Skip to content

Care, caregiving and health systems

TL;DR — Half the global economic burden of dementia is unpaid care: of US$1,313.4 billion in 2019, informal care accounted for US$651.4 billion (50%), direct social-sector costs US$448.7 billion (34%) and direct medical costs only US$213.2 billion (16%) (Wimo 2023, PMID 36617519). The people providing that care are measurably harmed — dementia caregivers differ from non-caregivers most on depression (g=0.58) and stress (g=0.55), with larger differences for dementia than for mixed caregiving samples (Pinquart 2003, PMID 12825775) — and pooled anxiety prevalence among informal dementia carers is 32.1% (95% CI 20.6–46.2) (Kaddour 2020, PMID 31409196). Two decades of care-model trials give a consistent, uncomfortable answer: structured multicomponent caregiver interventions work (REACH II reduced clinical depression from 22.7% to 12.6%, P=0.001) and collaborative primary care reduces behavioural symptoms (NPI difference −5.6 at 12 months, P=0.01), but the largest and most recent pragmatic trial found no difference between health-system dementia care, community-based dementia care and usual care on either behavioural symptoms or caregiver strain among 2,176 dyads (Belle 2006, PMID 17116917; Callahan 2006, PMID 16684985; Reuben 2025, PMID 39878968). Meanwhile the biomarker-and-antibody era demands infrastructure most systems do not have — after biomarker confirmation only about a quarter of amyloid-positive memory-clinic patients met appropriate-use criteria in one national cohort (Jeon 2025, PMID 41225766) — while diagnosis itself remains delayed by 31–44 months on average and differentially by ethnicity (Lin 2021, PMID 34091580).

The scale and shape of the burden

Measure Estimate Source
Global societal cost of dementia US$1,313.4 billion (US$23,796 per person) 2019 (Wimo 2023, PMID 36617519)
— informal care share 50% (US$651.4 billion) "
— direct social sector incl. long-term care 34% (US$448.7 billion) "
— direct medical 16% (US$213.2 billion) "
People with dementia in LMICs vs share of costs in HICs 61% vs 74% "
Direct health spending attributable to ADRD, 204 countries US$260.6 billion (95% UI 131.6–420.4) 2019 (Lastuka 2024, PMID 39150827)
Informal care cost (same model), 57% of total US$354.1 billion (95% UI 190.0–544.1) "
Projected direct spending, 9.4% of world health spending US$1.6 trillion (95% UI 0.6–3.3) 2050 (PMID 39150827)
US unpaid caregivers / hours provided >11 million people / 18.4 billion hours 2023 (2024 Facts and Figures, PMID 38689398)
Valuation of that unpaid care US$346.6 billion 2023 (PMID 38689398)
US payments for health, long-term and hospice care, 65+ with dementia US$360 billion 2024 (PMID 38689398)

The structural fact is that the majority of the cost sits outside any health budget, which means health-system cost-effectiveness analyses systematically under-count both the burden and any benefit that shifts care back into the formal sector. The same asymmetry appears geographically: 61% of people with dementia live in low- and middle-income countries while 74% of the measured costs are incurred in high-income countries — an artefact of what is counted as much as of what is spent (PMID 36617519).

Caregiver health

Comparison Effect Source
Caregivers vs non-caregivers, depression g = 0.58 Pinquart 2003, meta-analysis of 84 articles, PMID 12825775
" perceived stress g = 0.55 "
" self-efficacy g = 0.54 "
" general subjective well-being g = −0.40 "
" physical health g = 0.18 (significant but small) "
Dementia caregivers vs heterogeneous caregiver samples Larger differences for dementia caregiving "
Pooled anxiety prevalence, informal dementia carers 32.1% (95% CI 20.6–46.2), 10 studies, substantial heterogeneity Kaddour 2020, PMID 31409196

Caregiver burden is also a determinant of the care recipient's trajectory: caregiver burden was associated with nursing-home admission in a meta-analysis of 26 controlled studies, alongside poorer cognition, behavioural and psychological symptoms and ADL impairment (Toot 2017, PMID 27806743). Interventions aimed at caregivers are therefore not adjunctive — they act on the same pathway as the drugs on symptomatic and supportive therapy.

Care-model trials

Trial Design Result
REACH II (Belle 2006, PMID 17116917) 642 caregivers (212 Hispanic/Latino, 219 White, 211 Black/African American) in 5 US cities; 12 in-home and telephone sessions over 6 months addressing depression, burden, self-care, social support and care-recipient behaviours vs 2 "check-in" calls Significantly greater improvement in the composite quality-of-life outcome for Hispanic/Latino (P<0.001) and White (P=0.037) caregivers, and for Black/African American spouse caregivers (P=0.003). Clinical depression 12.6% vs 22.7% (P=0.001). No significant difference in institutionalisation at 6 months
Collaborative care (Callahan 2006, PMID 16684985) 153 older adults with AD and their caregivers randomised by physician to 1 year of interdisciplinary care management led by an advanced practice nurse, integrated in primary care, vs augmented usual care 89% of intervention patients triggered ≥1 behavioural protocol (mean 4 of 8). More received cholinesterase inhibitors (79.8% vs 55.1%, P=0.002) and antidepressants (45.2% vs 27.5%, P=0.03). Total NPI lower by 5.6 points at 12 months (P=0.01) and 5.4 at 18 months (P=0.01); caregiver NPI distress improved at 12 months and caregiver PHQ-9 at 18 months. No differences in Cornell depression, cognition, ADLs, hospitalisation, nursing-home placement or death; achieved without increasing antipsychotic or sedative-hypnotic use
WHELD (Ballard 2018, PMID 29408901) 847 nursing-home residents, 69 UK homes, person-centred care and social-interaction training plus antipsychotic review, 9 months QoL +2.54 (95% CI 0.81–4.28, P=0.0042); agitation CMAI −4.27 (P=0.0076); NPI-NH −4.55 (P<0.001); positive interactions +19.7% (P=0.03); cost saving; did not reduce antipsychotic use
D-CARE (Reuben 2025, PMID 39878968) 2,176 community-dwelling dyads at 4 US sites, randomised 7:7:1 to health-system-based care by an advanced practice dementia care specialist (n=1,016), community-based care by a social worker/nurse/therapist consultant (n=1,016), or usual care (n=144); 18 months No significant differences on either primary outcome. NPI-Q least-squares means 9.8 / 9.5 / 10.1 (MCID 2.8–3.2); health-system vs community 0.30 (97.5% CI −0.18 to 0.78); vs usual care −0.33 (−1.32 to 0.67). Modified Caregiver Strain Index 10.7 / 10.5 / 10.6 (MCID 1.5–2.3), all differences <0.3. Caregiver self-efficacy was higher for both active arms vs usual care (15.1 / 15.2 / 14.4; health-system vs usual 0.70, 95% CI 0.26–1.14; community vs usual 0.85, 0.42–1.29)

D-CARE is the most consequential negative result in this literature and deserves careful reading. It is large, pragmatic, ran to 18 months with >99% primary-outcome ascertainment, and included 20.6% Black or Hispanic participants. It found that two well-resourced dementia care models did not beat usual care on behavioural symptoms or caregiver strain — while both improved caregiver self-efficacy. Possible readings: usual care has improved since the 2006 trials; NPI-Q and caregiver strain are insensitive to what care programmes actually change; or the effective ingredient in REACH II (structured skills training targeted to an individual risk profile) is not what care-navigation programmes deliver. The trial does not license the conclusion that dementia care programmes are useless; it does mean their justification cannot rest on these two outcomes.

Diagnosis pathways

Recognition is late and unequal. Among 3,966 HRS participants aged 70+ with probable dementia linked to Medicare/Medicaid claims, missed or delayed clinical diagnosis occurred in 41% of non-Hispanic White, 46% of non-Hispanic Black and 54% of Hispanic participants, with mean estimated delays of 31.2, 34.6 and 43.8 months respectively, and more advanced dementia at the point of diagnosis in the two latter groups (Lin 2021, PMID 34091580). In community surveys in Indonesia and South Africa, fewer than 1% of people meeting dementia criteria had ever been diagnosed (0.2% and 0.5%) (Farina 2023, PMID 37278200).

This is the denominator problem for every downstream service: a treatment restricted to early-stage disease cannot reach a population diagnosed on average three years late. Plasma biomarker testing could compress that interval (fluid biomarkers), which converts a diagnostic bottleneck into a capacity problem.

The capacity problem created by disease-modifying therapy

The shift from syndrome-based dementia care to early, biomarker-guided treatment requires substantial adjustment of infrastructure and resources — identification of eligible patients, biomarker confirmation, treatment-delivery capacity, and surveillance MRI — and estimates of how many patients will qualify and whether health systems are ready are an explicit open issue (Heneka 2024, PMID 39549715). FDA approval of at-home subcutaneous lecanemab starting doses in July 2026 can reduce infusion-centre demand, but not the diagnostic, MRI or emergency-response components of that pathway (FDA, https://www.fda.gov/drugs/news-events-human-drugs/fda-approves-first-home-starting-dose-alzheimers-disease-treatment, directly fetched 2026-08-31).

The eligibility funnel is steep. Applying appropriate-use recommendations to 1,005 amyloid-positive attendees at a Korean memory clinic (from 2,726 total clinic visitors), 24.6% were eligible for lecanemab and 28.0% for donanemab — 9.1% and 10.3% of the original clinic population; more than a third of exclusions were MRI findings, mainly vascular lesions carrying ARIA risk; using demographically adjusted MMSE z-scores rather than raw thresholds raised eligibility to 38.4% and 33.7% (Jeon 2025, PMID 41225766).

Affordability is a separate constraint. Modelled value-based prices for lecanemab and donanemab range from US$254–9,434 and US$387–13,964 in high-income countries down to US$4–18 and US$9–32 in low-income countries, against QALY gains of 0.38 and 0.51 (Hoang 2026, PMID 42125995). See anti-amyloid immunotherapy.

Acute care and the end of life

People with dementia use acute services heavily and fare worse in them. Across 5,580,106 emergency admissions in England, people with dementia accounted for 13.9% of male and 18.6% of female admissions aged 65+; adjusted length of stay was 17% (95% CI 15–18) longer for men and 12% (10–14) for women, adjusted excess emergency re-admission risk was 17%, and the risk of dying within 30 days of discharge was roughly doubled while in-hospital mortality differed only slightly — a pattern the authors flag as requiring investigation (Reeves 2023, PMID 36888666).

The terminal phase is predictable and under-recognised. Following 323 nursing-home residents with advanced dementia for 18 months, 54.8% died; probabilities of pneumonia, febrile episode and eating problem were 41.1%, 52.6% and 85.8%, with adjusted 6-month mortality after those events of 46.7%, 44.5% and 38.6%; dyspnoea (46.0%) and pain (39.1%) were common; 40.7% underwent a burdensome intervention in the last 3 months of life. Where the proxy understood the prognosis and expected complications, the adjusted odds of a burdensome intervention were 0.12 (95% CI 0.04–0.37) (Mitchell 2009, PMID 19828530).

What palliative interventions achieve is more modest than that association suggests. A Cochrane review of nine studies (2,122 participants) found that changes to the organisation and delivery of care may increase comfort in dying (MD 1.49, 95% CI 0.34–2.64; 5 studies, 335 participants) but at very low certainty, probably have little effect on use of non-palliative interventions (RR 1.11, 0.71–1.72; moderate certainty), and may have little or no effect on advance-directive documentation. Advance care planning interventions specifically probably increase documentation of advance directives (RR 1.23, 95% CI 1.07–1.41) and discussions about goals of care (RR 1.33, 1.11–1.59), both moderate certainty, and may slightly increase concordance with goals of care (RR 1.39, 1.08–1.79, low certainty), with no detected effect on symptom management. No included study assessed adverse effects (Walsh 2021, PMID 34582034).

Care models are not interchangeable

Three trials clarify which components move which outcomes. The 780-dyad Care Ecosystem—telephone/internet navigation backed by an advanced-practice nurse, social worker and pharmacist—improved proxy-rated patient quality of life by 0.53 points (95% CI 0.25–1.30), reduced emergency-department visits by 0.14 (0.01–0.29), caregiver depression by 1.14 points (0.13–2.15) and burden by 1.90 points (0.08–3.89) at 12 months (Possin 2019, PMID 31566651). German nurse-led dementia care management reduced behavioural symptoms by 7.45 points (95% CI 3.81–11.08) and had a small, borderline burden effect (0.50 points, 95% CI −0.08 to 1.09); quality of life improved only among people not living alone (Thyrian 2017, PMID 28746708). In 250 distressed Hong Kong caregivers, component optimisation identified mindfulness—not every plausible component—as the active signal: depressive symptoms fell 2.13 points (95% CI 1.38–2.85) at 12 months, while behavioural-problem management transiently increased anxiety at six months (Kwok 2025, PMID 40094667).

Hospital discharge is a separate, and smaller, medication-safety problem than it first appears. Among 117,022 Medicare beneficiaries with dementia, 63% were already taking at least one psychotropic before admission, against which the authors call new prescribing at discharge relatively uncommon: 6.6% started at least one new class within seven days. What makes it consequential is persistence — 51% of new users were still filling the drug beyond 90 days — and that delirium predicted initiation (adjusted OR 1.8, 95% CI 1.7–1.9), turning an acute syndrome into a potential long-term exposure (Growdon 2023, PMID 36514208). A care model should therefore be judged not only by navigation or caregiver scales, but by whether it closes this transition-of-care loop.

Where the cost sits is a care-model variable, not a fixed property of the disease. Formal caregiving accommodation shifts spending out of unpaid care and into direct non-medical expenditure, raising direct costs to as much as 67.3% of the total; the AD-specific literature supporting this is thin, with only 12 of 5,536 screened reports meeting inclusion criteria (Tay 2024, PMID 37972428), and an earlier 27-study dementia review across 14 health systems likewise identified informal home care and nursing-home expenditure — not direct medical care — as the dominant drivers (Schaller 2015, PMID 25320002). The severity-graded cost ranges from both reviews, and why they must not be pooled, are tabulated on epidemiology and burden.

What the evidence supports

  1. Treat caregiver mental health as a target in its own right — the depression effect size is large and REACH II moved it (PMID 12825775; PMID 17116917).
  2. Expect care programmes to improve caregiver self-efficacy and process quality rather than behavioural symptoms or strain scores (PMID 39878968; PMID 16684985).
  3. In institutional settings, person-centred care and social interaction improve quality of life and agitation at lower cost (PMID 29408901).
  4. Advance care planning reliably changes documentation and conversations; whether it changes experience is much less certain (PMID 34582034).
  5. Plan acute-care pathways around the doubled 30-day post-discharge mortality rather than around in-hospital metrics (PMID 36888666).
  6. Any disease-modifying-therapy programme must be designed around the diagnosis delay and the eligibility funnel, not around the trial population (PMID 34091580; PMID 41225766).

Open questions

  • Why did D-CARE find no effect on behavioural symptoms or caregiver strain when earlier collaborative-care trials did — improved usual care, insensitive outcomes, or a different active ingredient (Reuben 2025, PMID 39878968; Callahan 2006, PMID 16684985)?
  • Is caregiver self-efficacy — the one outcome D-CARE moved — a valid target, and does it predict downstream outcomes such as placement or crisis use (PMID 39878968)?
  • Would REACH II's structured, risk-profiled caregiver skills training replicate at scale and in health systems outside the US (Belle 2006, PMID 17116917)?
  • What explains the doubled 30-day post-discharge mortality for people with dementia when in-hospital mortality is only slightly elevated (Reeves 2023, PMID 36888666)?
  • Can plasma-biomarker-enabled earlier diagnosis close the 31–44-month delay without generating demand that treatment-delivery and MRI capacity cannot meet (Lin 2021, PMID 34091580; Heneka 2024, PMID 39549715)?
  • What is the eligibility funnel for anti-amyloid therapy in health systems outside specialist memory clinics, where MRI and biomarker access differ (Jeon 2025, PMID 41225766)?
  • How should the 50% of dementia cost that is unpaid care enter cost-effectiveness decisions made by health payers who do not bear it (Wimo 2023, PMID 36617519; Lastuka 2024, PMID 39150827)?
  • Does any palliative-care intervention improve symptom experience rather than documentation in advanced dementia (Walsh 2021, PMID 34582034)?

References

  1. Wimo A, et al. The worldwide costs of dementia in 2019. Alzheimers Dement. 2023;19:2865-2873. PMID 36617519.
  2. Lastuka A, et al. Societal costs of dementia: 204 countries, 2000-2019. J Alzheimers Dis. 2024;101:277-292. PMID 39150827.
  3. Alzheimer's Association. 2024 Alzheimer's disease facts and figures. Alzheimers Dement. 2024;20:3708-3821. PMID 38689398.
  4. Pinquart M, et al. Differences between caregivers and noncaregivers in psychological health and physical health: a meta-analysis. Psychol Aging. 2003;18:250-67. PMID 12825775.
  5. Kaddour L, et al. Anxiety in informal dementia carers: a meta-analysis of prevalence. J Geriatr Psychiatry Neurol. 2020;33:161-172. PMID 31409196.
  6. Toot S, et al. Causes of nursing home placement for older people with dementia: a systematic review and meta-analysis. Int Psychogeriatr. 2017;29:195-208. PMID 27806743.
  7. Belle SH, et al. Enhancing the quality of life of dementia caregivers from different ethnic or racial groups: a randomized, controlled trial. Ann Intern Med. 2006;145:727-38. PMID 17116917.
  8. Callahan CM, et al. Effectiveness of collaborative care for older adults with Alzheimer disease in primary care: a randomized controlled trial. JAMA. 2006;295:2148-57. PMID 16684985.
  9. Ballard C, et al. Impact of person-centred care training and person-centred activities on quality of life, agitation, and antipsychotic use in people with dementia living in nursing homes: a cluster-randomised controlled trial. PLoS Med. 2018;15:e1002500. PMID 29408901.
  10. Reuben DB, et al. Health system, community-based, or usual dementia care for persons with dementia and caregivers: the D-CARE randomized clinical trial. JAMA. 2025;333:950-961. PMID 39878968.
  11. Lin PJ, et al. Dementia diagnosis disparities by race and ethnicity. Med Care. 2021;59:679-686. PMID 34091580.
  12. Farina N, et al. Comprehensive measurement of the prevalence of dementia in low- and middle-income countries: STRiDE methodology and its application in Indonesia and South Africa. BJPsych Open. 2023;9:e102. PMID 37278200.
  13. Heneka MT, et al. Passive anti-amyloid β immunotherapy in Alzheimer's disease — opportunities and challenges. Lancet. 2024;404:2198-2208. PMID 39549715.
  14. Jeon SY, et al. Eligibility for lecanemab and donanemab in Korea under appropriate use recommendations. Alzheimers Dement. 2025;21:e70875. PMID 41225766.
  15. Hoang MT, et al. Value-based prices of emerging disease-modifying therapies for Alzheimer's disease in 174 countries: a cost-effectiveness and threshold analysis. Alzheimers Dement. 2026;22:e71308. PMID 42125995.
  16. Reeves D, et al. Retrospective study of more than 5 million emergency admissions to hospitals in England: epidemiology and outcomes for people with dementia. PLoS One. 2023;18:e0281158. PMID 36888666.
  17. Mitchell SL, et al. The clinical course of advanced dementia. N Engl J Med. 2009;361:1529-38. PMID 19828530.
  18. Walsh SC, et al. Palliative care interventions in advanced dementia. Cochrane Database Syst Rev. 2021;9:CD011513. PMID 34582034.
  19. Possin KL, et al. Effect of collaborative dementia care via telephone and internet on quality of life, caregiver well-being, and health care use: the Care Ecosystem randomized clinical trial. JAMA Intern Med. 2019;179:1658-1667. PMID 31566651.
  20. Thyrian JR, et al. Effectiveness and safety of dementia care management in primary care: a randomized clinical trial. JAMA Psychiatry. 2017;74:996-1004. PMID 28746708.
  21. Kwok JYY, et al. Multicomponent intervention for distressed informal caregivers of people with dementia: a randomized clinical trial. JAMA Netw Open. 2025;8:e250069. PMID 40094667.
  22. Growdon ME, et al. New psychotropic medication use among Medicare beneficiaries with dementia after hospital discharge. J Am Geriatr Soc. 2023;71:1134-1144. PMID 36514208.
  23. Tay LX, et al. Economic burden of Alzheimer's disease: a systematic review. Value Health Reg Issues. 2024;40:1-12. PMID 37972428.
  24. Schaller S, et al. The main cost drivers in dementia: a systematic review. Int J Geriatr Psychiatry. 2015;30:111-29. PMID 25320002.