Key Takeaways
- In a large U.S. All of Us cohort, adults with a recorded ADHD diagnosis had a higher rate of subsequently recorded mild cognitive impairment or dementia than adults without a recorded ADHD diagnosis.
- Among 187,341 adults aged 50 or older, 6.7% of those with ADHD received a subsequent MCI or dementia diagnosis compared with 3.4% of those without ADHD.
- ADHD was associated with a 4.60-fold higher hazard of MCI or dementia in the primary model. The association weakened as researchers accounted for additional medical and psychiatric conditions, health care use, education, smoking, and other factors, but remained statistically significant in multiple analyses.
- Genetic liability to ADHD was also associated with the outcome. Participants in the highest fifth of the ADHD polygenic-risk score had a 35% higher adjusted hazard of subsequently recorded MCI or dementia than those in the lowest fifth. This does not establish causality or provide an individual risk prediction.
- The study does not establish that ADHD causes dementia. Reverse causation is an important concern because early cognitive changes could sometimes resemble or contribute to a late-life ADHD diagnosis.
- The study also does not show that ADHD medication increases dementia risk or that stimulant treatment prevents dementia. The researchers lacked sufficient lifetime treatment information to answer that question.
Introduction
A large U.S. cohort study found that adults with a recorded ADHD diagnosis had a higher rate of subsequently recorded mild cognitive impairment or dementia, although the observational findings do not establish that ADHD causes cognitive decline.
Attention-deficit/hyperactivity disorder is commonly associated with childhood and younger adulthood, but ADHD can persist throughout life. As people with ADHD age, researchers are increasingly examining whether the condition is also associated with cognitive health later in life.
A large study published October 7, 2026, in JAMA Psychiatry adds important evidence to that question.
Researchers analyzed electronic health records and genomic data from 187,341 adults aged 50 or older in the U.S. National Institutes of Health’s All of Us Research Program. Adults with a recorded ADHD diagnosis had a higher rate of subsequently recorded mild cognitive impairment, or MCI, or dementia than those without a recorded ADHD diagnosis. Higher ADHD polygenic-risk scores were also associated with a higher hazard of subsequently recorded MCI or dementia after the study’s specified statistical adjustments (Bauer-Negrini et al., 2026).
The results are striking, but they require careful interpretation. This was an observational study. It cannot establish that ADHD causes dementia, and several findings indicate that differences in health, psychiatric conditions, medical care, and the timing of ADHD diagnoses may explain at least part of the association.
What the Study Examined
The researchers conducted a retrospective cohort study using the All of Us Research Program.
Participants were at least 50 years old at cohort entry and had both electronic health records and genomic data available. The analyses were designed to examine incident mild cognitive impairment or dementia after cohort entry while accounting for the timing of ADHD diagnoses relative to cognitive impairment.
The final cohort included 187,341 adults. Their mean age at cohort entry was 56.4 years, and 57.3% were female.
A total of 3,026 participants, or 1.6%, had a recorded ADHD diagnosis. The researchers followed participants for an average of 10.26 years, although the duration of available electronic health records varied substantially between individuals (Bauer-Negrini et al., 2026).
ADHD and cognitive impairment were identified from electronic health records using prespecified algorithms based on diagnostic information recorded on separate dates. ADHD ascertainment also incorporated ADHD medication records. These approaches were intended to provide more robust case identification than relying on a single diagnostic code alone.
The primary outcome combined newly recorded diagnoses of MCI and dementia.
Researchers also calculated an ADHD polygenic risk score using genomic data. A polygenic risk score combines information from many genetic variants associated with a condition to estimate a person’s genetic liability relative to others in the study population. It is not a diagnostic test and does not determine whether an individual will develop ADHD or dementia.
What Researchers Found
During follow-up, 6,462 participants, or 3.4% of the cohort, received a new recorded diagnosis of MCI or dementia.
Among adults with ADHD, 203 of 3,026, or 6.7%, received a subsequent cognitive-impairment diagnosis.
Among those without ADHD, 6,259 of 184,315, or 3.4%, did so (Bauer-Negrini et al., 2026).
The incidence rates were 8.96 cases per 1,000 person-years among participants with ADHD and 3.29 per 1,000 person-years among those without ADHD.
In the primary adjusted Cox model, ADHD was associated with a 4.60-fold higher hazard of MCI or dementia compared with no recorded ADHD diagnosis.
The hazard ratio was 4.60 (95% CI, 3.99 to 5.29) (Bauer-Negrini et al., 2026).
That number should not be interpreted as meaning ADHD caused a 4.6-fold increase in dementia risk. A hazard ratio compares event rates over follow-up. In an observational study, the estimate can reflect differences between groups that statistical adjustment does not completely remove.
The incidence rates also should not be interpreted as lifetime probabilities of developing dementia.
The Association Weakened With Additional Adjustment
Adults with recorded ADHD differed substantially from those without a recorded diagnosis.
They had a greater documented burden of several psychiatric and medical conditions. Depression, for example, was recorded in 32.8% of participants with ADHD compared with 7.7% without ADHD. Anxiety disorders were recorded in 28.6% versus 6.5%, while substance-use disorders were recorded in 17.5% versus 6.2% (Bauer-Negrini et al., 2026).
The researchers therefore progressively adjusted their analyses for potential confounding factors.
Adding somatic medical conditions reduced the ADHD hazard ratio from 4.60 to 4.04 (95% CI, 3.50 to 4.66).
Accounting for neuropsychiatric comorbidities reduced it further to 3.19 (95% CI, 2.76 to 3.68).
In a more comprehensive model incorporating sex, race and ethnicity, cohort-entry year, APOE ε4 status, medical and neuropsychiatric conditions, health care utilization, educational attainment, and lifetime smoking history, the association remained but was again smaller: HR 3.08 (95% CI, 2.66 to 3.58) (Bauer-Negrini et al., 2026).
This attenuation indicates that measured differences between adults with and without diagnosed ADHD account for some of the initial association. It does not establish exactly how much of the remaining association is causal because residual confounding and other forms of bias can persist after statistical adjustment.
Genetic Liability to ADHD Was Also Associated With Cognitive Impairment
The researchers used genomic information as a complementary approach.
Participants in the highest fifth of the ADHD polygenic-risk score had a 35% higher adjusted hazard of MCI or dementia compared with those in the lowest fifth.
The adjusted hazard ratio was 1.35 (95% CI, 1.25 to 1.46) (Bauer-Negrini et al., 2026).
Unlike a clinical diagnosis made later in life, inherited genetic liability to ADHD cannot result from emerging symptoms of dementia. The researchers therefore considered the polygenic-risk analysis complementary evidence that is not subject to the same potential reverse-causation problem as a late-life ADHD diagnosis.
However, the genetic finding still does not establish that ADHD causes cognitive impairment.
ADHD-associated genetic variants may reflect shared liability with other psychiatric or neurodevelopmental traits or broader biological pathways influencing both ADHD and cognitive health. The polygenic-risk result therefore cannot identify a direct causal pathway from ADHD to dementia.
The genetic association was also considerably smaller than the association involving a recorded clinical ADHD diagnosis.
Reverse Causation Is an Important Concern
One of the study’s most important analyses involved the timing of ADHD and cognitive-impairment diagnoses.
Dementia can develop gradually, with subtle cognitive and behavioral changes beginning years before a formal diagnosis. In some older adults, emerging cognitive difficulties could potentially resemble ADHD symptoms or contribute to an ADHD diagnosis.
The researchers investigated this possibility by progressively requiring longer intervals between ADHD diagnosis and the later cognitive-impairment outcome.
The association became weaker as the required interval increased.
It remained statistically significant when the minimum interval reached eight years, with an HR of 1.44 (95% CI, 1.12 to 1.86).
At a nine-year interval, however, the association was no longer statistically significant: HR 1.28 (95% CI, 0.98 to 1.68) (Bauer-Negrini et al., 2026).
By nine years, 150 ADHD-cognitive impairment cases, or 74%, had been reclassified, substantially reducing statistical power.
The pattern is compatible with reverse causation, but it does not prove that prodromal cognitive disease caused the ADHD diagnoses. Longer-lag analyses also removed many cases, making the estimates less precise.
The authors therefore acknowledge that early cognitive disease being mistaken for or contributing to a late-life ADHD diagnosis cannot be excluded.
The Association Persisted in Other Sensitivity Analyses
The researchers conducted several additional tests.
When participants with ADHD were matched with comparison participants based on age, sex, race and ethnicity, cohort-entry year, APOE ε4 status, medical and psychiatric conditions, and health care utilization, the association weakened substantially but remained statistically significant.
In that matched analysis, ADHD was associated with a 2.26-fold higher hazard of cognitive impairment (95% CI, 1.75 to 2.92) (Bauer-Negrini et al., 2026).
Results also remained statistically significant when researchers examined MCI and dementia separately.
For MCI alone, ADHD was associated with an HR of 3.83 (95% CI, 3.27 to 4.47).
For dementia alone, the HR was 3.07 (95% CI, 2.12 to 4.44).
The polygenic-risk score was also associated with both outcomes separately.
These analyses make it less likely that the entire result arose from one particular statistical definition, but they do not eliminate residual confounding or establish causality.
How the Findings Compare With Earlier Research
The new findings fit a growing body of observational evidence connecting adult ADHD with later cognitive impairment.
A 2023 Israeli cohort study included 109,218 adults and followed participants for up to 17.2 years. During follow-up, dementia was diagnosed in 13.2% of people with adult ADHD compared with 7.0% of those without adult ADHD.
After adjustment for multiple potential confounders, adult ADHD was associated with a 2.77-fold higher hazard of dementia (95% CI, 2.11 to 3.63) (Levine et al., 2023).
The 2023 cohort began with participants who did not have an ADHD or dementia diagnosis, and adult ADHD was treated as a time-varying exposure. That design does not mean participants had been comprehensively assessed for ADHD throughout their lives.
The new All of Us study extends the evidence by examining both recorded clinical ADHD diagnoses and genetic liability within the same large population. It also accounts for APOE ε4 status, an important genetic risk factor for Alzheimer’s disease, along with health care utilization and multiple psychiatric and medical conditions (Bauer-Negrini et al., 2026).
Taken together, the studies support an association between adult ADHD diagnoses and later cognitive impairment. They do not establish whether ADHD itself contributes causally to dementia or whether shared genetic, psychiatric, behavioral, medical, and social factors explain some or much of the relationship.
What the Results May Mean
The study raises an important question about cognitive health in adults with ADHD, but it should not cause people with ADHD to assume that dementia is inevitable.
Most participants with ADHD in this cohort did not receive a diagnosis of MCI or dementia during the observed follow-up. The study also does not provide an individualized estimate of someone’s future dementia probability.
Several explanations could contribute to the association. These include shared genetic liability, psychiatric and medical comorbidities, differences in health care contact and diagnostic detection, health behaviors, and overlap between ADHD symptoms and early cognitive disease.
The study cannot determine the contribution of each explanation.
The distinction is particularly important because difficulties with attention, organization, memory, and executive function can occur in both ADHD and emerging cognitive disorders. A statistical association between the diagnoses cannot determine the underlying cause of such symptoms in an individual patient.
What About ADHD Medication?
The study does not answer whether stimulant treatment changes later dementia risk.
The researchers specifically identified this as an unresolved question because they did not have sufficient information about participants’ lifetime ADHD treatment.
Previous observational research has raised the possibility that the association may differ among people receiving psychostimulants. In the 2023 Israeli cohort, researchers did not observe a clear increase in dementia risk among participants with adult ADHD who received psychostimulant medication (Levine et al., 2023).
However, treatment was not randomly assigned, and differences between people who did and did not receive medication could influence the results. The earlier finding therefore does not establish that stimulant treatment protects against dementia.
The current findings should not be used to start, stop, or change stimulant or nonstimulant ADHD medication in an attempt to prevent dementia. Medication decisions should be made with the prescribing clinician because the new study did not evaluate lifetime treatment exposure or treatment effects.
Limitations
The observational design is the central limitation. Even extensive statistical adjustment cannot remove every difference between people with and without ADHD.
Reverse causation is particularly important. Some ADHD diagnoses recorded in later adulthood could potentially reflect symptoms arising during the long prodromal phase of cognitive disease. The weakening association in longer-lag analyses is compatible with that possibility, although it does not prove it.
Electronic health records also do not capture every diagnosis. The researchers noted that ADHD, cognitive impairment, and other conditions may be underascertained or misclassified in EHR data.
For most participants, available electronic health records began in midlife or later. The first ADHD diagnosis appearing in the record therefore may not represent when ADHD began or when it was first diagnosed clinically.
The mean age at first recorded ADHD diagnosis in the study was 53.3 years, but the authors explicitly caution that this does not necessarily represent the actual age of first diagnosis.
Health care utilization also differed between groups. People who interact more frequently with the medical system may have more opportunities for both ADHD and cognitive impairment to be recognized and documented.
Polygenic-risk scores have additional limitations. Their performance can vary across ancestry groups because the genetic studies used to construct them have historically included disproportionate numbers of people of European ancestry. The genetic estimate should therefore not be assumed to perform equally across all ancestry groups.
The researchers also lacked sufficient lifetime ADHD-treatment information to determine whether treatment influenced cognitive outcomes.
Finally, All of Us is a voluntary research cohort. Participants may differ from the broader U.S. population, limiting how precisely the results can be generalized to all adults with ADHD.
Final Thoughts
In a cohort of 187,341 U.S. adults aged 50 or older, a recorded ADHD diagnosis was associated with a substantially higher hazard of subsequently recorded mild cognitive impairment or dementia. Higher genetic liability to ADHD was also associated with the cognitive outcome, providing complementary evidence that the relationship warrants further investigation.
But the size of the clinical association became smaller as researchers accounted for psychiatric conditions, medical conditions, health care use, education, smoking, and other factors. Longer intervals between ADHD and cognitive-impairment diagnoses also weakened the association, leaving reverse causation as an important possible explanation (Bauer-Negrini et al., 2026).
The study therefore strengthens evidence that ADHD diagnoses and late-life cognitive impairment are associated. It does not demonstrate that ADHD causes dementia, predict whether an individual with ADHD will develop dementia, or establish that ADHD medication changes dementia risk.
The next challenge is to determine why the association exists, whether particular groups of adults with ADHD face greater risk, and whether modifiable health factors or treatment influence long-term cognitive outcomes.
Funding and Disclosures
The All of Us Research Program is supported by the National Institutes of Health Office of the Director. The paper also reports grants, fellowships, and other investigator support from organizations including the National Institute on Aging, Alzheimer’s Association, National Institute for Health and Care Research, and European Research Executive Agency. These sources did not necessarily fund every aspect of the study or every investigator in the same way.
The authors reported that the funders had no role in study design or conduct, data collection or management, analysis, interpretation, manuscript preparation or review, or the decision to submit the paper for publication (Bauer-Negrini et al., 2026).
Tharick Pascoal reported serving on an NIH-funded data safety monitoring board and previously serving on a Johnson & Johnson scientific advisory board. Guilherme Bauer-Negrini and Guilherme Povala reported Alzheimer’s Association grants during the study. Dana Tudorascu reported a University of Pittsburgh grant during the study and an American Psychiatric Association honorarium outside the submitted work.
Douglas Leffa reported ADHD-related conference travel awards and reimbursement for a lecture. Samuele Cortese reported research grants, honoraria from Medice, and reimbursement for educational and lecture-related activities. Luis Augusto Rohde reported research support, consulting fees, speakers’ bureau fees, educational or research support, and royalties involving multiple pharmaceutical companies and publishers, including companies marketing ADHD medications. No other disclosures were reported (Bauer-Negrini et al., 2026).
The 2023 study by Levine and colleagues reported no conflicts of interest. One author received funding from the Israeli National Insurance Institute, which had no role in the research or publication decision (Levine et al., 2023).
References
Bauer-Negrini, G., Leffa, D. T., Ferreira, P. C. L., Bellaver, B., Povala, G., Medeiros, M. S., Lussier, F. Z., Tudorascu, D. L., Cortese, S., Rohde, L. A., Molina, B. S. G., & Pascoal, T. A. (2026). Attention-deficit/hyperactivity disorder and late-life cognitive impairment. JAMA Psychiatry. https://doi.org/10.1001/jamapsychiatry.2026.3165
Levine, S. Z., Rotstein, A., Kodesh, A., Sandin, S., Lee, B. K., Weinstein, G., Schnaider Beeri, M., & Reichenberg, A. (2023). Adult attention-deficit/hyperactivity disorder and the risk of dementia. JAMA Network Open, 6(10), e2338088. https://doi.org/10.1001/jamanetworkopen.2023.38088