A new article published in BMC Public Health proposes a Bayesian Hierarchal Model to estimate the level and trend of subnational modern contraceptive prevalence rates (mCPR) using data from PMA2020 surveys in five sub-Saharan African countries: Burkina Faso, Ethiopia, Ghana, Kenya, and Uganda. Recognizing that most global monitoring efforts have relied on national level estimates of mCPR, and that there is growing recognition for the need for subnational mCPR estimates for de-centralized governments to take policy actions, PMA2020 researchers used small area estimation techniques to monitor trends in mCPR.
The article, entitled, “Subnational estimation of modern contraceptive prevalence in five sub-Saharan African countries: a Bayesian hierarchical approach”, was published on February 20, 2019.
The researchers used female interview data from four semi-annual PMA2020 survey rounds (rounds 1-4) conducted from 2013 to 2016 in Burkina Faso, Ethiopia, Ghana, Kenya, and Uganda to improve upon a previously published Bayesian model.
They used their model to calculate round-specific Bayesian estimates of mCPR by region for each of the five countries. Researchers also used this model to estimate subnational changes in mCPR from R1 to R4 in each of the countries.
Results from the analysis, which was led by Dr. Qingfeng Li, Technical Advisor for PMA2020, along with collaborating authors, include:
- Considerable narrowing of the uncertainty intervals around the Bayesian model estimates, compared to national estimates based directly on survey data;
- Substantial variations in the estimated subnational mCPRs for the five countries, with most experiencing upward trends across the four survey rounds; and,
- Trends in mCPR vary within regions in the countries across the four survey rounds.
Despite PMA2020 surveys being designed primarily to provide national estimates of mCPR, the model presented by the authors in this study provides an alternative opportunity to generate reliable subnational estimates using PMA2020 data on an ongoing basis. The generated subnational estimates will be used for local program planning and monitoring community demands for family planning.