Assessment of data quality used in direct 'estimation of infant and child mortality, KDHS 1998

Abstract

This study attempts to assess the quality of data used indirect estimation of infant and child mortality. The quality of data evaluated mainly focus on non-sampling errors, which are inherent in Kenya Demographic and Health Survey of 1998. The errors are usually difficult to evaluate statistically and at times difficult to identify and therefore bias any mortality estimate thus derived. The broader study hypothesis derived from the literature reviewed is that errors exist in Kenya Demographic Health Survey 1998 from various causes and have impact on estimates of both infant and child mortality. The specific hypotheses were formulated based on literature reviewed and the factors related to each kind of error. Some of the factors include socio-economic, demographic and cultural. Computing various indicators of flaws in data quality and checking for variations in those parameters ested the formulated study hypotheses. There are a number of methods used in data analysis, which includes the Myers' blended index, birth atios, and life table analysis. The 1998 Kenya Demographic-and Ifealth Survey data restricted to 23351 children born since 1960 regardless of their survival status. For life table probabilities of death, analysis was restricted to 5778 children born five years prior to the survey. All the 7881 women in the survey are sed in analysis of age reporting. Results for analysis of date of birth data indicates that there is usually marked displacement of birth rates by survival status of the child. It is usually more severe for dead than living children. This varies by region, education level, and place of residence, ethnicity and age of the mother. For accuracy of the age at death and the completeness of information on date of birth, variations also exist and these vary with socio-economic, demographic and cultural factors of the mother. The study has also important findings on the non-sampling errors in direct estimation of infant and child mortality. For instance, misreporting of date of birth and age at death have implications on mortality estimation and any imputation procedure used could either bias the estimates either upward or downward. The findings of this study have a number of implications for policy and further research. The study shows that there is need to redesign the current format of Demographic and Health Survey questionnaire particularly where births and deaths of children are concerned. This may include having more probing questions on those issues. To cater for the children, who have lost their mother, it is important to include paternal questions so that the fathers could respond to fertility questions in cases where the mother is dead. The study calls for training of interviewers in psychology so that they can make good judgment of responses provided by respondents.The results of the study call for further research on the plausibility of KDHS data based on external sources like the World Fertility survey the census and any other baseline studies. The study calls for future efforts to be directed at investigating peoples' perception to fertility and mortality questions as are asked the respondents. This will indicate peoples' fears and joys in discussing birth and death.