix 2.
14
4.1.2 Chronic diseases
In Nigeria, the NCDs are estimated to account for 28 percent of all deaths in 2013, as indicated in the chart below. The number of adults that lives with high risk factors to obtain chronic diseases is high: raised blood glucose (8-12 percent), raised blood pressure (38-41 percent) and obesity (5-9 percent). Nigeria has the highest number of people suffering from diabetes in Africa with 3.9 percent of the Nigerian population that has any form of diabetes (International Diabetes Federation, 2010). Research shows that Nigeria potentially loses substantial amounts in national income as a result of deaths from NCDs on labor supply and savings. In 2005, the estimated loss in national income from heart disease, stroke and diabetes in Nigeria is 0.4 billion USD, which implies a 0.23 percent loss in GDP due to NCD deaths in 2005 (World Health Organization, 2006).
Source: WHO (2013), NCD Country profiles.
Note: the dark blue triangles show the chronic noncommunicable diseases, the light blue triangles indicate the communicable diseases, however the chronic communicable diseases are not specified. *CVD stands for cardiovascular diseases
For the chronic communicable diseases as a group there are hardly any estimates available that show the burden to the economy. Focusing on the disease AIDS only, there are 3.4 million Nigerian people estimated to live with AIDS in 2013 (UNAIDS). This results in the fact that Nigeria has the second largest HIV/AIDS population in the world.
4.1.3 Healthcare system
Even though the burden of disease is high in Nigeria, the health sector is underfunded. In 2011 the total health expenditure per capita is 80 USD and private health expenditures are equal to 63.3 percent of that amount (World Bank). Governance across the health sector is very weak. In contrast with many developed countries, Nigeria has no social insurance scheme and take-up of health insurance is not compulsory. This has as a consequence that the coverage of most key preventive and curative health services is low in Nigeria. The National Health Insurance Scheme (NHIS) was launched by the Nigerian government in 1999 and has been the major initiative to expand health insurance in Nigeria. However NHIS still covers only 3 percent of the population (Dutta and Hongoro, 2013).
4.2 Supported by the Health Insurance Fund: a health insurance program
This health insurance program that is launched in 2009 by the HIF and some partners is called the Hygeia Community Health Plan (HCHP). This program is a combination of a demand and a supply side intervention. The demand intervention comprises the offering of subsidized health insurance to people that live in the targeted areas. To the targeted people – mostly farmers and their families from the Afon district – the benefit package provides coverage for the most common medical problems. It consists of outpatient primary healthcare, maternal and child health and chronic healthcare services and limited secondary health services. The supply side intervention comprises that in these same areas clinics are upgraded and contracted to deliver care to insured patients. The Kwara State government supports the program financially and encourages the community actively to make use of the program.
The insurance is structured so that people are enrolled on an individual and annual basis, however the program encourages family enrolment to minimize adverse selection. In the first year scheme members in Kwara State pay approximately 8 percent of the annual premiums themselves. This means that premium per person per year is 300 Nigerian Naira (NGN) which is equal to approximately 2 USD at that moment. The Kwara State Government contributes 60 percent and HIF accounts for the remainder (PharmAccess Database, 2012).
Access to healthcare is crucial for farmers and their families, as they depend on good health to carry out their daily work. HIF’s mission is to protect the wealth of low-income families from health-related risks.
4.3 Hypotheses
On the basis of the academic literature in the earlier chapters it is possible to formulate brief hypotheses with regard to the burden due the chronic diseases at baseline and with regard to the impact of the health insurance program.
The burden at baseline due to chronic diseases is expected to be significant. Self-reported health status will probably be lower for people with a chronic disease. In line with a bad health situation, people need more care, drugs and other medical services so we expect utilization to be higher for people with a chronic illness. However, a lack of money or low access to medical services, which is often the case in poor countries, could avoid utilization to be high. Due to their long-term illness people will be less able to do their normal daily activities, which probably influences individual income negatively. The expectation is that people with a chronic illness experience higher health expenditures than people who don’t have a chronic disease.
The introduction of the health insurance program has the expectation to create an improved situation for people with a chronic disease. The program should lower the total health expenditures as the coverage is expected to be sufficient and only a relatively small fixed premium needs to be paid. The program is expected to let the use of medical services increase because these services become better accessible and are of better quality. A successful health insurance program could therefore lower the number of people with any chronic disease as people are encouraged and facilitated by the program to seek care in an early stage.
4.4 Data description
The data are stemming from household surveys – a biomedical and socioeconomic questionnaire – in 2009 and 2011 (Kwara II Panel Data Set). The baseline survey (2009) took place before the start of the insurance program and the follow-up survey took place 2 years after the introduction of the program (2011). The sample consists of 1500 farming households. The Afon district, in which the health insurance program is implemented, is the treatment group. The rural households from the Eruku district function as the control group in. The number of households is lower in 2011 than at baseline due to migration and death. However, the number of individuals didn’t decrease that much because new members became part of the households and they are included in the follow-up survey as well (Gustaffson-Wright, 2010).
The program caused almost 30 percent of the people in the treatment group to be insured, compared to almost none in the control group which is in spite of the existence of the the National Health Insurance Scheme.
TABLE 1 SAMPLING
Control Treatment
Total (N) Share (%) Total (N) Share (%)
Sample (households) 600 900 Interviewed in 2009 (households) 571 892 Interviewed in 2009 (individuals) 2349 3629 Interviewed in 2011 (households) 488 790 Interviewed in 2011 (individuals) 2064 3484 Insured in 2009 (individuals) 21 0.89 15 0.41 Insured in 2011 (individuals) 7 0.34 1010 29.08 Insured in 2011 (households) 2 0.38 338 39.44
Source: Kwara II Panel data set, Nigeria surveys 2009 (baseline) and 2011 (follow-up).
Note: N stands for number of subjects. Insured households reflect the households in which at least 1 person is insured.
Tables 2A and 2B compare some key observable characteristics at baseline between both groups. TABLE 2A DESCRIPTIVE STATISTICS Control (mean) Treatment (mean) P-value Age 24.42 27.97 0.00*** (21.35) (23.72) Household size 5.44 2.47 0.01*** (2.47) (2.86) Income
Mean consumption per household per year 425,880 392,355 0.00***
(343,652) (295,598)
Mean consumption per capita per year 89,410 79,231 0.00***
(79,850) (61,858) Quintile 1 (poorest) 29,680 28,714 0.13 (6,926) (6,430) Quintile 2 48,012 47,615 0.91 (5,058) (4,816) Quintile 3 66,550 65,410 0.33 (5,726) (5,751) Quintile 4 95,971 92,151 0.05** (11,376) (10,783) Quintile 5 (richest) 195,629 169,609 0.01*** (120,624) (82,991) 18
TABLE 2B
DESCRIPTIVE STATISTICS
Control Treatment P-value
Employment
Self-employed 77.83% 78.75% 0.17
Employer 1.22% 2.03%
Part-time worker 0.22% 1.01%
Permanent paid employee 13.19% 7.34%
Temporary paid employee 2.33% 1.73%
Unpaid family worker 5.21% 9.13%
Religion Islam 59.81% 92.46% 0.00*** Christian 39.93% 7.32% Other or no religion 0.26% 0.22% Education Enrolment rate 96.91% 93.91% 0.01***
Less than Primary 10.04% 10.38% 0.13
Primary Complete 29.58% 39.75% Less than JSS 3.81% 3.00% JSS Complete 3.63% 2.05% Less than SSS 3.81% 3.00% SSS complete 26.47% 25.00% Diploma 9.34% 9.29% Degree 6.92% 0.68% Post-graduate 2.77% 0.55% Literate 51.41% 45.67% 0.00*** Health
Individuals reporting any chronic illness 6.49% 5.86% 0.32 Households reporting any chronic illness 20.84% 18.95% 0.37 Number of visits to hospital last year 1.34 1.06 0.56
10-step health ladder 9.50 9.29 0.00***
Total health expenditures per household 4477.2 4213.7 0.65
Number of observations 2349 3629
Source tables 2A and 2B: Kwara II Panel Data Set, survey 2009 (baseline).
Note: Standard deviations are stated in parenthesis. Income is stated in the Nigerian currency: Nigerian Naira (NGN). The exchange rate at baseline is 1 USD to 150.2 NGN. In all tables with regression results *, ** and *** stand for significance at 10, 5 and 1 percent level respectively. The enrolment rate contains the numbers of children between the age of 7 and 14. The level of education represents the highest level of education completed for all individuals older than 18. JSS stands for Junior Secondary School, SSS for Senior Secondary School. Number of visits to hospital is the mean for the people that did go to the hospital, not for the whole sample. The 10-step health ladder ranges from bad (0) to excellent health (10).
Even though the treatment and control group are carefully constructed, that is consistent with a difference-in-differences design, there are some significant differences between the two groups: age is on average three years higher in the treatment group; mean
income is significantly lower; religious practices are very different; and enrolment rate for schooling is a few percentage points lower in the treatment group, as is literacy. Important variables in this research such as ‘the reporting of any chronic illness’, ‘the number of hospital visits’ and ‘total health expenditures’ are not significantly different. In general people consider themselves very healthy; in the control group a bit more than in the treatment group.
4.5 The burden due to chronic illness
To possibly say something about the impact of offered health insurance on chronic illness related expenditures we need to know what the burden is at baseline for households and individuals due to chronic diseases. As health status and utilization are connected with health expenditures, the burden is going to be estimated on basis of three pillars: health status; utilization of medical services; and expenditures and costs. Table 3 shows the variables that will be used to estimate the burden on these three pillars. The column on the right shows which variables are also used to investigate the impact of the offered health insurance program (indicated by ‘x’); not all variables are available or suitable for the impact evaluation due to the design of the questions in the surveys.
TABLE 3
Burden at baseline Average treatment effect
Health status People self-reporting any chronic illness x
People diagnosed with any chronic illness x
Health status ladder
Utilization Number of hospital visits x
Number of nights spent in the hospital x
Expenditures and costs Days unable to do daily activities
Foregone main income
Total household health expenditures x
Note: Apart from the total health expenditures that reflect data on the household level, all other variables are constructed on the individual level.
Thus, in the following sections the burden at baseline is estimated. First the effects of chronic diseases on health status are discussed, followed by the effects on utilization and finally on the expenditures and costs. Section 4.6.1 contains the impact evaluation on expenditures and utilization, while 4.6.2 deals with the impact of the program on health status.
4.5.1 Health status
How to value someone’s health objectively is very hard. Tables 4 and 5 give insight in the prevalence of people that are diagnosed with chronic diseases and people that self-reported to have a chronic illness at baseline in the sample. Table 6 estimates the impact of any chronic illness – also with the distinction between self-reported and diagnosed chronic illness – on the self-reported health status, another proxy for health.
Table 4 shows the results of the medical examinations at baseline. Overweight, obesity, hypertension, HIV/AIDS and diabetes are in descending order present among the interviewed Nigerians. Apart from HIV/AIDS all chronic diseases become more prevalent the higher the age category.
TABLE 4
PREVALENCE DIAGNOSED CHRONIC DISEASES
Children Adults Elderly Total
Number of subjects 1601 2043 1123 4767
People with any diagnosed chronic illness 4.68% 22.61% 39.09% 20.47%
Medical examination Weight (KG) 26.58 57.32 57.53 41.77 Height (CM) 125.90 160.31 160.20 139.14 BMI 16.56 23.09 23.01 20.11 Overweight (BMI 25-30) 1.87% 13.05% 16.89% 8.81% Obese (BMI >30) 2.50% 6.33% 7.60% 4.67% Mild hypertension 0.54% 1.77% 9.26% 3.78% Severe hypertension 0.00% 1.25% 11.41% 4.06% Blood test
Hiv antibody positive 2.16% 3.22% 2.55% 2.96%
Diabetes 0.00% 0.94% 4.94% 2.16%
Source: Calculations based on Kwara II Panel Data Set, survey 2009 (baseline).
Note: Age categories are structured as follows: Underfives 0 to 5 year; children 5 to 15 years; adults 15 to 50 years; and elderly 50 years and older. Among under-fives medical examinations and blood tests are scarce. These data are therefore not presented in this table. The variable ‘people with any diagnosed chronic illness’ is constructed out of all chronic diseases that are displayed in the table. BMI is constructed with the formula: BMI = Weight(kg) / (length(m) * length (m)). The labeling of BMI groups are not trustworthy for children. It overestimates overweight for small children and it underestimates overweight for tall children. HIV tests are not 100% conclusive. Further tests need to be done to confirm the findings.
It is important to note that overweight and obesity are no real chronic diseases but that they ‘are major risk factors for a number of chronic diseases, including diabetes, cardiovascular diseases and cancer’ (World Health Organization, 2013). Hypertension lies somewhere in between; it is regarded as a major risk factor but also as a chronic disease. The
percentages of people diagnosed with diabetes and high risk factors such as raised blood glucose and obesity stemming from table 4 are similar to the percentages for the country as a whole. Raised blood pressure is a far smaller problem in this sample than it is on the country level. HIV is 1 percent less diagnosed in this sample compared to the country percentages (World Health Organization, 2011).
The self-reported baseline numbers in table 5 show a lower prevalence of chronic diseases than the objective measurements do. In general only 6.12 percent reports any chronic illness compared to 20.5 percent that is diagnosed with it. From the elderly it is almost 20 percent that reports chronic illness, compared to 40 percent that is diagnosed with it. These numbers however should be interpreted with caution, as they capture not completely the same chronic diseases. AIDS has not been reported often (only 2 times) while it has been diagnosed far more often (69 times) as a result of the blood tests. According to the self-reported data diabetes seems to be a smaller problem to the population than it is stemming from the objective measurements. The underreporting of hypertension is as expected because people often don’t know they have this problem and can live long without taking any notice. Section 4.6.2 explores possible explanations for the underreporting of chronic illness.
TABLE 5
PREVALENCE SELF-REPORTED CHRONIC DISEASES
Under-fives Children Adults Elderly Total P-value
Number of subjects 843 1634 2092 1147 5716 - Any chronic illness present 0.24% 0.86% 5.21% 19.60% 6.12% 0.00*** Diabetes 0.00% 0.00% 0.14% 1.13% 0.28% 0.22 Heart disease 0.00% 0.00% 0.33% 0.35% 0.19% 0.19 Asthma 0.00% 0.06% 0.53% 0.96% 0.40% 0.16 Hypertension 0.00% 0.00% 0.57% 3.22% 0.86% 0.08* Musculo-skeletal problems 0.12% 0.06% 1.67% 9.32% 2.52% 0.67 Epilepsy 0.12% 0.00% 0.05% 0.09% 0.05% 0.13 Allergy 0.00% 0.31% 0.48% 0.44% 0.35% 0.11 Sickle cell disease 0.00% 0.00% 0.05% 0.09% 0.03% 0.14 HIV/AIDS 0.00% 0.00% 0.05% 0.09% 0.03% 0.14 Physical disability 0.00% 0.18% 0.72% 1.48% 0.61% 0.19 Other 0.00% 0.12% 0.66% 2.59% 0.80% 0.45
Source: Calculations based on Kwara II Panel Data Set, survey 2009 (baseline).
Note: Age categories are the same as for table 4. The variable ‘any chronic illness present’ is constructed out of all diseases that are displayed in the table. P-value distinguishes prevalence of chronic diseases between young people (under-fives and children) and older people (adults and elderly).
Table 6 shows that health status decreases significantly for people that self-reported and that are diagnosed with any chronic illness. Adding control variables makes the effect smaller. If people are aware of any chronic disease it means that people value their health status less with 0.74 up to 1.09 point. If people are diagnosed with a chronic disease it probably means that very often these people were not aware of their chronic illness because the negative effect of a diagnosed chronic illness on health status is much lower. The impact of the objective diagnosis on health status becomes even insignificant when adding control variables. From the table it seems that self-reported health status is a considerable similar proxy for health as is self-reported chronic illness.
TABLE 6
CHRONIC ILLNESS AND HEALTH STATUS
10-step health ladder, from 0 (bad) to 10 (excellent) Self-reported any chronic illness -1.09*** -0.76*** -0.74***
(0.1) (0.1) (0.1)
Diagnosed any chronic illness -0.27*** 0.05 0.05
(0.05) (0.05) (0.05)
Age -0.01*** -0.01*** -0.01*** -0.01***
(0.00) (0.00) (0.00) (0.00)
Male 0.07 0.05 0.08* 0.06
(0.03) (0.03) (0.03) (0.03)
Upper two quintiles -0.1** -0.1***
(0.03) (0.03)
Never been injured 0.05 0.07
(0.04) (0.03) Literate -0.15*** -0.16*** 0.03 (0.03) _cons 9.44*** 9.67*** 9.89*** 9.42*** 9.69*** 9.83*** (0.02) (0.03) (0.06) (0.17) (0.03) (0.06) N 5721 5707 5693 5721 5707 5693
Source: Calculations based on Kwara II Panel Data Set, survey 2009 (baseline).
Note: Health status is reported on a scale from 0 to 10, from bad to excellent health. In this and all following tables N stands for number of subjects.
4.5.2 Utilization
Table 7 gives insight in the utilization of hospital visits and the number of nights spent in the hospital because of any chronic illness. The impact of chronic illness on both utilization variables is smaller for the objective measure of chronic illness. Independent of the control variables chronic illness does not have any significant impact on the number of hospital visits, neither on the nights spent in the hospital. Unfortunately there are many missing data for these
utilization variables, which lead to a relatively small sample size and that is at least one reason that would explain the insignificant results.
TABLE 7 UTILIZATION
Hospital visits Nights spent in hospital
Self-reported any chronic illness 0.38 0.32 9.33 6.91 7.99 7.05
(0.36) (0.43) (7.65) (7.38) (4.5) (4.63)
Diagnosed any chronic illness -0.16 -0.21
(0.15) (0.17)
Age 0.00 0.01 0.12 0.12
(0.00) (0.01) (0.06) (0.06)
Male -0.17 -0.23 7.23 7.77
(0.17) (0.19) (3.26) (3.33)
Upper two quintiles 0.37 0.35 -0.44 0.3
(0.18) (0.17) (3.22) (3.5)
_cons 1.12*** 0.90*** 1.24*** 0.94*** 7.37*** 0.35 6.35 -1.19 (0.08) (0.14) (0.13) (0.13) (1.2) (3.06) (1.2) (3.8)
N 118 118 118 118 115 115 115 115
Source: Calculations based on Kwara II Panel Data Set, survey 2009 (baseline).
4.5.3 Financial burden
This section explores the indirect costs and the total household health expenditures that are related to chronic illness at baseline.
Chronically ill people were asked how many of the past 30 days they were unable to do their regular activities because of their chronic condition. The burden due to any self-reported chronic illness is on average approximately 5 days for people that they were unable to do daily chores, for the elderly alone the average was 6 days at baseline. Probably, these numbers are an overestimation of the true lost days because they are self-reported (Abegunde et al., 2007).
Costs that exist due to a person in the household that is unable to do daily activities for several days, such as being less productive at work or experiencing higher risk of losing the job are difficult to calculate. However the income of the main job that people forego because of non-attendance at their work can be estimated. Table 8 presents that self-reporting of chronic illness has a negative but insignificant effect on main income of the past 12 months at baseline. Controlling for the fact that the person has not been ill before doesn’t seem to influence the results either. On the contrary, being diagnosed with a chronic illness does not have a negative effect on main income at all; surprisingly it shows to have a positive effect.
TABLE 8
FOREGONE MAIN INCOME
Main income (NGN) per year Self-reported any chronic illness -9620 -26951
(7118) (22178)
Diagnosed any chronic illness 12681 34260***
(6299) (6230)
Days unable to do chores -927 104
(804) (244)
Age -287 -439**
(158) (156)
Male 79535*** 85983***
(5542) (5786)
Has never been ill before 2228 2383
(5342) (5316)
_cons 103658*** 105398*** 98620*** 67178***
(3076) (24405.77) (2217) (11669)
N 2357 2237 2357 2237
Source: Calculations based on Kwara II Panel Data Set, survey 2009 (baseline).
Note: the assumption is made that a day that a person is unable to do daily chores, he/she is also absent from work. Main income is also a self-reported variable.
The next table deals with the total household health expenditures. Table 9 proves that if there are people in the household reporting any chronic illness that the household suffers from additional costs during the year up till baseline. This effect is not clear for households in which people are diagnosed but are not aware of their illness. Because of the unawareness, they probably don’t have the urge to search for medical care. Using the exchange rate of 1 USD to 150.2 NGN the total health expenditures increase on average with at least 14,8 USD due to any chronic illness in the household. The interpretation of this result is enhanced when becoming aware of the fact total household health expenditures per year are only 5732 NGN (38.2 USD) on average.
TABLE 9
TOTAL HOUSEHOLD HEALTH EXPENDITURES
Expenditures (NGN) per household per year Self-reported any chronic illness 2223** 2254***
(725) (720)
Diagnosed any chronic illness 944 970
(588) (586)
Upper two quintiles 1915*** 1906***
(573) (574)
_cons 4467*** 3464*** 4445*** 3441
(329) (320) (412) (401)
N 1462 1462 1462 1462
Source: Calculations based on Kwara II Panel Data Set, survey 2009 (baseline).
Note: total health expenditures (NGN) are self-reported for the household as a whole. Total health expenditures might include the value of items that the respondent received for free.
The addition of the dummy variable ‘being in one of the two upper consumption quintiles’ as a control variable does not change the impact much. Regressions for specific chronic diseases alone did not deliver any significant results.
4.6 Impact evaluation
In this section the impact of the offered health insurance program will be estimated with the difference-in-differences method. First, the impact on household expenditures and individual utilization is discussed. After that, the impact of the program on health status is estimated and the differences between the self-reporting and diagnosing of any chronic illness will be further explored. In the last section the findings will be related to the concept of adverse selection as to provide additional information with regard to the impact of offering health insurance on chronic illness related expenditures.
4.6.1 The impact of the health insurance program on health expenditures and utilization The difference-in-differences method allows us to construct an intention-to-treat (ITT) analysis. Hence, the focus is not only on the individuals that did buy insurance, but on all those people to which the insurance was offered. This means that the entire treatment group is examined capturing potential ‘spillover effects’ on the uninsured individuals in the treatment group as they might benefit from access to upgraded medical services or increased awareness. This means that the difference-in-differences method estimates the average treatment effect (ATE). Although possible selection effects are investigated in section 4.6.3, we do not take any selection effects into account in this section as it is precisely this average treatment effect that we are interested in because we like to know what the effect is of the offering of health insurance on the chronic illness related expenditures of the public in practice. Consequently, the effect of the program on the people that did actually enrol in the progam alone should be equal or larger than the average treatment effect; you can interpret the ATE therefore as a lower bound of the average treatment effect on the treated (ATET) (Gustafsson-Wright et al., 2013).
In this section the average treatment effect of the program is estimated on the financial protection and on the utilization of medical services. The first interaction term shows the ATE on people that self-reported a chronic illness at baseline (A) or that are diagnosed with a chronic illness at baseline (B). For the households that self-reported to have member(s) with chronic illness at baseline the insurance program decreased the total household health
expenditures on average with 3153 NGN (21 USD) and this result is statistically significant. For households consisting of members diagnosed with a chronic illness the program has a small and statistically insignificant impact on the household health expenditures. The other interaction term ‘time*treatment’ in both A and B shows that the program did lower the total household health expenditures for all households in the treatment group on average, though the significance of the results leave room for discussion. For both A and B the program did not have a significant effect on utilization. The addition of control variables to the regressions did not alter the results (not shown in the table).
TABLE 10 DIFFERENCE-IN-DIFFERENCES Financial protection (NGN per household) Utilization (per individual) Total health expenditures Hospital visits Nights in hospital
A: Self-reported any chronic illness
Self-reported chronic illness *Time *Treatment -3153** -0.2 -8.0
(1062) (0.3) (6.3) Time*Treatment -3555 -0.3 0.0 (1448) (0.3) (5.0) Time 3571** 0.2 0.6 (0 in baseline, 1 in follow-up) (1289) (0.3) (3.7) Treatment -70 -0.1 -0.2
(0 in control group, 1 in treatment group) (629) (0.2) (3.5)
Self-reported any chronic illness 2513** 0.2 5.9
(820) (0.3) (6.0)
_cons 4452*** 1.2*** 8.1***
(559) (0.2) (2.3)
N 2710 269 271
B: Diagnosed with any chronic illness
Diagnosed chronic illness*Time*Treatment 84 0.2 -6.9
(1117) (0.2) (4.1) Time*Treatment -3942* -0.4 0.9 (1539) (0.3) (5.4) Time 3375* 0.2 0.5 (0 in baseline, 1 in follow-up) (1295) (0.3) (3.8) Treatment -81 -0.1 -0.1
(0 in control group, 1 in treatment group) (639) (0.2) (3.4)
Diagnosed with any chronic illness 660 0.2 4.4
(726) (0.2) (3.3)
_cons 4362 1.3*** 7.6***
(689) (0.2) (2.2)
N 2704 269 271
Source: Calculations based on Kwara II Panel Data Set, surveys from 2009 (baseline) and 2011 (follow-up).
4.6.2 The impact of the health insurance program on health status
Tables 3 and 4 in this chapter showed that underreporting of chronic diseases is also the matter in this study. Research shows that the poor appear to underreport all kind of diseases. This is a problem as this is an indication that people do not receive necessary healthcare (Gustafsson-Wright et al., 2011). Table 5 showed that self-reported health status negatively depends on whether people self-report any chronic illness, but diagnosed chronic illness does not seem to have much impact on self-reported health status. Apart from the intangible costs of a lower health status, the total health expenditures for the households do increase with at least 2223 NGN (14.8 USD) if there are people reporting any chronic illness. This effect is not there for households consisting of people that are diagnosed with any chronic illness. The average treatment effect of the health insurance program on total health expenditures of households with members that self-reported a chronic illness was positive, namely lowering total household expenditures with 3153 NGN (21 USD). This is in contrast with the impact the program has on the household expenditures of households with members that were diagnosed with any chronic illness; there is no significant impact. This section explores the impact of the program on health status and investigates the differences between self-reporting and diagnosed chronic illness further.
There can be several factors that contribute to underreporting. Feelings of shame or embarrassment could make it difficult to report an illness, although this effect should be limited in this study since the questionnaires are handled anonymously and with care. Employment status might explain part of the bias as unhealthy circumstances or an intensive job can in some cases cause chronic illnesses, which would explain higher self-reporting. However, research shows that unemployed are found to report illness more than employed people, as the illness might be the reason for being unemployed (Butler et al., 1987). Another reason could be that people do not possess the necessary information about health, healthcare services or any related topic so that they simply don’t know how to detect an illness, even if it is in their very own body. This low level of information probably corresponds to a low level of education or literacy. Underreporting could also be related to low income because the poor experience a poor environment and might be unable to seek medical care and therefore have higher chance to develop diseases. The rich, however, are expected to have better information and therefore they could report relatively high numbers of chronic illness as well. This latter expectation, actually, is often confirmed in other studies (Butler et al., 1987; Sen, 2002; Escobar et al., 2010). To take some of these considerations into account, the estimates of the
average treatment effect of the program on health status in table 11 are among other things controlled for literacy and being relatively rich (upper two quintiles).
During the first two years of the health insurance program underreporting in the treatment group has decreased. Table 11 displays that the program has increased awareness about health status in the treatment group; the number of people that self-reported a chronic diseases did increase while the number of people that are diagnosed with chronic illnesses did not increase. As people are interviewed and tested for diseases they might have discovered any chronic illness, which would explain the reporting two years later. But even if people did not have any illness in the first place, they may have got more aware of chronic illnesses and their health in general, which would explain better detection and self-reporting of illnesses two years later.
TABLE 11
HEALTH STATUS AND AWARENESS
Self-reporting any chronic illness (0 for not reporting, 1 for reporting)
Being diagnosed with any chronic illness (0 if not, 1 for being diagnosed)
Treatment effect 0.56*** 0.48** 0.64*** 0.01 -0.07 -0.04 (0.15) (0.16) (0.17) (0.11) (0.11) (0.12) Time -0.15 0.43*** 0.07 -0.39*** 0.12 -0.15 (0.12) (0.13) (0.13) (0.08) (0.08) (-0.09) Treatment -0.17 -0.24 0.52*** -0.04 -0.06 -0.22* (0.11) (0.1) (0.12) (0.07) (0.07) (0.08) Literacy 1.28*** 0.25* 0.52*** -0.22*** (0.08) (0.1) (0.05) (0.06)
Upper two quintiles 0.69*** 0.4*** 0.60*** 0.36***
(0.07) (0.08) (0.05) (0.06) Age 0.05*** 0.04*** (0.0) (0.00) _cons -2.68*** -5.03*** -5.2*** -1.6*** -2.66*** -2.75*** (0.08) (0.19) (-0.18) (0.06) (0.11) (0.11) N 12685 10743 10743 12685 10743 10743
Source: Calculations based on Kwara II Panel Data Set, surveys from 2009 (baseline) and 2011 (follow-up).
4.6.3 Adverse selection in context
To extend the impact evaluation this section explores whether there is adverse selection in the health insurance program on certain characteristics. The treatment group as a whole took advantage of the total household health expenditures being lowered on average. But in the treatment group insurance up-take rose from almost 0 to almost 30 percent. Who are the people that choose to take the insurance then? Nigeria does not have a compulsory insurance scheme and thus is the impact depending on the insurance up-take too. This section tries to give an insight in this matter by looking for the factors that actually ‘select’ people into the health insurance program. This could be useful information about the potential this program has, as the impact of the program will be at its maximum when insurance up-take is as high as possible.
Table 12 shows one way to test for adverse selection. The results in the table show that there is no significant difference in utilization between groups. This means that people who are more exposed to risk and therefore make more use of medical services did not select themselves more into the insurance scheme than others, at least this study did not find proof for that to be the case.
TABLE 12 ADVERSE SELECTION Insured (mean) Uninsured (mean) P-value Hospital visits 1.00 1.22 0.23 (0.03) (0.12) Nights spent in hospital 7.5 9.09 0.59 (2.1) (1.64)
Source: Calculations based on Kwara II Panel Data Set, surveys from 2011 (follow-up). Note: For this table sample size (N) is equal to 156.
Table 13 shows the regressions that test for determinants of health insurance enrolment. A high burden due to chronic illness at baseline could motivate people to buy insurance. Factors as high utilization and much expenditure at baseline can therefore be considered as reasons for people to insure themselves. The table, actually, proves that this is not the case here. Another study also finds that health expenditures and utilization of medical services at baseline have little predictive power (Thornton et al., 2010). The same authors, who used data from Nicaragua to predict the take-up of health insurance, find that reporting a chronic disease is positively associated with rates of up-take. The likelihood of enrolment
increased by 4.4 percentage points for anyone that reported a chronic illness. This type of adverse selection is also encountered in Ghana (Giedon and Díaz, 2008). This is actually not confirmed in this study; in all regressions self-reporting of chronic illness at baseline does not have a statistically significant effect on the chance of being insured in 2011.
TABLE 13
DETERMINANTS OF INSURANCE ENROLMENT
Being insured in 2011
(0 for not being insured, 1 for being insured)
Baseline indicators
Self-reported any chronic illness 0.14 0.16 0.16 0.98
(0.15) (0.17) (0.17) (0.67)
Diagnosed with any chronic illness 0.26** 0.3*** 0.32*** -0.23 0.06
(0.1) (0.1) (0.1) (0.65) (0.1)
Never been ill or injured 0.3*** 0.3***
(0.09) (0.09)
More than one person chronic ill in
household 0.03 0.03
(0.19) (0.2)
High level of health expenditures 0.07 0.2 0.1
0.09 (0.6) (0.09)
Hospital visits 0.43
(0.21)
Nights spent in hospital 0.02
(0.02)
Upper two quintiles 0.51***
(0.08)
More years of education than primary school -0.23
(0.14)
Literate 0.05
(0.09)
Being head of household -0.19
(0.12) Age 0.01*** (0.00) Male -0.07 (0.09) _cons -1.5*** -1.53*** -1.83*** -1.85*** -2.4*** -1.93 (0.04) (0.04) (0.08) (0.09) (0.57) (0.15) N 4639 4639 4465 4465 82 4621
Source: Calculations based on Kwara II Panel Data Set, surveys from 2009 (baseline) and 2011 (follow-up). Note: This table shows the logistic regressions with variables from 2009 that are potential predictors of the chance of being enrolled in the health insurance program in 2011. The dummy variable ‘high level of health expenditures’ is constructed as follows: 0 stands for below mean spending in sample, 1 stand for above mean spending in sample.
Table 13 shows that being diagnosed with any chronic illness at baseline might increase the chance of being insured in 2011. This finding fits in earlier considerations that being diagnosed with any chronic illness increases awareness and thus increases self-reporting of chronic illness in 2011 (as proven in table 11) and this increased awareness could in turn increase the chance of enrolment. However, this effect becomes less clear when there is controlled for many other variables. Interestingly, table 13 indicates that people who never are injured or have been ill before at baseline have actually a higher chance to buy insurance. This is contradicting with the finding that the diagnosis of chronic illness is a determinant of enrollment too. These findings need further investigation in order to understand the dynamics.
Being relatively rich, thus being part of one of the two upper consumption quintiles, increases the chance of buying insurance. This is in contrast with the Nicaraguan study in which income had no significant effect on insurance take-up, although they do find a relation with control variables that are related with being wealthy (Thornton et al., 2009).
Being head of the household head does not influence the likelihood of buying insurance. Other studies do find that being the household head is a determinant of enrolment. Different characteristics are assigned to the household head that could explain these results. Studies from Ghana and Namibia show that households with better-educated heads are more likely to purchase insurance (Giesbert et al., 2011; Giedon and Díaz, 2008). Households with a sick household head have less chance to buy insurance as they lose income flow and therefore have difficulty in financing the insurance premium (Ito and Kono, 2010).
There is no proof in this case study for the level of information or ‘understanding’ to have influence on insurance up-take; neither being literate nor enjoying a higher level of education contributes to the chance of taking insurance. This is in line with Giné et al. (2008) who find no significant relationship between education and the up-take of insurance, but in contrast with other research (Butler et al., 1987; Sen, 2002; Escobar et al., 2010).
Overall shows table 13 only limited evidence that there is adverse selection in the health insurance program. We have reason to think that insurance enrolment can partly be explained by the following characteristics: being diagnosed with a chronic illness; on being in one of the two upper consumption quintiles; on age: and on never being ill or injured. Baseline health expenditures and utilization did not have any significant predictive power.
