The Dementia Mismatch

We are thinking about people with dementia

In 2021, 57 million people were living with dementia; nearly 10 million more develop it every year, and over 60% of them live in low- and middle-income countries.1 The global cost was US$1.3 trillion in 2019, and about half of that is informal care — family members who provide on average five hours of care and supervision a day, 70% of those hours by women.1,2

The WHO and Alzheimer's Disease International expect 78 million people with dementia by 2030 and 139 million by 2050;2,3 the Global Burden of Disease forecast published in The Lancet Public Health is higher still — 152.8 million by 2050 (uncertainty interval 130.8–175.9 million), a 166% increase on the 57.4 million of 2019.4 ADI estimates that 75% of people with dementia are not diagnosed — about 41 million people — and that the share reaches 90% in some low- and middle-income countries.5 Cost is expected to reach US$1.7 trillion by 2030, or US$2.8 trillion if corrected for rising care costs.2

Japan: more than ten million people when MCI is counted

The latest estimate commissioned by the Ministry of Health, Labour and Welfare (Kyushu University, Prof. Toshiharu Ninomiya, published May 2024) puts the number of people aged 65 and over living with dementia at 4.432 million in 2022 (12.3% of the age group), plus 5.585 million with mild cognitive impairment (15.5%) — together more than 10 million people, roughly one older adult in 3.6.7,8 The projections rise to 5.842 million with dementia in 2040 (14.9%) and 6.451 million in 2060 (17.7%), with MCI plateauing around 6.3 million.7 Young-onset dementia affects about 36,000 people (50.9 per 100,000 aged 18–64).8 For context, the 2015 New Orange Plan had expected about 7 million people with dementia by 2025 — one in five older adults;9 the new estimate is lower in headcount because cohorts are healthier, but the combined dementia-plus-MCI population is larger than anyone planned for.

Korea and the United States

Korea's 2023 national dementia epidemiology survey (published March 2025) found a prevalence of 9.25% among people aged 65+ and 28.42% for MCI; it projects 970,000 people with dementia in 2025 (9.17%) and 2.98 million with MCI, the one-million mark crossed in 2026, two million in 2044, and a peak of 2.33 million in 2059.19 The national dementia-management cost was ₩22.9 trillion in 2023 (0.95% of GDP), or ₩26.39 million per person per year.20 In the United States, 7.2 million people aged 65+ live with Alzheimer's disease; nearly 12 million family members and friends provide 19.2 billion hours of unpaid care valued at $413 billion, and formal care costs are projected at $384 billion for 2025.21 (The 2026 edition updates these to 7.4 million, 13 million, $446 billion and $409 billion.21)

Prevention is also personal: the 14 risk factors

The 2024 Lancet Commission estimates that 45% of future dementia could be prevented or delayed if 14 modifiable risk factors were eliminated; the two factors added in 2024 — high LDL cholesterol in midlife and untreated vision loss in late life — account for 7% and 2% respectively.6 No two people carry the same set, which is the first reason any serious prevention or care service has to be personal.

The lens: match and mismatch among four groups

Each of the four groups has a legitimate goal, and each goal is a proxy for the same thing: a good life for the person. But proxies drift. A family buys safety; a company builds what the buyer fears; a care facility buys what fits its workflow and budget; a professional rates needs the person does not feel. The diagram shows the four groups, what each tends to optimize for, and the six relationships between them. The numbers on the edges come from the evidence in the next section.

So I researched whether the mismatches exist. They do.

3.1 The person versus the family

The clearest evidence is the gap between how people with dementia rate their own lives and needs and how the people closest to them rate them on their behalf. In 1,330 dyads across European home- and institutional-care settings, people with dementia rated their quality of life (QoL-AD) at 33.2 on average, their proxies at 29.8, and item-level agreement was poor (κ < 0.20 for all but one item).22 In 266 US dyads with mild to moderate dementia, caregivers "consistently underestimated" the importance the person placed on all five values measured — autonomy, burden, control, family and safety — and over four years the caregivers' perception of what mattered to the person decreased further.23

Needs are rated differently too. In a UK study of 152 people with dementia living at home, the person reported on average 1.17 unmet needs, their carers 2.14, and the professionals 2.64; "daytime activities" were rated unmet for 14.5% of people by the person, 41.1% by carers and 50.7% by professionals, while the person's own most-cited unmet need was psychological distress (21.6%).24 A Dutch CANE study (236 people, 322 carers) found the same pattern: people with dementia report fewer (unmet) needs than their carers, with the most unmet needs in memory, information, company, psychological distress and daytime activities.25

The safety–autonomy tension is the sharpest form of this mismatch. In interviews about GPS tracking, carers gave "preference to patients' safety more than autonomy when they are responsible for the patients", and the main benefit identified was "the peace of mind of the caregivers themselves".26 When 296 family and professional carers were asked who should decide whether a person with dementia is tracked, the person themselves ranked third — after the spouse and the most-involved family carer.26

Japan's own public opinion shows the same structure from the other side. In the Cabinet Office's August 2025 survey (1,707 valid responses), the most common fear about developing dementia was "placing a physical or mental burden on my family" (74.6%), ahead of "losing memories of my family" (51.8%). Asked how they would want to live with dementia, 27.1% chose staying in the community with medical and care support and 15.1% living independently with some support, while 42.4% chose moving to a facility — and 27.5% of those gave as their reason "so as not to trouble the people around me". 35.6% still picture dementia as "unable to manage daily life, needing admission", and 75.3% did not know the Basic Act on Dementia exists.27

3.2 The person versus the companies

The second mismatch is between what is sold and what is used. The classic study of assistive technology found that 29.3% of all devices were completely abandoned, that abandonment peaked in the first year and again after five years, and that the strongest predictor was "lack of consideration of user opinion in selection".28 A systematic review catalogued 539 intelligent assistive technologies for dementia — growing by roughly 400% every five years — but concluded that "structural limitations to successful adoption" persist, "including partial lack of clinical validation and insufficient focus on patients' needs", and that a significant portion is designed without explicit ethical consideration.29

Reviews of the design process say why. Mobile technologies for dementia "support basic physiological and safety needs, whereas many pay little attention to higher-level needs such as self-esteem and agency"; "little attention seems to be paid to personhood".30 In development projects, people with dementia were mainly "objects of study and informants" rather than co-designers;31 the INTERDEM network's position paper notes that people with dementia want to be included, and a 2025 umbrella review of 112 systematic reviews still lists "personalization of applications" as the unsolved development challenge.31 In real households, "access to AT was driven by carers, with the majority of benefits being experienced by carers", and families improvised do-it-yourself devices because formal provision did not meet their needs;32 a review of 79 papers on surveillance technology in residential care found "little attention for the resident perspective".33

The commercial consequence in Japan is stark. The Japan Research Institute estimates that people with dementia or MCI living at home spend about ¥14.74 trillion a year (2025), yet products and services adapted to them are projected at only ¥98.7 billion in 2030 — a 1.7% "replacement rate" — and ¥531.3 billion (8.6%) in 2050.34 The money is there; the fit is not.

The MIND at Home baseline study in Baltimore (254 people with dementia living at home, 246 caregivers) shows that both sides are under-served, but in different domains — which is itself a design brief. 99% of the people with dementia had at least one unmet need (mean 7.7), concentrated in safety (90%; fall risk 70%, wandering risk 33%), general health care (63%), meaningful activities (53%) and legal/advance care planning (48%); 32% had never had a dementia evaluation. 97% of caregivers had unmet needs (mean 4.6), but theirs were referrals to resources (88%), dementia education (85%) and mental-health care (45%).46

3.3 The companies versus the professionals and carers

Japan's care-robot story is the cleanest example of a supply-side forecast meeting a demand side that did not buy. In 2013 METI projected a potential market for nursing-care and welfare robots of ¥123.9 billion in 2025 and ¥404.3 billion in 2035.35† Yano Research measured the actual care-robot market at ¥2.175 billion in FY2021 and forecast ¥3.625 billion for FY2025 — roughly one-thirtieth of the projected potential — and attributed growth prospects only to products whose "functions and prices have been revised to reflect user needs".35 (The broader care-DX software market was ¥54.58 billion in FY2024.35)

Adoption is finally rising, but for reasons that have little to do with the person with dementia. In the MHLW's 2025 survey of 6,875 establishments, 47.2% of residential and overnight services had introduced monitoring devices (30.0% in FY2022) and 56.4% care-operations support systems (10.2% in FY2022). Among the 1,292 non-adopters, the reasons were initial cost (67.9%), running cost (48.8%), the need for staff training (36.1%) and "difficulty choosing devices suited to the problems on our floor" (32.1%); 67.3% had never applied for a subsidy.36 What care workers actually do with their time explains the appetite: in an MHLW time study, direct care took 40.9% of an eight-hour shift (196.4 minutes), records 9.0% (43.1 minutes) plus 4.5% transcribing them, movement 18.5% and other indirect work 21.0%; after technology was introduced, documentation fell to 6.0%.37 Yet only 49.4% of establishments felt a lighter daytime burden and 44.6% at night.38

3.4 The professionals versus the families

Families want hands; the system is short of them. Japan had about 2.15 million care workers in FY2022 and needs about 2.40 million in FY2026 (+250,000, or 63,000 a year) and 2.72 million in FY2040 (+570,000).16 The job-openings-to-applicants ratio for care occupations was 3.97 in March 2025 against 1.16 for all occupations.39 65.2% of establishments report a staff shortage (83.4% for home-visit helpers), even though turnover fell to a record-low 12.4% — the problem is recruitment, not churn.38 Globally, the WHO projects a shortfall of about 11 million health workers by 2030.40

Meanwhile the family is older, fewer and still working. In 2022, in 63.5% of co-resident caregiving pairs both people were aged 65 or over (59.7% in 2019), and in 35.7% both were 75 or over; 63.1% of co-resident main carers of someone at care level 5 provided care "almost all day", and 74.5% of all-day carers were women.15 106,000 people left their jobs for caregiving in the year to 2022 (99,000 in 2017); 6.29 million people were caring for a family member, 3.65 million of them while employed.14 The mental-health cost is documented: across 17 studies and 10,825 caregivers of people with Alzheimer's, 34.0% had depression, 43.6% anxiety and 27.2% used psychotropic medication; spousal carers had 2.51 times the odds of depression.41 ADI's 2022 surveys found 54% of informal carers stressed "often or all of the time" (92% to some degree), while 37% of paid professionals were stressed often or always and 47% said they were not adequately compensated.42

3.5 The system itself

Behind the four groups sits a health system that too often never starts the conversation. 75% of people with dementia are undiagnosed; in ADI's 2021 survey of 1,111 clinicians and 2,325 people with dementia and carers, one in three clinicians believed nothing could be done, making diagnosis futile.5 In the 2019 attitudes survey of almost 70,000 people in 155 countries, almost 62% of healthcare providers thought dementia was a normal part of ageing, 40% of the public thought doctors and nurses ignore people with dementia, and 35% of carers had hidden the diagnosis.43 After diagnosis, 37% of people in high-income and 45% in lower-income countries were offered no support; 64% had no personalised care plan, and ADI estimates that up to 85% of people with dementia miss out on post-diagnosis care.42 Only a quarter of countries have a national dementia plan.2 Japan has its own "blank period": in a Sendai survey, 61.2% of families started long-term-care services within a year of diagnosis, but 17.1% took three years or more, and longer gaps meant heavier mental, physical and financial burden.44†

3.6 Where the interests match

The good news is that the four groups agree on the destination. 77% of US adults aged 50+ want to stay in their own home and 79% in their community;45 in Japan 49.3% would want to stay in the community versus 42.4% in a facility — and the facility choice itself is often motivated by not wanting to burden the family, so the underlying wish is shared.27 Both the person and the carer are almost universally under-served (99% and 97% with at least one unmet need46), which means a service that addresses both is wanted by both.

And personalised coordination demonstrably works. In the MIND at Home randomised trial (303 people, 18 months), care coordinators with "individualized care planning" cut the adjusted hazard of leaving home by 37% (HR 0.63, 95% CI 0.42–0.94), reduced unmet safety and legal needs, and improved self-reported quality of life.47 The money exists: people aged 50+ generated $12.5 trillion of US economic activity in 2024 — 43% of GDP.48 What fails is fit.

What I realized: personalization is the whole point

"Dementia" is not one condition and not one situation. The multimodal diagnostic model published in Nature Medicine in 2024 had to distinguish ten etiologies — Alzheimer's disease, Lewy body, vascular, prion, frontotemporal, normal-pressure hydrocephalus, systemic/external, psychiatric, traumatic brain injury and other — across 51,269 participants.84 Severity runs from CDR 0.5 to 3; behavioural symptoms are scored across twelve domains; the living situation ranges from a working daughter next door to an 80-year-old spouse caring alone; and the fourteen risk factors weigh differently in every life.6 A GPS tracker is relief in one household and an indignity in another.26 The MIND at Home result — 37% lower hazard of leaving home — came specifically from individualized care planning, not from a device.47

So AI has two jobs here, and today's products mostly do only the first:

So first, we need a marketplace where each person can choose

Choice without a matching layer is just noise. みんなの介護 lists 58,908 facilities,71 LIFULL介護 more than 57,000,72 Autumna in the UK over 26,000 providers,56 and Japan's 2025 "dementia-services chaos map" sorts the market into fifteen categories — from BPSD prediction and watch-over devices to insurance, inheritance and money management — with no layer that matches any of it to a given person.76 A marketplace for dementia services therefore has to do five things, in a loop:

The economics decide the ethics

Marketplaces already exist for elder care, and the oldest model — free for families, paid by the provider that wins the referral — shows exactly what goes wrong when the supply side pays. A Place for Mom is free to families and is paid by the facility that is chosen; it lists more than 15,000 communities and home-care providers (NBC counts roughly 14,000 — "less than half of the facilities on the market") and is contacted by about 250,000 families a year.54 In 2024 a Washington Post analysis found that more than a third of the facilities it recommended as "Best of Senior Living" across 28 states had been cited for neglect or substandard care in the previous two years, and that nearly 40% of families paid roughly $1,000 a month more than they had budgeted; the US Senate Special Committee on Aging opened an inquiry, and California's SB 875 now requires referral agencies to disclose whether, and when, facilities pay them.54,82 The lesson is not that marketplaces are bad; it is that a dementia marketplace must be paid by someone whose interest is the person's fit — Autumna charges providers listing subscriptions and takes no referral fees;56 Lottie monetizes leads but is moving to software (from £150 a month) and billing (£1 billion → £3 billion of care invoicing targeted by mid-2027);55 and the most durable navigators are paid by an institution — an employer, an insurer, or Medicare itself.

There are several ways to create it

Model 4 in detail: the CMS GUIDE Model is a publicly funded "navigator marketplace"

The US Centers for Medicare & Medicaid Services launched the GUIDE (Guiding an Improved Dementia Experience) Model on 1 July 2024 for eight years, with 390 organizations announced — 96 established programs from July 2024 and 294 new ones from July 2025 (the CMS page currently lists 292 participants, suggesting some attrition).63 Each participant receives a monthly Dementia Care Management Payment per beneficiary, tiered by complexity and by whether the person has a caregiver — in 2026, from $157 a month (low complexity, with caregiver, first six months) to $409 (moderate-to-high complexity, no caregiver), with lower "established patient" rates after six months and a separate $180/$70 tier for residential care communities — plus up to $2,500 a year of respite care and a one-time $75,000 infrastructure payment for new safety-net programs; a performance-based adjustment of roughly −2% to +6% applies from 2025.63 What it buys is exactly the loop in §05: a care navigator, a 24/7 support line, caregiver education and a personalised care plan. Japan's equivalent rails are the care manager (ケアマネジャー) and the community comprehensive support centre (地域包括支援センター), backed since April 2021 by the LIFE database — 53,370 facilities claimed LIFE-linked payments in April 202575 — and increasingly by AI care-plan tools such as ミルモぷらん and SOIN, the latter trained on 900,000 anonymized municipal records.74

What investors have backed so far

Examples — starting with Counsel Health

All Counsel figures are company-reported, from the Series A announcement and coverage.49

Other examples, by model

Model numbers refer to the five models in §06. Figures are as reported by the companies or the cited press; "†" marks claims I could not verify beyond the company's own site. Academic analog: Stanford's new $25M NIA-funded centre (Sept 2026) is building AI to personalize dementia interventions for agitation, sleep and depression.81

What I have to think about next: integrating all the medical data

The reference architecture: MIT's HAIM

The medical-AI literature already shows how heterogeneous data can be fused. The Holistic AI in Medicine (HAIM) framework from MIT's Abdul Latif Jameel Clinic for Machine Learning in Health (Soenksen, Ma, Zeng et al., npj Digital Medicine, September 2022) takes four data modalities — tabular, time-series, text and images — from 11 data sources, turns each into an embedding (a pre-trained transformer for notes, a pre-trained convolutional network for X-rays), concatenates them into one "fusion embedding" per patient, and trains a simple downstream learner (XGBoost). On HAIM-MIMIC-MM — 34,537 samples from 7,279 hospitalizations and 6,485 patients, built from MIMIC-IV and MIMIC-CXR — the team trained and characterized 14,324 models across every combination of inputs for 12 predictive tasks (ten chest pathologies, length of stay, 48-hour mortality). Multimodal models beat the best single-source models by 6–33% AUROC (pathologies 6–22%, length of stay 8–20%, mortality 11–33%), and Shapley values quantified how much each modality contributed to each task.83 The code is open (Apache-2.0).83

Dementia-specific evidence that more data types help

For dementia itself, the pattern repeats across very different data types. Xue and colleagues (Boston University, Nature Medicine 2024) combined demographics, personal and family medical history, medication use, neuropsychological assessments, functional evaluations and multisequence MRI from 51,269 participants in nine cohorts to classify thirteen labels (normal cognition, MCI, dementia and ten etiologies): micro-averaged AUROC 0.94 for the three-way stage call and 0.96 for the etiologies; in 100 cases, neurologists assisted by the model outperformed neurologists alone by 26.25%, and the model held an AUROC of 0.91 on an external cohort despite about 69% missing data — a property any real-world integration layer will need.84

Consumer devices carry signal too. In the Lilly–Evidation–Apple feasibility study (2019), 113 people aged 60–75 (31 with MCI or mild Alzheimer's dementia) used an iPhone, Apple Watch, iPad and a Beddit sleep monitor for 12 weeks; passive sensor data, mood and energy surveys and active tasks (dragging shapes, tapping, reading, typing) differentiated the symptomatic group "in ways not previously detected through common clinical screening tools".85 In the Intuition Brain Health Study (Biogen, Apple, Brigham; Nature Medicine 2025), 23,004 US adults used study iPhones and Apple Watches; a logistic-regression classifier on 205 features — keyboard metrics such as taps per minute, device and app use, speech detection, motion, heart rate, sleep and steps — identified MCI with a mean AUROC of 0.85 (sensitivity 80.2%, specificity 78.7%, accuracy 79.1%, negative predictive value 92.3%).86 Speech alone predicted progression from MCI to Alzheimer's within six years with 78.5% accuracy and 81.1% sensitivity in 166 Framingham participants,87 and a 2025 meta-analysis of 54 studies pooled speech-based MCI detection at 80% accuracy, 80% sensitivity and 77% specificity.88 Blood now joins the list: on 16 May 2025 the FDA cleared the first blood test for Alzheimer's diagnosis — Fujirebio's Lumipulse G pTau217/β-Amyloid 1-42 plasma ratio — for adults 55+ with cognitive symptoms in specialist settings; 91.7% of positives were confirmed amyloid-positive by PET or CSF and 97.3% of negatives negative, with fewer than 20% of 499 samples indeterminate.90 Japan filed the same test in November 2025; lecanemab (レケンビ), approved in Japan on 25 September 2023, requires amyloid confirmation by PET or CSF, MMSE ≥ 22 and CDR 0.5–1 before use.91 Gait matters: in 26,802 adults across 17 countries, "motoric cognitive risk syndrome" (slow gait plus a cognitive complaint) had a prevalence of 9.7% and roughly doubled the risk of incident cognitive impairment (HR 2.0) and dementia (HR 1.9).95 And the home is a sensor: Oregon's ORCATECH platform has collected continuous in-home data since 2004 in more than 480 homes, up to eight years per home,96 and an electronic pillbox study found adherence of 86.8% in cognitively intact elders versus 63.9% in those with mild decline.98

One caution from the literature: research is still imaging-heavy. A systematic review of 255 AI studies in dementia neuroimaging found structural MRI in about 71%, 71% relying on the single ADNI dataset, and few validating in an independent cohort.89 The real-world data in the table below — behaviour, speech, the home, care records, the person's own ratings — is exactly what research under-uses and exactly what a marketplace generates every day.

The full list: every type of data a personalized dementia service needs

Grouped the way HAIM would ingest it. "Modality" is the HAIM-style stream (tabular, time-series, text, image, audio). "Phone/Watch" marks what an iPhone and Apple Watch can capture today: yes natively via HealthKit or built-in features, partial via apps, research frameworks or third-party devices, no. "Lives in" names the system that holds it in Japan (and the US/EU equivalent where relevant).

How it would fit together

The questions this leaves open

Dementia is the mismatch problem of the ageing century: four groups who want the same outcome and keep buying, building and rating different things. Personalization is the fix, a marketplace is the mechanism, and data integration is the work. That is what I have to think about next.

Sources

All URLs were opened on 2 October 2026 unless marked †, which indicates a secondary report of a primary document that could not be opened directly. Figures in this note are quoted as published; where two sources differ, both are given in the text.

1. World Health Organization. Dementia — Fact sheet (data for 2021; cost for 2019). who.int
2. WHO. World failing to address dementia challenge — Global status report on the public health response to dementia, press release, 2 Sep 2021. who.int
3. Alzheimer's Disease International. Dementia statistics. alzint.org
4. GBD 2019 Dementia Forecasting Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050. Lancet Public Health 2022;7:e105–e125. PMC8810394
5. ADI. World Alzheimer Report 2021: Journey through the diagnosis of dementia (abridged). PDF · press release, 21 Sep 2021. alzint.org
6. Livingston G, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet 2024;404:572–628. doi · summary: Alzheimer Europe
7. 厚生労働省（研究代表者 二宮利治・九州大学）「認知症及び軽度認知障害の有病率調査並びに将来推計に関する研究」令和5年度老人保健健康増進等事業, 2024. mhlw.go.jp (PDF)
8. 「認知症施策推進基本計画」閣議決定, 2024年12月3日. mhlw.go.jp (PDF)
9. 厚生労働省「認知症施策推進総合戦略（新オレンジプラン）」概要, 2015. mhlw.go.jp (PDF)
10. 慶應義塾大学医学部（佐渡充洋ほか）「わが国における認知症の経済的影響に関する研究」2015. keio.ac.jp · 2060年推計: KOMPAS
11. 警察庁「行方不明者の状況」令和5年・6年・7年. R5 (PDF) · R6 (PDF) · R7 (PDF) · 2012–2013 values: nippon.com, facial-lykaon.com†
12. 厚生労働省「介護保険事業状況報告（暫定）」令和8年6月分. mhlw.go.jp (PDF)
13. 内閣府「令和8年版高齢社会白書」第1章第1節. cao.go.jp
14. 総務省統計局「令和4年就業構造基本調査 結果の概要」. stat.go.jp (PDF)
15. 厚生労働省「2022（令和4）年 国民生活基礎調査の概況」IV 介護の状況. mhlw.go.jp (PDF)
16. 厚生労働省「第9期介護保険事業計画に基づく介護職員の必要数について」2024年7月12日. mhlw.go.jp
17. 内閣官房「共生社会の実現を推進するための認知症基本法」資料. cas.go.jp (PDF)
18. 全国キャラバン・メイト連絡協議会「認知症サポーターの人数」令和8年6月30日現在. caravanmate.com
19. 보건복지부「2023년 치매역학조사 결과」2025년 3월 12일. korea.kr · 서울경제 보도. sedaily.com
20. 중앙치매센터「대한민국 치매현황 2024」(디멘시아뉴스 보도). dementianews.co.kr†
21. Alzheimer's Association. 2025 Alzheimer's Disease Facts and Figures. alz.org · 2026 edition. alz.org
22. Römhild J, et al. Inter-rater agreement of the Quality of Life-Alzheimer's Disease (QoL-AD) self-rating and proxy rating scale. Health Qual Life Outcomes 2018. PubMed
23. Reamy AM, Kim K, Zarit SH, Whitlatch CJ. Understanding discrepancy in perceptions of values. Gerontologist 2011. PubMed · four-year follow-up, 2013. PubMed
24. Miranda-Castillo C, Woods B, Orrell M. The needs of people with dementia living at home from user, caregiver and professional perspectives. BMC Health Serv Res 2013;13:43. BMC · 2010 companion paper. PubMed
25. van der Roest HG, et al. What do community-dwelling people with dementia need? Int Psychogeriatr 2009. PubMed
26. Landau R, et al. Families' and professional caregivers' views of using advanced technology to track people with dementia. Qual Health Res 2010. PubMed · Who should make the decision on the use of GPS? Aging Ment Health 2011. PubMed
27. 内閣府「認知症に関する世論調査」令和7年8月調査. gov-online.go.jp
28. Phillips B, Zhao H. Predictors of assistive technology abandonment. Assist Technol 1993;5:36–45. PubMed
29. Ienca M, et al. Intelligent assistive technology for Alzheimer's disease and other dementias: a systematic review. J Alzheimers Dis 2017. PubMed · Ienca M, et al. Ethical design of intelligent assistive technologies for dementia. Sci Eng Ethics 2018. Springer
30. Koo BM, Vizer LM. Mobile technology for cognitive assessment of older adults: a scoping review. JMIR mHealth uHealth 2019;7(11):e15122. JMIR
31. Span M, et al. Involving people with dementia in the development of supportive IT applications. Ageing Res Rev 2013. PubMed · Meiland F, et al. Technologies to support community-dwelling persons with dementia: INTERDEM position paper. JMIR Rehabil Assist Technol 2017. PubMed · Neal D, et al. Umbrella review. Digital Health 2025. SAGE
32. Gibson G, et al. The everyday use of assistive technology by people with dementia and their family carers. BMC Geriatr 2015. PubMed
33. Niemeijer AR, et al. Ethical and practical concerns of surveillance technologies in residential care for people with dementia or intellectual disabilities. Int Psychogeriatr 2010. PubMed
34. 日本総合研究所「認知症の人の消費・市場規模に関する推計」2025年2月21日. jri.co.jp (PDF)
35. 経済産業省ロボット市場予測（2013年; 産業タイムズ報道）. sangyo-times.jp† · 矢野経済研究所「介護ロボット市場に関する調査」2022. dreamnews.jp · 矢野経済研究所「介護DX市場」2024. yano.co.jp
36. 厚生労働省 社会保障審議会介護給付費分科会 第255回資料「介護テクノロジーの導入状況等」2026. mhlw.go.jp (PDF)
37. 厚生労働省「介護テクノロジー等による生産性向上の取組に関する調査及び効果測定事業」報告, 2026. mhlw.go.jp (PDF)
38. 介護労働安定センター「令和6年度 介護労働実態調査」結果の概要. kaigo-center.or.jp (PDF)
39. 厚生労働省 社会保障審議会福祉部会福祉人材確保専門委員会 資料, 2025. mhlw.go.jp (PDF)
40. WHO. Health workforce overview (projected shortfall of about 11 million health workers by 2030). who.int
41. Sallim AB, et al. Prevalence of mental health disorders among caregivers of patients with Alzheimer disease. J Am Med Dir Assoc 2015. PubMed
42. ADI. World Alzheimer Report 2022: Life after diagnosis. alzint.org · Newcastle University summary. ncl.ac.uk
43. ADI. World Alzheimer Report 2019: Attitudes to dementia. alzint.org
44. 認知症介護研究・研修仙台センター 平成29年度調査「空白の期間」(designing for dementia 紹介記事). designing-for-dementia.jp† · 日本老年医学会雑誌 63(1). J-STAGE
45. AARP. 2021 Home and Community Preferences Survey, press release, 18 Nov 2021. aarp.org
46. Black BS, et al. Unmet needs of community-residing persons with dementia and their informal caregivers: findings from the MIND at Home study. J Am Geriatr Soc 2013;61:2087–2095. Wiley · summary. ScienceDaily
47. Samus QM, et al. A multidimensional home-based care coordination intervention for elders with memory disorders: the MIND at Home randomized trial. Am J Geriatr Psychiatry 2014. PubMed
48. AARP. The Longevity Economy Outlook 2026. aarp.org
49. Counsel Health. Series A announcement, 16 Oct 2025. BusinessWire · Fierce Healthcare · company blog · seed round, Oct 2024. HIT Consultant · chronic-condition expansion, 7 Apr 2026. BusinessWire
50. Care.com. About. care.com
51. Honor / Home Instead. TechCrunch, 5 Oct 2021. techcrunch.com · Home Health Care News, Jul 2024. homehealthcarenews.com
52. Papa. Fierce Healthcare. fiercehealthcare.com · Refresh Miami, Aug 2026. refreshmiami.com
53. Sharecare acquires CareLinx, Aug 2021. sharecare.com
54. A Place for Mom. BusinessWire, 3 Feb 2026. businesswire.com · NBC News on the Senate probe, Jun 2024. nbcnews.com
55. Lottie. Accel-led Series A, Oct 2023. Yahoo Finance · CareMaster acquisition, 2 Oct 2026. tech.eu
56. Autumna. About us. autumna.co.uk
57. Elder. elder.org
58. Homage. Series C announcement, Sep 2021. homage.sg
59. Wellthy. TechCrunch, 11 May 2023. techcrunch.com
60. Cariloop. Series C, Apr 2024. cariloop.com · Grayce transition, Nov 2025. cariloop.com
61. Carallel. Series A, Dec 2022. carallel.com
62. KareHero. karehero.com†
63. CMS Innovation Center. GUIDE Model. cms.gov · Payment Methodology Paper (PY2026 rates). PDF · Model overview fact sheet. PDF · CMS blog. cms.gov
64. Rippl Care. GeekWire, Oct 2024. geekwire.com · Harbor Health acquisition, 10 Feb 2026. harborhealth.com
65. Remo Health and Cleveland Clinic Medicare ACO, 8 Jul 2025. clevelandclinic.org
66. Isaac Health. Fierce Healthcare, Aug 2025. fiercehealthcare.com
67. Harmonic Health. MobiHealthNews, Oct 2023. mobihealthnews.com
68. Together Senior Health and Rainier Health Network. PR Newswire. prnewswire.com
69. Ceresti Health. Home Health Care News, Aug 2025. homehealthcarenews.com
70. MapHabit. NIA commercialization release. maphabit.com
71. 株式会社クーリエ「みんなの介護」掲載施設数, 2025年5月. PR TIMES
72. LIFULL介護「LIFULL介護について」. kaigo.homes.co.jp
73. 株式会社メドレー「介護のほんね」と「みんかい」の統合について, 2025年10月27日. medley.jp
74. 株式会社ウェルモ「ミルモぷらん」. welmo.co.jp · 株式会社シーディーアイ「SOIN」. PR TIMES
75. 厚生労働省「科学的介護情報システム（LIFE）について」. mhlw.go.jp (PDF) · 収集項目（日本作業療法士協会資料, 2021）. jaot.or.jp (PDF)
76. Theoria technologies「認知症関連サービスカオスマップ2025」2025年1月. PR TIMES
77. トリニティ・テクノロジー株式会社 シリーズB資金調達, 2024年1月. PR TIMES
78. 케어링 (Caring). WowTale, 8 May 2026. wowtale.net
79. 케어닥 (Caredoc). 한국경제, Nov 2023. hankyung.com · 인더뉴스. inthenews.co.kr
80. Sensi.AI. TechCrunch, 26 Jun 2024. techcrunch.com · Series C, Oct 2025. CTech
81. Stanford Medicine. New AI centre for dementia care, Sep 2026. stanford.edu
82. California Senate Judiciary Committee. Analysis of SB 875 (Glazer) — referral-agency disclosure requirements for residential care facilities for the elderly. senate.ca.gov (PDF)
83. Soenksen LR, Ma Y, Zeng C, et al. Integrated multimodal artificial intelligence framework for healthcare applications. npj Digit Med 2022;5:149 (CC BY 4.0). nature.com · code. GitHub · project page. MIT Jameel Clinic
84. Xue C, Kowshik SS, et al. AI-based differential diagnosis of dementia etiologies on multimodal data. Nat Med 2024;30:2977–2989. nature.com
85. Lilly, Evidation Health and Apple. Feasibility study of digital biomarkers of cognitive impairment (KDD 2019), press release 8 Aug 2019. lilly.com · Chen R, et al. Apple ML Research
86. Butler PM, et al. Smartphone- and wearable-based digital biomarkers of cognitive impairment: the Intuition Brain Health Study. Nat Med 2025;31:829–839. nature.com
87. Amini S, et al. Prediction of Alzheimer's disease progression within 6 years using speech. Alzheimer's Dement 2024. Wiley · NIA summary. nia.nih.gov
88. Speech-based detection of mild cognitive impairment: systematic review and meta-analysis. Age and Ageing 2025;54(10):afaf316. OUP
89. Borchert RJ, et al. Artificial intelligence for diagnostic and prognostic neuroimaging in dementia: a systematic review. Alzheimer's Dement 2023. Wiley
90. US FDA. FDA clears first blood test used in diagnosing Alzheimer's disease, 16 May 2025. fda.gov
91. エーザイ「レケンビ」国内承認, 2023年9月25日. eisai.co.jp · 厚生労働省「最適使用推進ガイドライン レカネマブ」. mhlw.go.jp (PDF)
92. Jack CR Jr, et al. Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association Workgroup. Alzheimer's Dement 2024. Wiley
93. Mayo Clinic. Alzheimer's genes: Are you at risk? mayoclinic.org · Escott-Price V, et al. Common polygenic variation enhances risk prediction for Alzheimer's disease. Brain 2015. OUP
94. Apple Developer Documentation — HealthKit: HKQuantityTypeIdentifier · HKClinicalTypeIdentifier · HKStateOfMind · HKUserAnnotatedMedication · Apple Support, Fall Detection. support.apple.com
95. Verghese J, et al. Motoric cognitive risk syndrome: multicountry prevalence and dementia risk. Neurology 2014. PubMed
96. Oregon Center for Aging and Technology (ORCATECH), OHSU. Our technology. ohsu.edu · Lyons BE, et al. Pervasive computing technologies to continuously assess Alzheimer's disease progression and intervention efficacy. Front Aging Neurosci 2015. Frontiers · NIA CART initiative. nia.nih.gov
97. 象印マホービン「みまもりほっとライン」. zojirushi.co.jp · 中部電力ミライズ「テラシテ」2022年7月. chuden.co.jp · 尼崎市「認知症高齢者等GPS利用支援サービス」. city.amagasaki.hyogo.jp
98. Hayes TL, et al. Medication adherence in healthy elders: small cognitive changes make a big difference. J Aging Health 2009. OHSU
99. 課題分析標準項目（23項目, 令和5年10月改正）解説. hitori-cm.com† · 電子カルテ情報共有サービス（2025年開始）解説. pt-ot-st.net† · マイナポータルPHR API 連携（DeNA, 2023）. dena.com · NDB・介護DB連結解析（2020）. cbnews.jp
100. 内閣府「次世代医療基盤法の施行状況」2025年9月. cao.go.jp (PDF) · 個人情報保護法 要配慮個人情報 解説. businesslawyers.jp · European Commission. European Health Data Space Regulation (EU) 2025/327. europa.eu · HHS. HIPAA de-identification guidance. hhs.gov
101. Assessment instruments: MMSE and MoCA (Shirley Ryan AbilityLab) MMSE · MoCA · HDS-R（日本老年医学会）PDF · CDR (Knight ADRC) wustl.edu · ADAS-Cog (FDA) PDF · Zarit Burden Interview link · GDS-15 PDF · Lawton IADL PDF · MNA-SF PDF · DBD13 carenote.jp · UCLA Loneliness Scale v3 PDF · 認知症高齢者の日常生活自立度（厚労省）PDF · 要介護度と基準時間（長寿科学振興財団）tyojyu.or.jp · Dementia QoL PROMs (DOMS) dementiaresearch.org.au
102. Alzheimer's Society. This is me. alzheimers.org.uk · 厚生労働省「人生会議」(ACP). mhlw.go.jp · Heat and behavioural symptoms in dementia inpatients. J Am Med Dir Assoc 2004. ScienceDirect
103. Luz S, et al. Alzheimer's Dementia Recognition through Spontaneous Speech: the ADReSS Challenge. Interspeech 2020. ISCA (PDF)