Model-Based Small Area Estimation of Healthcare Outcome Integrating Census and Survey Data in Ghana: ‎Methodological Challenges and Policy Implications

Fiifi Amoako Johnson, University of Southampton
Hukum Chandra, University of Southampton
Nikos Tzavidis, University of Southampton
Sabu S. Padmadas, University of Southampton

The demand for subnational level estimates has grown considerably in recent years, especially in many African countries which increasingly rely on decentralized models of governance for effective resource allocation and planning. However, such estimates are difficult to obtain because of lack of reliable administrative data and limited census data. The Demographic and Health Surveys (DHS) are often the only source of data which collect reliable demographic, social and health information but these data are representative only at the national or regional levels. Also, the DHS data cannot be used directly to produce estimates at district or sub-district levels because of small sample size. In this paper, we illustrate the application of Small Area Estimation (SAE) techniques to derive district-level estimates of contraceptive use and unmet need for contraception in Ghana, integrating data from the 2003 Demographic and Health Survey and the 2000 Population and Housing Census. We further discuss the development and challenges of SAE methodology, reflecting on binary and count data which are often used in estimating demographic and health indicators as well as addressing problems related to zero-inflated data. Finally, we discuss the relevance and use of model-based small area estimates for designing policy and programme interventions in settings where local area data are non-existent.