Baseline Cognition, Not Age or Education, Predicts Daily Function Decline in Dementia

For millions of families caring for a loved one with dementia at home, one of the most agonizing uncertainties is not the diagnosis itself but the trajectory that follows it. Will independence erode slowly over many years, or will the ability to dress, bathe, and manage a household collapse within months? A new systematic review and meta-analysis published in European Geriatric Medicine offers the most rigorous attempt yet to answer that question, and its central finding is strikingly simple: among the many factors clinicians have proposed as predictors of functional decline, only one—baseline cognitive function—can currently be trusted to forecast how fast activities of daily living will deteriorate in community-dwelling older adults with dementia.

The study, led by Johannes Riis of Aalborg University and Aalborg University Hospital in Denmark, together with colleagues including Dorte Melgaard, Victoria Gunmalm, Martin Grønbech Jørgensen, and Stig Andersen, set out to identify independent prognostic factors associated with decline in activities of daily living, commonly abbreviated ADL. These activities span the essential tasks of self-care, from eating and grooming to toileting and transferring, and their loss marks the point at which dementia shifts from a cognitive condition to a disabling one. Functional decline drives caregiver burden, predicts nursing home admission, and shapes the timing of difficult decisions about long-term care, which is why reliable prognostic markers have long been sought.

To assemble the evidence, the team searched five major databases—PubMed, EMBASE, Cochrane CENTRAL, CINAHL, and PsycINFO—from their inception through October 2025. The eligibility criteria were strict: only longitudinal studies following community-dwelling adults aged 65 or older with an established dementia diagnosis qualified. The screening process was enormous in scale. The researchers sifted through 6,915 records, examined 100 full-text articles in detail, and ultimately included 19 reports derived from 13 distinct studies covering a combined 9,490 participants. Two reviewers independently screened studies, extracted data, and judged risk of bias, following the PRISMA 2020 reporting framework and registering the protocol in advance with PROSPERO under identifier CRD42024502839.

Methodological rigor was central to the design. Risk of bias was assessed with the Quality in Prognostic Studies tool, a validated instrument developed specifically for prognostic factor research, and the certainty of the pooled evidence was graded using the GRADE system, which rates confidence in findings from high to very low. Meta-analyses were only attempted for predictors reported in at least three studies with sufficient homogeneity to justify statistical pooling. That constraint proved decisive, because although the included studies reported a remarkably wide range of candidate prognostic factors, only a handful had been examined consistently enough to combine: age, sex, educational attainment, living alone, and global cognitive function measured with the Mini-Mental State Examination.

The results of the pooled analyses upend several common assumptions. Age, often assumed to accelerate functional deterioration in dementia, showed no significant effect, with a standardized mean difference of 0.0 and a 95 percent confidence interval of 0.0 to 0.1, at moderate certainty. Years of education fared no better: the pooled estimate was 0.00 with a confidence interval of −0.02 to 0.02, also at moderate certainty. In other words, neither being older nor having spent fewer years in school meaningfully changed how quickly daily functioning declined once dementia was established. The finding challenges the intuitive notion that chronological age is a reliable yardstick for disease progression in this population.

Baseline cognition, by contrast, emerged as the single dependable predictor. Better global cognitive function at the start of follow-up, as measured by the MMSE, was associated with slower subsequent decline in activities of daily living, with a standardized mean difference of −0.05 and a 95 percent confidence interval of −0.06 to −0.04, rated at moderate certainty. Although the effect size appears modest in statistical terms, its consistency across studies gives it practical weight: a patient who scores higher on cognitive testing at diagnosis can be expected to retain functional independence longer than a patient with equivalent diagnosis but poorer cognition. This aligns with a broader literature linking cognitive and functional trajectories in dementia, but the new analysis is the first to establish the association through a formally graded synthesis of longitudinal community-based studies.

Two further factors—sex and living alone—could not be interpreted with confidence. The evidence for both was rated as very low certainty and inconsistent across studies, meaning the pooled data could not determine whether men and women, or people living alone versus with others, decline at different rates. The authors conclude that only baseline cognition can currently be recommended for predicting functional decline in dementia, while age and sex may be of limited importance, and that substantially more research is needed to improve person-centered prognostication.

The review is as notable for what it exposes about the research field as for what it confirms. Risk of bias was frequently judged high for participant selection, in 16 of 19 reports, and for attrition, in 14 of 19, often because of poor reporting rather than demonstrable flaws. This pattern matters because prognostic studies are particularly vulnerable to selection effects: cohorts recruited from memory clinics may differ systematically from the broader population of people with dementia living at home, and loss to follow-up is rarely random, since the sickest participants are often the first to drop out. The authors’ transparent accounting of these weaknesses, and their decision to grade certainty rather than simply pool everything, models the approach that prognostic factor research has been urged to adopt under frameworks such as PROGRESS.

The heterogeneity problem also deserves attention. Studies measured functional decline with different instruments, over different follow-up periods, and adjusted for different covariates, which is precisely why only five factors could be meta-analyzed at all. Candidate predictors that clinicians might consider important—comorbidities, polypharmacy, behavioral symptoms, physical performance measures such as balance, caregiver characteristics, and treatment effects—appeared in individual studies but could not be synthesized. The Danish team’s conclusion that the evidence base is limited and heterogeneous is a call to action: without standardized measurement and better reporting, the field cannot build the multivariable prediction models that would allow genuinely individualized forecasting.

For clinicians, caregivers, and health planners, the practical message is nonetheless actionable. When counseling a family after a dementia diagnosis, the strongest available evidence supports using baseline cognitive performance, not age or education, to gauge the likely pace of functional loss. That information can inform realistic planning for home support, respite care, and eventual transitions in care arrangements, while avoiding the fatalism that an advanced age alone might invite. At the same time, the review tempers expectations: with moderate certainty for only one predictor and very low certainty for others, prognostication in dementia remains an inexact science. The 9,490 participants whose data underpin this analysis have clarified what is known, and just as importantly, mapped the terrain of what remains unknown. Closing those gaps will require the kind of high-quality, well-reported longitudinal research that this review shows is still too rare.

Subject of Research: Prognostic factors for functional decline in community-dwelling older adults with dementia

Article Title: Prognostic factors for functional decline in community-dwelling older adults with dementia: a systematic review and meta-analysis

Article References: Riis, J., Melgaard, D., Gunmalm, V., Petersen, K. S., Jørgensen, M. G., & Andersen, S. (2026). Prognostic factors for functional decline in community-dwelling older adults with dementia: a systematic review and meta-analysis. European Geriatric Medicine. https://doi.org/10.1007/s41999-026-01613-9

Image Credits: AI Generated

DOI: 10.1007/s41999-026-01613-9

Keywords: dementia, functional decline, activities of daily living, prognosis, systematic review, meta-analysis, older adults, cognition, MMSE, geriatrics, community-dwelling, risk of bias