Researcher brief

Eight leadership conclusions

Interpretive synthesis based on NHATS estimates. These conclusions are not causal findings.

Methods

Data source, measures, and analytic approach

NHATS is an annual, nationally representative study of Medicare beneficiaries ages 65 and older. It began in 2011 and periodically refreshes its sample to support national trend estimation. The chart book restricts each annual round to adults aged 72 and older to ensure population comparability across years.

The annual analytic samples, excluding nursing-home residents, ranged from 3,303 to 6,320 respondents. The corresponding samples, including nursing-home residents, ranged from 3,384 to 6,552. Weighted cross-tabulations account for the complex survey design.

Table 1. Key measure families and interpretive questions

Source: NHATS measure definitions (Freedman and Cornman, 2026) and author synthesis.

Table 2. Evidence hierarchy used in this research

The hierarchy prevents the proposed strategy from being mistaken for a published causal finding.

Quality control

Verification ledger

Every relative change asserted in the report was recomputed from the published endpoints. The ledger is shown rather than silently corrected, so a reader can see which figures are reproducible and which involve a judgement.

How the interactive series were produced

Six figures in the report are line charts that print only their endpoint labels. Rather than approximate the intervening years, each series was traced from the published image: gridlines were located by colour, the value scale was calibrated from those gridlines, and each series was then read column by column at the year positions.

The trace was validated against the printed labels. Every traced endpoint reproduces its published value, including all five 2022 dementia labels (24.3, 15.4, 8.9, 5.3 and 4.6), all five 2023 digital labels (84.6, 80.6, 68.2, 52.7 and 34.1), and the end-of-life series at 82.8, 83.1 and 67.4. Endpoint values in the dashboard are the published figures. Intervening years are traced and should be read as close reconstructions rather than as published estimates.

Analytic procedure in the report: extract all summary time series and detailed table totals; calculate percentage-point change as latest minus baseline and relative change as that difference divided by baseline; treat p-values as evidence about the source’s linear trend test rather than as proof of operational importance; inspect discontinuities, plateaus, and reversals that a single linear statistic could obscure; compare subgroup patterns; and separate published estimates from author interpretation.

Empirical findings

Function is increasingly maintained through adaptation

The share fully able to perform all seven self-care and mobility activities without help, difficulty, or device use declined by 5.1 percentage points, from 26.6% in 2011 to 21.5% in 2023. Over the same period, successful accommodation with devices increased by 6.2 points, from 30.8% to 37.0%.

The result is not straightforward functional deterioration. A growing share of older adults moved from unassisted full ability to successful device-supported performance while avoiding researched difficulty or the need for personal assistance. This is precisely the type of adaptation that conventional disability counts can miss.

How self-care and mobility are being carried out

Share of adults ages 72 and older, excluding nursing-home residents. Categories are hierarchical and sum to approximately 100% within a year.

Figure 2. Self-care and mobility composition, 2011 and 2023. The distribution shifted toward successful accommodation with devices and away from full ability.

View the published figure

Where the adaptation did not reach

Age and subgroup patterns temper the positive interpretation. Successful accommodation increased for adults ages 72 to 84, but not for those ages 85 and older. Men and women improved, as did non-Hispanic White and Black adults, but the trend was not significant for Hispanic or other race and ethnicity groups. White adults experienced a decline in assistance, while other groups did not. Technology diffusion was substantial but not universal.

Empirical findings

Household activities: the support-gap signal

The household-activity series provides the report’s strongest evidence of a support mismatch. Difficulty increased from 12.9% to 18.0%. Receipt of assistance declined from 30.3% to 25.4%. Household unmet need rose from 6.8% to 8.2%.

The coherence of these movements matters. Rising difficulty alone might reflect changes in expectations or research methods. Declining assistance alone might reflect improved capacity. Rising difficulty, declining assistance, and increasing adverse consequences together support a more concerning interpretation: some household needs are being absorbed by the older person, delayed, performed with difficulty, shifted to technology, or left unmet.

Household difficulty rose while assistance declined

Percent of adults ages 72 and older, 2011 to 2023, excluding nursing-home residents.

Figure 3. Household-activity difficulty, assistance, and unmet need. The opposing trajectories of difficulty and assistance form the central support-gap signal. The 2023 unmet-need value is 8.2% in Detailed Table 4.

View the published figure

Table 3. The household support gap in endpoint terms

Source: Freedman and Cornman (2026). The relative change is the author’s calculation and does not indicate a causal effect.

Derived indicator

The size of the group needing support did not change. Its coverage did.

The household measure is a three-category hierarchy: fully able, has difficulty, or receives assistance. Those three shares sum to 100 in both years. That arithmetic carries a finding the endpoint table does not state directly.

In 2011 the group that was not fully able totalled 43.2% of the population. In 2023 it totalled 43.4%. The population needing household support was essentially unchanged. What changed was the share of that group receiving help, which fell from 70.1% to 58.5%.

Why declining assistance is ambiguous

Reduced assistance can be positive when better capacity, effective devices, or accessible environments make help unnecessary. It can be negative when help is unavailable, unaffordable, mismatched, unstable, or shifted to an overextended family member.

This distinction should change measurement. A payer that counts hours of help without measuring task completion may reward more services even when adaptation is effective. A system that counts lower service use as success without measuring unmet consequences may reward abandonment. The appropriate outcome is safe, preferred, reliable task completion with the least burdensome effective mix of personal help, technology, and environmental adaptation.

Empirical findings

Capacity improved, but the gains are uneven

The share in the 2011-defined high physical-capacity quartile increased from 19.7% to 24.6%, while the low-capacity quartile declined from 27.8% to 21.0%. Because the quartile thresholds are anchored to the 2011 distribution, these changes represent improvement relative to the baseline population rather than a mechanically fixed 25% in each quartile.

Individual performance tests broadly confirm the direction. The share with low grip strength fell by 9.8 percentage points, low balance by 8.4 points, low chair-stand performance by 7.2 points, and low walking performance by 6.7 points. Low lung airflow did not improve significantly.

Physical capacity improved, with a pandemic measurement gap

Percent in 2011-defined capacity quartiles. Performance tests were unavailable in 2020 and partly unavailable in 2021.

Figure 4. Physical-capacity trends. The missing performance-based years are left visible rather than bridged. A simple line between 2019 and 2022 could imply a continuity that was not observed. The uninterrupted self-reported series provides a useful, but not identical, comparator.

View the published figure

Dementia

Probable dementia among adults aged 72 and older outside nursing homes declined from 11.9% in 2011 to 8.1% in 2022, a 3.8-point and 31.9% relative decline. When nursing-home residents were included, prevalence declined from 14.4% to 9.7%. Significant age-specific declines occurred among adults ages 80 to 84, 85 to 89, and 90 and older, but not among those in their seventies.

Dementia prevalence declined primarily at older ages

Percent with probable dementia by age group, 2011 to 2022, excluding nursing-home residents.

Figure 5. Dementia prevalence by age group. Absolute prevalence remained strongly age-graded even as the oldest groups improved. The series ends in 2022 because classification requires information from two survey rounds.

View the published figure

A companion NHATS analysis found declining prevalence and incidence through 2021 but also sharply increased mortality among people with dementia at the onset of the pandemic. Compositional shifts explained 27% of the full-period prevalence decline, compared with 45% of the pre-pandemic decline (Freedman and Cornman, 2024). Prevalence is jointly shaped by incidence, survival, diagnostic classification, education, vascular and other risk exposures, and pandemic mortality.

Sensory capacity

Poor hearing declined from 17.8% to 12.7%, with improvement among both sexes and in three of four regions. Poor vision changed from 10.7% to 9.5%, but the overall linear trend did not meet the chart book’s significance threshold. Vision improved among women, adults ages 90 and older, and non-Hispanic Black adults.

The asymmetry between hearing and vision is operationally important because sensory function affects fall risk, communication, cognitive assessment, informed consent, portal use, medication instructions, and telehealth accessibility.

Empirical findings

Community residence has become the dominant care platform

Among adults outside nursing homes, traditional community residences increased from 84.2% to 86.5%, while residential care outside nursing homes declined from 7.9% to 6.4%. When nursing-home residents were included, nursing-home residence declined from 4.4% to 2.2%, a fall of 50% relative to its 2011 level.

This is a profound reallocation of where disability is managed. Care work once concentrated in institutions is increasingly dispersed across private residences, senior housing, outpatient clinics, emergency departments, home health, hospice, family networks, and paid home-care agencies. Institutional decline may reflect preference, policy, financing, supply, eligibility, mortality, and improved accommodation. It should not be equated with the elimination of complex needs.

Figure 6. Residential setting and home modification. The community is expanding as the site of aging, while environmental adaptation is concentrated in bathing and toileting. Residential-setting estimates have different inclusion rules as labeled.

Housing form and the flat mobility line

Housing type changed modestly. One-level and accessible multi-level houses remained broadly stable, while limited-access multi-level houses increased, mobile-home residences declined, and multi-unit residences declined. The increase in limited-access multi-level housing is counterintuitive for an aging population and may create stair, entry, and emergency-egress challenges not captured by modification counts.

Home modification was more dynamic. Any modification increased from 72.8% to 81.3%. Bathing modifications rose from 66.1% to 74.4%, and toileting modifications from 39.3% to 50.5%. Mobility-related modifications, such as ramps, elevators, or stair glides, remained nearly flat, rising only from 12.4% to 12.8%.

Relatively common interior bathroom adaptations have diffused readily. Structurally intensive mobility adaptations have not. Cost, ownership, building form, and structural complexity remain the plausible barriers, and each of them is addressable by a financing or contracting pathway rather than by clinical advice.

Empirical findings

Digital participation has become part of late-life functioning

Electronic communication changed faster than any disability or housing indicator in the chart book. Emailing or texting increased from 29.5% in 2011 to 73.0% in 2023, a 43.5-point and 147.5% relative increase. Going online for purposes other than email increased from 30.9% to 68.1%. Online household activities increased from 16.1% to 54.9%, and online health activities from 12.7% to 39.1%. Social networking increased from 15.0% in 2013 to 38.0% in 2023.

These are not peripheral lifestyle trends. Banking, shopping, prescription ordering, insurance administration, clinical communication, scheduling, results review, transportation, and social connection increasingly require digital interaction. Digital capability therefore affects whether an older adult can perform instrumental activities of daily living, navigate healthcare, and remain socially connected.

Electronic communication increased at every age

Percent emailing or texting, by age group, 2011 to 2023.

Figure 7. Emailing or texting by age group. Adoption rose in every age group. In 2023, 84.6% of adults ages 72 to 74 communicated electronically compared with 34.1% of adults ages 90 and older.

View the published figure

Adoption is not equity

The aggregate increase conceals large 2023 gradients. Email or texting ranged from 84.6% among adults ages 72 to 74 to 34.1% among those ages 90 and older. It reached 91.2% among adults with graduate degrees but 32.8% among those without a high-school diploma. Rates were 77.8% among non-Hispanic White adults, 54.3% among non-Hispanic Black adults, and 47.3% among Hispanic adults. Metropolitan and non-metropolitan rates were 74.1% and 67.8%.

Figure 8. Digital stratification in 2023. All displayed groups experienced significant gains, yet current access remains highly unequal. The figure uses email or texting because it is available across the full period and for detailed subgroups.

Digital exclusion is multidimensional. A recent scoping review identifies resource, skill, and motivational exclusion, with consequences including technology anxiety and broader social exclusion (Ge et al., 2025). Digital health literacy depends on age, education, health, device access, experience, family and social support, and interface design (Shi et al., 2024). NHATS-based work also shows that older adults with vision impairment have lower odds of having and knowing how to use common devices (Thomas et al., 2024).

The remaining non-user population is not a residual group that can be ignored. It is disproportionately older, less educated, more likely to have functional or cognitive limitations, and more likely to need assistance. Leaders should measure successful completion, not portal activation. Telephone, in-person, proxy, caregiver, accessible-device, language, and low-literacy pathways should be treated as designed channels rather than exceptions.

Empirical findings

Participation recovered, but the end-of-life experience deteriorated

Participation in valued activities remained near 92% from 2011 through 2019, fell sharply to 79.1% in 2020, and recovered to 92.5% by 2023. Participation restrictions moved from 22.9% in 2019 to 17.0% in 2020 and then returned to 23.4% in 2023.

The simultaneous decline in participation and researched restrictions in 2020 appears paradoxical only if restriction is treated as the inverse of participation. In NHATS, restriction reflects health or functional constraints on valued activities. A widespread social shutdown may have reduced participation for reasons not attributable to the person’s health or functioning.

Participation recovered after the 2020 shock

Percent participating in, or restricted from, valued activities, 2011 to 2023.

Figure 9. Participation and participation restriction. The 2020 discontinuity illustrates why context and measure definition are essential. Participation recovered by 2023, but the series does not prove that all social, psychological, or functional consequences were resolved, and it does not reveal frequency, meaning, accessibility, or whether the same people regained the same activities.

View the published figure

End-of-life care

Hospice use in the last month of life increased from 48.3% in 2013 to 60.5% in 2022, while nursing-home deaths declined from 25.8% in 2012 to 15.8% in 2022. Home death was broadly stable across endpoints, and hospital death fluctuated. NHATS-linked research indicates that hospice is associated with better ratings of several end-of-life outcomes among older adults with dementia (Harrison et al., 2022).

However, proxy-rated excellent or very good care declined from 82.8% in 2012 to 67.4% in 2022, and from 83.1% in 2019 to 67.4% in 2022. Uptake and quality diverged. The series cannot identify whether the decline reflects workforce strain, symptom management, communication, care coordination, place of care, pandemic restrictions, bereavement context, hospice timing, or changes in expectations. It does show that a service-use measure cannot substitute for an experience measure.

Figure 10. End-of-life care and location of death. Hospice use rose, nursing-home deaths fell, and excellent or very good care declined. The findings require separate governance of access, timing, coordination, symptoms, respect, communication, and family experience.

Stratified analysis

The national average conceals distinct equity patterns

Age

Age gradients remain large even where trends improve. Adults aged 90 and older had a 24.3% dementia prevalence in 2022, and only 34.1% used email or texting in 2023. They did not share in the significant increase in device-mediated self-care and mobility accommodation observed among those under age 85. Household unmet need reached 11.7% for the 90-and-older group in 2023, and self-care or mobility unmet need reached 32.4%, although the latter did not show a significant linear trend. Aggregate improvement should not dilute planning for the oldest old.

Sex

Most trends were similar for men and women, but important exceptions emerged. Difficulty with self-care and mobility increased among men but not women. Assistance with self-care and mobility declined among women but not men. Household unmet need increased among men but not women. Women experienced declines in residential care other than nursing homes and in excellent or very good end-of-life care, while hospice use increased among women. These patterns merit investigation of survival, living arrangement, caregiving availability, disability severity, and service access rather than simplistic gender explanations.

Race and ethnicity

Many significant trends were concentrated among non-Hispanic White and Black adults, partly reflecting sample size and statistical power. Device-mediated accommodation increased among these groups, but not among Hispanic or other race and ethnicity groups. Only non-Hispanic White adults experienced significant declines in self-care and mobility assistance and poor hearing. Only non-Hispanic Black adults experienced a significant decline in poor vision. All racial and ethnic groups increased their digital activity, yet substantial 2023 digital gaps remained. The evidence supports universal design with targeted support, not a one-size-fits-all diffusion assumption.

Education

The educational distribution shifted upward over the period, a compositional change that can improve aggregate indicators of physical, cognitive, and digital well-being. At the same time, education remained one of the strongest digital gradients: emailing or texting ranged from 32.8% among adults without a diploma to 91.2% among those with graduate education. Household unmet need increased significantly among adults without a diploma, reaching 12.8% in 2023. Health systems that move administrative work online without literacy-sensitive support may convert educational inequality into access inequality.

Region and metropolitan status

The South experienced more significant trends than other regions, reflecting both population scale and substantive change. Unmet self-care or mobility needs increased in the Midwest, while household unmet needs increased in the Northeast. Physical-capacity improvement was largely metropolitan.

The non-metropolitan pattern is especially consequential. Unmet need for self-care or mobility increased from 12.6% to 16.5% in non-metropolitan areas while declining from 14.6% to 12.7% in metropolitan areas. Household unmet need more than doubled from 4.7% to 10.1% in non-metropolitan areas, while it did not significantly trend in metropolitan areas. Electronic communication increased in both settings but remained lower outside metropolitan areas.

Non-metropolitan gains coexist with a widening care gap

Percent in 2011 and 2023. Trend significance is reported in the detailed NHATS tables.

Figure 11. Metropolitan and non-metropolitan divergence. Non-metropolitan adults achieved major digital gains but experienced significant increases in unmet need. This is a classic example of progress in one infrastructure domain coexisting with deterioration in another.

View the published figure

Table 4. Selected 2023 equity signals

Source: Detailed Tables 3, 4, 6, and 12 in Freedman and Cornman (2026). Comparisons are descriptive and not adjusted causal effects.

Interpretation

Disability is being redistributed, not erased

Why the findings appear contradictory

A prevalence-based reading produces contradictions: physical capacity improves while the share fully able declines; assistance declines while household difficulty increases; nursing-home residence falls while the number of people aging in the community rises; digital participation expands while digital inequities persist; hospice use rises while rated quality falls.

The contradictions largely dissolve when the outcomes are placed in a system of person-environment fit. Capacity is only one input. An older adult may have adequate grip, balance, or cognition yet struggle with medication management, transportation, meal preparation, stairs, insurance forms, or caregiver coordination. Conversely, a person with limited capacity may function well when a device, an accessible bathroom, a reliable helper, and a usable digital channel reduce task demand. The unit of analysis should be the functioning system around the person.

The redistribution pathway

Stages one through four are grounded in NHATS trends. Downstream risk and the required operating model are evidence-informed interpretations requiring local validation.

Competing interpretations

The report’s preferred interpretation integrates rather than dismisses these explanations.

The future pressure is demographic

Current improvements do not eliminate future demand. Recent NHATS-based projections estimate that late-life care needs will rise about 23% from 2022 to 2040, with roughly 7.7 million older adults potentially needing care but lacking a family caregiver and approximately 14 million experiencing unmet care needs. The principal driver is population growth at the oldest ages, not a collapse in family commitment (Freedman et al., 2026a).

A separate 2026 analysis found increasing limitations and unmet need among adults ages 65 to 74 even as older age groups improved, suggesting that future cohorts may not simply inherit the favorable trends observed among today’s oldest adults (Freedman et al., 2026b).

Author-developed conceptual model

Adaptive Aging Infrastructure Framework

The framework integrates six domains that NHATS often presents separately. Its purpose is not to create a premature composite score. It is to prevent leaders from optimizing one domain while degrading another. Safe, equitable independence is the outcome of fit across all six domains.

Figure 12. Adaptive Aging Infrastructure Framework. The framework connects capacity, accommodation, environment, care sufficiency, digital access, and goal-concordant life. It is a governance model, not a validated diagnostic instrument.

Explore the six domains

Table 5. Framework domains and measurement implications

Author-developed framework. Measures are illustrative and require validation, ownership, and specification of denominators.

Model

Prevalence and burden are different questions

Most of this report’s favourable findings are rates. Capacity plans are built in people. When the population at the oldest ages grows, a falling rate and a rising caseload are not contradictory; they are the expected result.

NHATS-based projections estimate that late-life care needs will rise about 23% between 2022 and 2040, driven principally by population growth at the oldest ages rather than by a collapse in family commitment (Freedman et al., 2026a). This model applies that arithmetic to any denominator you can defend locally.

Model

Adaptive Aging Infrastructure readiness

Twenty-four items across the six framework domains. The overall score is a geometric mean rather than an average, because the framework’s central claim is that fit is governed by the weakest domain. A programme that is excellent in five domains and absent in the sixth does not deliver five-sixths of safe independence.

From evidence to operating model

Leadership implications

Action architecture

Table 6. Adaptive Aging Infrastructure Scorecard

A useful scorecard links population need, service response, and experienced consequence. It should avoid a single composite index until reliability, weighting, and unintended incentives are tested.

Table 7. Implementation roadmap

The sequence is intentionally small-pilot first. It does not assume that every domain can be automated or owned by a health system.

Recommended 90-day pilot

Table 8. Pilot outcome set

Proposed evaluation set. The pilot should be registered internally before launch, with definitions frozen for the initial cycle.

Scientific restraint

Limitations and research agenda

Table 9. Principal limitations

Table 10. Priority research questions

Leadership synthesis

Conclusion

The NHATS trends document substantial progress. Older Americans are more likely to successfully use assistive devices, less likely to fall into the lowest physical-capacity group, less likely to live with probable dementia, less likely to report poor hearing, more likely to live in the community, more likely to have home modifications, and far more likely to participate digitally. Participation in valued activities also recovered after the 2020 disruption.

The progress should not obscure the central warning. Household difficulty increased as assistance declined, household unmet need rose, non-metropolitan unmet need worsened, and excellent or very good end-of-life care declined even as hospice use expanded. The architecture supporting independence has become more distributed and more dependent on systems that healthcare traditionally treats as peripheral: caregivers, accessible homes, digital access, transportation, community services, and cross-sector coordination.

The strategic question is no longer whether an older adult is simply independent or dependent. It is whether capacity, accommodation, environment, care sufficiency, digital access, and valued life goals fit together reliably. A person can remain at home and still be unsafe. A person can use a portal and still fail to obtain care. A person can enroll in hospice and still experience poor coordination. A person can use a walker and still lack grocery or medication support.

Audit table

Appendix A. Detailed trend catalog

Table A1 consolidates major total-population endpoints from the chart book. It is not a substitute for the source’s detailed subgroup tables. Baseline and latest years are 2011 and 2023 unless the interpretation column indicates otherwise.

Source: Detailed tables in Freedman and Cornman (2026). The source reports the p-values.

Measure guide

Appendix B. Operational definitions

Condensed from the chart book’s methods and measures appendix. Consult the source for full item wording, hierarchical classification, and imputation rules.

Implementation reference

Appendix C. Evidence-to-action map

Action implications are author recommendations, not effects estimated by NHATS.

References

The reference list prioritizes the controlling NHATS source, peer-reviewed NHATS analyses, and recent systematic or scoping reviews directly relevant to the interpretation.