Network-Adjusted Empirical Bayes Estimation of Age-Specific Adult Mortality from DHS Sibling Survival Histories: Evidence from Nigeria
This paper introduces a network-adjusted empirical Bayes framework to correct visibility bias and sampling variability in Nigerian sibling survival histories, revealing that adult mortality declined between 2018 and 2024 and is significantly lower than previously estimated by conventional methods and international reference series.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In many parts of the world, keeping an official record of every death is difficult. When hospitals and government offices do not capture these events, doctors and policymakers are left guessing about how many adults are dying and at what ages. This gap in knowledge makes it hard to know if a country's health system is improving or failing. To fill this void, researchers often turn to household surveys, asking people to list their brothers and sisters and report who is still alive and who has passed away. This method, known as sibling survival history, relies on the memory of living relatives to reconstruct the lives of those who are gone. However, this approach has a hidden flaw: it tends to see some people more clearly than others. Families with many surviving sisters are more likely to be interviewed, meaning their members are over-represented in the data, while people from smaller families or those with no living sisters remain invisible. This uneven visibility can distort the picture of mortality, making trends look different than they truly are.
A new study focused on Nigeria, a nation where fewer than one in ten deaths are officially registered, set out to fix this distortion. The researchers took data from the 2024 Nigerian Demographic and Health Survey and applied a sophisticated statistical correction to account for the fact that some people are harder to see than others in these surveys. They combined this correction with a technique that smooths out the natural randomness that occurs when counting a small number of deaths in specific age groups. By doing so, they created a clearer, more reliable map of adult mortality across the country. When they applied this same rigorous method to data from 2018 to compare the two time periods, they found that adult mortality in Nigeria has genuinely declined, particularly among people in their mid-thirties and older. This decline was not just a statistical illusion caused by the survey method; it was a real improvement that held up even when the researchers tested their results against other major global estimates.
The researchers began by gathering information from nearly 37,000 women who were interviewed in the 2024 survey. These women provided details on more than 189,000 siblings, creating a massive dataset that included over 3,400 reported deaths. The first challenge was to correct the "visibility" bias. In a standard survey, a person with many living sisters is more likely to be reported than someone with few or no sisters. This skews the data because the survey ends up seeing a disproportionate number of people from larger families. The team developed a method to adjust for this, essentially weighing the data so that every person in the population counts equally, regardless of how many sisters they have left alive. This adjustment changed the numbers slightly but significantly, proving that the old way of counting had been missing or miscounting certain groups.
Once the visibility was corrected, the researchers faced a second problem: the data was too sparse to be perfectly precise. When you break down the deaths into small groups based on age and sex, some groups have very few deaths recorded. In these small groups, a single extra death can make the rate look wildly high or low just by chance. To solve this, the team used a statistical smoothing technique. Imagine trying to guess the average height of people in a room where you can only see a few individuals; you would use the general pattern of height to make a better guess for the ones you see less clearly. The researchers did something similar, using the overall pattern of mortality to stabilize the estimates for the smaller, noisier groups without erasing the real differences between age groups. This process, which they repeated hundreds of times to ensure the results were robust, produced a final set of mortality rates that were both corrected for bias and smoothed for reliability.
The results of this refined analysis revealed a clear story of improvement. In 2024, the estimated death rates for adults were lower than they were in 2018 across all age groups and for both men and women. The decline was most pronounced for people in their late thirties and forties, where the drop in mortality was large enough to be statistically distinct from random chance. For example, the estimated death rate for men aged 45 to 49 fell by nearly 37 percent, and for women in the same age group, it dropped by nearly 40 percent. These findings suggest that the health of Nigerian adults is genuinely getting better, likely due to factors such as improved access to treatment for diseases like HIV and broader public health gains, though the study itself does not pinpoint the exact causes.
The study also tested its own conclusions by comparing them with other major global estimates. The researchers found that their new, corrected numbers were consistently lower than the estimates provided by the United Nations and the Global Burden of Disease study. This gap suggests that previous models may have overestimated the number of deaths in Nigeria, perhaps because they relied on older or less precise data. Interestingly, the study highlighted a specific age group where the old and new methods disagreed on the direction of the trend. For young women aged 15 to 19, the traditional method suggested mortality was rising, while the new, corrected method showed it was falling. The researchers identified this young age group as the most difficult to measure accurately because it has the highest number of people who are "invisible" to the survey method, reinforcing the importance of their new correction technique.
Ultimately, this research offers a more trustworthy way to count adult deaths in countries where official records are missing. By fixing the bias that favors larger families and smoothing out the noise of small numbers, the study provides a clearer view of who is dying and when. The evidence points to a real, positive shift in Nigeria's adult mortality, with the most significant gains seen among those in the middle of adulthood. The authors note that while their method is powerful, it is not perfect; the youngest age groups remain difficult to measure with complete certainty, and the data covers a period that included the global pandemic, which may have temporarily affected death rates. Nevertheless, the framework they developed can be applied to similar surveys in other countries, offering a path toward better health surveillance and more informed policy decisions for millions of people living in regions without complete death registration.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.