The burden of cognitive impairment

Abstract

INTRODUCTION

Over 10% of older American adults (age ≥ 65) experience cognitive impairment (CI) from Alzheimer's disease and related dementias (ADRD). An additional 15%–22% experience mild cognitive impairment (MCI), often preceding ADRD and in earlier life. However, CI burden estimates including MCI and early‐onset CI (age < 65), remain limited.

METHODS

Using microsimulation, we estimated CI burden (ADRD and/or MCI) over remaining life‐years for Americans (aged 51/52) who developed CI overall, by disease progression, onset age, and sociodemographic subgroups. Burden equaled the difference in quality‐adjusted life‐years (QALY valued at $150K/year) and economic outcomes between a hypothetical intervention preventing CI and status quo.

RESULTS

Per capita lifetime CI burden was $124K, driven by lost QALYs. Aggregate CI burden was $627B, 59% from ADRD, 41% from MCI. Early‐onset CI carried the largest per capita burden ($376K), ∼50% from MCI. Individuals with less education, progression to ADRD, or from minoritized racial/ethnic subgroups also carried disproportionate burden.

DISCUSSION

MCI constitutes 41% of CI burden, disproportionately impacting individuals with early‐onset.

Highlights

• The lifetime burden of cognitive impairment (CI) was estimated at $124,000 per capita, totaling $627 billion in aggregate—with 59% attributed to Alzheimer's disease and related dementias (ADRD) and 41% to mild cognitive impairment (MCI).

• Early‐onset CI (before age 65) carried the highest per capita burden at $376,000, with these individuals living with MCI for about 13 years—over 7 years longer than those with late‐onset CI.

• Significant disparities exist in CI burden, with racial/ethnic minorities and those without a high school degree experiencing disproportionately higher burden. Despite representing just 36% of the CI population, non‐White individuals accounted for 47% of the aggregate CI burden.

1. BACKGROUND

Cognitive impairment (CI) represents an escalating fiscal, societal, and public health crisis as the global population continues to age. About 10% of older American adults (age > 65) experience CI related to Alzheimer's disease and related dementias (ADRD or “dementia,” including other etiologies like vascular dementia or frontotemporal dementia), ¹ , ² , ³ , ⁴ , ⁵ , ⁶ and an additional 15%–22% experience mild cognitive impairment (MCI; inclusive of cognitively impaired, no dementia), characterized by noticeable, but less severe cognitive dysfunction. ¹ , ² , ⁴ , ⁶ , ⁷ , ⁸ , ⁹ Most research thus far has focused on ADRD burden among adults older than 65, currently estimated between $215–321 billion annually. ¹ , ² Evidence on MCI burden is limited ¹⁰ , ¹¹ despite being prodromal to ADRD in 40%–60% cases ¹² , ¹³ and consistently linked to poor quality of life, ¹⁴ , ¹⁵ high medical expenditures, and unpaid caregiving costs. ¹⁶

CI burden estimates inclusive of MCI are even scarcer across subgroups defined by age of onset, disease progression, and sociodemographics. For example, early‐onset CI (EO‐CI) (age < 65) currently affects 1‐6% of individuals with ADRD ⁶ , ¹⁷ , ¹⁸ , ¹⁹ and 7%–10% with MCI. ¹⁹ , ²⁰ , ²¹ However, its lifetime burden remains understudied despite being the fastest growing subtype, having doubled in prevalence between 2013 and 2017. ²² Those with early‐onset ADRD are more likely to experience more rapid cognitive decline, and may die younger relative to those with late onset (after age 65), ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ all of which disproportionately affects racial/ethnic minorities and less‐educated groups. ⁶ , ²⁹ , ³⁰ It is unknown whether and to what extent such findings apply to MCI.

Using microsimulation, we estimated the individual and aggregate burden of CI (ADRD and MCI) for a nationally representative sample of Americans aged 51–52 years in 2016 over each individual's remaining life‐years, both overall and by subgroups defined by disease progression (MCI that does vs. does not progress to ADRD), age of CI onset (beginning before vs. after age 65), and sociodemographics. We defined CI burden as the improvement in health and economic outcomes that would result in the absence of any CI onset.

2. METHODS

This study was approved by the institutional review board at the University of Southern California with a waiver of informed consent. Reporting of this study followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline. ³¹

2.1. Future elderly model overview

We utilized the Future Elderly Model (FEM), an individual‐level dynamic microsimulation, to predict health and economic outcomes over the lifetime of a synthetic cohort of 51‐ and 52‐year‐old Americans. The FEM projects person‐level outcomes such as cognitive function, disease conditions, mortality, disability, functional status, nursing home admission, caregiving, and economic outcomes using nationally representative data from the Health and Retirement Survey (HRS) and other surveys like the Medical Expenditure Panel Survey (MEPS) and the Medicare Current Beneficiary Survey (MCBS). The FEM employs multivariate models at the individual level in 2‐year intervals, adjusting for a comprehensive set of individual characteristics and behaviors, with weights ensuring national representativeness. FEM has been previously used to predict cognitive function, along with other chronic disease and disability burdens. ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ Its mortality and quality‐of‐life projections have been validated, ³⁷ , ⁴⁰ and the model has been compared with other ADRD‐focused evaluations. ⁴¹ Detailed FEM documentation, including model validations, is available in the Supplementary Appendix.

In this study, we used FEM to compare a hypothetical intervention that permanently prevents CI onset with the status quo (existing care) (Figure 1A). We estimated biennial transition probabilities between health and economic states and simulated life trajectories for individuals diagnosed with CI under each scenario. See estimates of transition probability models in eTables 1.1‐1.7. By comparing outcomes between the status quo and the intervention scenario, we estimated the CI burden among individuals older than age 50. Further details are provided below.

FIGURE 1.

Approach and descriptive statistics for a simulated cohort of adults ever experiencing CI. (A) Schematic representation of simulation, outcomes, and burden calculation. (B) Descriptive statistics of simulated cohort of Americans aged 51–52 years in 2016 and their projected lifetime outcomes at status quo (additional subgroups in eFig3.1). Transition probabilities used in the simulation are estimated on a longitudinal, binennial 2000‐2016 HRS sample of adults older than 50 years. All CI (cognitive impairment) includes individuals who developed CI at some point in their lifetime: MCI only (i.e., terminal MCI), MCI followed by ADRD (i.e., MCI with ADRD); all mutually exclusive. Terminal MCI refers to individuals who die with MCI without progressing to ADRD; MCI with ADRD refers to individuals who experience MCI which progresses to ADRD; all mutually exclusive. EO‐CI includes individuals with CI onset (MCI or ADRD) prior to age 65 and LO‐CI includes individuals with CI onset at or above age 65; all mutually exclusive. Subgroups are defined by simulating life trajectories to label participant lifetime outcomes at baseline. All costs are in USD $2020 with 3% annual discount rate. ADRD, Alzheimer's disease and related dementias; CI, cognitive impairment; EO‐CI, early‐onset cognitive impairment; MCI, mild cognitive impairment; USD, United States dollars.

2.2. Data and study population

We used data from the 2000–2016 longitudinal HRS, a nationally representative study of Americans aged 50 years and older. Our estimation sample included 26,607 HRS participants, accounting for 131,820 person‐waves. Additionally, our simulated cohort included 1301 HRS participants aged 51–52 years in 2016 living either in the community or in nursing homes, with complete data on cognitive function. Unpaid caregiving (hours of help) is measured directly in HRS, while formal medical costs (total, Medicare, Medicaid, out‐of‐pocket) are assigned using MCBS models for Medicare enrollees and MEPS models for non‐enrollees, supplementing HRS to provide comprehensive burden estimates. More details on HRS data can be found elsewhere. ⁴²

2.3. Defining cognitive impairment

We used the modified version of the Telephone Interview for Cognitive Status (TICS) in the HRS to obtain respondents’ cognitive function described as dementia, MCI, or no cognitive impairment (NCI). Cognitive status was determined using information provided by the survey's respondents TICS score or a response by a proxy (typically a spouse or other family member). ⁴³ To avoid underestimating the prevalence and incidence of CI, ⁴⁴ proxies were used to include severely cognitively impaired individuals—who comprised less than 10% of the sample in each wave—in the survey (see eTable 2.1). For each survey respondent, we summed TICS scores from three cognitive assessments: immediate and delayed word recall (0–20); counting down from 100 by 7′s test score (0–5); and backward count (0–2). The cutoff points were 0–6 for dementia, 7–11 for MCI, and 12–27 for NCI. ⁴⁰ , ⁴⁵ Imputation was used for missing values (about 3%). ⁴⁶ Proxies do not answer the cognitive battery on behalf of the respondents but assess their memory performance and instrumental activities of daily living (ADL). Interviewer's assessment of subjects’ cognitive limitations was also included. See eTable 2.1 for details on self‐ and proxy responses and imputation across waves. Using this information from proxies and informants, cognition scale (range 0–11) consisted of (1) the number of limitations in instrumental ADL (iADL) (0–5); (2) interviewer impairment rating (0 implies no cognitive limitations, 1 Implies some cognitive limitations, 2 implies cognitive limitations); and (3) a proxy informant's rating of the respondent's memory (from 0 [excellent] to 4 [poor]). Cutoff points for proxy scores were: 0–2 for NCI, 3–5 for MCI, and 6–11 for dementia. ⁴⁷ , ⁴⁸ These scores are then combined with TICS categorization. ³⁹ For someone to be categorized as having persistent MCI, we required MCI in two consecutive waves or MCI in one wave followed by dementia or death in the following wave. We assumed that once a respondent has MCI based on these two criteria, they either remained in that state forever or progressed to dementia. To identify persistent dementia, we required one wave with dementia and evidence of continued CI (either MCI or dementia) in a subsequent wave, or at least one wave with dementia before death. Others have used similar measures. ³⁹ , ⁴⁹ , ⁵⁰ , ⁵¹

We divided respondents with CI into subgroups in several ways. First, by disease progression: individuals with MCI who never progressed to ADRD were classified as having terminal MCI, while those who progressed from MCI to ADRD were classified as MCI with ADRD. Second, by age of onset: individuals whose CI began before age 65, were categorized as having early‐onset CI (EO‐CI) versus those with onset at age 65 or older as late‐onset CI (LO‐CI). Third, by sociodemographics, including sex (male vs female), race/ethnicity (White vs. non‐White), and educational attainment (with vs. without a high school degree).

RESEARCH IN CONTEXT

1. Systematic review: Cognitive impairment (CI), including Alzheimer's disease and related dementias (ADRD) and mild cognitive impairment (MCI), is an increasing fiscal, societal, and public health issue as populations age. While 10% of older adults over 65 experience CI related to ADRD, 15%–22% have MCI, a precursor to ADRD. Most research has focused on the economic and healthcare burden of ADRD, but the burden of MCI remains understudied, despite its impact on quality of life and healthcare costs. Early‐onset CI (EO‐CI; before age 65) has also seen a sharp rise, particularly in underrepresented populations.

2. Interpretation: Using microsimulation, we estimated the burden of CI for a nationally representative sample, offering insights into the impact of MCI and EO‐CI across subgroups defined by disease progression, age of onset, and sociodemographics. We found that MCI accounts for a significant portion of the CI burden and disproportionately impacts those with early onset. Lower‐educated, dementia, and racial/ethnic minority groups also experienced larger burden.

3. Future directions: Findings highlight the need for early intervention and targeted support for at‐risk populations, especially those with EO‐CI and MCI.

2.4. Modeling cognitive impairment

Conditional on survival, FEM uses probit models to estimate two‐year incidences of cognitive state separately (e.g., from NCI to MCI, and from MCI to dementia, as defined above). We require that a person with normal cognition either remains cognitively normal or progresses to dementia indirectly through MCI, but not directly, as these instances are rare (< 5%). We model transitions to cognitive states as a function of a previous cognitive state, and of both fixed and time‐varying components using HRS for Americans aged 50+. The fixed components included sex, race/ethnicity, and education. The time‐varying components were updated biennially in the simulation and included age, chronic health (heart disease, diabetes, stroke, lung disease, cancer, or hypertension), body mass index (calculated as weight in kg/height in meters), and smoking behavior as measured in a previous wave. Chronic health is coded as a binary outcome, and measured as answers to questions like “have you ever smoked” or “has a doctor ever told you were diagnosed with diabetes….”. See Supplementary Appendix for MCI and dementia incidence estimates in eTable 1.2.

2.5. Modeling other economic and health outcomes

To understand the dynamic interplay of cognition and other outcomes, FEM uses estimates of self‐reported outcomes that determine cognitive state (i.e., smoking, body mass index, or chronic health as described in the CI model above); and of outcomes that were affected by cognitive function directly or indirectly. Cognitive state as a categorical variable enters as a key covariate in models of health status, functional status measured as limitations in ADL or iADL, ⁵² nursing home residence and its duration, unpaid caregiving hours, employment status and type (out of labor force vs. unemployed; part‐time vs. full‐time), earnings, and mortality. Outcomes like medical expenditures (Medicare, Medicaid, private, and out‐of‐pocket) are indirectly influenced by cognitive state. We used ordinary least squares for continuous outcomes, ordered probit for categorical outcomes, and probit models for binary outcomes, employing first‐order Markov processes of a 2‐year transition controlling for factors measured at baseline. Supplementary Appendix section A1, describes transition models in detail.

2.6. Simulation

Using transition probabilities estimated from 2000–2016 HRS, we simulated life trajectories for Americans aged 51–52 years in 2016 to identify the cohort prevalence of CI. This synthetic cohort was based on HRS respondents from the 2016 survey, with adjustments for demographic changes (age, race‐ethnicity, and sex), and other factors such as obesity, smoking, hypertension, diabetes, earnings, and medical costs. For adults ever developing CI, we compared projected lifetime outcomes under no intervention (e.g., status quo or existing care) to a counterfactual scenario with zero CI incidence.

To calculate standard errors and 95% confidence intervals (95% CIs), we used a non‐parametric bootstrap method to resample data from the FEM transition models, generating 50 sets of transition models. The simulation was run 150 times using these different transition models, and the resulting outcomes were used to calculate 95% CIs (based on 1.96*standard deviation) around the mean.

2.7. Estimating the burden of cognitive impairment using simulation

Our primary outcome of interest was CI burden per capita, defined by the average net social value of full prevention of CI among those with CI. Net social value was calculated as the difference between the net present value of life‐time total health gains [defined as a change in quality‐adjusted life‐years [QALYs] adjusted for total economic costs incurred above age 50 after the hypothetically reducing CI incidence by 100% (Figure 1A)]. We reported monetary values in 2020 USD, discounted 3% annually. ⁵³ We also calculated CI burden that accrues prior to age 65 and the burden specific to ADRD and MCI. We examined how burden varies by disease progression, age of onset, sex, race/ethnicity, and education. All FEM simulations and analyses were performed using C++ and STATA MP 16.0 ⁵⁴ between May 2020 and October 2022.

Below, we define each element used to estimate the burden of CI, obtained through simulation.

2.7.1. Per capita health gains

Our primary measure of health gains was a change in QALY valued at $150,000 per additional QALY. ⁵⁵ , ⁵⁶ QALY is a widely‐used measure of healthy life‐years encompassing several dimensions of health (e.g., ADL, iADL, obesity, smoking, and marital status), including life expectancy and has been regularly used in cost‐effectiveness literature related to CI. ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ In this study, we calculate QALYs using the Multi‐Attribute Health Status Classification System:

Health Utilities Index Mark 3 (HUI‐3), which also includes cognition as a health dimension. ⁶⁰ , ⁶¹ See Supplementary Appendix for transitions models. Our other outcomes of interest were age of dementia onset and entry to nursing home, life‐expectancy (i.e., the average age at death across individuals in each cohort), and disability‐free life expectancy (defined as average years lived without any ADL or iADL limitations, or nursing home residency). Similarly, we examined dementia‐ or CI‐free life expectancy, and nursing‐home–free life‐years.

2.7.2. Per capita total economic costs

We defined economic costs as the sum of total formal (medical expenditures including doctor and hospital visits and nursing home stay) and unpaid caregiving costs (i.e., from family and friends), and loss of gross annual earnings from labor force participation. To estimate unpaid caregiving costs, we multiplied the number of hours of unpaid caregiving received from co‐resident spouses and non‐spouse caregivers by the average wage rate paid to a home health aide hired through an agency ($15 per hour). ⁶²

After calculating the per capita net social value, we assessed the overall effect of the hypothetical CI prevention at the population level by multiplying the average per capita net social value with the number of individuals ever diagnosed with CI overall and within each subgroup of interest.

3. RESULTS

3.1. Modeled cohort

Of simulated Americans aged 51–52 years in 2016 (N = 1301), 59% experienced CI during their lifetime (Figure 1B, eFigure 3.1). Of these, 44% experienced only MCI (terminal MCI) and 56% progressed from MCI to ADRD; CI onset occurred early for 20%. Those who progressed to ADRD and those who experienced EO‐CI were both more likely to be non‐White and without a high school degree compared to their terminal MCI and LO‐CI counterparts. However, those who progressed from MCI to ADRD were more likely to be female with lower prevalence of chronic conditions at baseline while those with EO‐CI were more likely to be male with higher prevalence of chronic conditions.

3.2. Population prevalence of CI

Weighted to be nationally representative, CI prevalence was 17% (MCI 10.8%, ADRD 6.2%) among adults aged between 51 and 100 years in HRS 2000–2016. Among those aged between 65 and 100 years (N = 81,486 person‐years), the CI prevalence was 25% (MCI 14.9%, ADRD 10%). See eTable 2.2 for a comparison of our ADRD prevalence estimates by age group with previous studies.

3.3. Status quo

An average 51 to 52‐year‐old adult with CI had remaining life expectancy of 36.1 years (95% CI: 35.7–36.5) of which 23.8 years (95% CI: 22.8–24.9) were healthy (defined by QALY), 6.7 years (95% CI: 6.5–7.0) were with MCI, and 4.0 years (95% CI: 3.8–4.1) with dementia (Figure 2). This individual spent 11.4 years (95% CI: 11.1–11.8) with disability, required unpaid caregiving for 6.4 years (95% CI: 6.0–6.9), and nursing home residency for 2.0 years (95% CI: 1.7–2.3), all of which overlapped toward end‐of‐life. Of all subgroups, individuals with EO‐CI experienced the poorest health outcomes under status quo followed by those with terminal MCI, without a high school degree, and people from racial and ethnic minority groups (eFig3.2). For example, individuals with EO‐CI lived 12.3 (95% CI: 11.8–12.8) fewer total years and 10.3 (95% CI: 9.7–10.9) fewer QALYs compared to those with LO‐CI. MCI and dementia occurred 22.5 years (95% CI: 22.1–23.0) and 15.9 years (95% CI: 15.4–16.3) earlier and lasted 6.7 years (95% CI: 6.3–7.1) and 3.6 years longer (95% CI: 3.2–3.9) for EO‐CI. EO‐CI was also associated with an earlier onset of disability, unpaid caregiving, and nursing home residency. All life‐years are reported undiscounted.

FIGURE 2.

Life trajectories of per capita health outcomes from age 51–52 until death for status quo (SQ) and PP of cognitive impairment for All‐CI, and Terminal MCI, MCI with ADRD, EO, and LO subgroups. Additional subgroups in eFig 3.2. Average outcomes include (1) total life expectancy; (2) duration with MCI (gray), ADRD (black), unpaid caregiving (medium red), nursing home (dark red), and disability (light red); (3) quality‐adjusted life‐years (dark green); and, (4) life expectancy free of CI, (white, gray text), unpaid caregiving (white, medium red text), nursing home (white, dark red text), and disability (white, light red text). All life‐years are reported undiscounted. ADRD, Alzheimer's disease and related dementias; CI, cognitive impairment; EO, early onset; LO, late onset; MCI, mild cognitive impairment; PP, permanent prevention; SQ, status quo.

3.4. Per capita burden of CI

The per capita lifetime burden of CI for Americans aged 51–52 years in 2016 who at some point developed CI was $124K (95% CI $101K–$148K). Compared to those who never developed CI under a hypothetical permanent prevention scenario, those who develop CI under the status quo lost 3.0 years of total life expectancy (95% CI: 2.8–3.2), 2.5 quality‐adjusted years (95% CI: 2.2–2.8), and 13.6 CI‐free years (95% CI: 13.3–14.0) (Figures 2, 3A). They experienced disability onset 2.1 years earlier (95% CI: 2.0–2.3), though with a 0.8‐year shorter duration (95% CI: 0.7–1.0). Unpaid caregiving began 2.6 years earlier (95% CI: 2.5–2.8) but lasted 0.4 fewer years (95% CI: 0.2–0.5). Nursing home residence occurred 3.2 years earlier (95% CI: 3.0–3.4) and lasted 0.2 more years (95% CI: 0.1–0.3). Altogether, these outcomes led to $48.6K (95% CI: $42.4K–$54.8K) in reduced lifetime medical spending, but this was offset by $4.07K (95% CI: $1.82K–$6.33K) in added unpaid caregiving costs, $6.40K (95% CI: $5.03K–$7.77K) in lost earnings, and $162K (95% CI: $139K–$185K) in QALY losses. All life‐years are reported undiscounted, and all monetary values as discounted at 3% annually.

3.5. Aggregate burden of CI

The aggregate lifetime CI burden was $627B (95% CI: $511B–$743B): 57% ($357B) specific to ADRD in adults older than 65, 2% ($1.5B) specific to ADRD burden accrued between ages 50–65 years, and 41% ($254B) specific to MCI (age > 50), respectively (Figure 3B).

FIGURE 3.

Burden of CI. (A) Per capita and (B) aggregate burden of CI for All‐CI and Terminal MCI, MCI with ADRD, EO‐CI, and LO‐CI subgroups. Additional subgroups in eFig3.3‐3.4. 95% confidence intervals reported. Discounted QALY gains between status quo and permanent prevention (used to calculate burden) are as follows: All‐CI = 1.1 years (95% CI: 0.9–1.2); Terminal MCI = 0.8 years (95% CI: 0.7–0.9); MCI with ADRD = 1.3 years (95% CI: 1.1–1.5); EO‐CI = 2.9 years (95% CI: 2.4–3.3); LO‐CI = 0.6 years (95% CI: 0.5–0.7). (B) Insert: Shares of aggregate CI burden that occurs between ages 50–65 and after age 65 by MCI and ADRD. 95% CI, 95% confidence interval; ADRD, Alzheimer's disease and related dementias; CI, cognitive impairment; EO‐CI, early‐onset cognitive impairment; LO‐CI, late‐onset cognitive impairment; MCI, mild cognitive impairment; QALY, quality‐adjusted life‐year.

3.6. Subgroup burden of CI

Subgroup‐specific disparities in CI burden by disease progression, age of onset, race/ethnicity, education, and sex (Figure 3, eFig3.3): Of all subgroups, individuals with EO‐CI experienced the greatest lifetime per capita CI burden at $376K (95% CI: $315K–$436K), followed by those without a high school degree at $204K (95% CI: $167K–$242K), those whose MCI progresses to ADRD at $176K (95% CI: $144K–$207K), and non‐White individuals at $161K (95% CI: $131K–$193K). Though primarily driven by QALY losses of $429K (95% CI: $366K–$492K), EO‐CI burden also included $30K (95% CI: $24K–$36K) in lost earnings. Despite never progressing to ADRD, individuals with terminal MCI experienced 1.7 year loss in QALYs, valued at $119K (95% CI: $99K–$139K).

On aggregate, individuals with EO‐CI, MCI with ADRD progression, without a high school degree, and non‐White represented 20%, 56%, 13%, and 36% of the total CI simulated population, yet they accounted for 60%, 79%, 21%, and 47% of the subgroup‐specific aggregate lifetime burden (eFigure 4), respectively. Individuals with terminal MCI and EO‐CI experienced the highest shares MCI‐specific burden, with 100% and 46% of their total burden, respectively, attributable to MCI.

4. DISCUSSION

In this nationally representative study of older US adults, we estimated that an average per capita lifetime CI burden was $124K, or $627B in aggregate. Burden was primarily driven by loss in healthy life‐years consistent with prior work showing loss in life‐years due to CI. ⁶³ Our estimates of CI burden are higher than those of other costly chronic conditions such as obesity ⁶⁴ and diabetes, ⁶⁵ underscoring the need for better CI prevention, screening, and management strategies.

Our findings suggest that prior ADRD burden estimates in adults older than 65 could underestimate the total CI burden by 43%, largely due to the exclusion of individuals with MCI. Though MCI may not cause functional limitations at first, it costs $254B in aggregate. We show that MCI contributes to significant losses in life expectancy, healthy life‐years, and high formal and unpaid caregiving costs regardless of its progression to ADRD. While a lack of systematic screening paradigms, ⁶⁶ phenotypic heterogeneity, ⁷ , ⁸ , ⁹ and variations in definitions ⁶⁷ make early identification of MCI difficult, our burden estimates add to a growing body of work focused on prodromal states of CI when cognitive deficits and functional impairment are not as evident. ⁶⁸ This is especially important considering increasingly MCI‐focused inclusion criteria of investigational therapeutics. ⁶⁹

We find that MCI is particularly costly for those who experience CI early. Those with early CI onset lived with MCI for 13.1–7.38 years longer than those with LO‐CI on average. Indeed, at $375K per capita, we estimated that adults with early onset carried the highest lifetime burden of all subgroups. Their aggregate burden was 60% of the total burden, despite only representing 20% of the CI population. These results are not surprising: compared to those with LO‐CI, those with EO‐CI had more comorbidities at a given age, needed more unpaid caregiving, and CI drove greater loss in healthy life‐years and earnings. These findings align with the fact that while most adults older than 65 with CI are retired, many with EO‐CI are of working age, and lack access to existing medical care, home care or community programs designed for much older adults. ⁷⁰ This is concerning because, even though the number of people affected by early‐onset ADRD, for instance, is currently low, its diagnosis in ages between 30 and 65 has experienced 131% growth between 2013 and 2017. ²² , ⁷⁰ Thus, the EO‐CI burden is expected to grow.

Individuals with EO‐CI and those who progressed from MCI to ADRD were more likely to be from minoritized racial/ethnic groups or without a high school degree—two populations we also found to carry larger CI burden than their counterparts. Prior studies have found associations between lower educational attainment and MCI or ADRD, ⁶ , ²⁹ and that American adults older than 65 with dementia and/or MCI are more likely to be Black or Hispanic descent. ⁶ , ²⁹ , ³⁰ Others have found that Black persons experienced earlier onset of CI and bear a disproportionate share (one‐third) of ADRD costs. ³⁰ Our results are aligned with these findings in that non‐White individuals experienced 47% of the aggregate CI burden; suggesting that these subgroups would benefit significantly from medical breakthroughs in preventing or treating CI, particularly novel therapeutics such as aducanumab ⁷¹ or lecanemab ⁷² or behavioral interventions. ⁷³

Our study has limitations. First, cognitive status measurement using TICS is less precise than clinical assessments and does not distinguish between etiologies of CI. However, TICS was developed and validated by Langa and Weir (2010) and has been since used and validated by many studies. ³⁹ , ⁷⁴ This score has also been shown to be highly sensitive in distinguishing people with and without dementia, including diverse populations, ⁷⁵ , ⁷⁶ , ⁷⁷ and was shown excellent specificity for differentiating dementia from MCI. ⁷⁶ Thus, while not a replacement for face‐to‐face evaluation, TICS offers an easily applicable and brief option for dementia or severe CI detection. To reduce the measurement error using this score, we used a conservative measure of CI prevalence, requiring at least two subsequent diagnoses of CI, so our burden estimates may represent a lower bound. We found MCI and dementia prevalence at around 15% and 10% for Americans aged 65+, respectively, comparable with prior estimates using a different approach. ⁶ , ⁷ , ⁸ , ⁹ Second, FEM assumes that transition probabilities from current to future health status depend on current health status only and do not account for prior history. Third, parameters are estimated from relatively small samples for subgroup analyses, but we assess statistical uncertainty using bootstrap methods. Fourth, our life‐time burden estimates may not be directly comparable to annual cross‐sectional estimates, though they are mostly nationally representative of older adults. Fifth, while the bias introduced by proxy data cannot be predicted, its impact is likely minimal, as proxies account for less than 10% of our data.

5. CONCLUSION

We find that CI imposes a substantial and uneven lifetime burden, with a significant share attributable to MCI. Prior estimates likely understate this burden by excluding MCI and younger adults, whose losses in life expectancy, healthy years, and economic outcomes are considerable. The burden is especially high for those with EO‐CI, progression to ADRD, and individuals from racially minoritized and socioeconomically disadvantaged or groups. These disparities highlight the need for earlier detection and broader access to interventions, particularly as MCI becomes a key focus of emerging therapeutics. Accounting for the full spectrum of CI is essential to reducing its growing impact on individuals and society.

This study involves secondary analysis of data from the Health and Retirement Study (HRS) and was granted exemption from the University of Southern California social and behavioral Institutional Review Board. HRS protocols, informed consent, and data use policies are available at https://hrs.isr.umich.edu.

All data, methods, and models used in this manuscript are in the public domain.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

Supporting information

Supporting Information

Supporing Information

Gracner T, Chaturvedi R, Nguyen PG, et al. The burden of cognitive impairment. Alzheimer's Dement. 2025;21:e70436. 10.1002/alz.70436