Page 51 - IJPS-3-2

Multilevel analysis of infant mortality and its risk factors in South Africa

Within-country variation in child mortality has been well documented, with rates often varying substantially across
different regions and social groups (Moser, Leon and Gwatkin, 2005; Mosley and Chen, 1984). It is said that although
global or national level results are vital for assisting policy makers to better prepare for the emerging health needs
of populations, they constitute a less effective guide for refocusing health priorities because efforts to reduce health
disparities would be more successful if they are based on evidences from lower administrative units (Heuveline, Guillot
and Gwatkin, 2002). This research, therefore, studies infant mortality at national, provincial and municipality levels
to highlight concentration at lower levels of geography that in turn underscores the ineffectiveness of national level
indicators for monitoring progress in health achievement.
The objective of this research is to analyze infant mortality using hierarchical model and to identify important risk
factors for infant mortality in South Africa. The study uses the 2011 South African census data and all the risk factors are
defined at three levels: Individual, municipal and province level. A three-level logistic regression model is fitted using
data on children born twelve months before the census date where child, municipality and province are the first, second,
and third levels.
Factors affecting infant mortality are investigated by fitting multilevel logistic regression models in order to quantify
the impacts of socioeconomic factors, including poverty and inequality, on infant mortality. The hypothesis is that there
are significant spatial variations of child mortality, which are associated with socioeconomic differentials in the country,
and hence multilevel modelling helps to measure the impacts of the risk factor at different administrative levels.
Mortality of children has been explained by different theories such as the social and economic explanation, the public
health explanation, and the Mosley and Chen analytical framework. Mosley et al., (1984) identified a set of proximate
determinants that directly have an impact on the morbidity and mortality of children. The factors are then grouped into
5 sets: Maternal factors (age, parity and birth interval), environmental contamination (air, food/water/fingers, skin/soil/
inanimate objects, insect vectors), nutrient deficiency (calories, protein, micronutrient, vitamins and minerals), injury
(accidental, intentional); the last set is personal illness control (personal preventive measures, medical treatment). It
is within this framework that many studies on child mortality and its correlates have been carried out. This study also
follows this framework but based on the following classification of the determinants: New-born demographics (Boco,
2010; Hill and Upchurch, 1995; Kembo and Ginneken, 2009; Mustafa and Odimegwu, 2008), Maternal factors (Boco,
2010; Hobcraft, McDonald et al., 1985; Kabir, Islam, Ahmed et al., 2001; Kembo and Ginneken, 2009; Omariba, Beaujot
et al., 2007), socioeconomic factors (Bawah and Zuberi, 2005; Cleland, 1990; Hobcraft, 1993; Mustafa and Odimegwu,
2008; Sastry, 1996; Wagstaff, 2000), environmental factors (Bartlett, 2005; Kabir, Islam, Ahmed et al., 2001; Kazembe,
Clarke and Kandala, 2012; Kembo and Ginneken, 2009) and HIV/AIDS (Dorrington, Johnson, Bradshaw et al., 2006;
Ng’weshemi, Urassa, Usingo et al., 2003; Wang, Liddell, Coates et al., 2014; Zaba, Marston and Floyd, 2003). These
factors are defined at individual level, which could be child, mother or household, municipal level or province level
depending on the nature of the variable as well as the availability of data.
2. Data and Methods

2.1 Data
The study used data from the 10%-unit record of the 2011 de facto population and housing census of South Africa (StatsSA
2014). Children born within 12 months before the census date are considered for this research. The sample data, among
other things, consisted of data on children’s demographics,general health functioning, income, educational attainment,
employment status, fertility, household characteristics and mortality variables. After removing missing, unknown and
inconsistent cases, there are 86 877 (un-weighted) children with valid survival status for analysis. These children can be
viewed as they are nested in a structure under the 234 municipalities and 9 provinces of South Africa. The mortality status
of these children was the outcome measure of the study, and it was assumed that the rate of under-reporting of births in
the past 12 months is the same as that of under-reporting of deaths in the past 12 months so that the effect on mortality
estimates is negligible.
HIV prevalence rates were taken from the 2012 South African National HIV Prevalence, Incidence and Behaviour
Survey conducted by the Human Science Research Council (HSRC, 2014). More details of the survey can be found
elsewhere (HSRC, 2014)

2.2 Risk Factors Considered in the Study
The individual level risk factors considered are: sex, age and birth order of the child; age, racial group, marital status
and employment status of the mother as well as the living standard of the household where the infant resides. Living
standard was computed by constructing an index from different variables which are supposed to be related with the living

44 International Journal of Population Studies | 2017, Volume 3, Issue 2