Counting dementia has always been a problem of apples and oranges. One country diagnoses in a memory clinic, another in a village survey; one study follows people for three years, another for fifteen; one uses one definition of dementia, the next uses something close but not identical. Stack those numbers on top of each other and you get a global figure that nobody fully trusts. What a team led by the Centre for Healthy Brain Ageing at UNSW Sydney has done, working through the Cohort Studies of Memory in an International Consortium (COSMIC), is take the raw participant-level data from 22 population-based longitudinal cohorts in 18 countries and territories across six continents, 38,418 people in all, and run them through the same statistical machine. The result, published in JAMA Neurology on October 12, 2026, is an estimated 11.4 million people worldwide developing dementia in 2023, and a projection of 24.7 million new cases a year by 2050.
The underlying arithmetic is modest and worth stating plainly, because the headline number is built from it. Across 287,632 person-years of follow-up, 3,001 participants developed dementia, giving an age- and sex-standardized incidence of 9.7 new cases per 1,000 person-years at risk (95% confidence interval 7.89 to 11.97). Translated into something a family can hold: out of a thousand people of the ages studied, followed for a year, roughly ten would receive a dementia diagnosis. Baseline recruitment in these cohorts began between January 1990 and December 2011, follow-up data ran through the end of 2021, and the analysis itself was done between June 2024 and June 2025. The lead author, Dr Ashleigh Vella, framed the point of the exercise as being that "we can compare dementia incidence across a geographically diverse group of populations using the same statistical approach."
Two divides open up inside that single global rate. Age-standardized incidence was 10.6 per 1,000 person-years in women (8.61 to 13.11) against 8.6 in men (6.94 to 10.72). The gap by national income was wider still: 13.6 per 1,000 person-years in low- and middle-income countries (9.67 to 19.24) compared with 8.3 in high-income countries (6.57 to 10.54). Those confidence intervals overlap, and the authors are explicit that the error bands around the low- and middle-income estimates are the wider ones, particularly at older ages, because fewer cohorts contribute. This is an observational pooling of cohort data. It describes where and in whom new cases are being recorded. It does not establish why, and nothing in it tells any individual woman, or anyone living in any country, what her own odds are.
The 2050 figure deserves the same care. It was produced by applying the study's age- and sex-specific rates to the projected 2050 population, assuming incidence stays constant over time. That is a deliberate, conservative modelling choice rather than a forecast of worsening brain health: it says that if people in each age group keep developing dementia at today's observed rates, sheer population ageing more than doubles the annual count. It cuts both ways. If prevention lowers rates, the real number lands below 24.7 million. If risk factors worsen, above. Senior author Professor Perminder Sachdev tied the income-level finding to capacity, noting that "these regions are experiencing rapid population ageing while often having the fewest resources for prevention, diagnosis and care."
For a family already living with a diagnosis, an incidence paper can feel like weather reported from a satellite: true, vast, and not obviously about you. But incidence is the number that health systems build on. Prevalence, the more than 57 million people worldwide currently living with dementia, tells a government how many people need support today. Incidence tells it how many new assessments, scans, diagnostic appointments, carer-support places and memory services it will need to open next year, and the year after. A harmonized benchmark means a health ministry in a middle-income country can, for the first time, plan against a figure derived in part from populations like its own, rather than borrowing one estimated in Western Europe. The study's own authors flag the obvious hole in that promise: only a minority of the contributing cohorts come from low- and middle-income settings, which is precisely where the projected growth is concentrated.