Expanding the South African Rapid Mortality Surveillance to cover provincial mortality

• Main Author: Maseko, Tafadzwa
• Other Authors: Dorrington, Robert
• Format: Dissertation
• Language: English
• Published: Faculty of Commerce 2019
• Online Access: http://hdl.handle.net/11427/30003

Mortality estimates are useful for evaluating health status for a country. These estimates, especially for a country heavily affected by HIV epidemic and at the sub-national level, are a foundation in formulating health strategies and policies to reduce premature deaths and improve quality of life. However, estimating consistent levels and trends in mortality for a developing country like South Africa with incomplete vital registration and errors in censuses or survey data is difficult. This research examines whether one can combine National Population Register (NPR) data and Vital Registration (VR) data to produce reliable estimates of specific indices of mortality at the provincial level. In addition, the same approach as used in 2015 RMS report is applied to registered maternal deaths, neonatal deaths, and neonatal deaths captured by the District Health Information System (DHIS) to derived reliable estimates of MMR and NMR at the provincial level. The quality of NPR data and VR data is evaluated and since NPR data does not include the whole population, it is adjusted for incompleteness relative to VR data. It is found that there are some problems in the VR data such as VR death misplacement between the provinces and missing VR deaths in most recent years, which makes extrapolation of past trends in the completeness of NPR relative to VR into the future difficult. Suitable assumptions are made to correct for these anomalies and NPR data together with VR data are further adjusted for the general under-registration. Estimates of child mortality from Pillay-van Wyk, Laubscher, Msemburi et al. (2016) and estimates of adult mortality from Dorrington and Timæus (2017) are used to derive the estimates of completeness of death registration. The estimates of mortality rates produced from combining NPR and VR data appear to be sensible, showing some internal and external consistency. However, the estimates of MMRs produced from VR data as well as NMR from VR and DHIS data shows that there is a great deal of uncertainty around the estimates of these mortality indicators.