There is a lack of high-quality data on levels, trends, and age patterns of under-5 mortality (U5M) in sub-Saharan Africa. Civil registration and vital statistics (CRVS) systems are incomplete or non-existent across the region, which has led to a reliance on alternative data sources and model-based estimates. Though the available information indicates that there has been substantial improvement to child health over the past decades, the region remains confronted with the highest rates of U5M in the world. Accurate measurement will be essential to tracking and accelerating gains in U5M, and providing the evidence to guide effective health policy making. This thesis examined bias and uncertainty in U5M estimates for sub-Saharan Africa from different data sources. Of primary interest were data collected on pregnancies, pregnancy outcomes, and early mortality in Health and Demographic Surveillance Systems (HDSS). This work was conducted through four studies which form the main chapters of this thesis. In the first three studies, HDSS monitoring of pregnancies and newborns was evaluated through cross-site and external comparisons with survey and clinical data. The analyses were performed across a number of populations in unique epidemiological settings and took into account site-specific characteristics and data collection protocols. In the final study, the direction of inquiry was reversed, and the prospective data of HDSS were used to shed light on bias in U5M estimates from retrospective maternity history questionnaires in high-HIV settings. This thesis supports existing calls to improve data collected on perinatal and U5M in sub-Saharan Africa. HDSS sites are valuable resources for this purpose, often serving as the only population-based data systems in countries lacking complete CRVS. Collectively, the findings contribute to improving understanding of U5M in sub-Saharan Africa and our ability to accurately measure it in the context of population-based surveillance systems.