AI for Dementia Caregiver Training Could Reduce ER Visits and Transform Neurological Care at Home

It's 2 a.m. in a suburban kitchen in Raleigh, North Carolina. A 58-year-old woman, Linda, is standing across from her mother, who has Alzheimer disease and is convinced that Linda is a stranger who has broken into her house. Linda's mother is agitated, pacing, raising her voice. Linda has no training in de-escalation. She has no framework for understanding that this is a behavioral and psychological symptom of dementia (BPSD), not a personal rejection. She has a pamphlet from the neurologist's office, a phone number for the Alzheimer's Association helpline, and a growing sense that she is failing at something nobody taught her how to do.

This scene repeats itself thousands of times a night across the United States, and it is one of the most consequential failures in neurological care that nobody talks about.

The Training Gap for Dementia Care at Home

Running one of America's largest dementia care platforms at CareYaya, I've seen this gap up close. We've mobilized over 50,000 health care students nationwide as caregivers for seniors with dementia, and the number one piece of feedback from families is not about scheduling or cost – it's that they themselves have no idea what they're doing.

According to the AARP and National Alliance for Caregiving's 2025 report, only about 11% of family caregivers report receiving any formal training in Activities of Daily Living, and only 22% receive training for medical or nursing tasks, despite more than half performing those tasks at home.¹ When you narrow the lens to dementia caregiving specifically, the picture is similarly bleak: 1 analysis found that roughly 1 in 12 informal dementia caregivers had received any caregiving training at all.²

Let that sink in. We're talking about nearly 12 million Americans who provided 19.2 billion hours of unpaid dementia care in 2024, valued at $413.5 billion.³ These are not casual volunteers. They are performing complex medical work like managing medications, monitoring for delirium, navigating fall risks, handling incontinence care, and responding to behavioral crises including agitation, paranoia, wandering, and resistance to care – all with essentially zero preparation.

Measurable Clinical Consequences from Lack of Caregiver Training

For NeurologyLive readers, beyond a societal problem, I think this is a clinical one with neurological consequences that show up in your emergency departments and on your hospital floors.

A 2025 study published in the Journal of Applied Gerontology examined over 426,000 older home health patients with dementia and found that 83% experienced an unmet need for caregiving. Patients with high cognitive symptom severity had 80% higher odds of unmet needs due to lack of caregiver capacity.⁴ In a separate large cohort study published in JAMA Network Open in 2025, higher caregiver burden was directly associated with increased likelihood of an emergency department revisit within 30 days after discharge.⁵

The mechanism is quite straightforward: untrained caregivers hit a behavioral or medical crisis they can't interpret, panic, and call 911. The patient ends up in the emergency department, where the disorientation of an unfamiliar environment frequently worsens BPSD symptoms, leading to longer stays, higher costs, and worse outcomes. Total dementia care spending is projected to hit $384 billion in 2025 and approach $1 trillion by 2050.³ A meaningful share of that spending traces back to preventable utilization triggered by caregiver knowledge gaps.

And the toll on caregivers themselves can now be quantified. The Alzheimer's Association reported in 2025 that dementia caregivers are 30% more likely to smoke, 27% more likely to develop hypertension, and significantly more likely to delay their own medical care.⁶ Nearly 60% have at least 1 modifiable risk factor that increases their own chances of developing dementia. The stress moves beyond just emotional to also physiological, and it compounds.

Why Traditional Training Models Don't Scale

To be sure, evidence-based caregiver training programs exist and they work. REACH II and REACH VA, the Savvy Caregiver program, and Tele-Savvy have demonstrated real improvements in caregiver mastery, self-efficacy, and well-being in randomized controlled trials.^7,8 Digital and telehealth caregiver interventions, in recent meta-analyses, show meaningful short-term improvements in burden and stress.⁹

But 2 realities block scale:

Time and logistics. Traditional programs are multisession, synchronous, and scheduled – exactly the format that is hardest to attend when caregiving intensity is at its peak. A daughter working full-time and managing her mother's middle-stage Alzheimer disease doesn't need a 6-week evening course. She needs to learn how to respond to a bathing refusal at 7:45 AM on a Tuesday.

Personalization. Dementia symptoms, triggers, culture, family dynamics, and home environments vary wildly. Behavioral and psychological symptoms of dementia affect up to 90% of patients over the course of illness¹⁰, but the triggers are highly individual: pain, fear, overstimulation, constipation, medication effects, environmental changes. One-size curricula struggle to address this variability.

Even the WHO's iSupport, a self-guided online program designed for global scale, has shown mixed effectiveness results in rigorous trials – an instructive reminder that content alone is not the same as engagement plus personalization plus coaching.¹¹

YayaGuide: AI-Powered Micro-Learning for Dementia Caregivers

This is the gap AI is uniquely positioned to close, not by replacing clinicians, but by scaling something health care has never reliably delivered: just-in-time, personalized training for the person doing the care.

YayaGuide is our answer to this problem. Developed initially through the Johns Hopkins AI and Technology Collaboratory for Aging Research (AITC) and backed by an NIH/NIA grant (R43AG094339), YayaGuide is an AI-enabled, personalized caregiver training platform built around micro-learning and conversational coaching.^12,13

The practical idea is that family caregivers don't need the full scope of a certified nurse aide curriculum. Most can't afford the time or the thousands of dollars those programs cost. What they need is a targeted competence pathway – the specific skills relevant to this stage of dementia, this home setup, these behavioral triggers, these comorbidities, and this caregiver's learning pace. Think of it as the Duolingo of dementia care training.

AI enables 5 capabilities here that are genuinely difficult to scale otherwise:

1. Micro-learning that fits real caregiving life. Training has to work in 3-minute windows between tasks. YayaGuide delivers short, adaptive modules and a conversational agent that reinforces key skills and checks understanding, not unlike the spaced-repetition approach that makes language-learning apps so effective.

2. Personalization to the caregiver and the patient. The platform adapts training based on caregiver-reported scenarios and measured competencies, adjusting what comes next based on what the caregiver has demonstrated they've learned. The same "agitation" event might require completely different training depending on whether the underlying driver is pain, overstimulation, or a medication adverse effect.

3. Evidence-based behavioral management coaching. Modern dementia care increasingly emphasizes nonpharmacologic first-line approaches for BPSD — because they're safer and often effective when properly applied.¹⁴ Frameworks like DICE (Describe-Investigate-Create-Evaluate) help caregivers systematically identify triggers and test targeted interventions.¹⁵ AI can walk a caregiver through the DICE framework conversationally, in the moment, without requiring a scheduled appointment with a specialist.

4. Psychosocial skill-building, not just task checklists. The hardest parts of dementia caregiving are relational: de-escalation, validation therapy, preserving the patient's dignity, avoiding power struggles, and navigating family conflict about driving, finances, safety, and independence. Training needs to address communication patterns and caregiver self-regulation – areas where conversational AI coaching can rehearse scenarios repeatedly and without judgment.

5. A bridge from clinic plans to home reality. Clinicians give care plans; families live them. AI tools can translate "watch for delirium" into concrete checklists, red-flag thresholds, and rehearsal scenarios – helping caregivers decide when to call the clinic, when to adjust the environment, and when the ED is truly necessary.

A Realistic Claim About Costs and Utilization

I want to be honest with readers here, because overpromising is a problem in health tech. It's tempting to claim that caregiver training will slash hospitalizations. The more honest position is that the mechanism is strong. Caregiver skills reduce crises, improve adherence to care plans, and reduce burden – all factors tied to emergency department revisits and poor care transitions.⁵ Some training programs for paid caregivers have shown promising reductions in utilization and costs.¹⁶ But broad reviews of health service interventions in community dementia populations have found mixed results on acute hospital use.¹⁷

This is exactly why NIH-funded development matters. We need rigorous evaluation, not just hype. YayaGuide is being built with measurement infrastructure from the start, not as a consumer app released into the wild, but as a thoughtful tool designed to generate evidence about what works, for whom, and under what conditions.

Guardrails and What Responsible AI Caregiver Training Must Include

As I discussed in my previous NeurologyLive article on social robotics and embodied AI in neurology care, when AI touches neurological populations, it must be held to careful standards.¹⁸ For caregiver training specifically, that means:

1. Scope control: Training and coaching, not diagnosis or prescribing.
2. Evidence grounding: Modules aligned with established best practices, updated as evidence evolves.
3. Escalation logic: Clear "call your clinician" and "call 911" decision thresholds built into every interaction.
4. Privacy and consent: Caregivers share intimate home-life details with these platforms. Data safeguards are non-negotiable.
5. Bias and accessibility: Multilingual support, low-literacy design, and cultural competence are not optional add-ons – they are requirements for any platform that aims to reach the populations most affected by dementia caregiving disparities.

I believe that AI should not replace neurologists or the multidisciplinary care team. But it can extend what clinicians wish they could provide: frequent reinforcement, behavioral rehearsal, and practical support between visits.

The Moment for Neurology to Lead

The NIA convenes its 2026 Dementia Care and Caregiving Research Summit next month. Washington state is rolling out the country's first universal long-term care insurance benefit this year, and seven other states are weighing similar programs.¹⁹ The policy infrastructure for supporting caregivers is finally being built.

But insurance without training is like handing someone a prescription without instructions. If we're serious about keeping patients with dementia safely at home, which is where 80% of care already happens, then caregiver training must be embedded in every long-term care policy, every discharge plan, and every neurology practice workflow.

NeurologyLive readers are in a unique position here. You are the clinicians delivering those diagnoses. You see the fear in families' eyes when they realize what comes next. Right now, we hand them a pamphlet and hope for the best. AI gives us a chance to replace hope with learning – personalized, evidence-based, and available at 2 a.m. when the crisis actually hits.

Linda, the dementia caregiver in Raleigh, deserves better than a pamphlet. The 12 million Americans doing this work right now deserve better. And the patients they care for – your patients – will be the ones who benefit most when we finally take caregiver training as seriously as we take the diagnosis itself.

REFERENCES
1. AARP & National Alliance for Caregiving. Caregiving in the U.S. 2025. AARP. Published 2025.
2. Kim J, et al. Receipt of caregiver training among dementia caregivers. J Appl Gerontol. Published 2023.
3. 2025 Alzheimer's disease facts and figures. Alzheimers Dement. 2025;21(4). doi:10.1002/alz.70235
4. Burgdorf JG, Wolff JL, Barrón Y, Amjad H. Dementia Severity Associated With Unmet Caregiving Needs During Skilled Home Health Care. J Appl Gerontol. 2026;45(1). doi:10.1177/07334648251332232
5. Germain N, et al. Caregiver burden and 30-day ED revisits. JAMA Netw Open. 2025.
6. Alzheimer's Association. New Analysis Finds Dementia Caregivers More Likely to Report Modifiable Risk Factors. Published October 14, 2025. Accessed February 10, 2026. https://www.alz.org/news/2025/dementia-caregivers-modifiable-risk-factor-analysis
7. Nichols LO, et al. Translation of a dementia caregiver support program in a health care system — REACH VA. Arch Intern Med. 2011;171(4):353-359.
8. Hepburn K, et al. Tele-Savvy: an online psychoeducation program for dementia caregivers. J Gerontol Nurs. 2021.
9. Lumini MJ, et al. Digital interventions for dementia caregivers: a systematic review and meta-analysis. Int J Environ Res Public Health. 2025.
10. Cerejeira J, et al. Behavioral and psychological symptoms of dementia. Front Neurol. 2012;3:73. doi:10.3389/fneur.2012.00073
11. WHO iSupport Programme. Effectiveness evaluations in randomized controlled trials. World Health Organization. Published 2024.
12. Johns Hopkins AITC. Hopkins AITC Announces Awardees of Third Funding Round. Whiting School of Engineering. Published 2024. https://engineering.jhu.edu/news/hopkins-aitc-announces-awardees-of-third-funding-round/
13. NIH/NIA. Award R43AG094339: AI-Enabled Personalized Training for Caregivers of Older Adults with ADRD. https://reporter.nih.gov/project-details/11182438
14. Kales HC, Gitlin LN, Lyketsos CG. Assessment and management of behavioral and psychological symptoms of dementia. BMJ. 2015;350:h369.
15. Kales HC, Gitlin LN, Lyketsos CG. Management of neuropsychiatric symptoms of dementia in clinical settings: recommendations from a multidisciplinary expert panel. J Am Geriatr Soc. 2014;62(4):762-769.
16. Reuben DB, et al. The University of California at Los Angeles Alzheimer's and Dementia Care program for comprehensive, coordinated, patient-centered care. Health Aff. 2019;38(4):631-638.
17. Godard-Sebillotte C, et al. Impact of health service interventions on acute hospital use in community-dwelling persons with dementia: a systematic literature review and meta-analysis. PLoS One. 2019;14(6):e0218426.
18. Shah NK. How Social Robotics and Embodied AI Could Reshape Neurology Care. NeurologyLive. Published December 15, 2025. https://www.neurologylive.com/view/how-social-robotics-embodied-ai-could-reshape-neurology-care
19. NPR. Policy relief for family caregivers seems stalled out. But there are signs of change. Published December 30, 2025. https://www.npr.org/2025/12/30/nx-s1-5660935/family-caregiver-support-policies-states

Neal K. Shah is a health care researcher specializing in caregiving, workforce innovation, and artificial intelligence. He is an NIH-funded Principal Investigator on the YayaGuide AI for Caregiver Training project which he started at Johns Hopkins, and the co-Principal Investigator of the University of Pennsylvania's Counterforce Health AI project to help patients and clinics appeal health insurance denials. Neal also serves on North Carolina's Steering Committee on Aging. He is CEO of CareYaya, Chairman of Counterforce Health and the author of Insured to Death: How Health Insurance Screws Over Americans - And How We Take It Back.