Novel methods for estimating adult mortality using a survey

ABSTRACT Accurate estimates of adult death rates are critical to good science and health policy. In most high-income countries, adult death rates are directly calculated since civil registration systems record every death. But in low- and middle-income countries (LMICs)--including most of Asia and Africa–civil registration systems are weak. Estimates from high quality surveys could be a

promising source of information about adult mortality in these countries, but unfortunately, decades of research have revealed that estimating adult death rates from a survey is extremely challenging. A major obstacle is that current methods require many thousands of interviews per survey. Large surveys are prohibitively expensive, and they make it difficult to innovate and

improve methods over time. As a result, there is a critical lack of evidence about adult mortality in LMICs. This is a problem: without accurate adult death rates, researchers cannot evaluate the impact of policies intended to confront deadly pandemics such as AIDS and COVID-19, quantify the pace of population aging, produce population projections that inform social and economic

policy, or even directly measure life expectancy. This project will develop a new set of statistical and data collection tools make it possible to collect information about adult mortality using surveys with moderate sample sizes (n ≈ 1,000-2,000), thereby greatly expanding the potential sources of data available to understand adult mortality in LMICs. Aim 1 will develop new

statistical methods, and apply them to already-collected pilot data; estimated death rates will be compared to a gold standard, allowing for errors and accuracy to be calculated for each method. Aim 2 will collect new qualitative data to help explain why the best-performing methods were successful in this setting (as revealed by Aim 1); and help generate hypotheses that will form

the basis for a larger-scale, multi-site test of these methods in the future. The findings will form the basis of a new website and support other tools for disseminating our results. The results will produce tools that can be used to increase the amount of evidence available about adult death rates around the world.