Modeling multivariate health and demographic outcomes in low- and middle-income countries using sample registration and complex survey data

Abstract

Improving the health of the most disadvantaged populations requires obtaining reliable health and demographic indicators in order to inform policy and interventions. Unfortunately, the low- and middle-income countries with the largest burden of disease and disability also tend to have the least comprehensive data. Calls for more and higher quality data collection have led (and continue to lead) to additional sources of information, namely sample registration systems and household surveys. These new data beget the development of statistical methods to produce the highest quality estimates. While the majority of current methods center on modeling univariate outcomes, improved estimates may be attained by borrowing strength across related outcomes via multivariate modeling techniques. This dissertation develops multivariate modeling frameworks using data from sample registration systems and surveys with complex sampling designs in an attempt to improve estimation of key indicators compared to modeling univariate outcomes separately.