An estimate without an accompanying measure of its error is as useful as a bicycle without wheels. A large number of administratively produced coverage estimates for health indicators are not accurate with some producing estimates exceeding 100%. None have a measure of their own error. Yet, we cannot monitor and evaluate health and other social programs and policies without quantitative knowledge of coverage indicators. What is sometimes ignored is that other sources of data that can be incorporated into our estimation techniques to bolster administrative estimates.
This webinar recording introduces a new analytic method, which Low- and Middle-Income countries can use to improve the key estimators for program monitoring and evaluation. It suggests that whilst so many countries are investing in DHIS-2 to strengthen their health systems, it behooves them and the Bank to also ensure they have a means to determine their risks when using DHIS-2 or HMIS estimates to plan or assess the status of their programs. By adding this new technique to the development agenda, Ministries of Health could potentially prevent waste by knowing the accuracy of administrative estimates, and increasing their accuracy by combining them with a probability sample to create a combined or hybrid estimate. An extra advantage of this method is that it makes use of samples that have been collected for other purposes. Often statistical samples have already been obtained for other reasons. Having served their primary purpose, they are often archived or discarded. This technique presents a new approach to make use of such data at minimal extra cost.
With Gates Foundation’s support, Joseph Valadez of the Liverpool School of Tropical Medicine, and Marcello Pagano of the Harvard T.H. Chan School of Public Health, developed such a method to anneal probability surveys carried out in Benin and Madagascar with administrative data from Child Health Days. They explain Statistical Hybrid Prevalence Estimation and their findings in the Proceedings of the National Academy of Sciences, Dec 5, 2018, DOI 10.1073/pnas.1810287115. They also successfully applied this method in Bihar, India, and Nepal using LQAS surveys. Other survey methods can also be used.
Joseph Valadez, Chair of International Health, Liverpool School of Tropical Medicine
View materials from the webinar: