Urban rural differnetials of infant mortality in Kenya using a shared frailty model
Njagi, Purity M
MetadataShow full item record
Kenya targets to reduce infant mortality to 26/1000 live births by 201s.The Kenya Demographic Health Survey (KDHS) 2008/09 shows that compared to the 2003 KDHS, the Infant Mortality Rate (IMR) improved to 52 from 77 per 1000 live births. However this reduction was achieved with an inverse in the trend whereby infant mortality rate for the first time in Kenya was higher in urban areas than rural areas. Although studies have focused on urban rural differentials in Kenya, most have utilised logistic regression and standard survival models. This study therefore set to examine the implications of unobserved heterogeneity on parameter estimates of urban rural differentials in infant mortality in Kenya. The study is conceptualised using the Mosley and Chen framework on child mortality. It utilises data from the KDHS 2008/09, the variables analysed include socio economic variables (Mother's education, Wealth quintile and region of residence), demographic variables (Mother's age at birth and birth order) and biological variables (Preceding birth interval and birth size). Standard Log normal Accelerated Failure Time model was used for analysis, shared frailty model was further fitted to account for unobserved heterogeneity; frailty was considered at household level. Results indicate that although infant mortality was seen to be higher in urban areas compared to rural areas, this difference was not significant. In urban areas birth size was the only significant factor that influenced infant mortality whereas in rural areas region of residence, preceding birth interval, birth order and birth size were significant factors. Consistent with other studies AFT model was seen to underestimate the negative parameters while overestimating the positive parameters. Recommendations emanating from this study are that, frailty provides a more precise estimate of parameters compared to standard AFT models. In addition, programs or interventions targeted at reducing infant mortality should seek to address factors by urban and rural differentials so as to address specific needs of the population in those regions. It would also be paramount to study other factors that act to influence infant mortality especially in urban areas.
University of Nairobi, Kenya