The cost of dementia in the United States in 2026

Abstract

INTRODUCTION

Comprehensive cost measurement is essential for an effective policy response to societal dementia costs.

METHODS

Using dynamic microsimulation, the Health and Retirement Study, and other national data, we quantified the 2026 cost of dementia in the United States.

RESULTS

In 2026, 5.7 million (95% confidence interval [CI] [5.6, 6.0]) US adults aged 51 and older are living with dementia, supported by 5.2 million (95% CI [4.9, 5.5]) care partners. Total costs are $818 billion (B, 95% CI [759, 866]), driven by quality‐of‐life losses for persons with dementia ($320B, 95% CI [269, 363]) and care partners ($15B 95% CI [6, 25]). Unpaid care ($237B, 95% CI [220, 253]), earnings losses ($23B), and out‐of‐pocket costs combined with quality‐of‐life losses account for 80% of costs and are borne by families. Governments cover 70% of healthcare costs ($222B, 95% CI [209, 237]).

DISCUSSION

The costs of dementia fall on families, highlighting limited policy and work supports. Treatment innovation may increase medical costs but reduce caregiver burden and improve quality of life.

Highlights

• The costs of dementia in the United States in 2026 are $818 billion.

• Quality‐of‐life losses are the largest driver of dementia's total burden.

• Individuals and families bear over three times the cost versus health systems.

• Methods enable analysis of treatment, care, and policy innovations on future costs.

1. BACKGROUND

The 2011 National Alzheimer's Project Act (NAPA) and subsequent National Plan to address Alzheimer's disease and related dementias sparked a coordinated effort across federal agencies to address the growing Alzheimer's disease crisis in the United States. These efforts have generated increased funding and research on prevention and treatment and on improving diagnosis and care for persons living with dementia. As a result, today we have a deeper understanding of disease processes and prevention and risk reduction strategies. ¹ , ² There are new US Food and Drug Administration‐approved disease‐targeting therapies for Alzheimer's disease and blood‐based biomarkers to aid diagnosis. ³ , ⁴ , ⁵ , ⁶ New models of caring for persons with dementia, such as Guiding an Improved Dementia Experience (GUIDE), are reimbursed by the Centers for Medicare and Medicaid Services (CMS), ⁷ and the importance of family caregivers is better recognized and supported through federal‐ and state‐funded programs such as the National Family Caregiver Support Program and Medicaid's home and community based services. Population aging as well as future prevention, detection, treatment, and care advances will impact the size of the population with dementia – dementia that is due to a variety of underlying pathologies (all‐cause). They will also change the costs of dementia and how the costs are distributed across the population and between families and the public sector.

Knowledge about the economic impact of dementia and how costs are distributed across the population is incomplete. Prior research on the economic impact of dementia considered a limited set of costs, including direct costs such as medical and long‐term care costs (e.g., nursing home, home health) and some indirect costs such as the valuation of family's and friends’ unpaid time caring for persons living with dementia. ⁸ , ⁹ Early estimates of costs based on nationally representative data were between $157 billion and $307 billion in 2010, with the difference driven largely by how the value of hours of unpaid care by family and others is quantified. ⁸ , ¹⁰ The impact of dementia, however, goes beyond medical costs and the value of caregivers’ time and includes quality‐of‐life impacts and current and future earnings losses, for both persons living with dementia and their care partners and caregivers. ¹¹ , ¹² Dementia is not equally distributed across demographic groups. Higher rates of dementia have been recorded among persons with low levels of education and Hispanic and Black persons. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ Comprehensive estimates of the societal costs of dementia and how these are distributed across the US population have not been quantified and are necessary for targeting resources to maximize impact.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed US literature estimating the economic impact of dementia using traditional sources (e.g., PubMed, Google Scholar). Few studies assessed population‐level medical, long‐term, and informal care costs, and none quantified societal costs including earnings losses and quality‐of‐life declines.

2. Interpretation: Comprehensive cost estimates show that the largest share of dementia's burden arises from quality‐of‐life losses among persons living with dementia. Combined with out‐of‐pocket expenses, unpaid care by family and friends, and related earnings and quality‐of‐life losses, the economic toll is more than three times higher for individuals and families than for health systems and governments.

3. Future directions: Annual cost estimates will clarify how cost levels and distribution across types of costs are changing over time. The methods developed support analyses of how prevention, treatment, care, and policy advances may affect future costs.

Quantifying the total cost of dementia to the US population now and in the future is needed to understand who bears what costs and how changes in population health, prevention, treatments and care, and policies are affecting the level and distribution of costs over time. This study used nationally representative, longitudinal data from the Health and Retirement Study (HRS), CMS and other national sources, and dynamic microsimulation methods to quantify the costs of dementia in 2026 for the US population. Rich, national‐level data supported estimates that represent the US population, and dynamic microsimulation methods captured how individual and family circumstances, health, and economic status interacted to impact different types of costs and vary across different populations. We quantified different types of costs for a more complete assessment of the impact of dementia including those due to medical and long‐term care, caregiving by family and friends, and impacts on quality of life and earnings experienced by both persons living with dementia and their care partners. Along with cost estimates from a societal perspective, we quantified costs to the healthcare system, with separate estimates for spending by Medicare, Medicaid, other payers, and out‐of‐pocket spending by persons with dementia and their families. This study advanced our understanding of who bears what types of costs and developed the methodological tools to track the evolution of impacts over time and across heterogeneous persons to inform policies and decision‐making about resource allocation and priorities to reduce the burden of dementia for the US population.

2. METHODS

2.1. United States Cost of Dementia Model overview

We developed the US Cost of Dementia Model (USCDM) to quantify annual, comprehensive costs of dementia in the United States and track changes over time. The costs of dementia estimated fall broadly into three groups: direct (medical and long‐term care), indirect (value of unpaid care hours and earnings loss), and intangible (quality of life).

USCDM is a dynamic microsimulation model that simulates the life course of individuals. It captures how an individual's risk factors, morbidity, and mortality progress. Key to the USCDM are dynamic models of cognitive impairment (cognitively unimpaired, mild cognitive impairment [MCI], dementia), changes in functional limitations, onset of other chronic health conditions (e.g., diabetes, vascular diseases), changes in work status and earnings, and unpaid care received by and provided to family and friends. USCDM uses nationally representative, individual‐level, longitudinal data from the HRS from 1998 to 2020 supplemented with other national surveys. USCDM builds upon prior models, the Future Elderly Model (FEM) ¹⁹ and the Alzheimer's disease FEM, both of which have been extensively validated internally and externally. ¹⁹ , ²⁰ , ²¹ Longitudinal data from survey respondents capture individual heterogeneity in health, health care, and long‐term care use and spending, employment, and other economic factors. The modeling of the dynamics of health over time allows for competing risks of the development of other diseases and death in the assessment of costs of dementia among heterogeneous persons. ²² , ²³ , ²⁴ The model development was informed by a series of discussions with a panel of persons living with MCI and care partners and caregivers of persons living with dementia and a separate set of discussions with a panel of public health leaders and healthcare providers. Qualitative assessment of the meetings’ transcripts provided insight into costs associated with living with dementia and caring for persons living with dementia. Key measures and methods are described in what follows. A separate methods study provides comprehensive detail on the model, data, measures, methods, and validation analyses. ²⁵

The USCDM study – including panels of persons with lived experience and public health and care experts – was reviewed and approved by the University of Southern California Institutional Review Board under expedited review.

2.2. Data and study population

The primary data source for the USCDM is the HRS. The HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and conducted by the University of Michigan. The HRS is a biennial, nationally representative survey of Americans 51 years and older that provides rich, longitudinal data on health conditions and functional limitations, healthcare use, and spending on health care and long‐term care, caregiving provided, and care received from family and others, as well as data on employment, income, wealth, and insurance coverage. ²⁶ We used biennial data from 1998 through 2020. We used measures from the RAND HRS data files when available. ²⁶ , ²⁷ The RAND HRS Longitudinal File is an easy‐to‐use dataset based on the HRS core data. This file was developed at RAND with funding from the National Institute on Aging and the Social Security Administration. We supplemented the HRS with annual data on healthcare use and spending from the Medical Expenditure Panel Survey (MEPS) and the Medicare Current Beneficiary Survey (MCBS). The combination of datasets provided more comprehensive information than any single national survey. Data from the National Health Interview Survey (NHIS) were used to inform population prevalence of chronic conditions of persons ages 51 and 52 in future years, so that cost of dementia estimates are representative of the US population over age 50 in each year costs are assessed.

2.3. Dementia

The cognitive functioning of HRS respondents ages 51 and older is assessed at each wave using an adapted version of the Telephone Interview for Cognitive Status (TICS). ¹⁴ , ¹⁶ , ²⁸ , ²⁹ , ³⁰ The TICS assessment is completed by the respondent whenever possible. If the respondent is unavailable or unable to participate (e.g., due to cognitive or physical limitations), a proxy – typically a spouse or close family member – provides an informant‐based memory evaluation, responding to a standardized set of questions about the respondent's current level of and changes in memory and cognition. Dementia in the USCDM is assigned based on these scores following the Langa–Weir Classification of Cognitive Function, as a categorical variable with three cognitive states: normal cognition; cognitive impairment, not dementia; and dementia. ³¹ We modeled the cognitive status of individuals over time using ordered probit models adjusting for demographic characteristics, health behaviors, and health conditions. Model estimates are provided in Table SA1.

2.4. Medical and long‐term care costs

Medical and long‐term care costs are from data from the Medicare Current Beneficiary Survey (MCBS) for persons enrolled in Medicare and the Medical Expenditures Survey (MEPS) for those who are not. We estimated total medical and long‐term care spending of persons with dementia. Thus, healthcare costs are interpreted as those of persons with dementia and not as costs primarily due to dementia. We also estimated separate models of spending from different payers: Medicare, Medicaid, other payers, and out‐of‐pocket spending. The cost models are estimated using individuals’ demographics (age, sex, race), education, marital status, disease conditions, cognitive status and functional status, nursing home residence, and whether respondents are in their last year of life. Estimates from medical and long‐term care models are presented in Table SA2. Model covariates are common across MCBS, MEPS, and HRS. Model parameter estimates are used to predict health and long‐term care spending of respondents in the HRS with dementia. Aggregated population healthcare expenditures are aligned by matching population totals with data from the National Health Expenditure Accounts.

2.5. Cost of care received from family and others

HRS respondents indicate how many hours and days of personal care they receive from family and friends for activities of daily living (ADLs) – getting across a room, dressing, bathing, eating, getting in/out of bed, and using the toilet. Respondents also indicate the number of hours they received help in a typical month from family or friends for instrumental ADLs (IADLs) – preparing meals, shopping for groceries, making telephone calls, taking medications, and managing money. These measures are converted to a yearly number of hours of care received among HRS respondents with dementia. We estimated two probit models of informal care hours received: one for any hours received and another for full‐time care if receiving any care. Ordinary least‐squares (OLS) regression is used to estimate the number of hours if receiving care but not full time (Table SA3). Among those receiving full‐time care, the annual number of hours is 8541 summed across all of an individual's caregivers.

We calculated the dollar value of unpaid care hours based on the number of hours of care received by HRS respondents with dementia and assigned a dollar value based on a replacement rate. The replacement rate method used in the USCDM was an hourly value of $34.92 for each hour of care received, which is the assumed hourly rate based on national averages if a paid caregiver were hired for care (July 2025). ³² An alternate method for valuing hours of care based on an opportunity cost of time approach is detailed in the companion methods paper.

2.6. Quality‐of‐life loss

Quality‐adjusted life years (QALYs) are measured using Health Utility Index 3 (HUI3) for HRS respondents in 2000. The HUI3 measures aspects of health, including vision, hearing, speech, ambulation, dexterity, emotion, pain, and cognition (Table SA4). It is mapped to a single score ranging from −0.36 to 1, where 0 represents death, negative scores are worse than death, and 1 is perfect health. ³³ To determine the loss in quality of life associated with living with dementia, we simulated a counterfactual scenario in which people with dementia in 2010 instead had MCI and did not progress to dementia between 2010 and 2026. The difference in the total number of QALYs in this counterfactual scenario with the status quo scenario in 2026 is multiplied by $150,000, the value of the loss of one QALY.

To assess the impact on the quality of life of care partners to persons with dementia, we estimated a model of depressive symptoms using the eight‐item Center for Epidemiological Studies Depression scale (CES‐D). We estimated the effects of being a care partner or caregiver and hours of care provided on CES‐D and subsequently CES‐D on health utility (HUI3). See Tables SA5 and SA6 for model estimates. We monetized utility scores to compute QALYs of a dementia caregiver, separately for caregivers who are spouses of persons with dementia and not spouses (e.g., adult child of a parent with dementia). We used USCDM to simulate a counterfactual scenario such that the person with dementia being cared for does not acquire dementia and quantified QALYs of the caregivers under this scenario. The QALY loss due to dementia caregiving is the difference in caregivers’ QALYs. The annual population‐level quality‐of‐life loss for dementia caregivers is calculated by multiplying the total number of spouse and non‐spouse dementia caregivers and the respective per‐capita QALY loss estimate. Refer to the methods study for further details.

2.7. Annual earnings loss among persons living with dementia

To estimate earnings loss due to dementia of persons living with dementia, the USCDM used earnings data, which are composed of wage income, bonuses, overtime pay, commissions, tips, second jobs, military reserve earnings, and professional practice or trade income. We estimated earnings separately by full‐time status using prior work and earnings characteristics, health conditions, and functional limitations, among other variables (Table SA7). Dementia impacts earnings through work and full‐time or part‐time work status; these models contain variables similar to those for earnings, fully interacted by MCI and dementia. To determine the loss in earnings associated with living with dementia, we simulated a counterfactual scenario in which people with dementia in 2010 instead had MCI and did not progress to dementia between 2010 and 2026 and calculated the total earnings difference with baseline in 2026.

2.8. Annual earnings loss of dementia care partners and caregivers

We estimated earnings loss of care partners and caregivers using data collected on HRS respondents who provide care to parents and parents‐in‐law as reported by HRS respondents. Respondents report on whether any living parent or parent‐in‐law has Alzheimer's disease or other dementia and other social and demographic characteristics of each parent such as age and education. HRS respondents report if they provide care to a parent or parent‐in‐law for ADLs and the number of hours for hours more than 100 per year. Using longitudinal data from HRS waves 2012 to 2020, we estimated a probit model of onset of a parent's dementia and a two‐stage model: any care to parents and amount of care hours (above 100 annual hours). Model covariates include demographic, health, and economic characteristics of the caregiver (HRS respondent) and characteristics of the parents, including co‐residency with adult child, widowhood, sex of parent, education, homeownership, number of children, and whether the parent has dementia. We then used a probit model to estimate the likelihood of caring for a parent(‐in‐law) with dementia and a probit model for the impact of respondents’ caregiving on labor force participation. Finally, OLS regression is used to estimate the impact of work on earnings. Table SA7 provides model estimates.

To determine the earnings loss associated with dementia caregiving, we simulated a counterfactual scenario such that parents of respondents did not have and do not acquire dementia. In this scenario, at the start of the simulation, respondents work and co‐reside with parents at the same rate as respondents who do not have parents with dementia. Earnings are calculated under this counterfactual scenario. The difference between the baseline and counterfactual scenario is the earnings loss due to dementia caregiving in 2026.

2.9. Statistical analysis and uncertainty

Markov transition models of dementia onset and changes in other physical and functional health measures, work status and earnings, and caregiving relied on use data from the 1998 to 2020 waves of HRS. The microsimulation begins with a nationally representative population of Americans ages 51 years and older in year 2010 and continues to 2026 (Figure 1). The USCDM has a 2‐year cycle that mimics the frequency of the underlying survey data. At each 2‐year timestep, individual characteristics and outcomes are updated based on predictions from the transition models, and subsequently, models of medical and long‐term care costs (and by payer), quality of life, and care hours utilizing these updated individuals measure are estimated, and then costs are calculated. Simulations predict outcomes for each 2‐year cycle and are calculated for 2026.

FIGURE 1.

Simulation transitions and cohort replenishing.

Individual‐level costs of persons living dementia in 2026 for each of six cost measures are aggregated across all persons living with dementia using HRS survey weights for population representativeness. Aggregates of each cost are summed for 2026 to quantify the total costs of dementia in the United States in 2026. In addition, we calculated costs separately for non‐Hispanic White, non‐Hispanic Black, and Hispanic persons living with dementia. All costs are reported in 2025 dollars.

All simulation outcomes are averages across 50 Monte Carlo repetitions. Uncertainty around the mean was evaluated through a non‐parametric bootstrap method, involving 201 re‐samplings of the survey observations used in re‐estimating the transition models and 30 Monte Carlo repetitions for each bootstrap (6030 total simulations). Outcomes were calculated for each repetition and averaged within each bootstrap. The confidence intervals were created from 2.5th and 97.5th percentiles for each outcome. Data preparation, analyses, and simulations were performed using SAS 9.4, Stata MP 19.0, and C++.

3. RESULTS

3.1. Sample characteristics

Table 1 shows characteristics of HRS respondents, pooled over survey waves 2000 to 2020, and used for model estimation. The average age is 69.5 years, and 57.8% are female. Non‐Hispanic White persons, non‐Hispanic Black persons, Hispanic persons, and persons of other races are, respectively, 69.8%, 16.7%, 10.7%, and 2.8% of the sample population. The highest level of education achieved is high school for 55.1% of respondents. Dementia is present in 7.1% of the sample, 19.5% have at least one ADL, and 16.8% have at least one IADL. The prevalence of certain health conditions is as follows: heart disease: 24.6%, stroke: 9.2%, hypertension: 58.7%, diabetes: 22.6%, cancer: 14.9%, and lung disease: 9.9%.

TABLE 1.

Sample summary statistics.

Characteristic	Estimate
Mean age (SD)	69.5 (10.1)
Female (%)	57.8
Education (%)
Less than high school	20.5
High school graduate	34.6
Some college or more	44.9
Marital status (%)
Married/partnered	62.5
Widowed	21.3
Single	16.2
Working for pay (%)	34.0
Race and ethnicity (%)
Non‐Hispanic Black	16.7
Hispanic	10.7
Non‐Hispanic white	69.8
Other race	2.8
Cognitive status (%)
No cognitive impairment	76.1
Cognitive impairment, not dementia	16.8
Dementia	7.1
Other health conditions (%)
Heart disease	24.6
Stroke	9.2
Hypertension	58.7
Diabetes	22.6
Cancer	14.9
Lung disease	9.9
Any activities of daily living limitation (%)	19.5
Any instrumental activities of daily living limitation (%)	16.8

Note: Survey years 2000 to 2020 pooled, unweighted.

3.2. Population results

In 2026, 5.7 million (95% CI [5.6, 6.0]) persons ages 51 and older are living with dementia in the United States, 5.1 million (95% CI [4.9, 5.3]) are ages 65 and older (Table 2). The number of care partners/caregivers of persons with dementia is 5.2 million (95% CI [4.9, 5.3]) with 4.3 million (95% CI [4.1, 4.6]) care partners being either an adult child, other family member, or friend and 868,000 (95% CI [801,000, 930,000]) partners are spouses of persons living with dementia. Together, persons with dementia received 6.8 billion (95% CI [6.3, 7.2]) hours of care from spouses, other family members, and friends. Table SA8 provides unrounded estimates.

TABLE 2.

Numbers of persons living with dementia, their care partners, and annual hours of care received.

Persons with dementia and care partners	Estimate (95% CI)
Total persons living with dementia (ages 51 and older)	5.7 M (5.6 to 6.0)
Persons living with dementia (ages 65 and older)	5.1 M (4.9 to 5.3)
Total care partners/caregivers	5.2 M (4.9 to 5.5)
Care partners/caregivers (other family, friend of person living with dementia)	4.3 M (4.1 to 4.6)
Care partners/caregivers (spouse of person living with dementia)	868K (801 to 930)
Total hours of care received from care partners/caregivers	6.8B (6.3 to 7.2)

Note: Number of caregivers and relationship to recipient of care are reported by the recipient of care. Billion (B), Million (M), Thousand (K).

The cost of dementia in the United States in 2026 is $818 billion (B) (95% CI [759, 866]) (Figure 2, Table 3). Medical and long‐term care costs for persons with dementia are $222B (95% CI [209, 237]). The value of hours of unpaid care provided by family and others is $237B (95% CI [220, 253]), and these care partners and caregivers also experience quality‐of‐life loss of $15B (95% CI [6, 25]) and annual forgone earnings of $9B (95% CI [0, 17]). Persons living with dementia also experienced earnings losses of $14B (95% CI [7, 21]). The largest portion of costs is due to reductions in quality of life for persons with dementia, $320B (95% CI [269, 363]). Table SA9 provides unrounded estimates.

FIGURE 2.

2026 cost of dementia in United States (2025 US$).

TABLE 3.

Total costs in 2026 of persons with dementia in the United States by type of cost and payer (in 2025 dollars).

Cost category	Estimate (billions of US$) (95% CI)
Medical and long‐term care
By payer	222 (209 to 237)
Medicare costs	110 (100 to 121)
Medicaid costs	44 (37 to 50)
Out‐of‐pocket costs	46 (41 to 52)
Other payers	23 (17 to 29)
Unpaid care from care partners	237 (220 to 253)
Foregone earnings of persons living with dementia	14 (7 to 21)
Forgone earnings of care partners	9 (0 to 17)
Quality‐of‐life loss for persons living with dementia	320 (269 to 363)
Quality‐of‐life loss for care partners	15 (6 to 25)
Quality‐of‐life loss of spousal caregivers	3 (2 to 4)
Quality‐of‐life loss of other (non‐spouse) caregivers	12 (3 to 21)
Total	818 (759 to 866)

Nearly half ($110B; 95% CI [100, 121]) of the $222B in medical and long‐term care costs are borne by Medicare (Table 3). Spending by Medicaid is $44B (95% CI [37, 50]) and together with Medicare account for about 70% of medical and long‐term care costs. Out‐of‐pocket costs borne by individuals and families is $46B (95% CI [41, 52]) and similar in magnitude to Medicaid spending. Spending on persons with dementia by other payers (e.g., private insurance plans) is another $23B (95% CI [17, 29]).

Figure 3 reports the percentage of total costs of dementia by race and ethnicity alongside the population percentage. Non‐Hispanic White persons and “other” race persons with dementia represent about 65.5% of estimated total costs, which is slightly more than their share of dementia population (60.0%). Non‐Hispanic Black and Hispanic persons living with dementia account for a lower share of total costs, 17.3% and 17.2%, as compared to their percentage of the dementia population, 20.7% and 19.4%, respectively.

FIGURE 3.

Percentage of total costs and dementia population by subgroup.

4. DISCUSSION

In the United States in 2026 there are 5.7 million persons living with dementia, and the total cost of dementia is $818B. This includes $222B medical and long‐term care costs of person living with dementia and $237B in costs associated with the unpaid care received by persons living with dementia by 5.2 million family and friends. This study added to prior cost estimates new measures of quality‐of‐life loss for persons with dementia ($320B), care partners’ quality‐of‐life loss ($15B), earnings loss among person living with dementia ($14B), and care partners’ loss of earnings ($9B). Measuring population‐level quality‐of‐life loss for persons living with dementia and their care partners and caregivers can inform healthcare payers about the value of treatments and other types of health and care interventions for persons living with dementia when making coverage decisions. The substantial quality‐of‐life loss estimated in this study points to the potential value of not only treatments but policies and services that may help persons living with dementia maintain independence.

While public healthcare systems support much of the medical costs of persons with dementia, families shoulder over 20% of medical and long‐term care costs of persons with dementia. Earnings losses for persons with dementia and their caregivers are often hidden, and by quantifying their magnitude, this study provides a more complete understanding of the income impact on families and underscores the need for policies that support care partners and caregivers including but not limited to financial supports, flexible work policies, and expanded options for long‐term care.

Costs are not distributed equally across populations. Non‐Hispanic Black and Hispanic persons account for slightly larger shares of the overall dementia population (34%) relative to the percentage of total costs attributed to them. The disproportionately lower cost share may reflect lower use of medical and nursing home care, given the higher use of unpaid family care among Hispanic and Black families. ³⁴ These differences may reflect cultural preferences or differential access to formal care; however, we do not test potential explanations. Moreover, we do not assume that higher healthcare use and spending reflect higher‐quality care. Illuminating differences, however, may support further research into this area for guiding policy.

Total costs included medical and long‐term care spending, unpaid caregiver costs, and quality‐of‐life and earnings losses and provided a measure of the overall economic burden of dementia on society. It quantified the real‐world resource burden on families, health systems, and society. It captured how dementia shapes medical care, long‐term care, and family caregiving in ways that cannot be easily disentangled from co‐occurring conditions. This approach enables meaningful comparisons of disease burden across conditions and supports better‐informed targeting of resources.

Prior national estimates of costs of dementia did not include costs associated with quality‐of‐life losses or impact on persons with dementia and caregivers’ work and earnings. ⁹ , ³⁵ The comparable types of costs measured (e.g., medical, long‐term care, and family caregiver hours) in our analysis vary from prior studies due to estimated dementia prevalence numbers, types of costs quantified, data sources, and methods for valuing the hours of care from family and others. ¹⁰ , ²⁹ , ³⁶ Nandi et al. ⁹ projected the costs of dementia that included medical and long‐term care costs and care provided by family and others to be $580B in 2025 (in 2020 US$). This is approximately equivalent to $722B in 2025 US$. Estimates of direct medical and long‐term costs were $252B in 2020$ (95% CI [184, 326]) or about $314B in 2025$ (95% CI [229, 405]), and although the estimates reported in this new study ($222B) are lower, they are within a 95% CI. Despite level differences, both Nandi et al. and this study report the value of unpaid care to be just over half of the sum of value of unpaid care and medical and long‐term care costs. There are differences in data and methods. Nandi et al. used self‐reported medical and care expenses in HRS, and the medical and long‐term care costs reported in this study used data from MCBS and MEPS and were calibrated using National Expenditure Accounts. Nandi et al.’s valuation of informal care according to replacement rate cost was $328B in 2020 US$, or about $408B in 2025 US$ and was also higher than estimates reported in this study ($237B in 2025 US$). Differences in estimated costs associated with care provided by family and others may be attributed to assumptions about how very low, e.g., <1 h/week, or very high, e.g., 24 h, reported care hours were considered. The Alzheimer's Association's 2025 estimates of medical and care costs and value of informal care are also higher than those produced by USCDM, about $384B and $413B (in 2025 US$), respectively. ³⁷ Similarly, the proportion of costs due to informal care is just over half, while the higher‐level difference is driven in part by a higher estimated number of persons with dementia – 7.2 million ages 65 and older. We reported 5.7 million persons ages 51 and older, of which 5.1 million are ages 65 and older. The difference is likely driven by data sources. The Alzheimer's Association's prevalence estimate used data from the Chicago Health and Aging Project, a longitudinal study from 1993 to 2012 of urban populations collected from three neighborhoods on the South Side of Chicago, and projected prevalence based on demographic‐specific Census projections for future years. ³⁸ Finally, we used a QALY value of $150,000 per year; however, other studies used a range of QALY values, often between $100,000 and $200,000 per year. ³⁹ , ⁴⁰ Given the large size of the dementia population, a QALY value 50% higher or lower would change the reported value proportionally; however, the interpretation of a meaningful impact remains unchanged.

Providing national estimates of different types of costs as well as total costs and tracking them over time is critically important given the rapidly aging population and resulting increase in the number of persons with dementia. US Census projects a more than doubling of the US population ages 85 and older by 2050. ⁴¹ Age is a strong predictor of dementia with prevalence rates equal to 2.8% at age 65, 16.0% at age 85, and 22% at ages 90 and above. ²⁰ The financial impact of caring for persons living with dementia on healthcare systems and government is matched by that on families who provide the majority of care for persons living with dementia and pay out of pocket for long‐term care services and support. The lack of universal long‐term care policy and financing mechanisms leaves many families vulnerable to high out‐of‐pocket spending, with only a Medicaid safety net for those families who exhaust most of their financial resources. Tracking the costs of dementia in the United States over time is also critically important because of the rapidly changing landscape of pharmaceutical treatments that can change the natural history of dementia and new diagnostics that will change what we know about who has what type of dementia and when pathologies emerge. These and other breakthroughs will alter the number of persons with dementia and the composition of the population with dementia and shift costs across persons and payers in uncertain ways.

The study had limitations. Number of persons living with dementia was measured in the HRS using TICS, and, despite not distinguishing between etiologies, the score has been shown to be highly sensitive in distinguishing people with and without dementia in a nationally representative sample. The number of caregivers was based on reports by persons receiving care and caregivers more often report providing help than care recipients report receiving it. ⁴² The omission of the population younger than age 51 sacrifices little generality, since dementia and expenditures on the disease generally occurs at older ages. Moreover, we accounted for care partners of all ages that are reported by respondents who received care. However, estimates of quality‐of‐life loss and earnings loss for non‐spouse caregivers were based on HRS respondents at least age 51 who provided care to a parent with dementia, so we assumed the impacts on earnings or quality of life were quantitatively the same for caregivers younger than age 51. Furthermore, restricting the sample to respondents with a living parent for model estimation resulted in large confidence intervals around parameter estimates. Despite a more comprehensive measure of costs that includes; quality‐of‐life decrements and impacts earnings, there are other costs including but not limited to those related to legal and financial planning expenses and those related to housing modifications and housing transitions. ⁴³ , ⁴⁴ Costs to the Medicaid program are likely underestimated as there are state‐specific programs that support persons with dementia and the spending on these services is not fully captured by the data used for the study.

A separate paper with methodological details and publicly available code and data sources supports the transparency and reproducibility of estimates by the research community and provides results on model validation assessments. It also describes the historical development of this methodology and how dynamic microsimulation has served as an important tool for guiding policymakers. By way of example, the federal government recently announced increased access to anti‐obesity medication through expanded Medicare coverage – a policy shift informed by research using dynamic microsimulation to demonstrate the health and economic benefits of such coverage. ⁴⁵ The dynamic microsimulation methods utilized in the study support analyses for understanding how advances in prevention and treatments, detection tools, care models, and changes to policies impact costs and for whom. These insights will be necessary for prioritizing investments and mobilizing resources to reduce the economic costs of dementia.

The results of this study underscore the large and multifaceted economic impact of dementia in the United States in 2026. As the population ages and the prevalence of dementia continues to rise, the costs – spanning medical care, long‐term care and services, family caregiving, quality of life, and work and earnings impacts for persons with dementia and their care partners – are rapidly growing. These comprehensive estimates highlight the urgent need for action to address gaps in long‐term care financing, expand support for caregivers, and invest in effective prevention, treatments, and care models in order to address these growing costs and in sustainable ways. Understanding the full scope of dementia's economic toll is a critical step toward shaping the health and social policies that will be critical to reducing the burden of dementia for individuals, families, and society.

CONFLICT OF INTEREST STATEMENT

Dr. Dana Goldman reports grants received in the last 3 years from the American Heart Association, Alexion, Amgen, Biomarin, Blue Cross Blue Shield of Arizona, Blue Cross Blue Shield of Massachusetts, BrightFocus, Bristol Myers Squibb, California Hospital Association, Cedars‐Sinai Health System, Charles Koch Foundation, CommonSpirit, Edwards Lifesciences, Gates Ventures, Genentech, Gilead Sciences, Incyte, Johnson & Johnson, Lilly, National Institute on Aging, National Institute of Diabetes and Digestive and Kidney Diseases, Novartis, Pfizer, RA Capital, and Sarepta Therapeutics; personal fees from Edwards Lifesciences and GRAIL. He is a co‐founder of EntityRisk and holds equity in the company. The remaining authors have nothing to disclose. Author disclosures are available in the supporting information.

Supporting information

Supporting Information

Supporting Information