An individual American’s dementia risk at any given age has fallen by roughly two-thirds since 1984, according to Duke researchers writing in JAMA. The number of Americans living with dementia is going up anyway. Both statements are true, and holding them together is the whole point of the paper.
The work is a Viewpoint — a short analysis piece rather than a new study with new data. Eric Stallard, an actuary, along with Svetlana Ukraintseva and Murali Doraiswamy, argue that widely quoted US projections contain a methodological error: they never adjusted for a birth cohort effect. Each successive generation arrives at 75 or 85 with systematically less dementia than the generation before it, the paper reports.
Before going further, one thing needs saying plainly, because most coverage of this research skips it. The headline figure is not entirely measured. The survey underpinning it — the National Long Term Care Survey — ended in 2004. The 1984-to-2024 decline is twenty years of observation plus twenty years of extrapolation at an assumed 2.7% annual rate, a modelling choice the authors state openly in the paper. That is a defensible actuarial judgement. It is not the same as having counted.
What the dementia risk numbers actually show
Within the measured window the trend is strong and consistent. Age-specific prevalence fell at 2.5% to 3% a year across four decades — a 67% cumulative decline, with a confidence interval of 61% to 73% — and, unusually, it never levelled off.
That surprised the man who found it. “The overall declines in prevalence rates for dementia observed during 1984–2004 continued during 2004–2024 at almost the same rate—dropping in half every 25 calendar years—with no slowdown at the end of the observation period,” Stallard said. “I had originally expected the rate of decline to slow down substantially over the past decade.”
The cohort pattern is the mechanism. “Each successive birth cohort exhibited systematically lower dementia prevalence rates at each given age than prior birth cohorts,” Stallard said. One US study of 3,010 people found incidence 77% lower for those born between 1932 and 1941 than for those born three decades earlier. A meta-analysis of 49,202 people across seven population studies in the US and Europe found incidence falling 13% per decade between 1988 and 2015. Declines show up in Sweden, the Netherlands, the UK and France too.
The co-author’s own warning
Doraiswamy, a professor of psychiatry and geriatrics at Duke, pre-empted the misreading his own paper was likely to generate.
“That said, people should not take away from our research that overall dementia rates are going to decline. In fact, overall dementia rates will continue to rise as people live longer, new risks (such as obesity and diabetes) emerge and dementia is detected at earlier stages using more sensitive brain scans or blood tests.”
The paper itself puts it in one line: age is the dominant risk factor, and barring effective prevention, the number of people with dementia will rise as the population ages.
The arithmetic is unforgiving. The US population over 80 doubles between 2025 and 2050; the over-95s triple. Against that, the Alzheimer’s Association counts 7.2 million Americans aged 65 and over living with Alzheimer’s dementia today and projects 13.8 million by 2060. Even Stallard’s more optimistic re-projection does not produce a fall — just a smaller rise. “Rather than a doubling of the number with dementia over the next 25 years, the increase would be on the order of 10% to 25%,” he said.
Why a generation arrives healthier
Doraiswamy’s list of candidate drivers is mundane and mostly not medical: “over the last 4 decades, education levels have risen, smoking rates have dropped and treatment of cardiovascular risks and hearing loss has improved all of which may explain why dementia rates dropped.” He immediately adds the counterweight — “rising rates of obesity, diabetes and sedentary lifestyle may lead to increased dementia rates in the future.”
Roughly 40% of dementia risk may be modifiable, he notes, which is why the cohort question matters at all. If the decline is made of schooling, blood-pressure pills and fewer cigarettes, it is not a law of nature. It is a set of social conditions that can reverse.
There is also a distributional catch the cheerful version of this story omits: the paper flags populations where prevalence held steady or rose, including in Japan and among African Americans in the US, and calls for more racially diverse samples. The gains are real, but they have not been evenly handed out. Research into treatment continues on other fronts, including unexpected results with psilocybin in an Alzheimer’s patient.
What to do with two true numbers
For a person, the finding is genuine good news: your dementia risk at 80 is meaningfully lower than your grandparent’s was at 80. For a health system, it changes almost nothing — there will still be roughly twice as many patients in 2060 as there are now.
The Duke team’s actual claim is narrower than the celebration around it, and more useful. Projections that ignore the cohort effect overstate the coming burden, perhaps badly. Correcting them does not make the burden go away. It makes the number honest, and the number is still going up. Full details are in the Duke summary of the research.
