For decades the eye exam sat in its own quiet corner of senior health: a new prescription, a cataract referral, a reminder to come back in a year. Nobody filed it under brain health. Then a team led by Joshua Ehrlich took the 2018 round of the Health and Retirement Study, a probability sample of 16,690 US adults aged 50 and older, and ran vision impairment through the same arithmetic used for hypertension and smoking. The result, published in JAMA Neurology in 2022, was small and stubborn: a population attributable fraction of 1.8 percent, which the authors translated into more than 100,000 prevalent US dementia cases that might potentially have been avoided with healthy vision. In the same 12-factor model, all the risk factors together accounted for an estimated 62.4 percent of cases, with hypertension highest at 12.4 percent.

A population attributable fraction is not a prediction and not a diagnosis. It is a thought experiment with a number attached: if this exposure were removed entirely from the population, and if it were genuinely causing what it appears to be associated with, this is the share of cases that would go with it. The 2022 analysis was cross-sectional, a single snapshot in time, and it leaned on self-reported vision. That last detail matters, because when a different team measured eyes rather than asking about them, the number moved.

In September 2024, Jason R. Smith and colleagues at Johns Hopkins, with Ehrlich among the co-authors, published a cross-sectional analysis in JAMA Ophthalmology using the National Health and Aging Trends Study: 2,767 community-dwelling Medicare beneficiaries who completed tablet-based vision and cognitive testing in 2021. Distance acuity, near acuity and contrast sensitivity were each measured directly. The attributable fraction for having at least one vision impairment came out at 19 percent, with a confidence interval running from 8.2 to 29.7, wide enough to be honest about. Contrast sensitivity impairment, the ability to pick a grey letter off a paler background, carried the largest share at 15 percent, ahead of near acuity at 9.7 percent and distance acuity at 4.9 percent. The authors were explicit that this does not prove cause and effect. They also noted that these impairments frequently come from conditions that are treatable, and often reversible, such as uncorrected refractive error and cataract.

That is why this particular statistic travelled. Most items on a dementia prevention list ask for years of behaviour change. Vision is different in kind: a pair of glasses, a cataract operation, a clinic appointment that was postponed and never rebooked. The 2024 Lancet Commission on dementia prevention, intervention and care added vision loss to its list of modifiable risk factors, a decision reported at the time by the Lancet's own announcement and by professional bodies including the American Optometric Association. The eye care sector has kept the message running since: the International Agency for the Prevention of Blindness carried a piece under the heading of untreated vision impairment as a modifiable dementia risk factor in March 2026.

For a family sitting with a diagnosis, or watching a parent stop reading the paper, the useful part of this evidence is narrower than the headlines but not nothing. Poor vision makes cognitive testing harder, shrinks the world a person moves through, and can look like withdrawal when it is simply an inability to see. None of the studies here tested whether restoring sight prevents dementia; that trial has not been done. What they establish is that vision belongs in the conversation, and that an eye examination is a reasonable thing to raise with a clinician rather than a thing to leave for next year. The National Institute on Aging, which funded the 2022 work, framed it the same way when it reported the finding: existing, cheap, underused eye care is the intervention already sitting on the shelf.