Employment instability linked to poorer later-life cognition in U.S. study

Employment histories may help explain differences in cognitive health as people age, according to research announced by Columbia University on September 23. The study linked unstable employment and some pathways out of paid work with poorer later-life cognitive performance, adding evidence that the conditions surrounding working life deserve attention in healthy-aging research.

The analysis, published in Alzheimer’s & Dementia, examined associations rather than testing an intervention. It cannot establish that maintaining employment protects the brain, or that an individual would reduce dementia risk by delaying retirement.

Following work histories across 16 years

Researchers used the U.S. Health and Retirement Study and its Harmonized Cognitive Assessment Protocol. The analysis included 2,418 people and connected employment and perceived job-insecurity patterns recorded from 1998 through 2014 with cognitive testing and impairment status in 2016.

The team grouped similar employment histories separately for women and men, identifying six clusters for each. Weighted, multivariable-adjusted regression models then compared cognitive outcomes across those groups. This approach captured sequences of work experiences instead of relying on employment status at a single point.

Among women, several trajectories—including self-employment, early retirement, remaining outside the labor force and moving from secure employment into retirement—were associated with lower memory and executive-function scores than secure employment. Reported differences across these comparisons ranged from 0.13 to 0.32 standard deviations below the reference group. Among men, inconsistent employment was associated with poorer performance across most cognitive domains.

These are differences between groups on cognitive assessments. They do not measure how many years a particular person’s brain has aged or establish an individual prognosis.

Why employment could matter

The findings fit a broader research framework linking work with cognitive health through several overlapping pathways. A review published in Alzheimer’s & Dementia in July described how jobs can provide intellectual demands, social contact and material resources, while also exposing workers to stress and physical hazards.

That framework makes a simple employed-versus-retired comparison difficult to interpret. Leaving a demanding or harmful job may have different consequences from an involuntary job loss that disrupts income and social connections. The review also noted evidence that retirement can benefit cognition in some circumstances, particularly after physically demanding work.

Columbia’s announcement presents employment continuity as a potential focus for prevention research and policy. Whether measures that improve job stability also improve cognitive outcomes remains a question for further investigation.

The limits of the evidence

The central concern is that health and employment influence each other. People with emerging illness or cognitive difficulties may leave work earlier or experience less stable employment. A separate 2025 study of employment histories and dementia explicitly identified this healthy-worker effect as a possible explanation for observed associations.

Statistical adjustment helps address measured differences, but it does not turn employment histories into randomly assigned exposures. The new analysis also relates a long period of work experience to cognitive outcomes assessed in 2016; those comparisons should not be presented as direct measurements of each group’s rate of cognitive decline.

For healthspan research, the practical contribution is a more detailed account of the social circumstances accompanying cognitive aging. Establishing prevention benefits will require evidence that can better separate employment conditions from prior health, resources and the reasons people leave paid work. This study supports that research agenda without establishing a preferred retirement age.