Fewer than 3% of adult patients with a diagnosis of dementia or cognitive impairment had a caregiver registered for shared access to their electronic health record portal account, according to a research letter published online August 24, 2026 in the Journal of the American Geriatrics Society. The finding came from an internal audit at a large academic health system in the Midwest that uses the Epic electronic health record and its MyChart portal, conducted by researchers including Lauren Southerland and Courtney Hebert of Ohio State University and Jennifer Wolff of Johns Hopkins University.
The researchers, using a de-identified data query tool inside their own system in December 2025, first looked at all patients with active portal accounts regardless of diagnosis. There, shared access was common for children, with 89% of patients under 18 having a registered caregiver, but that figure fell to under 1% once patients turned 18. The team then asked whether a dementia or cognitive impairment diagnosis changed that pattern for older adults, reasoning that caregivers of patients with memory problems would need portal access more, not less. Shared access was higher in that group across every age band, the letter reported, but it stayed below 3%, a difference the authors describe as real but small against the scale of unmet need.
The result lines up with findings from other health systems. A 2023 research letter in JAMA Internal Medicine, drawing on five years of records at Johns Hopkins, found that among patients 65 and older with diagnosed dementia, 71.20% were registered portal users but only 10.41% had a caregiver with shared access, and those who did logged roughly twice as many portal sessions and sent nearly twice as many messages as those without. A separate 2024 chart review and interview study in JMIR Aging, conducted across two health systems in Maryland and Colorado, likewise found that caregivers of people with dementia are often not consistently documented in the record at all, which the authors said makes it harder for systems to know whom to offer access to in the first place.
None of the three studies were designed to say why shared access remains rare, only that it does. The Ohio State authors note their audit relied on a single institution's Epic build and a SNOMED code for dementia or cognitive impairment, not a formal diagnosis confirmed by a clinician, and covered a six month window. The Johns Hopkins portal-use letter and the Maryland-Colorado chart review each describe a single or two-site academic health system as well, not a national sample, and none of the three studies tracked whether the absence of shared access changed a patient's clinical outcomes.