Assistive Technologies

The World Health Organization defines assistive technology broadly: products, systems and services that maintain or improve functioning in areas including cognition, communication, mobility, self-care, hearing and vision. More than 2.5 billion people are estimated to need at least one assistive product.¹ That scale matters, but it does not turn any connected object into a universal solution. Access, affordability, fitting, training, maintenance and follow-up are part of the technology too.²

A memory problem after brain injury is not the same as progressive dementia, visual impairment or everyday overload. People with the same diagnosis may value different routines, and needs can change. Selection therefore begins with a meaningful activity: making breakfast, taking the right medicine, reaching an appointment or walking outside.

Marketing usually shows the moment a reminder appears or a sensor notices movement. Daily usefulness depends on what happens before and after that moment.

1. Personal goalA desired activity, routine or freedom defines success.
2. Cue or signalA label, alarm, sensor or person notices the relevant moment.
3. Supported actionThe next step is clear, possible and easy to initiate.
4. ConfirmationThe system distinguishes a prompt from what actually happened.
5. ResponseA person or service acts only when the agreed threshold is met.
6. ReviewThe user and supporters check benefit, burden and changing needs.

A spoken medication reminder, for example, can show that a cue was delivered. It cannot show that the correct tablet was swallowed. A door sensor can show that a door opened, not why. A motion sensor can show a period without detected movement, not whether someone is asleep, reading beyond its range, away from home or in difficulty. Sensors observe proxies; software assigns meaning; people carry the consequences.

Build backwards from the last link. If nobody can receive a night-time alert, more sensors may create false reassurance. The user also needs a way to dismiss errors; serious escalation should not rest on one ambiguous signal.

External aids do not “make the brain lazy.” They move part of a task into the environment so attention and memory can be used elsewhere. Cognitive-rehabilitation guidance after traumatic brain injury supports personalised compensatory strategies, including external aids, with structured training and attention to real-life goals.⁶ Evidence is not interchangeable across conditions, however. A Cochrane review after stroke found short-term improvement in self-reported memory problems but uncertain long-term or functional effects.⁷

1. Make the environment do the remembering.Give essential objects one visible home. Use a labelled tray by the door, a large clock and day display, a whiteboard, a paper checklist, photographs on cupboards or the next step placed where the action happens.
2. Add one familiar digital cue.A phone alarm, calendar, timer or voice reminder may work when the device is already carried, charged and understood. Use concrete text—“Put blue folder in bag”—rather than “Prepare.”
3. Link the cue to a routine.Time alone may be a poor trigger. A checklist beside the kettle or a prompt after the front door closes can fit the context better, provided the trigger is dependable and not intrusive.
4. Add confirmation only when it solves a real problem.A tick box, acknowledgement button, dispenser log or supporter call adds information but also effort and surveillance. Define what the confirmation really proves.
5. Seek specialist assessment for complex or changing needs.An occupational therapist, neuropsychologist, speech and language therapist, rehabilitation clinician, pharmacist or assistive-technology service can match strategies to cognition, communication, vision, hearing, movement and environment.

Good prompts are recognisable, specific and actionable. They arrive early enough to help, but not so often that they become background noise. They use more than one channel only when that helps the individual: sound plus a visual message may improve noticeability, while simultaneous sound, vibration and flashing may be overwhelming. The W3C’s cognitive-accessibility guidance favours familiar patterns, clear steps, visible labels, error recovery, limited distraction and processes that do not themselves rely on memory.⁸

Trial prompts in the real place and routine, not only beside a helper. Training should be paced, repeated and use the same words and symbols as the interface.

Medication support is a safety system, not just a notification feature. The appropriate level depends on the regimen, consequences of a late, missed or repeated dose, the person’s dexterity and understanding, packaging, changing prescriptions and who is available to help. NICE recommends an individualised approach to adherence: understand the person’s concerns and practical barriers rather than assuming non-adherence is a simple failure to remember.⁵

Each step is different. A smart cap may record opening; a dispenser may release one compartment; a camera or supporter may observe part of the routine. None automatically proves ingestion, correct technique or a safe clinical decision. Conversely, an unopened container does not prove a missed dose if medication was prepared elsewhere. Logs are clues, not a complete medication record.

A large randomised trial of three simple reminder devices found no improvement among 53,480 adults with low adherence to one to three long-term oral medicines; a cue alone may not resolve practical, motivational or clinical barriers.²⁵ Reconcile the medication list with a pharmacist, name who fills and checks any dispenser, practise a missed-dose scenario and keep a current backup list. High-risk regimens may need human or professional support. WHO likewise treats medication harm as a systems problem, not merely an individual lapse.⁹

Object tags can help find keys or bags, especially beside a fixed storage habit. Results depend on radio technology, compatible networks, surroundings, battery and attachment; “precision finding” is not a universal accuracy guarantee.

Location devices worn or carried by a person raise a different set of questions. They may support safer walking and reduce uncertainty for some people living with dementia, but they are also surveillance tools. Available assessments describe plausible benefits—more freedom, faster location and reduced supporter anxiety—alongside incomplete evidence and important ethical and implementation questions.¹² A systematic review of the ethics literature found recurring tensions around autonomy, privacy, benefit, harm and justice; tracking can support or erode each of them depending on design and use.¹³

A location plan needs more than a geofence

• Meaningful agreement: involve the person early, explain the purpose plainly and revisit consent as circumstances change.
• Defined access: list who can see current or historical location, how access is revoked and which third parties receive it.
• Practical reliability: test charging, coverage, mobile data, attachment and the route used in real life.
• Proportionate boundaries: a late return may first trigger a call, not police. Avoid using continuous history merely because the app makes it available.
• Backup identification: an accessible contact card, agreed community plan and recent photograph may still matter when electronics fail.

A smart home can automate a lamp, show a reminder, notice a door or relay an alarm. Its value comes from a comprehensible routine, not device count. Reviews describe promising uses, but technologies and study quality vary; evidence that these systems reliably keep people with dementia at home remains uncertain.¹⁰¹¹

Automate low-risk, reversible actions first: route lighting, an information display or one command for several lamps. Keep labelled physical controls. Essential functions should not depend on an app, a remembered phrase or one administrator.

Matter, Thread and the limits of compatibility

Matter is an IP-based interoperability standard for compatible connected-home products; Thread is a low-power IPv6 mesh protocol some devices use as transport.¹⁷¹⁸ Those labels do not guarantee every feature across every ecosystem, clinical effectiveness, accessibility, privacy or long-term support. Check the exact device, controller, network and platform features.

Falls, fire, carbon monoxide, water leaks, extreme temperature, open doors and unattended cooking require different engineering and response paths. A motion sensor cannot substitute for a smoke alarm; a watch algorithm cannot prevent a fall; a connected notification is not useful if the recipient is asleep or far away. For serious hazards, choose equipment certified for the exact intended purpose and installed and maintained according to local requirements.

Fall technology deserves restraint: devices and tests vary, evidence for people with dementia or MCI is weak, and false alarms can dominate real deployments.²⁶²⁷ An alert complements—not replaces—assessment of medicines, vision, footwear, strength, balance and hazards. NICE guidance favours individualised prevention rather than a gadget alone.¹⁴

Write the response as a short decision tree

1. Verify what the system knows.

   Which sensor fired, when, and with what confidence? Is there a second signal or direct call?

2. Contact the person in the agreed way.

   Use a speaker, telephone or video only with prior agreement. Allow enough time for mobility, hearing and communication needs.

3. Escalate to a named human.

   The responder needs current contact details, access instructions, relevant risks and authority to act.

4. Use emergency services for emergencies.

   Local criteria and symptoms matter. Automation should not delay urgent help or flood services with unverified routine events.

5. Record and review.

   Was it a true event, a near miss, a device fault or a bad rule? Adjust the environment and plan, not just the threshold.

6. Repair trust after a false alarm.

   Explain what happened, correct access or alert burden and involve the user before re-enabling monitoring.

A prototype may detect a staged event yet fail in daily life. A device liked during a short study may be abandoned when batteries, subscriptions and false alerts accumulate. Match each claim to the outcome actually studied.

1. Technical functionDid the sensor or rule operate under defined conditions?
2. UsabilityCould intended users understand, set up and operate it?
3. Daily adoptionWas it still used correctly outside the demonstration?
4. Functional benefitDid a meaningful activity become easier, safer or more self-directed?
5. Life impactDid participation, wellbeing, care burden or harm improve—and at what cost?

Ask what a percentage means: were falls staged, did adherence mean opening a container, was independence a questionnaire or time at home, were intended users included, and how many stopped using the system?

The pragmatic ATTILA trial is instructive: among 495 people with dementia, a fuller assistive-technology and telecare package did not extend time living independently, reduce safety incidents or improve caregiver wellbeing compared with basic support.²⁴ That does not mean every aid is useless; it shows why a plausible mechanism is not proof of a broad life outcome. Prefer reviews, professional guidance, standards, regulator records and transparent comparative studies to testimonials. Check the exact model, intended use and jurisdiction.

Installation day is a poor test: attention is high, batteries are full and a helper is present. A trial should show whether support survives ordinary life. Start with one goal and add complexity only to solve a remaining problem.

1. Describe one meaningful outcome.

   Use observable language: “arrives at the community lunch with less prompting,” not “improves cognition.” Ask the person what would feel better.

2. Map the current routine.

   Watch where the cue, decision or action fails. Note noise, lighting, fatigue, dexterity, hearing, vision, language, internet and who is nearby.

3. Choose the smallest intervention.

   Try placement, labelling or a familiar alarm before a new account and hub. Compare total cost, subscription, consumables, return terms and support life.

4. Co-design the prompt.

   Use the user’s words, preferred channel and timing. Decide how to snooze, dismiss, correct and ask for help without embarrassment.

5. Practise the complete response.

   Test true events, false alerts, no response, power loss and an unavailable supporter. Confirm that authorised responders can enter safely.

6. Review benefit and burden.

   After a realistic period, compare the target activity, errors, anxiety, interruptions and work shifted to others. Keep, simplify, change or remove.

Professional assessment is valuable when errors have serious consequences, needs interact or cognition is changing. UK occupational therapy, memory, rehabilitation, community-alarm and pharmacy services cover different problems; routes and funding vary. Do not assume a consumer product is reimbursed from a news story.

The UN disability-rights convention connects independent living with equal choice, community inclusion and support—not technological control.³ Someone can value safety and still refuse a bedroom camera, route history or family access to every alert. Seek the least intrusive design that meets the purpose.

Capacity is decision-specific and can fluctuate; diagnosis alone does not remove it. In England and Wales, the Mental Capacity Act presumes capacity, requires practical support and distinguishes an unwise decision from inability to decide. If another person must act, best-interests and least-restrictive principles apply; rules differ elsewhere.¹⁵ An app’s “supporter” role is not legal authority.

Health, biometric and monitoring data may receive heightened protection, but duties depend on operator, purpose and jurisdiction. GDPR principles include purpose limitation, minimisation, accuracy, storage limitation and security.¹⁶ US consumer-app data is not automatically covered by HIPAA merely because it concerns health; FTC rules may still apply.²⁰²¹

A short security baseline

• Use unique passwords and multi-factor authentication where it is accessible; document a safe recovery route that does not depend on one person’s memory.
• Install security updates, replace unsupported products and check how long the maker commits to updates. “Works today” is not a lifecycle plan.
• Keep the home network and router supported. Review guest, installer and third-party integration access after setup.
• Prefer products that disclose vulnerabilities, support secure updates, protect stored and transmitted data and allow safe reset. NIST’s consumer-IoT baseline provides useful capabilities to ask about.¹⁹

Support can fail quietly: dead battery, moved sensor, changed Wi-Fi, locked account, closed cloud service or a responder assuming someone else acted. “No alert” may mean nothing happened—or the observation path broke. Systems should report their health clearly.

Redundancy must be independent, not two messages through one dead phone. Use a paper medication list, physical switch, local alarm and second authorised responder. The higher the risk, the clearer the fallback and maintenance record.

False alarms disturb sleep, strain relationships, desensitise responders and prompt abandonment. Improve placement, rules and context rather than simply changing sensitivity. Review false alerts and known misses with the user.

The most capable system is not the one with the most sensors. It is the one that makes a valued activity more possible, communicates clearly, asks only for necessary data and remains understandable on an ordinary day. It lets the person correct it, pause it and keep a workable alternative.

Begin with one goal. Find the precise point where the routine breaks. Try the lightest support, then test the entire chain from cue to response. Be exact about what was confirmed. Rehearse power, network and human failure. Review both benefit and burden, and remove technology when it stops serving the person.

That is a more demanding standard than “smart.” It is also a more hopeful one. Memory aids and connected homes can create genuine cognitive room, safer routines and greater freedom—but only when intelligence remains distributed among the person, their environment, trusted relationships and tools that know their limits.

1. Assistive technology. World Health Organization fact sheet.
2. Global report on assistive technology. WHO and UNICEF, 2022.
3. Convention on the Rights of Persons with Disabilities. United Nations; see Articles 12, 19 and 20.
4. Dementia: assessment, management and support. NICE guideline NG97.
5. Medicines adherence. NICE guideline CG76.
6. INCOG 2.0 Guidelines: Memory. Cognitive rehabilitation after traumatic brain injury, 2023.
7. Cognitive rehabilitation for memory deficits after stroke. Cochrane evidence review.
8. Making Content Usable for People with Cognitive and Learning Disabilities. W3C Working Group Note.
9. Medication Without Harm. World Health Organization patient-safety initiative.
10. Smart-home technologies supporting people with dementia. Systematic review, 2021.
11. Technologies to Increase Freedom for People Living With Dementia. Evidence report, 2023.
12. GPS Locator Devices for People With Dementia. CADTH horizon scan.
13. Electronic tracking devices in dementia care. Systematic review of ethics literature.
14. Falls: assessment and prevention. NICE guideline NG249.
15. Mental Capacity Act Code of Practice. Applies to England and Wales.
16. General Data Protection Regulation. Official EU text.
17. Matter overview. Connectivity Standards Alliance.
18. What is Thread? Thread Group overview.
19. IoT Device Cybersecurity Capability Core Baseline. NISTIR 8259A.
20. Consumer health information: HIPAA and the FTC Act. US Department of Health and Human Services.
21. Health Breach Notification Rule. US Federal Trade Commission guidance.
22. Smart Home Technologies for Enhancing Independence of Living. Systematic review, 2025.
23. Guiding the Ethics of Locator Devices in Dementia Care. Stakeholder framework, 2026.
24. Assistive technology and telecare to maintain independent living. ATTILA randomised trial.
25. Effect of Reminder Devices on Medication Adherence. REMIND randomised trial.
26. Digital technologies to prevent falls in dementia or MCI. Systematic overview.
27. Real-world accuracy of wearable fall detection. Small deployment study.