CHILDHOOD MORTALITY AND POVERTY IN KENYA: EVIDENCE FROM 2015/2016 KENYA INTEGRATED HOUSEHOLD AND BUDGET SURVEY

DECLARATION

This is to declare that this project is my original work and has not been presented to any other university for any kind of academic award.

Signature……………………………              Date ………………………………….

KONGO SCHOLASTICA KINGI

Q56/6912/2017

This project has been submitted with my approval as the university supervisor

Signature…………………………………     Date…………………………………………

Prof. Alfred O. Agwanda

Population Studies and Research Institute

University of Nairobi

Signature …………………………………… date ……………………………………….

Dr. Anne Khasakhala

Population Studies and Research Institute

University of Nairobi

Table of Contents

DECLARATION.. 2

LIST OF TABLES. 6

LIST OF FIGURES. 7

CHAPTER 1.0. 9

INTRODUCTION.. 9

1.1         Background Information. 9

1.2         Problem statement. 10

1.3         Research Questions. 11

1.4         Objectives of the Study. 11

1.5         Justification. 11

1.6         Scope. 13

1.7         Limitations. 13

CHAPTER 2.0. 13

2       LITERATURE REVIEW… 13

2.1         Introduction. 13

2.2         Trends in Childhood Mortality. 14

2.2.1          Childhood Mortality in Africa. 15

2.2.2          Analytical Framework. 18

2.3         ESTIMATION OF CHILDHOOD MORTALITY. 20

2.3.1          Techniques to estimate child mortality. 21

2.4         Summary of Operational Framework used. 27

2.5         Definition of variables. 27

CHAPTER 3.0. 29

3       METHODOLOGY. 29

3.1         Introduction. 29

3.2         Source of Data. 29

3.3         Method applied. 29

3.4         Data quality. 30

3.4.1          Assessment of data quality. 31

3.4.2          Whipple’s Index. 31

3.4.3          Myers Blended Index. 32

3.5         Data Analysis. 33

3.5.1          Method of Estimating Under Five Mortality. 33

3.5.2          Method of Estimating Mean Age of Childbearing. 38

3.5.3          Method of Estimating the relationship between Under Five Mortalities and Absolute Poverty Level 39

3.6         Assumptions. 39

CHAPTER FOUR: 41

4       LEVELS OF UNDER-FIVE MORTALITY BY COUNTY. 41

4.1         INTRODUCTION.. 41

4.2         Mean age of childbearing by county. 41

4.3         Differentials in Proportion of Children Dead By Women. 43

Under Five Mortality Rates. 45

4.4         Regression Analyisis. 47

4.5         Correlation between Under-5 Mortality Rate and Poverty. 50

4.6         Discussion. 51

5       Reference. 53

LIST OF TABLES

Table 1: Levels of Infant and under-5 mortality rates in Kenya, 1962-2014  8

LIST OF FIGURES

Figure 1: framework showing the operation of the five groups of proximate determinants on the health dynamics of a population.

Figure 2: Modified Conceptual Framework

CHAPTER 1.0

INTRODUCTION

1.1         Background Information

Childhood mortality is one of the key indicators used in determining both the general and national socio-development index. In the past ten years, the international community has set and periodically reviewed child mortality targets (Ferrarini and Norström, 2010). Under-five years’ old mortality has regularly been renewed as part of the Sustainable Development Goals (SDGs) pioneered by the United Nations (United Nations Department of Public Information, 2008). Among the 17 sustainable goals aims, the third one is concerned with the reduction of neo-natal mortality. The objective of this third target is to see neo-natal death reduce to 12 per 1000 live births. It also aims to reduce mortality for children below five years to less than 25 per 1000 live births (Pillay, 2019). Mortality is a significant indicator of a society’s wellbeing. There have been many social and economic developments in the 21st century that have an impact on social services. Lately, there has been a general decrease in mortality rates in the world partially due to improvements in standards of living, sanitation, advancements in employment, provision of healthcare and sanitation facilities, availability of safe water, as well as the provision of affordable housing (Romani & Anderson 2002).

Child death is a huge burden in Africa where most people are young (WHO 2005; World Bank, 2006). Childhood mortality varies across the globe, but it is highest in sub-Saharan Africa and South Asia (Heckmann, 2015). For instance, more than 50 percent of childhood deaths occurred in sub-Saharan Africa in 2017. Further, Southern Asia accounted for 30 percent of global childhood mortality in the same year. UNIGME (2018) found that 38 percent of all childhood deaths occur in countries listed among the least developed in the world. High childhood deaths in sub-Saharan Africa may be owed to an increase in fertility in this region. Childhood deaths in the region have grown from 30 percent in 1990 to 50 percent recorded in 2017. Projections indicate that sub-Saharan Africa will be accounting for 60 percent of global childhood deaths by the year 2050 (Renschler et al., 2015).

Childhood mortality rates in Kenya differ greatly depending on the region where a child is born. The fewest deaths in the under-five years demographic occur in the central province, where there are 42 deaths in 1000 live births. Conversely, Nyanza recorded the most number of under-five years of mortality level, with as many as 82 deaths per 1000 live births (Mwangi and Murithi, 2015). Nairobi region also has high childhood mortality rates as it ranks second among the regions with high childhood mortality within Kenya.

Several World Health Organization publications have focused on the issue of childhood mortality. This includes a WHO report in 2005 which highlighted the causes of childhood mortality. The report linked childhood deaths to several conditions (World health report, 2005). While progress is being made in fighting child mortality, far too many children do not live to age of five. UN IGME (2018) reported around 5.4 million deaths among children below the age of 5 in 2017 alone. Almost 50% of the deaths took place in Africa.

1.2         Problem statement

Estimation of U5MR is highly dependent on data from vital registration systems containing national reported birth and death histories. However, most of the developing countries have little or no records on such data yet many childhood deaths occur in those nations. Data to estimate U5MR is instead provided for by the full or even the summary birth histories. The Kenya Demographic and Health Survey 2014 is the latest source of data on the Full and Summary Birth Histories. The purpose of this study is at determining whether data obtained through the KDHS 2014 is relevant enough to give plausible estimations of the U5M mortality at county level in Kenya. This study aims at estimating the under-five mortality rate in Kenyan counties in which these statistics will come from. Therefore, a comparison of the U5M across different counties is made possible. This finding is based on the latest available data. Also, the association between poverty levels and childhood mortality in the counties is also determined.

1.3         Research Questions

These research questions will be vital in addressing the issues mentioned above:

1. What are the levels of under-five mortalities at the county level in Kenya?
2. Are the estimates of U5M at county levels plausible from KIHBS data?

1.4         Objectives of the Study

This study will focus on how poverty impacts childhood mortality rates in Kenya at the counties or sub-national level.

The specific objectives are:

1. To estimate U5M at National and county levels using partial-birth history based on the KIHBS data.
2. To determine the quality of KIHBS data.
3. To examine the extent to which data from KIHBS is sustainable for estimating U5M at the county level.

1.5         Justification

This study is useful in reporting childhood mortality in the Kenyan context. The study provides useful data that can be used to evaluate the effectiveness of government programs such as family planning and free maternal healthcare. Child and infant deaths are significant contributors to total mortality rates, although both occur relatively early in life. However, childhood mortality rates are significantly higher in third world countries and are caused by factors that differ from those that cause death later in life. Measurements of childhood mortality are vital indicators of the socio-economic status of a nation or region. Information on mortality can be used for research and planning purposes. Mortality statistics are utilized for health planning, population projection, assessment of the failure or success of specific service- provision, among other uses. Mortality statistics are also forerunners for epidemiological research on the causes of high mortality in some regions. Policymakers believe that fertility can be reduced by reducing childhood mortality. Children are a form of social security in developing countries and losing one is devastating.

According to Chandrashekar (1972), most old people depend on their children as a matter of family and filial obligation. Consequently, many families in developing countries opt to have many sons to increase the chances of one surviving. This implies that reduced child mortality may be effective in lowering birth rates in third world countries. The findings of this study can be used to identify issues of public health that contribute to high childhood and infant mortality rates. For instance, the study can be utilized to convince the government to pool resources and come up with effective solutions in reducing child mortality in Kenya.

Up-to-date data and information on the trends and causes of child deaths are vital in guiding global strides aimed at improving child survival. It is a priority under the Sustainable Development Goals (SDGs) to end preventable deaths of infants and children under the age of five years. Under the developmental goals, a target of reducing child mortality to as low as 25 cases per 1000 live births is set to be achieved when there is quality data to know the trends and focus areas. Similarly, to obtain reliable time series estimates of under-five mortality rates depends on the data collected in both censuses and general surveys.

The complete recording of deaths in the death recording systems in all countries is vital to achieving the SDG child mortality reduction goals. However, most developing countries have incomplete death registrations, especially on the under-five cases. The high cases of child mortality in the WHO African countries are a testament to the lack of quality prevention programs that should arise from forecasts and estimates produced by updated and quality data. This study is essential in realizing the gap by using available data sources and determining whether they are plausible in the estimation.

1.6         Scope

This study focusses on the estimation of the under-five mortality at the county level in Kenya. The study uses data from KDHS 2014 to determine the mortality rate of children who are 5 years or below in age. In so doing, this study determines whether the available sources provide data that can provide plausible estimations.

1.7         Limitations

1. Specific birth rates were not available hence were borrowed from published data of the DHS 2014.
2. The study worked on the assumption that estimates would change much yet the time difference between surveys was large.

CHAPTER 2.0

2           LITERATURE REVIEW

2.1         Introduction

This section gives an overview of the previous literary works on child mortality and poverty. The chapter provides the framework on the topic for the specific case study comprising of the main research focus. The context of the review of previous research studies is set by providing: the purpose of the review for the research’s case study, remarks on different methods which have used in the estimation of the under-five mortality in the world and also specific to developing countries, and how the scope of the work is presented in this chapter.

The trends and determinants of childhood mortality are discussed in this chapter with a broader perspective of African nations. Hodge and Jimenez-Soto (2013) suggest that the globe has seen a significant decline in child mortality. SDG countries have seen a 50 percent decline in child mortality since 1990. Contrastingly, UNIGME (2018) reports that the majority of child deaths occur in sub-Saharan Africa. Similarly, projections of the UN indicate more than half of the world’s under-five deaths will be occurring in sub-Saharan Africa by 2050 (Okiro and Ayieko, 2018). Research studies need to be undertaken to determine the trends, causes, and mitigation strategies against child mortality in less-developed countries (Okiro and Ayieko, 2018). The procedures for the direct and indirect approaches in estimation under five mortality rates are discussed in this chapter.

2.2         Trends in Childhood Mortality

The world has made notable progress in reducing child and infant mortality. Millions of children have made it to five years since efforts targeting child mortality started in the 1990s. Under-five mortality rates have declined drastically from 93 deaths in each 1,000 live births in 1990 to 39 in every 1,000 live births today (Hodge and Jimenez-Soto, 2013). This is a 58 percent reduction. In 1990, 1 in 11 children would not reach the five-year milestone. Currently, only 1 in 26 children die before their fifth birthday. A 50 percent reduction in under-five mortality has been recorded in the majority of SDG countries since 1990. Under-five deaths have reduced by at least 75 percent in 74 countries since 1990 (Hodge and Jimenez-Soto, 2013). Most impressively, some 33 countries classed as low-income countries were able to reduce under-five mortality by 75 percent in the same period. By 2017, deaths among under-five children had reduced to 5.4 million per year from 12.6 million in 1990. On average, 34,000 children were dying daily in 1990, but this number has dropped to 15,000 children in 2017 (UNIGME, 2018)

2.2.1        Childhood Mortality in Africa

While other regions across the globe have attained impressive progress towards minimizing the under-five mortality, this has not been the case in the majority of Africa. The region still reported the highest under-five mortality, with 76 deaths in every 1,000 live births in 2017. Besides, 1 in every 13 children will die before attaining the age of five. In contrast, only 1 in 185 children die before their 5^th birthday in most of the high-income nations (Kyei, 2012). This means that child mortality is 14 times higher in Africa in comparison to developed countries. In Australia and New Zealand, child mortality is as low as one child in every 263 (UNIGME, 2018). The majority of child deaths occur in two of the most underdeveloped regions in the world. In 2017, sub-Saharan Africa accounted for 50 percent of childhood deaths, while southern Asia accounted for another 30 percent of deaths occurring before the age of five (Okiro and Ayieko, 2018). Least developed countries accounted for the majority of under-five deaths, with 38 percent of deaths occurring in these countries. This has an impact on the increase in child population and a shift in demographics. UN projections indicate that 60 percent of under-five deaths will be occurring in sub-Saharan Africa by the year 2050. Gender disparities in childhood mortality rates are almost zero. In the past, boys had higher probabilities of dying before attaining the age of five (Okiro and Ayieko, 2018). However, the probabilities of girls dying in some countries are now higher than that of boys. Countries with these peculiar child mortality rates patterns are mostly on Western and Southern Asia.

Despite the recent interventions in Nigeria, the Under-Five Mortality Rates (U5M) remain high. According to the NDHS or Nigeria Demographic Health Survey conducted in 2013, A U5M of 128 deaths in every 1000 live births was witnessed, which meant that on average, one per every eight Nigerian children dies after his 5^th  birthday. This is 21 times the expected rate in first world countries. Further, for two decades that is (1990 to 2005), child mortality rates have significantly dropped by 57%. That is averagely a reduction from 126 to 69 deaths in a population of 1000 births. U5M fell by an approximate percentage of 49%, which means it dropped from 213 to 109 deaths (Opiyo & Sawhney, 2014). Deaths of children arise from risk factors and diseases that are preventable. Disease morbidity and mortality among children results from malaria, chronic malnutrition, Acute Respiratory Infections (ARI) and diarrhea. Endogenous status of an infant, the quality of antenatal care, post-partum and whether during delivery assistance was offered have been linked to the causes of death for the first 28 days of a human being. As they grow older, especially from 11 months onwards, a child’s death is linked to nutritional practices, environmental factors, the status of the household and health behavior (Opiyo & Sawhney, 2014).

Morakinyo (2017) asserts that factors like breastfeeding practices, fertility behavior, ethnicity, religion, child’s sex, maternal education and place of residence, among others, have also been contributing towards children’s mortality. Been from the richest quintiles’ households reduced chances of U5M by 0.092, 0.050 and 0.057 in 2003, 2008 and 2013 in comparison with children from poor backgrounds (Morakinyo, 2017). Under-five years of deaths were also prevalent among children from uneducated mothers. Mother’s education, her wealth status, her religion, her residence zone, her weight at birth, her media use, the type of cooking oil she used and the type of toilet she used were primary predators of under-five mortality in many surveys conducted throughout the years. Some child health factors are limited to some countries. The Nigerian Demographic & Health Survey (NDHS) indicated that sources of drinking water, a child’s sex, age of the mother and marital status are vital determinants of child mortality. While considerable gains have been made in many sub-Saharan countries, they are not enough to attain the MDG 4 goal. The new targets under the SDG call for countries to renew their focus on the children. To achieve these targets, the drivers of controllable deaths in infants and children have to be closely monitored. This study will provide useful knowledge that can be the foundation of efforts to reduce controllable deaths of infants and under-fives. The study will include population-based statistics on the drivers and trends of infant, neo-natal and under-five mortality in over a decade in Kenya. In 1999, Kenya recorded a spike in childhood deaths in the background of a steady decline as indicated by data collected in Kenya Population and Housing Censuses between 1962 and 1989.

Figure 1: framework showing the operation of the five groups of proximate determinants on the health dynamics of a population.

Source: (Mosley and Chen, 1994)

2.2.2        Analytical Framework

Studies by Osita et al. (2015), Omolo (2014) and Mwangi et al. (2015) depict that child survival depends on economic, social, environmental and cultural conditions. The studies highlight key aspects such as mothers’ personal hygiene, education, water supply, health status, education, household economic stability and health expenditure directly affect infant mortality.

According to Caldwell (1979), there was an association between child survival and maternal education in Nigeria. Also, the study established that higher education levels among mothers reduced infant mortality rate via factors such as improved number of hospital delivery, changing the old family relationships and enhanced antenatal care among pregnant mothers. The study concluded that changing care and feeding practices result in improved health-seeking that is stirred via a mother’s education.

Further, Medrano et al. (2000) and Kosvsted et al. (2003) established that maternal education and religion are vital in measuring the degree of health knowledge in a household and with better knowledge and accommodative religion good health for a child is attainable. Conversely, Beenstock and Sturdy (1990) argued that the mother’s education had a very minimal impact on childhood mortality in Africa.

In their study Mwangi and Murithi (2015) they established that childhood mortality was low in Central, Eastern, Coast, Rift valley and Nairobi while it was high in North Eastern, Nyanza, Western and Eastern provinces. Kabubo-Mariara et al. (2012) established that children in rural areas are more subjected to poverty hence are likely to die compared to those living in cities and urban areas. This finding was later supported by Mwangi and Murithi (2015) when they concluded that child mortality is higher in rural areas than in cities due to inadequate health facilities.  Osita et al. (2015) also supported this view by arguing that there was a need to improve child health care in rural areas in an attempt to minimize childhood mortality.

Both Kaldawei and Pitterle (2011) and Elmahdi (2008) identified malnutrition, injury and diseases as the primary factors impacting childhood mortality. The duo agreed that breastfeeding was a key factor in measuring child mortality. However, they could not establish the difference between rich and poor mothers since the rich mothers are extremely busy to have enough time for their babies while the poor mothers are unable to afford good medical care and healthy nutrition for the child and also themselves.

As per Jones et al. (2006), tetanus, diarrhea, Pneumonia, bacterial infections and premature deliveries are the major factors that lead to childhood mortality in India. Omolo (2014) argued that children delivered in public health centers experienced a lower mortality rate compared to those delivered in private health centers in Kenya. Contrary, Mwangi and Murithi (2015) established that children delivered in private health centers had a lower mortality rate than those delivered in public hospitals. They argued that private hospitals enjoy better health workers, drugs and facilities than the public health facilities.

Hosseinpoor et al. (2005) argued that there was a need for more interventions towards sanitation and the environment as a way to minimize childhood mortality. Besides, Alves and Belluzzo (2005) established that the mortality rate in Brazil was influenced by hygiene at the environment and household levels. Finally, Uddin et al. (2009) stated that infant mortality was high among mothers who had not attended antenatal visits across Bangladesh.

Hill et al. (2001) in their study concluded that HIV/AIDS epidemic was among the top causes of increased childhood mortality. This was contrary to the earlier research works that argued that the main causes of child mortality were demographic and socioeconomic factors.

2.3         ESTIMATION OF CHILDHOOD MORTALITY

Data Sources

Mortality rates can be estimated in various ways depending on the source of data. Most of the organizations apply a common method in estimating Under-five mortality across countries to obtain valid comparisons. Data are obtained from censuses and surveys and administrative sources such as vital registration. When using vital registration, mortality rates are obtained from an abridged life table having a standard period that uses age-specific deaths and mid-year populations. Census and survey data on mortality of children take two different ways: the full birth history, women are queried the birth date of each child they have borne, whether the children are alive, and otherwise, the age at death; and the summary birth history, women are asked the number of children they have ever borne and those that have died. Either approach results in respective child mortality estimates referring to some time frame prior to the date of the survey.

Life tables are used to calculate childhood mortality. There are two main categories of lifetables: abridged (period) life tables and model life tables. Model life tables are developed on empirical studies of age-specific death patterns. They are commonly used to calculate demographic parameters for nations having limited data. The model life tables give the relative probabilities at specific ages. A model life table also gives different levels of mortality corresponding with life expectancies at birth. A desired characteristic such as under-five mortality of the corresponding life table is found by observing the table at the particular level in the model life table. The tables use stable population data to illustrate relationships among different variables. Stable populations give the ultimate effects of a fertility and mortality schedule.

The application of the model life tables assumes that the age-mortality pattern of the studied population is similar to that of the life tables. Besides, population projections employing the factor  assume that the age distributed within every five-year interval of the population that is stationary is the same as that of the projected population. One major limitation of using model life tables is that people in a population are from different cohorts containing varying mortality experiences, whereas information from various cohorts is as if combined into one table. The disparities in the trends mortality across cohorts can affect life table values and, therefore, provide excess mortality measures.

2.3.1        Techniques to estimate child mortality

According to past literature, the two main approaches used in estimating the mortality rates are the direct and indirect methods. The direct techniques of estimating the rates utilize data regarding the date of birth of children, the status of their survival, the dates when deaths occur, or the ages of the deceased children at death (Kovsted et al. 200). On the other hand, indirect methods employ information on the status of survival of children to particular cohorts of mothers. The cohorts could be identified typically with age or the time since the mothers’ first birth. The direct methods use data that are obtained from specifically designed surveys containing histories of birth and vital statistics (Uddin et al. 2009). However, the vital statistics systems are reported to be deficient in developing countries. The indirect methods can use data collected from censuses and other general surveys (Uddin et al. 2009). This study cites the procedures to estimate child mortality documented in DHS publications (Sullivan et al., 1994; Rutstein and Rojas, 2006)

Direct Estimation

Direct approaches used to estimate child mortality basing on household surveys use birth or pregnancies’ histories to obtain data to calculate the indicators. Information on each birth or pregnancy of the respondent include:

• Year and month of birth for each child;
• The sex of every child;
• Whether the child is alive or dead;
• The age of each child who is alive;
• Date or age at death of each dead child;
• Outcome of each pregnancy, such as, still birth, miscarriage, live birth.

The collection of birth or pregnancy histories is done in full or truncated forms. The full histories include all live births and pregnancies by the woman being interviewed by the survey date. The data is collected chronologically from the first pregnancy to the last. Truncated histories include pregnancies or births that have occurred during a certain fixed period of time. The data is collected reverse chronologically from the date of the survey. The birth history information are arranged in data files having each record to describe a single birth. Other vital variables are the sample weight and the date of the interview.

Direct estimation variants

The three essential variants of the direct approaches for computing childhood mortality include:

1. An approach basing on vital statistics whereby the number of deaths of children who are below age 12 months in a specific time frame are divided by the total births in the same period.
2. An approach involving a true cohort life table whereby a division deaths of children below 12 months from a particular cohort of births by the total births in that cohort.
3. An approach basing on a synthetic cohort life table whereby probabilities of mortality for small age segments basing on real cohort mortality are aggregated into common age segments.

Calculation Algorithm

The calculation algorithm used in direct methods tabulates the numerator first then followed by the denominator. Death is tabulated by identifying the lower limit of the age group ( ) that the child belonged at the time of death ( ) and ( ) is the period in which the child occurred. In particular, ( ) . The algorithm will then tabulate the denominator representing the exposure experienced by the child in his/her life. It follows the particular child from the time of birth to the child’s death (if the child died before age 5 years), then to age 5 years (if the child is 5 or older or death occurred at age 5 or older) or to the interview date (if the child is less than 5 years). The algorithm determines the time period reached by every age group the child was alive. Identification of whether the upper bound of the age group is in the same or the following time period is done. After tabulating the numerators and denominators based on the age group and time period, calculation of component probabilities of death in each age group ( ) for each time frame is done by the division of numerators by the denominators. The under-five mortality rate ( ) is calculated as follows:

Where x represents the age groups 0, 1-2, 3-5, 6-11, 12-23, 24-35, 36-47, 48-59.

Indirect Estimation

Indirect estimation techniques use indirect data from the recorded number of children born and those surviving or dying. There are two types of indirect methods which include model-based technique obtained from Brass (Brass and Coale 1968), and the empirical approach developed by Rajaratnam et al. (2010) at IHME. The Brass method utilizes the mother’s age in approximating the child’s average length of exposure to a risk of dying. The three variants of the indirect methods are:

1. Estimation of child mortality categorized by the mother’s age (AGE).
2. Estimating child mortality categorized by the duration of marriage (DOM).
3. Estimating child mortality categorized by the time since the first birth (TSFB).

The data needed for the indirect estimation is for:

1. Children ever born
2. The total children who have ever been born, or
3. The average number of children who have ever been born.
4. Children who are surviving
5. The whole number of children who are surviving, or
6. The average number of children who are surviving

• The proportion of the surviving children

1. The number of dead children, or
2. The average number of dead children, or
3. The proportion of the dead children
4. The number of women
5. The overall number of women including those t1hat have never married (AGE)
6. The number of women who have ever been married (DOM)

• The number of women who have given birth (TSFB).

Calculation algorithm

The steps in computing the indirect estimates are as follows:

1. Sum the number of children ever to be born and those surviving by age group, or marriage duration or the period since first birth.
2. Compute the mean parity per woman by group:

Where is the parity for age group ,  represents the number of children who have ever been born for age group , and  represents the total number of women that particular age group.

3. The proportion of dead children by group is calculated:
4. Multipliers are calculated then followed by the probabilities of dying which are computed as shown below:

Whereby  represents the probability of dying at age x exactly.

5. The reference period is calculated and finally the values of for each group are transformed into , , and for every reference point. The transformation is made possible by applying linear interpolation between the levels of the model life table. TheU5MR is then calculated by:

Assumptions of the indirect estimation

• The respondent reports accurate data on CEB and CS.
• There is knowledge on the fertility and mortality shapes.
• The levels of fertility and mortality have been constant in the last 15 years.
• Mortality conditions are homogeneous- namely, similar mortality risks are exposed to children given birth to by women in different ages, duration of marriage or the period since the first birth.

A common weakness of the indirect approaches is that censuses and general surveys are not designed to collect data specifically for estimation of mortality hence have reports of dead children omitted (Uddin et al. 2009). Elsewhere, there are cases where stillbirths and live births have both been used to answer the question on the number of children ever born resulting in the overestimation of mortality rates. This study uses the indirect estimation variant basing on the age of women to estimate under five mortality. According to Hill and Figueroa (1999), there is a high mortality experienced by children born to mothers in younger age groups such as 15-19 and 20-24. The woman’s age variant has a limitation of overestimating the mortalities in the younger age groups.

2.4         Summary of Operational Framework used

This study used a three-phase operational framework. Phase 1 entailed a preliminary study where the key concepts and background of the study were defined and a sketch of the entire research methodology was established. The second phase entailed the theoretical study, where different pieces of literature were reviewed and associated with the research gap in line with the research objectives. This part involved relating different concepts, theories and frameworks that are associated with childhood mortality rate and absolute poverty. Finally, phase 3 involved the practical study that is the actual data analysis or processing, the establishment of findings and discussion of the results after the study. This part also involved giving recommendations and drawing conclusions.

2.5         Definition of variables

Childhood mortality: entails death among children aged zero days to 5 years. It consist of Child mortality (q4), Infant Mortality and Under-5 Mortality. However, in this research work, q5 will be used to represent childhood mortality.

Under-Five (U5M or q5): the chances of dying between zero days and 5 years after birth.

Child mortality (qi): the chances of dying between 1 and 5 years after birth.

 Socio-economic factors: the current communal and individual conditions or status of relevance such as mother’s poverty levels, education, residence and region.

Absolute Poverty: also known as the overall poverty lines for rural and urban areas in monthly adult equivalent terms as Ksh. 3,252 and Ksh. 5,995, respectively.

Region of Residence: refers to the 47(forty-seven) counties that are Kenya is administratively divided into.

Place of Residence: distinguishes between rural and urban areas.

CHAPTER 3.0

3                                                                METHODOLOGY

3.1         Introduction

This chapter defines the structure of the research procedure to be used is presented. It includes the scope, sampling procedure, data quality issues encountered and methods of data analysis that have been utilized to arrive at the study’s results.

3.2         Source of Data

The data uses in this study was secondary data that was sourced from the 2015/2016 Kenya Integrated Household Budget Survey (KIHBS). Data utilized in the study included statistics on child deaths and childbirths in 12 months between August 2016 and September 2015. The survey design deliberately captured a family range of health status indicators such as demographics, household income, expenditure and consumption patterns, among other indicators.

3.3         Method applied

The indirect method of calculating mortality rates was used to establish the levels of under-five mortality rates. This method operates under the assumption that mortality and fertility patterns have remained consistent in the period of the study. The study required three sets of data, including the number of women by age group, children dead by the age of mother (CD) and Children Ever Born (CEB) by the age of the mother. From the 3 sets, this method changes the percentage of dead among the CEB to these women into conventional measures of mortality.

The parameters estimated from it are q1, q2, q3, q5, qlO, ql5 and q20. However, only q5 is considered in this study.

The DHS survey illustrates a typical estimation of the mortality rates for five years, i.e., 0-4 years, 5-9 years. These estimates are given by women aged 15-49 based on their births and infants deaths. Naturally existing is the bias arising from incomplete and data that is not fully representative. No data is available for dead women since only surviving women aged between 15-49 years are interviewed. There will be bias on the estimates of mortality in case a difference is seen between children born to surviving women and those born to women who were dead at the time of the survey. However, any method of estimating childhood mortality by mothers that are dependent on retrospective reporting is biased.  Women above 40 years of age are left out of the survey and cannot contribute information on the exposure of deaths of their little ones for periods before the interview. This censoring of information and the resulting potential for bias becomes more severe as mortality estimates are made for periods more distant before the survey. To reduce the methodological limitations of child mortality in this report, a period of 15 years is restricted before the survey is conducted.

3.4         Data quality

Errors are prone to data of any kind, especially those arising from faulty respondent recall and history of births. Another compromise in the data quality is event omission where children who died after surviving for a few hours or days; this data is not always included in the survey figures.  In some cultural setups, purposive underreporting is associated with emotional events. This error is related to faulty respondents. The completeness of reporting deaths is investigated by using internal consistency to find out whether there was an underreporting of deaths. Errors resulting from Respondent’s recall can also result in misreporting of date of births and age at the death of children. In KIHBS, the maximum age for the women there is 49 years. This means that the birth history data for previous periods are confined to mothers who are younger at the time of birth. Since the mother’s age at the time of giving birth affects the survival of the child, mortality rates could be biased since only the younger women are interviewed. Further, the data of birth history is limited to the experience of children for mothers who are alive. The fact is that children of dead mothers may be subject to greater mortality risks. The more time in history dates back to a great extent; the greater is the proportion of high-risk children who are represented by data of the birth history.

3.4.1        Assessment of data quality

Women in different parities did not report the data on CEB and CD. This leads to biases in the denominator and numerator downwards when computing the proportion of CD and numerator downwards when computing the average number of CEB to a woman. To correct this, proportions from the 2009 Census will be used to calculate the missing number of children ever born and children dead assuming that they remained constant over the years.

3.4.2        Whipple’s Index

This index determines age heaping or digits preference on either the last digits 0, 5 or even a combination of both digits in age ranging between 23 and 62. This index changes between 100, standing for no preference for zero or five, and five hundred demonstrating that only digits 5 and 0 were reported. This method excludes extremely old and early childhood age brackets that are affected by different reporting errors rather than age preference. The mean values of the Whipple’s index vary between 100 and 500 based on the assumption that the total number of individuals either rises or declines linearly with an increase in age. The index is developed to determine the degree of preference for individuals with ages that end in 5 and 0. It is calculated in single years as a ratio of individuals aged between 25 and 60 with their ages ending in 5 or 0 as a proportion of 1/10^th or 1/5^th of the total number of individuals between 23 and 62 randomly (Kyei, 2018). In the event there is avoidance or dislike of ages ending in 5 or 0, then this index ranges from 0 to 100.

Table 1: UN Score for testing

Source: Shyrock and Siegel, 1976, Methods and Materials of Demography

From the results, the Whipple’s index for the females who reported ages ending with zero is 97.98, while those who reported ages ending with five is 114.95. The data for ages ending with zero is highly accurate while that ending with five is approximated.

3.4.3        Myers Blended Index

Myers Blended Index is computed to assess the level of misreporting and heaping of children’s birth year and women’s age which has a significant influence in the calculation of Total Fertility Rate. The measure is a variant of the Whipple Index. The measure shows whether the data has improved and quality has been achieved in different surveys.

This index is used for individuals with age greater than 10 years and it indicates the deficit or excess of persons with age ending in every 10 digits calculated as percentages (Pal et al., 2014). Myers blended index assumes that the target population is evenly distributed across different age brackets.

Table 2:  Calculated Myers Blended Index for Female Population

From the results above, there is an indication that the female age data had some anomalies. There existed age heaping at ages with particular terminal digits at the end, demonstrating avoidances or preferences in reporting the ages.  Ages ending with 1, 4 and 9 were avoided and this choice is not random, but people/response choose ” attractive” ages like those ending with “5” or “0” (Susuman et al., 2012). High anomalies in age recorded data for a woman might result from the reporting of proxy age data by a man who tends to be the head of household in most of the developing countries.

3.5         Data Analysis

3.5.1        Method of Estimating Under Five Mortality

The Brass indirect method of estimation, and in particular the Trussell variant for West model life tables, was utilized to determine the mortality rates for under-fives. The Brass method uses three main categories of information: the aggregate female population, the total number of children ever born, and the number of children dead.

Nature of data Required

Children ever born and dead

The information on the number of children ever born and those that are dead are obtained in a census or survey by asking women in the age range of (15 to 49) on their experience in childbearing. The set of questions asked include:

Set 1. How many children have you ever born alive?

Set 2. How many children have you ever born alive?

How many are alive?

Set 3. How many children alive do you have?

How many children did you have who later died?

Set 4: How many children do you have living with you?

How many of your children live elsewhere?

The Brass technique employs the technique that data consisting of children ever born and dead being categorized by the mother’s age. The classification is done using the traditional five-year age groups which run from 15-19, to 45-49. The age-groups are utilized in purposes of tabulation hence the data can be used for estimation.

The female population of reproductive age

The reproductive age of females used for the estimation is in the range (15-49) irrespective of their marital status. The information from the category often is a source of errors since the approach assumes that the data represents all women in the accepted range. The following section gives steps of computation using the Trussel version.

Procedure of Computation

Step1. Calculating average parity per woman

Average parity is the mean number of children who have ever been borne by women in an age-group of five years.

Step 1

Where  represents the average parity of females in the age group .   represents the children given birth by the women in the age group , and  represents the total number of females in the age group regardless of their marital status. Similarly, the value includes women who did not give their response to the question of children ever borne. The inclusion of the group is in line with the assumption that they have no children. Parity values are required for the age groups 15-19, 20-24, and 25-29.

However, calculations of the rest of the age groups can be calculated to assess the quality of the raw data.

Step 2. Calculating proportions of dead children to those living

The proportion is obtained by the ratio of the children who are dead to the total number of children who are ever born.

Whereby  represents the proportion of children who are dead for women falling in the age group , while  represents those children who have died as reported by the women.  represents the total number of children that the women have ever borne.

Step 3. Calculating multipliers,                  

The equation to estimate the multipliers according to the Trussel version is:

Whereby

The measure  which is the probability of dying exactly at age  is associated to the proportion of dead children  multiplied by a factor . The factor is calculated by the coefficients , , and  and the parity ratios  and . The coefficients are obtained by a regression analysis of model cases that are simulated.

Step 4. Calculating probabilities of dying by age , .

The calculation of  and separately for every age group provides the estimates of  by getting their product.

Step 5. Calculating reference dates for  and .

The reference time  can be obtained for each  from step 4 when conditions of steady mortality variation with the change in time. It is represented as the number of years prior to the survey and is estimated by using coefficients utilized on parity ratios. The estimation equation for the reference time is given by:

The calculated values of  can be changed in actual dates by getting a difference of the values from the reference date of the survey which are expressed as decimals.

Step 6. Converting to a common index

The fourth and fifth steps give estimates of  for ages of  of 1,2,3,5,10,15,20 and that for  which is the count of years prior to the survey. Each value of the estimated  is changed into a single measure in order to facilitate the comparisons within and between the data and also to analyze the trends. The common index proposed for the purpose is the probability of dying by age five .  Using of infant mortality is not recommended since the estimates of  are sensitive to the pattern of mortality underlying the varying models.

The values of  that correspond to the model-life-table family can be utilized to perform the needed conversions. The actual conversion is obtained by the linear interpolation between values that are tabulated. Supposing that an estimation of  denoted by  is converted to a corresponding value  where . For a model-life-table family, it is vital to find the levels of mortality with  values that are enclosing the estimated value, . Therefore,  and  can be identified from the table of annex I levels such that

Where  and  are the values of the model of  for levels  and . On the other hand,  is the value that is estimated. The desired common index is thus obtained by

Where  is the interpolation factor given as follows.

Given that the data on the number of children ever borne and children who have died are combined for both sexes, the values of the model  should be retrieved from the tables of the combined sexes in  annex I. Conversely, if the data are separate, the values of  should be sex-specific and the converting to a common index employs values of the model  from the relevant sex tables in annex I.

Step 7. Interpretation

After obtaining the seven estimates for each group  of the chosen common index, they are then plotted against time. The  values can be changed into reference dates as noted in step 5, and the values of the chosen index can be plotted against the dates. Graphical representation of the results is necessary to determine consistency and identify the general pattern of the estimates.

3.5.2        Method of Estimating Mean Age of Childbearing

The mean at first birth is the mean value of the age at which women give birth to their first child having gone through their reproductive years. Estimates of the mean age that women have their first birth are provided for countries with vital statistics that are accurate. However, less-developed countries like Kenya have incomplete vital statistics hence the estimates are not readily available. The estimates are, therefore, essential in assessing the commencement of childbearing in less developed. Similarly, the estimates are used to show the trends in the onset period of child bearing. The following equation is used in the estimation of the measure.

Where = the average childbearing age at time

= The proportion of women not having borne a child at age  and time

= proportion of women never to have given birth at

The estimates of  are obtained with a variant of the DHS standard approach used in estimating the proportion by single year instead of the five year age differences. However, the estimates could be subject to errors at ages higher than 40 since proportions of childless women at the ages are often less than 0.05 and similarly, there are usually fewer respondents. As a way to reduce the error effects of the average age at fifth birth,  is set at 40 years and  is approximated as the average of  values for single years between 35 and 45.

3.5.3        Method of Estimating the relationship between Under Five Mortalities and Absolute Poverty Level

The relationship between under-five mortality and absolute poverty level was estimated using the Pearson correlation. This measures the strength of association and the direction of the relationship between the two variables. The assumptions are that the two variables are normally distributed and the data is equally distributed about the regression line.

3.6         Assumptions

1. The rates of childhood deaths and fertility have not changed in the past.
2. Infant mortality and age patterns of fertility in the population are presented by model patterns utilized in establishing the method.
3. At no time will the children mortality fluctuate by 5 year age group of their mothers.
4. No relation exists between a mother’s survival in a population and the mortality risk of children.
5. Any alterations in the death rates of children in the previous years must have been slow and without a specific direction.
6. The effects of HIV/AIDs have already been incorporated in the estimation of mortalities.

CHAPTER FOUR:

4           LEVELS OF UNDER-FIVE MORTALITY BY COUNTY

4.1         INTRODUCTION

This chapter presents the results of the levels of under-five mortality at national and county level, by the residence and its relationship with absolute poverty.  The sections discuss the mean age of childbearing and its variability, the proportion of children dead by women, the under-five mortality by county and residence and the relationship between under-five mortality and absolute poverty measures.

4.2         Mean age of childbearing by county

Table 4.1 below summaries the mean age of childbearing that was used to calculate the childhood mortality by county.

Table 4.1: Mean Age of Childbearing by Residence/Region

Source: Computed from KDHS 2014, KNBS

Table 4.1 shows the mean age of childbearing by region or residence. The overall mean age of childbearing in Kenya is 24.63. The region with the lowest mean age of childbearing is Marsabit (25.96), followed by Nyamira (26.08) then Nairobi City (26.63). The counties with the highest MAC are Bungoma (29.12), Kilifi (29.04), Lamu(28. 89), and Garissa (28.83).  Differentials in the MAC are normally consistent with the fertility patterns. However, the ASFR for age groups 45 – 49 may be slightly biased due to truncation. At times the 40- 44 age groups are also affected.

4.3         Differentials in Proportion of Children Dead By Women

Table 4.2 below provides a synopsis of the differentials in the proportion of children dead by women.

Table 4.2: Differentials in the Proportion of Children Dead By Women

Source: Computed from KIHBS 2015/2016, KNBS

From Table 4.2, the average proportion of children dead by women in Kenya is 0.0530, with urban residence recording a mean of 0.0433, while rural areas are recording an average of 0.0572. In both the urban and the rural areas, the highest proportion of children dead by women were recorded in women aged between 45-49 years at 0.0748 and 0.0886 respectively. However, the lowest proportion of children dead by women in urban was recorded in women aged 15-19 years at 0.0226, and for rural areas was recorded in women aged 20-24 years at 0.0207.

Overall, the counties that recorded the highest average proportion of children dead by women were Migori, Mandera and Siaya at 0.1441, 0.1277 and 0.1191 respectively. Conversely, the regions that recorded the lowest were Laikipia, Meru and Nyeri at 0.0207, 0.0264 and 0.0276 respectively.

For the youngest age of 15-19, the regions that had the highest regional differentials in the proportion of children dead by women were Kiambu, Muranga and Nairobi at 0.3316, 0.2663 and 0.1545. Contrary, the regions that had the lowest proportion of children dead within the same age group were Samburu, Kajiado and Turkana at 0.0133, 0.0288 and 0.3316 respectively.  

A keener look at the table shows that majority of the counties have higher proportions of dead children for women in the age group 15-19 and 45-49. According to Manual X, there is an expected high number of child deaths among the two extreme age-groups. In the 15-19 age group, there is likely to be little prenatal care hence a high number of deaths of children. Similarly, several vulnerabilities in the age-group of 45-49 could lead to giving birth to underweight children which may cause deaths. Therefore, the data from KIHBS could be sustainable according to the mother’s age-groups experiencing higher proportions of dead children by women.

 Under Five Mortality Rates

The information presented in this section shows the probabilities of deaths by age five in different regions across Kenya. The results indicate a substantial difference between the counties.

Table 4.3: Under- Five Mortality Rates

Source: Computed from 2015/ 2016 KIHBS, KNBS

4.4         Regression Analyisis

The demographic transition states that there is a close relationship between TFR and childhood mortality

Using data from the DHS StatCompiler on the Total fertility rate and the Under-five mortality among 31 countries from Sub-Saharan Africa, a significant regression model is found to be

Below are the regression statistics obtained.

Table 4.4: Regression statistics

Source: Excel

The correlation coefficient between the two measures is 0.568709 implying that there is a relatively strong linear relationship between U5M and TFR in Sub-Saharan Africa. The R square value is 0.32343 indicating that at least 32.343 percent of the variation in the values of U5M can be explained by the model. Therefore, the model can be used to estimate the U5M values for the counties in Kenya. Below is a table showing the corresponding values of U5M calculated using the regression equation on the Total fertility rates of Kenyan counties.

Table 4.5: Predicted U5M using a regression model.

Source: TFR are from DHS 2014, U5M computed by author

Comparing Tables 4.3 and 4.5, there are significant differences in the number of U5M. For example, data from KIHBS shows that Mombasa county has 71 U5M while the information obtained after employing a significant regression model shows that the county has 49 U5M. The same can be noted for the other counties including the county hosting the capital city, Nairobi. Therefore, it is concluded that data from KIHBS regarding U5M is not plausible.

On average, the under-five mortality in Kenya is 48 deaths per 1000 live births, with the average in urban at 41 and in rural at 52 per 1,000 live births. The value is much lower compared to other values computed in other studies i.e., 2009 KPHC and 2014 KDHS. The 2009 KPHC recorded 79 deaths per 1000 live births, while the 2014 KDHS recorded 52 deaths per 1000 live births. The current rate suggests a 39% decrease from the 2009 census rate.

The highest under-five mortality is recorded in Migori, Madera and Wajir at 150, 131 and 106 respectively. The lowest under-five mortality is recorded in Meru, Bomet, Samburu and Laikipia at 12, 16, 17 and 18 respectively. Traditionally high childhood (IMR and U5MR) mortality rates are recorded in regions like Nyanza, Coast, Western and North Eastern (2009 KPHC, KNBS). However, some areas in Eastern and the Rift Valley have shallow mortality indicators that are inconsistent with subsequent development indicators across these regions. This points to probable under-reporting of actual deaths in such areas.

All counties have recorded a decline in the under-five mortality compared to the 2009 census. In the 2009 census, counties with the highest U5MR were Siaya (227), Kisumu (182), Migori (173) and Homabay (170). Currently, Siaya has 98 deaths per 1000 births, while Kisumu, Migori and Homabay recorded 59,150, 99 deaths per 1000 live births respectively.

4.5         Correlation between Under-5 Mortality Rate and Poverty

Table 4.4 below shows the correlation between under-five mortality and the overall poverty estimates.

Table 4.4 Correlation between Under-5 Mortality Rate and Poverty

The results show there is a low positive association or correlation between the Under-5 mortality rate and poverty. A Pearson correlation coefficient of 0.234 is statistically insignificant at p< 0.05 since the P-value, in this case, is 0.102, which is more than 0.05. Hence, there is an insignificant positive relationship between the Under-5 mortality rate and poverty across Kenya.

4.6         Discussion

Based on the findings above, this study establishes that urban regions experience lower proportion of children dead by women with a mean of 0.0433 compared to the rural areas with an average of 0.0572. This is in agreement with the findings of Kabubo-Mariara et al. (2012), who established that children in rural areas are more subjected to poverty hence are likely to die compared to those living in cities and urban areas. Osita et al. (2015) also supported this view by arguing that there was a need to improve child health care in rural areas in an attempt to minimize childhood mortality.

The highest proportion of children dead by women in both urban and rural regions were recorded in women aged between 45-49 years at 0.0748 and 0.0886 respectively. These findings were supported by Molitoris (2016), who argued that older mothers from poor families are affected by a range of health issues. These issues included malnutrition, anemia, and damaged reproductive systems, while women past their 40s may have bodies that are physically depleted, especially if they have given birth to many children; hence, they are likely to give birth to unhealthy children.

The lowest proportion of children dead by women in urban regions were recorded in women aged 15-19 years at 0.0226, and for rural areas were recorded in women aged 20-24 years at 0.0207. On the contrary, Cogneau and Rossi (2019) argued that very young mothers who have not attained full biological maturity have the highest chances of complications related to pregnancy and rank among the age group recording the highest under-5 mortality rate.

Mwangi and Murithi (2015), established that childhood mortality was low in Central, Eastern, Coast, Rift valley and Nairobi while it was high in North Eastern, Nyanza, Western and Eastern province and from the results above the counties from those regions depicts that the findings were true.

Finally, this research work established that there is an insignificant positive relationship between the Under-5 mortality rate and poverty across Kenya. The findings are in agreement with both Kaldawei and Pitterle (2011) and Elmahdi (2008), who identified malnutrition, injury and diseases as the primary factors impacting childhood mortality. However, they could not establish the difference between rich and poor mothers since the rich mothers are extremely busy to have enough time for their babies while the poor mothers are unable to afford good medical care and healthy nutrition for the child and also themselves. The two sets of scholars concluded that there was some positive relationship between childhood mortality and poverty, but the relationship was statistically insignificant.

5           Reference

Amina, I. (2008). Child Mortality and Poverty in Kenya: Levels, Patterns and differential. Unpublished Masters of Arts Research Paper. University of Nairobi

Beenstock, M. & Sturdy, P. (1990), “The determinants of infant mortality in regional India,” World Development, 18 (3):443-453.

Caldwell, J. C. (1979) “Education as a factor in mortality decline: An examination of Nigerian data.” Population Studies Journal, 33(4): 395-413.

Cogneau, D., & Rossi, P. (2019). Malaria Control and Infant Mortality in Africa. SSRN Electronic Journal. doi: 10.2139/ssrn.3457313

Doctor HV, Weinreb AA: (2003): Estimation of AIDS Mortality by Verbal Autopsy in Rural Malawi. AIDS: 2003 Nov 21; 17(17) 2509-13. PubMed

Elmahdi, H. (2008), “Socio-economic determinants of infant mortality in Kenya: Analysis of Kenya DHS 2003” Journal of humanities & social sciences, Volume 2(2):4-14.

Ferrarini, T., & Norström, T. (2010). Family policy, economic development and infant mortality: a longitudinal comparative analysis. International Journal of Social Welfare, 19. doi: 10.1111/j.1468-2397.2010.00736.x

Hill, K., and M-E. Figueroa (1999). Child Mortality Estimation by Time Since First Birth. Johns Hopkins Population Center Papers on Population WP 99-05. Baltimore, Maryland: Johns Hopkins Population Center. Available from http://www.jhsph.edu/popcenter/publications/pdf/WP99-05.pdf.

Hill, K., Bicego, G., & Mahy, M. (2001). Childhood mortality in Kenya: An examination of trends and determinants in the late 1980s to mid 1990s. Johns Hopkins Population Center Working Paper.

Hodge, A., & Jimenez-Soto, E. (2013). Determinants of childhood mortality. Bmj, 347(nov08 1). doi: 10.1136/bmj.f6632

Hobcraft J. N. et al. (1984) Socio-economic factors in infant and child mortality: Across National Comparison. Population Studies (London), Vol. 38, No. 2 (July) PP. 193- 223.

Hosseinpoor, A. R., Mohammad, K., Majdzadeh, R., Naghavi, M., Abolhassani, F., Sousa, A., & Vega, J. (2005). Socio-economic inequality in infant mortality in Iran and across its provinces. Bulletin of the World Health Organization, 83(11), 837-844.

Kabubo-Mariara, J. Karienyeh, M & Kabubo, F. (2012), “Child survival, poverty and inequality in Kenya: does physical environment matter?” African Journal of SocialSciences 2(1): 65-84.

Kaldewei, C., & Pitterle, I. (2011). Behavioral Factors as Emerging Main Determinants of Child Mortality in Middle-Income Countries: A Case Study of Jordan. New York: DESA Working Paper, (103).

Kenya National Bureau of Statistics (KNBS) 2007. Kenya Integrated Household Budget Survey- 2005/6. Ministry of Planning and National Development, Nairobi, Kenya

Kenya National Bureau of Statistics (KNBS) 2017. Kenya Integrated Household Budget Survey- 2015/16. Ministry of Planning and National Development, Nairobi, Kenya

Kenya National Bureau of Statistics (KNBS) and ICF Macro (2010) Kenya Demographic and Health Survey 2014. Calverton, Maryland: KNBS and ICF Macro.

Kovsted, J., Pörtner, C. C., & Tarp, F. (2002). Child health and mortality: Does health knowledge matter?. Journal of African Economies, 11(4), 542-560.

Kyei, K. A. (2012). Determinants of Childhood Mortality in South Africa: Using Categorical Data Modeling. Journal of Human Ecology, 37(1), 47–56. doi: 10.1080/09709274.2012.11906448

Kyei, K. (2018). Assessment of Anxiety and Depression Among Breast Cancer Patients in Ghana. Journal of Global Oncology, (4_suppl_2). doi: 10.1200/jgo.18.37000

Mariara et al., (2012) “Child Survival, Poverty and Inequality in Kenya: Does Physical Environment Matter?”, African Journal of Social Sciences. 2 (1): 65-84.

Medrano, P., Rodríguez, C., & Villa, E. (2008). Does a mother’s education matter in a child’s health? Evidence from south africa1. South African Journal of Economics, 76(4), 612-627.

Mekonnen, Y., Tensou, B., Telake, D. S., Degefie, T., & Bekele, A. (2013). Neo-natal mortality in Ethiopia: trends and determinants. BMC Public Health, 13(483).

Mutunga, C. (2004). Environmental determinants of child Mortality in Kenya. Unpublished Master of Arts Research Paper. The University of Nairobi.

Molitoris, J. (2016). Feeling the Squeeze: The Effect of Birth Spacing on Infant and Child Mortality During the Demographic Transition. SSRN Electronic Journal. doi: 10.2139/ssrn.2773587

Mosley, W. and Chen, L. (1984), “An analytical framework for the study of child survival in developing countries” Population and Development Review, 10:25-45

Mwangi D and Murrithi D(2015) “Determination of infant and child mortality in Kenya using Cox- proportional hazard model” American Journal of Theoretical and Applied Statistics  4(5):404-413.

Okiro, E. A., & Ayieko, P. (2018). Childhood mortality during conflicts in Africa. The Lancet, 392(10150), 804–805. doi: 10.1016/s0140-6736(18)31373-4

Omolo, S. (2014), “Socio-economic determinants of under-five mortality in principal cities of East Africa Community: A case study of Nairobi, Dar-es-salaam and Kigali.” Unpublished Masters of Arts Research Paper.The University of Nairobi.

Opiyo, C. O., & Sawhney, M. (2014). Determinants of the Recent Rise in Childhood Mortality in Sub-Saharan Africa: Evidence from Kenya Demographic and Health Surveys, 1990 – 2003. African Population Studies, 28(0), 1182. doi: 10.11564/28-0-566

Osita, P. (2015), “Studying the impact of household economic and community variables on child mortality.” Population and Development Review, (10): 215-235.

Pal, J., Mukhopadhyay, B., & Gupta, S. (2014). Myers’ Blended Method: An Alternative Approach. Journal of Scientific Research and Reports, 3(18), 2459–2465. doi: 10.9734/jsrr/2014/10673

Paxton C & Schady, N (2004). Child Health and the 1988-1992 Economic Crisis in Peru, Policy Research Working Paper No.3260. The World Bank, Washington, D.C.

Preston, S.H (1975). The Changing Relationship between Mortality and Level of Economic Development. Population Studies, 29(2), 231-248.

Pillay, T. (2019). Parent-Carer Education: Reducing the Risks for Neonatal and Infant Mortality. Neo-natal Medicine. doi: 10.5772/intechopen.82786.

Qin, D. (2017). Research on Improvement of College Students Trust in Counselors under the Guidance of qFive Development Conceptsq. Proceedings of the 2nd International Conference on Contemporary Education, Social Sciences and Humanities (ICCESSH 2017). doi: 10.2991/iccessh-17.2017.261

Renschler, J. P., Walters, K. M., Newton, P. N., & Laxminarayan, R. (2015). Estimated Under-Five Deaths Associated with Poor-Quality Antimalarials in Sub-Saharan Africa. The American Journal of Tropical Medicine and Hygiene, 92(6_Suppl), 119–126. doi: 10.4269/ajtmh.14-0725.

Romani, J.H. & Anderson, B.A. (2002). Development, Health, and the Environment: Factors Influencing Infant and Child Survival in South Africa. Cape Town: HSRC Publishers

Simmons, G. B., and Stan Bernstein, 1982.―The Educational Status of Parents and Infant and Child Mortality in Rural North India‖, Health Policy and Education, Vol. 2 349-67.

Shultz, T (1984): studying the Importance of Household Economic & Community Variables on Child Mortality; Population and Development Review 10: 215-235.

Spencer, N (2005). Health Consequences of Poverty for Children. End Child Poverty. Available at www.endchildpoverty.org.uk.

Sullivan, J. M., S. O. Rutstein, and G. T. Bicego (1994). Infant and child mortality. DHS Comparative Studies, No. 15. Calverton, Maryland: Macro International Inc. Available from  http://www.measuredhs.com/pubs/pub_details.cfm?ID=23

Rutstein, S. O., and J. G. Rojas (2006). Guide to DHS Statistics. Calverton, Maryland: ORC Macro. Available from http://www.measuredhs.com/pubs/pdf/DHSG1/Guide_DHS_Statistics.pdf

Uddin, M., Hossain, M., & Ullah, M. O. (2009). Child Mortality in a Developing Country: A Statistical Analysis. Journal of Applied Quantitative Methods, 4(3), 270-283.

United Nations Department of Public Information. (2008). End poverty 2015, Millennium development goals, high-level event on the Millennium Development Goals, United Nations Headquarters, New York, 25 September 2008 = Éliminer la pauvreté 2015, objectifs du Millénaire pour le développement, réunion de haut niveau sur les objectifs du Millénaire pour le développement, siège des Nations Unies, New York, 25 septembre 2008. New York (N.Y.).

World Bank (2006). World Development Report: Equity and Development. Washington, D.C.

World Health Organization (2005). The World Health Report 2005: Make every mother and child count.

Yaya, S., Uthman, O. A., Okonofua, F., & Bishwajit, G. (2019). Decomposing the rural-urban gap in the factors of under-five mortality in sub-Saharan Africa? Evidence from 35 countries. BMC Public Health, 19(616).

Younger S, (2001) Cross-Country Determinants of Declines in Infant Mortality: A Growth Regression Approach. CFNPP Working paper no. 130.