If 1984 was 2024

The aging of many populations, which is expected to continue the coming decades, has led to concerns about future increase in age-related conditions like dementia. Stallard, Ukraintseva and Doraiswamy (2025) challenge this with respect to the United States, and argue that declining age-specific dementia prevalence may offset the increase driven by changes in the age pyramid. They use data from the National Long Term Care Survey (NLTCS), for age groups from 65–69 years to ≥95 years, 1984 and 2004, and extrapolate dementia prevalence to 2024 using data for 2000–12 from the Health and Retirement Study (HRS) and data for 2011–21 from the National Health and Aging Trends Study (NHATS).

In their model, prevalence of dementia has declined within all the included age groups. People in the age groups from 75–79 to 90–94 years had roughly the same dementia prevalence in 2024 as 10 year younger persons in 1984, which, for example, corresponds to a prevalence decrease from about 30 percent to about 10 percent in the age group 85–89 years. In the oldest ≥95 age group, the relative decrease is somewhat less marked, from nearly 60 percent to about 35 percent. They speculate that the decrease may be due to factors such as better vascular health, higher education, and lower burden of infections in younger cohorts. Moreover, they note that decreasing age-specific dementia incidence also has been reported from other countries, including Sweden (Wolters et al. 2020).

These trends, with decreasing dementia prevalence and incidence, may be contrasted with the increase in dementia and neurodegenerative disease as underlying cause of death in many high-income countries, which I have written about before. For example, one may look at the charts in my post 4 July 2025, showing the rapid increase in the share of these causes in United States and Sweden, in contrast with an Eastern European country like Lithuania, where they have remained uncommon. It is rather obvious that these trends largely have resulted from changes in rules for selecting underlying cause of death and reporting practices, which probably have been self-reinforcing, so that, for example, the more doctors see dementia reported as a common cause of death, the more apt they become to write it on death certificates.

However, one might wonder what the cause of death pattern in a country like Sweden would have looked like in the past, if dementia had been recognized and given priority as a cause of death the way it is today. One could model the dementia mortality \(\m_{p,a,s}\) for a population \(p\), age group \(a\), and sex \(x\), using the relative prevalence change from a reference population \(r\), \(\rr_{r,a,s}\) and multiply with the mortality rates for the reference population.

\[\m_{p,a,s}=\rr_{r,a,s}\times\m_{r,a,s}\]

These back-projected mortality rates for \(p\) may then be divided by the total mortality to obtain projected shares of dementia deaths, because total mortality does not change in the modeling and all deaths have exactly one underlying cause. Then, it is also possible to derive a probability of eventually dying from dementia given a life table, using the methods I discussed in my 26 April post.

The R script 2026-08-04-1984.R in the directory postdata/2026-08-04-1984 under the blog repository contains one such modeling for the Swedish 1984 population, with the Swedish 2024 population as reference population, using my R package morr, with cause-specific mortality data from WHO (2026) and life tables from University of California, Berkeley and Max Planck Institute for Demographic Research (2026). It is assumed that the prevalence ratio for dementia between 1984 and 2024 is 3 in the age groups from 60–64 to 85–89, 2 in the age group 90–94 and 1.5 in the age group ≥95, for both women and men.

In 1984, the probability at birth of dying from dementia or neurodegnerative disease, according to the morr definition¹ would be 2.0 percent for women (the data frame ndegse_f84h), and 1.2 percent for men (ndegse_m84h). For 2024 it was 17.2 percent for women (ndegse_f24h), and 11.1 percent for men (ndegse_m24h).

ndegse_f24h |> select(Age, caldx)
# ...
#     Age caldx
#   <int> <dbl>
# 1     0 0.172

However, if we model a mortality to prevalence ratio like 2024, as I have described, the probability of dying from dementia would be somewhat higher in 1984 than 2024, despite the lower life expectancy: 19.8 percent for women (ndegse_f84p) and 11.9 percent for men (ndegse_m84p). This also implies that deaths from other causes, corresponding to nearly 20 percent of total mortality for women in 1984, and about 10 percent for men, would have to be subtracted if the total number of deaths is to remain constant.

References