Public Health Expenditure And Health

Using a utility maximization approach as developed by Grossman, the results revealed that health expenditure does not affect health outcomes in Kenya. The factors that affect health outcomes include: distance to nearest health facility (5km or more) and other household income. This implies that increasing public health expenditure does not lead to reduced maternal mortality rates.

Since the other determinants (access to medical facility and other household income) significantly affect the health outcomes, the government needs to put measures in place to ensure that women can easily access health facilities and sensitize them to ensure that they deliver in health facilities and attend antenatal care.

This study did not include some important variables that affect maternal mortality rates like the impact of cultural practices such as female genital mutilation (FGM), preference of certain types of health care providers (including traditional and herbal medicine) and earlier marriages. Therefore we suggest that in future, studies in this field should give attention to these variables.

CHAPTER ONE: INTRODUCTION

1.1 Background information

Health is the extent to which an individual or group is able to cope with the interpersonal, social, biological, and physical environments (World Bank, 2004). Health is therefore a resource for everyday life, not the objective of living. It is a positive concept embracing social and personal resources as well as physical and psychological capacities. Health financing is a key input in the provision of quality healthcare.

Governments have always had a prominent role in overcoming public health risks and this is a major area of concern in less developed countries like Kenya (Scott, 2001). The provision of good health satisfies one of the basic human needs and contributes significantly to maintaining and enhancing the productivity of the people (Owino, 1997). Public expenditure on health services therefore is a key investment in human capital and plays a catalytic role in the growth of the economy by enabling people to achieve their full potential and lead productive lives. In recognition of the importance of human health, one of the Kenya government’s major goals since independence has been to achieve adequate and good-quality health care for all citizens (GOK, 1965).

To address health outcomes in developing countries such as Kenya, UNICEF advocates for increased public expenditure on health (UNICEF, 2006). Therefore, many countries in developing countries have increased their health expenditure over time. For example, to achieve better health outcomes, Kenya has increased its health expenditure from Kshs. 11.9 billion in 2000 to Kshs. 20 billion in 2004 representing a 30% increase as shown in Figure 1.1 (GOK, 2007). But more resources alone may not necessarily lead to better health outcomes because health care expenditure is only one of the many factors that contribute to health outcomes, considering the fact that these resources may be channeled to various projects that may not directly influence health outcomes. The link between government health expenditures and health outcomes may therefore not necessarily be present. First, an increase in public health expenditures may result in a decrease in private health expenditures; a household may divert its funds towards other uses once the government increases its provision of basic health care. Second, the incremental government expenditures may be employed on the intensive rather than the extensive margin. An example of intensive expenditures would be if expensive and low productivity inputs are bought with marginal funds in which case the impact of these expenditures on health outcomes may be small. Third, even if extra funds are applied extensively to health care (e.g. more staff at hospitals, adequate stocking of medications), but complementary services, both inside and outside the health sector, are not there (e.g. lack of roads or transportation to hospitals and clinics, subsidized prices for medication, etc.) the impact of extra government health expenditures may be little or none (Wagstaff, 2002a).

In addition to health expenditure, Kenya also joined hands with other one hundred and eighty eight countries in a global effort to improve health outcome and reaffirmed its commitment to the united Nations Millennium Development goals (MDGs). Three of these millennium development goals are directly related to health. These are to (i) reduce child mortality, (ii) Improve maternal health and (iii) combat HIV/AIDS, malaria and other diseases.

Despite these global and local interventions in health, performance of Kenya’s health sector in terms of maternal mortality has remained as high as four hundred and eighty eight maternal deaths per 100,000 live births in 2008/9, an increase from four hundred and fourteen per 100,000 live births in 2003, five hundred and ninety per 100,000 in 1998 (KDHS, 2008-09). Figure 1.2. Most maternal deaths are due to causes directly related to pregnancy and childbirth, unsafe abortion and obstetric complications such as severe bleeding, infection, hypertensive disorders, and obstructed labor (KDHS, 2008-09). Improving maternal health being one of the eight Millennium Development Goals (MDGs) adopted at the 2000 Millennium Summit, and with only three years left until the 2015 deadline to achieve the MDGs, closer examination of maternal mortality levels is needed to inform planning of reproductive health programmes and to guide advocacy efforts and research at the national level. These estimates are also needed at the international level, to inform decision-making concerning funding support for the achievement of this goal.

Therefore this study focuses on the relationship between health expenditure and health outcomes in Kenya more particularly, how public health expenditure impacts on maternal mortality rates and other determinants of health outcomes.

Figure 1.1 Public Health expenditure trends in Kenya

Source: Kenya Demographic Health Survey 2008/09

Figure 1.2 Trends in maternal mortality: 1990-2008

Source: Kenya Demographic Health Survey 2008/09

1.1.2 Public Health expenditure in Kenya

Adequate resources are critical to sustainable provision of health services. The government remains the major financier of health care, meeting nearly half of the national health recurrent expenditure. The Kenya policy framework of 1994 identified several methods of health services financing, including taxation, user fees, donor funds, and health insurance. These methods have evolved into important mechanisms for funding health services in the country.

The GOK funds the health sector through budgetary allocations to the MOH. However, tax revenues are unreliable sources of health finance, because of macroeconomic conditions such as poor growth, national debt, and inflation, which often affect health allocations. The government therefore works closely with development partners to raise money for the health sector. Donor contributions to the health sector have been on the increase, rising from eight percent of the health budget in 1994-95 to sixteen percent in the fiscal 2001/2002. In some years, donor contributions accounted for over ninety percent of the development budget of the MOH (Ministry of Health, 2006).

According to the 2001-2002 national health accounts (NHA), as cited by Wamai (2009) Kenya spends 5.1% of its GDP on health. He cited that the health budget had grown significantly from Ksh15.2 billion in Fiscal 2001/02 to Ksh34.4 billion in Fiscal 2008/09. He added that the proportion of overall government expenditure that the government spent on health declined over the same period from 9% to 7.9% in Fiscal 2006/07.

In 1992 a cost-sharing system was introduced to leverage more resources for health services (Collins et al, 1996). Revenue from the cost-sharing system increased exponentially from Ksh60 million in Fiscal 1992/93 to over Ksh1, 468 million in Fiscal 2005/06. However, the revenue’s overall share of total health expenditure for Fiscal 2005/06 was just 6.4% of the MOH’s total spending (MOH, 2007).

Figure 1.3: Overview of Kenya’s health budget, FY2002 – 2008 ( US$ million)

Source: Health Policy Initiative analysis of Ministry of medical services, 2008

Figure 1.4: Absolute value of Total Health Expenditure (THE) by financing source 2001-2010

Source: Kenya National Health Accounts 2009/10

Reviews of public expenditures and budgets in Kenya show that total health spending constitutes about eight percent of the total government expenditure and that recurrent expenditures have been consistently higher than the development expenditures, both in absolute terms, and as a percentage of the GDP.

Government financing of health expenditure is about sixty percent of what is required to provide minimum health services, implying that healthcare delivery in Kenya is under-funded (KHDR, 1999). This is accentuated by inefficiency of the system, including lack of cost-effectiveness in service delivery. However, preliminary information from Kenya’s national health accounts shows that the financial contributions of households (out of pocket expenses) exceed those of the government. (Collins et al. 1996)

The per capita expenditure therefore falls short of the Government of Kenya’s commitment to spend fifteen percent of its total budget on health, as agreed in the Abuja Declaration. The under-financing of the health sector has thus reduced its ability to ensure an adequate level of service provision to the population (Collins et al. 1996). The national health concern therefore is the extent to which additional health expenditure on specific care programmes like maternal health will promote /increase benefits of the patients through improved outcomes in health (decline in maternal mortality rates).

1.1.3 Maternal healthcare in Kenya

Improving maternal health is one of the eight Millennium Development Goals (MDGs) adopted at the 2000 Millennium Summit. The two targets for assessing progress in improving maternal health are reducing the maternal mortality ratio (MMR) by three quarters between 1990 and 2015, and achieving universal access to reproductive health by 2015. With only three years left until the 2015 deadline to achieve the MDGs, closer examination of maternal mortality levels is needed to inform planning of reproductive health programmes and to guide advocacy efforts and research at the national level. These estimates are also needed at the international level, to inform decision-making concerning funding support for the achievement of this goal.

Good maternal health is crucial for the welfare of the whole household, especially children who are dependent on their mothers to provide food and care. Prevention of the death of a mother is the single most important intervention for the health of a child.

Women are intensely vulnerable to the effects of costs incurred during childbirth. User fees can be especially high for emergency or technological procedures such as caesarean section, sometimes reaching catastrophic amounts, which push families into poverty (Graham and Newell, 1999). Many women often leave the hospital before they are well enough for discharge because they cannot pay for the care they have received. User charges add to the costs of transport and companion time, which can be substantial for those living far from facilities. The time spent looking for cash can also delay access to emergency life-saving care in facilities (Kunst and Houweling, 2001).

In sub-Saharan Africa, one in sixteen women die in pregnancy or childbirth (WHO, 2001). An estimated ten to twenty million women develop physical or mental disabilities every year as a result of complications or poor management (Ashford, 2006). The long-term consequences are not only physical, but are also psychological, social, and economic. Despite the commitment expressed with the Millennium initiative, maternal health has not been given financial priority internationally. Safe motherhood programmes compete for funding with other priorities such as tuberculosis, malaria and HIV/AIDS.

1.2 Statement of the problem

In Kenya, as in most Sub-Saharan African countries, health care expenditure has steadily increased over time, therefore making its containment a major issue for successive governments. The existence of a large public deficit and the need to reduce it drastically

to comply with the requirements of the AU has added importance to controlling health care expenditure.

Financing health care has remained a challenge to the Government of Kenya for a long time. There seems to be very low investment in health by the government, and inappropriate allocation of resources within the government health budget. In Kenya, health is a basic human right and therefore the health situation in Kenya remains a significant concern for the policy makers. The cost of health care, especially maternal health is a heavy burden on households. While health financing has undergone numerous reforms, more changes are needed to ease the burden of maternal health care costs on households in a bid to increase utilization and subsequently improve the health status of the population.

In Kenya, like in most developing countries, maternal health care program encompass a medical condition that is regularly associated with death. The maternal mortality rates are very high. The major concern in this study is therefore the change in patient improvement due to additional expenditure on maternal health care (reduced maternal mortality rates). It analyzes whether increasing health care expenditure towards maternal health care program will reduce the maternal mortality rates.

1.3 Objectives of the study

The broad objective of this study is to analyze the relationship between health care expenditure and maternal health outcomes in Kenya. The specific objectives of this study are:

To identify the determinants of maternal health in Kenya.

To investigate the relationship between government expenditure on maternal health care and maternal health outcomes

To make policy recommendations based on study findings

1.4 Significance of the study

A key factor that has contributed to the declining health outcomes has been the decline in annual real per capita government budget to the health sector. As noted earlier, the actual expenditures fall below budgetary allocations. With respect to this, policy makers are highly interested in the relationship between expenditure on public health and the resultant health outcomes/benefits. The issue is whether extra spending on maternal health leads to better maternal health outcomes.

From a policy perspective, this study can help set priorities on resource allocations across specific program of care. For instance it can help the government to know whether additional expenditure on maternal health care will reduce maternal mortality rates in the country. The government is able to set its priorities right whether to invest more on these specific care programme or to reduce its expenditure given the severe budgetary constraints. It also gives policy makers some guidance on appropriate cost containment measures that will help improve health system performance in Kenya. It is also very useful at the international level, to inform decision-making concerning funding support for the achievement of the fifth millennium development goal.

This study also adds to the existing literature on the relationship between health care expenditure and health outcomes, determinants of health outcomes and how health outcomes can be improved.

CHAPTER TWO: LITERATURE REVIEW

2.1 THEORETICAL FRAMEWORK

Healthcare is an intermediate good that has no intrinsic value in itself but has value in its contribution (along with other inputs such as environmental and social factors) towards production of health itself. Health, or in general, health status, refers to measures of the physical and emotional well-being of an individual or a defined population. The quantity of healthcare ‘product’ produced by a healthcare ‘firm’ is referred to as its output. The ultimate output of the health sector is health. Healthcare therefore can be viewed as any other good or service in which each individual maximizes utility subject to a budget constraint.

The basic economic theory of production provides a basis on the linkage of health expenditure and health outcomes. The theory suggests that there are many ways inputs can be used in various proportions to produce outputs (Wolfe, 2002). Inputs refer to the resources needed to carry out a process or provide a service. Inputs required in healthcare are usually financial, physical structures such as buildings, supplies and equipment, personnel, and clients while output refers to the direct result of the interaction of inputs and processes in the system; the types and quantities of goods and services produced by an activity, project or program. The use of inputs in healthcare leads to outcomes. (Cremieux et al. 1999).

Health production theory highlights the manner in which health care as an input is related to health as an output. In this theory, health is the output measured in terms of improved health status such as reduced mortality, morbidity or achieving health related millennium development goals while inputs consist of the number of trained health professionals, the number of school years completed, residential place, the proportion of GDP spent on health and the government health expenditure in the health sector (Desai, 1998).

Health production theory utilizes the health production function which is the change in health status affected as an approximate matter by changes in the consumption of various health services effective in improving health. The production function summarizes the relationship between inputs and outputs with health status being the dependent variable (function of healthcare) dependent on population’s social and environmental factors, policy variables and country specific effects inclusive of biological endowment, and lifestyle.

Many studies on this subject have adopted Grossman’s (1972) model of health production which views each individual as both a producer and a consumer of health. It regards health as a commodity which the individual will wish to consume and maximize, subject to his/her budget constraints, in conjunction with a number of endogenous and exogenous variables which have an impact on individual’s health. Within this model, income and educational level play an important role as explanatory variables.

In Grossman’s model, he regards health care as both a consumption good that provides direct satisfaction and utility, and as an investment good, it provides satisfaction to individuals indirectly through reduction in sick days, increased wages and increased productivity. In this case, health can be viewed as a stock which degrades over time if there are no investments in the individual health, and that health is taken as a sort of capital. Investing in health may seem costly as individuals must trade off resources and time that may be devoted to health, unlike other goals. These factors are also used in determining the optimal level of health that is needed by an individual. The model therefore makes predictions on the likely effects of health care price changes and other goods, outcomes in labor market such as technological changes, wages and employment.

In the Grossman model, at the optimal level, health investment occurs where the marginal cost of health capital is equal to the marginal benefit. Over time, health is likely to depreciate at a certain rate which may be denoted by δ. The consumer faces an interest rate which may also be denoted by r. By adding these variables, the health capital marginal cost can be calculated as under:

δ

In this case the health capital marginal benefit is the rate of return from this capital in both non market and market sectors. In this model, the health stock at optimal level can be caused by factors such as education, wages and age. The theory further advocates that investing in health should be combined with other factors which are crucial in order to produce new health, which in the long run may offset the process of deterioration in the stock of health.

Medical scientists could argue that only effective medical care should be universally available (OHE, 1979). The government therefore may resort to explicit rationing which is not only to set limits on total expenditure for care, but also to develop mechanisms to arrive at more rational decisions as to relative investments in different disease specific programmes, and the establishment of certain minimal uniform standards. This rationing does not guarantee mothers to equal access to appropriate maternal/medical care. Treatment is still within the postulate that the doctor will do his best with the resources available to him but there is no such constraint on those resources as government decides (OHE, 1979).

This study looks at maternal health as the “output” of an aggregate production which utilizes variables such as public health expenditure, access to government medical services and household incomes as the “inputs”. The assumption is that for reasons associated with diminishing returns and the adverse effects of certain variables after an initial positive outcome, the relationship is expected to be nonlinear (Nixon and Ulman, 2006).

2.2 EMPIRICAL LITERATURE

Health status are commonly measured using four major indicators, maternal mortality, mortality rate in infants, mortality rate for under five and life expectancy at birth (Akinkugbe et al. 2009); (Gupta et al. 1999); (Wang,2002); (Imam et al. 2003). Other measures of health outcomes/status used include preference of cancer or circulatory diseases, disability adjusted life years, quality adjusted life years, fertility indicators and achievement of other health related millennium development goals. Similarly, government health expenditure, GDP per capita, female literacy, number of physicians, immunization coverage, urbanization and calorie intake among others are some of the most used explanatory variables (Wolfe, 1986); (Wang, 2002); (Or, 2000b); (Caldwell, 1990) and (Filmer et al. 1999).

Most studies have used cross-sectional analysis (Bokhari et al.2007); (Imam et al. 2003); (Anyanwu et al. 2008); (Gani et al. 2009); (Wang, 2002); (Nixon and Ulman, 2006) and (Martin et al. 2009). Some have used panel data (Gupta et al. 1999) and (Or, 2000b), while Akinkugbe et al. (2009) used time series analysis to estimate the relationship between the public health expenditure and health outcomes. To solve the problem of autocorrelation in cross sectional analysis, heteroskedasticity test was done, corrected standard errors for panel data analysis while augmented Dickey Fuller tests were used to test for stationarity in all studies using time series data.

All studies reviewed used health expenditure as one of the explanatory variables except Wang, (2002) who looked at it in a different perspective. According to him, demand for electricity, access to piped water and sanitation and female education increases health expenditure but it does not increase public health expenditure in improving health outcomes.

Most studies indicated that public spending contributes significantly to health status improvements (Filmer et al. 1999); (Abel Smith, 1963); (Kiymaz et al. 2006); (Ester et al. 2011); (Gakunju, 2003); (Bokhari et al.2006); (Anyanwu et al. 2005); (Cremieux et al. 1999); (Nixon and Ulman, 2006) and (Blendon et al. 2006). For example, Filmer et al. (1999) used data from the early 1990s and estimated multivariate regression model of child mortality on per capita income, government health expenditure and other controls. They found that there was significant correlation between child mortality and income per capita.

Some studies however indicated that public health expenditure alone as a determinant of health is inadequate (Ogbu and Gallagher, 1992); (Castrol-leal et al. 1999); (Gupta et al. 2003); (Anderson and Frogner, 2005); (Hitris and Posnet, 1992); (Caldwell, 1986); (Dor et al. 2007) and (Cochrane et al. 1978).

In estimations, different methods were used by different authors. Generally two main methods were used: generalized least squares and the ordinary least squares. However, other methods have also been used. For example, Bokhari et al. (2006) and Gupta et al. (1999) used two stage least squares because of the instrumental variables used to address the problem of reverse causality and measurement errors in the variables. Anyanwu et al. (2005) used Robust Ordinary Least Squares as the baseline specification and robust two stage least squares to control for endogeneity and reverse causality. Bhalotra (2007) used the linear probability model.

Particularly, Flippi et al (2006) took a broader perspective on maternal health and drew attention to the economic and social vulnerability of pregnant women. They called for action to reduce maternal mortality rates by channeling more resources towards maternal healthcare, improving on human resources and information. They used maternal mortality ratio (by cause) as the major indicator and recommended that research is needed on how to finance health services and ensure equitable access to generate more evidence.

While examining the association of the socio-demographic characteristics of women and the unobserved hospital factors in Kenya, Magadi et al. (2001) used multilevel logistic regression. The results showed that the probability of maternal mortality depends on both observed factors that are associated with a particular woman and unobserved factors peculiar to the admitting hospital. The individual characteristics observed to have a significant association with maternal mortality include maternal age, antenatal clinic attendance and educational attainment. The hospital variation was observed to be stronger for women with least favorable socio-demographic characteristics. For example, the risk of maternal death at high-risk hospitals for women aged thirty five years and above, who had low levels of education, and did not attend antenatal care is about two hundred and eighty deaths per a thousand admissions. The risk for similar women at low-risk hospitals is about four deaths per a thousand admissions.

In a study carried out on health care services and sources of revenue in six countries from Western Europe and North America, Abel Smith (1963) found that health care expenditure was associated with reduced life expectancy and increased mortality rates. In a similar study carried out in the year 1967 involving twenty nine countries he found that the level of national income was associated with improved health status and that the demand for healthcare increased in countries with declining mortality. Abel Smith’s studies laid down foundation for the development of methodologies for tracking health expenditures in both private and public sectors.

While investigating the factors that are associated with infant mortality in Sub-Saharan Africa, Ester et al. (2011) carried out an ecological multi-group study using a bi-variate and multi-variate analysis with infant mortality rate as the dependent variable. They used a linear regression model between infant mortality rate and the correlated indicators (social security expenditure and government expenditure per-capita on health). This study revealed, in the multi-variate analysis, three factors associated with the IMR: a higher social security expenditure on health as a percentage of the general government expenditure on health, a higher per-capita government expenditure on health and a higher number of children under five years of age with diarrhea receiving oral dehydration therapy indicated a lower IMR.

During the examination of the effectiveness of public social spending on education and health care in several African countries, Castro-Leal et al. (1999) reviewed the benefit incidence of government spending in Cote d’ivoire, Guinea, Kenya, Madagascar, South Africa and Tanzania. Their study found that public expenditures on health were not sufficient especially on the poor to reduce mortality rates. On the other hand, Gupta et al. (2003) used cross-country data for over seventy developing countries to assess the relationship between public spending on health care and the health status of the poor. Their findings confirmed that the poor have significantly worse health status than the rich. The results however suggested that increased public spending alone will not be sufficient to significantly improve health status.

Another study carried out on the health effects of per capita income and public expenditure on social services in Kenya, proved that per capita income had been very influential in determining health status. The study found that expenditures on education and health care improved health status at a great margin. It further established that per capita income was significantly linked to the levels of mortalities, and that some of the negative trends in health status could have been attributed to unfavorable growth and insufficient social spending on health (Manyala, 2000). In his findings income elasticities were all statistically significant, current income had the expected effect on life expectancy but not on infant mortality. He further found that if mothers are malnourished and are in poor state of health, their infant will inherit part of this poor health, and therefore will be at greater risk of mortality relative to infants of healthy mother.

A comparative study by Wagstaff (2002a) that focused on forty two developing countries used child mortality due to malnutrition and diarrhea as the health outcomes/indicators. Wagstaff (2002b) treated government health expenditure as an exogenous variable and found that it did have a statistically significant (negative) coefficient. The study used a simple stylized theoretical model rationalizing the health-income relationship and found that public spending on health care had a larger impact on child mortality among the poor than among the non-poor population.

In his study on health and schooling investments in Africa, Schultz (1999) found that health status rose with increased public spending on health services. He also argued that the health status will tend to decline with a rise in relative prices of health inputs such as salaries of medical personnel, cost of drugs and other medical supplies, relative to prices of nutrients that help fight infections and disease. He also found that levels of education were correlated with lower mortality rates. The relationship between mothers’ education and mortality rate was stronger than the fathers’. He recommended that an additional year of schooling to the mother especially in low-income countries was associated with a five to ten percent reduction in mortality rates.

On his analysis of the factors determining health status in Kenya, Gakunju (2003) found that government expenditure on public health was noteworthy in shaping individual health status. He also found that government health expenditure influences health status with over a long time. This actually implies that the government investment and spending in the health sector have had a major effect on the health of the people. He also acknowledged a number of factors as being important in resolving the health problem Kenya such as: Per capita income, individual access to doctors, HIV/AIDs prevalence, literacy level for women, Child immunization coverage and spending/investment by the government in the health sector. His study majorly used the central government expenditures on health to explain the status of health among the people.

While studying the major effects and determinants of expenditure on health in the developed countries, Hitris and posnett (1992) did a cross-section analysis and found no evidence of relationship between healthcare expenditure and mortality rates. They found that reduced healthcare expenditure resulted into reduced life expectancy and increased death rates among children less than a year old. On the other hand, Bokhari et al. (2006) found evidence linking public expenditure and per capita income to mortality rates.

A study carried out on Turkey analyzed the relationship between health care expenditure and GDP and population growth. Kiymaz et al, (2006) used annual time series of private and public data and employed a multivariate co-integration technique to analyze their data. They found evidence of long term multivariate co-integration relationship between health care expenditure and GDP and population growth rate.

Another study also examined the link between healthcare spending and health outcomes across ten Canadian provinces over fifteen year period 1978-1992. Their estimated regression equation consisted a mixture of potentially endogenous/independent variables (such as the number of physicians, health spending, alcohol and tobacco consumption, and expenditure on meat and fat) and exogenous /dependant variables (such as income and population density). Cremieux et al. (1999). In his simultaneous analysis he found a strong correlation between health spending and outcomes. He found that a greater percentage of families living below the poverty level were associated with greater infant mortality even though life expectancy was unaffected. There was an established relationship between a country’s overall health and its citizens’ longer life. He however found no such relationship in infant mortality. It limited both specification bias and data heterogeneity.

While analyzing international health care organizations and financing, Dor et al, (2007) carried out a research on financing end stage renal disease which is a medical condition of chronic kidney failure. They did a comparative review of case studies in twelve countries in which almost all countries, end stage renal disease is publicly financed. They considered mortality rates for males and females as the outcome. They found no evidence of a relationship between higher spending and better outcomes. They analyzed the link between resources and incentives and outcomes. They used input price parity in their methodology and assumed that all countries had the same cost so even these used the cost from the US.

Nixon and Ulman (2006) also while trying to examine the correlation between expenditure on health care and health outcomes/benefits in various countries of the European Union over the period 1980-1995 using life expectancy and infant mortality as the dependent variable and lifestyle environmental and occupational factors as independent variables. They conducted an econometric analysis using a fixed- effect model on a panel data set. They found out that healthcare expenditure was considerably linked to hefty improvements in infant mortality but only slight in life expectancy.

In his process of trying to compare between parametric and non parametric estimation techniques in estimating the relationship between income and health expenditures, Martio (2003) used a three time series cross section data and carried out OLS estimates and locally weighted scatter plot smoothing (LOWESS) technique in his regression. He found that health expenditure elasticity depends not only on the level of analysis but also the range of income and economic development an economy finds itself at; as incomes rise health expenditures become more income inelastic. The implication is that health care needs do not necessarily consume a rising share of national output. As income rise, the health expenditure to GDP ratio is likely to stabilize.

Blendon et al, (2006) seek public opinion of the Americans’ view about their health care system. A majority of American wanted more money to be spent on healthcare by the government due to rising costs of medicines. During their study, they queried Americans especially about the overall national spending on health care and government’ spending on national health care. The majority of respondents, (57%) thought that the United States was spending too little on health care. A majority of them believe that hospitals charges the price of prescription drugs were unreasonably too high especially where someone has a chronic illness and therefore felt that the government should spend more to subsidize these.

Wilton and Smith (2002) wanted to assess the econometric evidence of the impact of budgetary models using pooled cross sectional time series data for the United Kingdom (UK), United States of America (USA), New Zealand (NZ), Germany and Netherlands. They used physician and pharmaceutical expenditure models in the data analysis and the econometric results suggested that the influence of GDP expenditure growth was positively correlated with primary care expenditure, suggesting that as countries experience faster rate of growth, more can be afforded to the provision of health care services.

Caldwell (1986) examined cross-national differences in child and infant mortality. His study was majored in poor countries and he established that there was an enormous gap between the apparent potential of public spending to improve health status and the actual performance. He also found that public spending appears to explain little differences across countries in infant and child mortality which are well explained by economic and social factors. He included dummy and hypothesized that these countries will have higher mortality. He also found that both aggregate and household size show that higher levels of female education are associated with better health status.

In a study carried out in England, Martin et al. (2009) used program budgeting data for two hundred and ninety five primary care trusts to examine the link between spending and outcome. The study had shown that health care expenditure had a positive effect on outcomes in five of the care program that they investigated (that is, for cancer, circulation problems, respiratory problems, gastrointestinal problems and diabetes). In their study they used budgeting data for 2006/07 and mortality data for the period 2004-06. The estimates confirmed that the marginal cost of a life year saved was quite low and that the findings were not confined to cancer and circulation problems. It provided evidence that expenditure on the various disease categories yielded quite consistent benefits in terms of life years saved. In particular, they used 2SLS models to investigate the relationship and found a far more marked influence of healthcare spending on health outcomes than is often indicated by more conventional analysis.

Cochrane et al. (1978) on their study on health service as an input and mortality as an output applied regression analysis to the statistical relationship between mortality rates and per capita consumption of inputs such as health care provisions by doctors, nurses, acute hospital beds, pediatricians and midwives. They found that the indicators of health care were generally not associated with outcomes (in this case mortality rates).

While estimating the relationship between the level of life expectancy and infant mortality rate, Bidain and Ravallion (1997) used a multivariate regression to explain health outcomes without any health sector variables. They found a consistent impact of public spending for the poor unlike the non-poor. According to them, reductions without reallocation of resources would affect the poor. The impact on the poor versus the non-poor was also not a constant parameter but it depended on the composition and efficacy of public spending. They estimated a multivariate regression model.

Anyanwu et al. (2005) did an econometric analysis of health expenditures and health outcomes in forty seven African countries between 1999 and 2004. They used infant mortality and under five mortality as the health outcomes. They found that health expenditures had a statistically significant effect on infant mortality and under five mortality.

On the other hand, Amagnionyediwe (2009) evaluated the impact of government health expenditures on the poor in Nigeria. From the descriptive analysis the study found that the health status of the average citizen and the condition of health infrastructure had not improved appreciably despite government spending (though with little fluctuations) on this sector. Thus, he concluded that there is the need for the public sector to, not only, improve its health care expenditure but also put into productive use the available funds in the health sector. Also, the result suggested that public spending on health had a consistent and significant influence on child mortality and therefore government health care spending should be made more productive and accessible. This should not solely be on increasing the number of health care facilities, as this does not necessarily translate to increase in the health status of the population, emphasis should be on the various ways of improving health care facilities, as these will enhance both the scope and quality of health care services. Furthermore, government resources need to be reallocated towards health intervention designed to respond primarily to the health needs of the poor. Government should also ensure that health interventions reach their intended beneficiaries. He used child mortality rates as the health indicators and initial primary school enrollment ratio, public spending on health per capita and mean consumption per person as explanatory variables. He used secondary data and used OLS to estimate his regression equation.

2.3 OVERVIEW OF LITERATURE

As demonstrated in the above literature, issues of healthcare expenditure have been of interest for decades. On examining the countries studied, it can be seen that the majority studied various combinations of developing countries of Central and South America. Studies based specifically on African countries are very limited.

In terms of modeling techniques, most studies used cross-sectional analysis (Bokhari et al.2007); (Imam et al. 2003); (Anyanwu et al. 2008); (Gani et al. 2009); (Wang, 2002); (Nixon and Ulman, 2006) and (Martin et al. 2008). Some used panel data (Gupta et al. 1999 and Or, 2000b), while Akinkugbe et al. (2009) used time series analysis to estimate the relationship between the public health expenditure and health outcomes. To solve the problem of autocorrelation in cross sectional analysis, heteroskedasticity test was done, corrected standard errors for panel data analysis while augmented Dickey Fuller tests were used to test for stationarity in all studies using time series data. In some cases the modeling incorporated shift dummies to account for fixed effects within the sample, for example, in investigating heterogeneity due to country-specific effects or the impact of health care system or social insurance.

The principal results showed that health expenditure was a significant explanatory variable for most health outcomes examined (Filmer and pritchet, 1999); (Abel Smith, 1963); (Kiymaz et al. 2006); (Ester et al. 2011); (Gakunju, 2003); (Bokhari et al.2006); (Anyanwu et al. 2005); (Cremieux et al. 1999); (Nixon and Ulman, 2006) and (Blendon et al. 2006). Other studies found that income was a significant explanatory variable while others did not find health expenditure to be significant when controlling for income (Ogbu and Gallagher, 1992); (Castrol-leal et al. 1999); (Gupta et al. 2003); (Anderson and Frogner, 2005); (Hitris and Posnet, 1992); (Caldwell, 1986); (Dor et al. 2007); and (Cochrane et al. 1978). From the review of the literature, no conclusive evidence appears to exist regarding the contribution of healthcare expenditure on health outcomes. This may be due to country-specific characteristics.

The studies reviewed have used different variables to establish the link between healthcare expenditure and health outcomes (infant mortality rate, maternal mortality rate, morbidity etc). Healthcare expenditure in Kenya has been of interest, considering the fact that there have been rapid increases in healthcare spending in over time. Most of the studies reviewed in Kenya are confined to the impact of GNP per capita on health status (Gakunju, 2003 and Magadi et al. (2001). This study will attempt to arouse more interest in this area. This study will also make comparisons between its results and the results of other similar studies done before especially Kargbo (2010).

CHAPTER THREE: METHODOLOGY

3.1 CONCEPTUAL FRAMEWORK

The figure below conceptualizes how public spending influences health outcomes(maternal health) besides other factors like household income and access to public medical services.

Figure 3.1: Conceptual Framework

What is maternal health? What is its value to the country? E.g. attainment of MDGs

Micro-economic evaluation at treatment level e e.g. cost-benefit analysis on the amount used to deliver maternal care.

What influences maternal health? (Other than health care) Occupation, level of education, income, access to medical care etc.

Public spending on maternal health

Health status/outcomes

Medical and biological progress

Increased government expenditure on maternal healthcare will lower maternal mortality rates. However, establishing the causal relationships between health expenditure and health outcomes is complex considering the fact that other socio demographic and clinical factors contribute to maternal mortality.

3.2 MODEL SPECIFICATION

Based on the literature in section two, more interest is focused on infant mortality and under-five mortality rates as the indicators/measures of health outcomes. This study uses maternal mortality rate as an indicator or measure of health outcomes because it is a key indicator of a country’s progress in improving health, and it forms the basis for one of the United Nations’ millennium development goals. This will be useful as a process indicator in guiding policy development towards the attainment of the fifth millennium development goal.

The study uses a combination of utility maximization methods as earlier introduced by Grossman (1972) model on the correlation between health status and other socio-economic factors. In this utility maximization model it is assumed that individuals wish to maximize utility from a stream of health services, in order to achieve good health status. Maternal mortality rate in this case is the dependent variable while independent variables are public health expenditure, access to medical facility and doctors (proxied by distance to nearest health facility (5km or more)] and other household income. The functional model is specified as:

MMRi =expenditure on health (EH), household income (HHI), diagnosis by medical worker in hospital (DMWH), place of delivery, (PD) error term)……….(i)

Where MMR is the maternal mortality rate.

3.2.1 The model specification

The model can therefore be specified as:

MMRi = EHi+ HHIi+ AMi+ εui………………………………………. (ii)

Where MMR is a measure of health outcomes in Kenya (maternal mortality rate); EH is the expenditure on health; HHI is the other household income; AM is the access to medical facility and ui is the stochastic variable.

,, and are the respective coefficients (parameters to be estimated).

Equation (ii) above is transformed to logs to linearize the equation since past studies have shown a non-linear relationship between health status and its determinants (Nixon and Ulman, 2006) and to allow comparisons with earlier results. The equation can be transformed as:

lnMMR = β0+β1lnEH+β2lnHHI+β3lnAM +ui……………………………………..(iii).

3.3 HYPOTHESES

The hypotheses are formulated as follows:

Null hypothesis H0: β1, β2, β3, = 0, Maternal Mortality rate is not related to public health expenditure, access to medical facility and doctors or other household income.

Alternative hypothesis H1: β1,β2,β3 ≠ 0 Maternal Mortality rate is dependent on public health expenditure, access to medical facility and doctors and other household income.

3.4 Descriptions of the variables used and justification for their use

Maternal Mortality rate (MMR): Is the number of maternal deaths in a given period per 100 000 women of reproductive age during the same period of time. This study has taken pregnancy related illness in seventy Kenyan districts to proxy this variable due to availability of data on this indicator.

Public expenditure on health (EH): This is the government expenditure on healthcare. It is the recurrent and capital spending from government (central and local) budgets, thus the actual value of money allocated for use in the health sector. It is a good measure of the actual government’s performance in terms of improving health outcomes.

Other household income (HHI): Is a measure of the combined incomes of all people sharing a particular household or place of residence. Is the combined gross income of all the members of a household who are at least fifteen years old and older. The variable was chosen because it is taken as a good measure of standard of living and thus determines the kind of healthcare that a mother is likely to seek at a certain level of income

Access to health facility and doctors: Is the distance to nearest health facility, in this case 5km or more.

3.4 Data management

The study used stata statistical package to analyze data because it is fast, accurate and easy to use. This was done by regressing health outcomes (proxied by pregnancy related illnesses) on the explanatory variables that influence health outcomes (public health expenditure, distance to nearest health facility (5km or more), other household income, diagnosis by medical worker in hospital and place of delivery). Ordinary least squares (OLS) method of estimation was used to find the value of the parameters.

Cross-sectional data from seventy districts for twelve months was used for the analyses. The period of study was chosen because of the availability of data during the same period. The data was obtained from the Kenya Integrated Budget Household Survey and various publications from the government ministries of Finance, planning and of Health.

CHAPTER FOUR: ANALYSIS OF FINDINGS

4.1 Introduction

This chapter presents results in different sections. The first section presents basic exploratory analysis while the other section has results of regression analysis. Data from all the districts in Kenya (totaling to seventy districts) were used in the study.

Description of the data used in this analysis

The Kenya Integrated Budget Household Survey (KIHBS), 2006, was collected by KNBS in seventy districts in Kenya from about 13,430 households for twelve months which helped to inform planners on poverty, living standards among other social economic indicators. The respondent households are cluster sampled from all districts and 861 rural and 482 urban clusters. The variables selected from KIBHS (2006) for our study are pregnancy related illness, expenditure on health, distance to nearest health facility (proxied by access to hospitals and doctors) and household income. Quality control measures cut across research design, sampling methods to include data collection and reporting which assures accuracy of final database.

Summaries and descriptive statistics

The study conducted exploratory checks including summaries, normality tests and white test for heteroskedasticity. Baringo, Samburu, Mandera, Kutui and Turkana districts have some of the lowest percentage of births in hospitals yet still the pregnancy related morbidity is below 1%. But there is one alarming case that is Wajir district. The district has the lowest proportion of hospital births (at 3.1) and highest record of pregnancy related illness (3.1%). So for Wajir, it is almost a forgone case that increased investments in formal health care provision may lower maternal deaths.

Correlation between the percentage of deliveries that were attended by a doctor and the percentage of hospital births was 0.72, which is very high. Use of the two variables together as covariates can cause multicollinearity. The study therefore, opted to use hospital births instead of doctor attended births since the former made more practical sense in terms of accessibility (ratio of doctors to patients is very low countrywide). Besides, hospitals are a better indicator of health care investment than availability of doctors given that they are durable and immobile as well as health infrastructural assets indicator.

Concerning distance to nearest health facility, the KIBHS survey revealed that, countrywide, only 11.3 per cent of Kenyans traveled a kilometer or less to reach their nearest health facility. At the same time, almost half travel for 5 kilometers or more. Indeed, the descriptive analysis (Table 4.1) reveals that, 55.5% travel for at least five kilometers [1] . The best rated is Nyeri district with only 10% of households located a similar distance (clearly better than Nairobi at 20.2%) between them and health facilities. worst off districts are Marsabit, Tharaka, Bomet, Kuria, Malindi and Lamu with 100, 100, 96, 92, 89 and 81 percent of households located at least 5km away from health care centers respectively. Income level of household was measured by other HH Income besides salary from employment, agriculture and business sources. Machakos district records highest proportion of households with extra incomes (34.6%) whereas Wajir has the lowest (0.4%).

Exactly 0.42 percent of respondents had reported cases of pregnancy related illness. There were several districts with zero cases whereas Wajir district had the highest prevalence of pregnancy related morbidity at 3.1 percent. To capture expenditure on health, the study used proxy variables (diagnosis by a doctor and access to hospital for delivery). Nationally, a quarter of respondents were assisted by a doctor to deliver. Concerning distance to the nearest medical facility, 54% of respondents had to travel a distance of at least 5 kilometers to access medical care. About 10% of households received other income besides salary or agricultural income. Such incomes (say social transfers) are an indication of economic need. Place of delivery 28% of households had their delivery from a hospital environment.

Table 4.1: Summaries for the general sample and sub-samples with pregnancy illness of over 0.5% and below respectively

Mean

Mean for pregnancy illness dummy =1

Mean for pregnancy illness dummy =0

Pregnancy related illness

.42

1.32

.06

Distance to nearest health facility

25.77

62.34

51.81

Other income of household

54.95

7.82

11.55

Diagnosis by a medical worker in hospital

10.42

30.36

23.76

Place of delivery is a hospital

28.28

25.95

29.3

Source: Analysis from KIBS 2006

When we separate the districts with extreme prevalence of pregnancy illness (over 0.5%) from those with mild frequency, we realize that households in the former category are located a greater distance from health centers, they receive less transfers (other income), and they have less deliveries in a hospital even though they report higher diagnosis by doctors.

Figure 4.1: Comparison of various aspects

Source: Analysis from KIBS 2006

Except for proportions of households which were assisted by a doctor and those who had their delivery in a hospital, all other variables had a statistically significant difference at 10% level of significance. This means that, districts that have higher pregnancy related morbidity also have higher diagnosis rates by medical workers in hospitals, have less incomes, and longer distances between households and health care centers.

Results of the OLS regression

Given the continuous nature of the dependent variable, and also that the variables are found to be normally distributed whereas error terms adopt uniform variance, we choose to conduct an Ordinary Least Squares. The results of the White test are shown below.

White’s test for Ho: homoskedasticity

against Ha: unrestricted heteroskedasticity

chi2(14) = 18.17

Prob > chi2 = 0.1989

Cameron & Trivedi’s decomposition of IM-test

—————————————————

Source | chi2 df p

———————+—————————–

Heteroskedasticity | 18.17 14 0.1989

Skewness | 11.03 4 0.0263

Kurtosis | 2.87 1 0.0903

———————+—————————–

Total | 32.07 19 0.0307

The null (Ho) and alternative (Ha) hypotheses of the white test are such that:

Ho:, implying presence of constant variance in error terms

Ha: , denoting absence of constant variance in error terms

The criteria for evaluation is such that, if the calculated Chi- square statistic is greater than critical (or if p> χ2), we may reject Ho. In this case we sustain Ho of constant variance; hence we do not have a problem of heteroskedasticity.

Table 4.2: Regression of pregnancy related illnesses against covariates

Estimated Coefficient

t-statistic

Significance

Level

P>|t|

Constant/Intercept

-.022

-0.06

0.95

Distance to nearest health facility

.006

1.82

0.07

Other income of household

-.015

-1.27

0.21

Diagnosis by a medical worker in hospital

.016

2.66

0.01

Place of delivery is a hospital

-.007

-1.40

0.167

N = 65; F (4, 60) = 4.38; Prob > F = 0.0036; R2 = 0.2260; Adj R2 = 0.1744

Source: Analysis from KIBS 2006

The F statistic is 4.38, significant at 1% level, 4 numerator and 60 denominator degrees of freedom. F-statistic is a measure of joint significance; therefore, the prevalence of pregnancy related morbidity is affected by distance to the nearest health facility, other incomes of household, access to medical worker for diagnosis and access to hospital as a place of delivery.

Thus we are able to substantiate our earlier stated hypotheses in favor of the alternative: Alternative hypothesis H1:, , Maternal Mortality rate is dependent on the explanatory variables. Specifically, higher expenditure on health, higher household incomes and greater access to medical facilities reduces maternal mortality.

The estimation achieved an adjusted R-squared of 0.1744 which means that from the population, 17.4% of variance in pregnancy related illnesses are determined by changes in the covariates [2] .

Of all explanatory variables, only distance to nearest health facility and diagnosis by a medical worker in hospital are significant. specifically, for every extra percentage in number of households living 5 or more kilometers away from a medical facility, the maternal mortality increases by .006 percent, ceteris paribus. The implication is that distance-hampered access to medical facilities increase maternal mortality. Also, for every extra percent of households that are diagnosed by medical workers in hospitals leads to 0.016 percent rise in maternal mortality, other things held constant. This is a finding that seems contrary to expectation unless one argues that, hospital health care providers are more apt in diagnosis of pregnancy illnesses than those in smaller health centers.

The effect of income is to reduce maternal mortality. Thus a percentage increase in proportion of households that receive extra incomes leads to a -.015% decrease in maternal mortality, ceteris paribus. Similarly, delivering in hospitals reduces maternal mortality. Hence a percentage increase in households that go to hospitals for delivery leads to a -.007% decline in maternal mortality. Finally the intercept denotes the natural incidence of maternal mortality when the effect of other variables is assumed zero. The only undesirable thing is that both income and hospital deliveries are not statistically significant. However it only means that the estimated coefficients are not consistent in variant samples. Rather, it doesn’t prove that the variables in the model are useless parse.

CHAPTER FIVE: SUMMARY, CONCLUSION AND POLICY IMPLICATION

5.1 Summary

This study aimed at investigating the correlation between expenditure on health care and the resultant health outcomes/patient benefits in Kenya. Based on a background of low investment in health relative to other sectors and budgetary items (such as recurrent spending), the study undertook to examine the effect of government expenditure on maternal healthcare on maternal mortality rates and to identify other determinants of maternal mortality rates apart from public health expenditure. Literature reveals that investing in human capital, specifically health sector, reduces maternal mortality for such economies. Such health care investments range from private sources to government sources for different countries. Variables such as income, and distances traveled for health care access are also suggested as suitable influences of maternal mortality. The study develops an empirical relationship between maternal mortality (proxied by pregnancy related illnesses), incomes, access to hospitals and doctors and distances travelled to medical facilities. Analysis was done by descriptive, comparison of means and OLS. The next section discusses implications of findings on policy.

Discussion and conclusion

Distance to the nearest health facility is a significant determinant of pregnancy related illnesses. Hence it stands out as an important variable influencing maternal mortality. Similar findings have been arrived at by Mwabu and Wangombe (1998) using cross sectional primary data in Kenya. From our findings over half of the sampled households travel for at least five kilometres to arrive at the nearest health care centre. It is alarming to note that 100% of households in Marsabit and Tharaka Districts cover even larger distances. It is notable that transport infrastructure in most interior/remote districts of Kenya (Like Marsabit and Wajir) are almost totally absent. In addition, the transport costs incurred to arrive at the nearest health facility is an extra challenge that households have to bear, more so those located in excess of five kilometres from the health centres.

The study also revealed that diagnosis by hospital medical practitioners as opposed to other health care centres (dispensaries, clinics etc) is not only significant but also has direct impact on pregnancy related morbidity. Our point of view is that access to professionals in higher level facilities can help to diagnose and even contain maternal mortality. But could there be instances where medical practitioners are available but drugs are not? How about availability of tools, equipment and technology? At the same time there is huge variance in terms of access to facilities and medical resource persons. From the study findings, Kirinyaga district had the highest access (70%) while Samburu and Kajiado reported the lowest (8%). Such im