Lifespan exposures to rural–urban conditions and later‐life cognitive function

Abstract

INTRODUCTION

Limited research exists on life‐course rural–urban residence and cognitive functions.

METHODS

This study examines associations between rural–urban residence during childhood, adulthood, residential mobility (childhood to adulthood), and later‐life cognitive outcomes among U.S. adults ≥65 years of age (N = 3073) from the Health and Retirement Study‐Harmonized Cognitive Assessment Protocol.

RESULTS

Linear and logistic regression showed that childhood rural residence was associated with lower memory (standardized beta (β) = –1.68, 95% Confidence Interval (CI) [–2.39, –0.97]), executive function (β = –1.20, 95% CI [–1.84, –0.55]), and language fluency (β = –0.86, 95% CI [–1.64, –0.08]). Individuals living in rural areas in childhood had a higher risk for mild cognitive impairment (MCI; odds ratio [OR] = 1.31, 95% CI [1.02, 1.69]) and dementia/MCI (OR = 1.30, 95% CI [1.04, 1.62]). Compared to lifelong urban residents; residents who lived in rural areas during childhood and/or adulthood had lower cognitive function and a higher risk of MCI.

DISCUSSION

Rural residence throughout the lifespan, especially during childhood, is linked to poorer later‐life cognitive outcomes, underscoring the need for early targeted health care interventions to address rural–urban disparities.

Highlights

• Few studies have examined the timing of rural residence on cognitive health.

• Rural residence across the lifespan is associated with lower cognitive function.

• Childhood rurality is particularly associated with lower cognitive function.

1. BACKGROUND

In the United States, ≈20% of the population lives in rural areas, ¹ whereas nearly 43% of the global population lives in rural areas. ² Compared to urban‐residing people, those who live in rural areas face an elevated risk of experiencing poor cognitive functioning, ³ delayed detection of dementia, ⁴ , ⁵ and death attributed to dementia. ⁶ As reported in a recent study, between 1999 and 2019, rural areas in the United States had a higher prevalence of dementia and experienced the highest increase in dementia‐related mortality, compared to urban areas. ⁷ Rural–urban disparities in dementia risk may be due to lower education attainment, occupational complexity, greater engagement in unhealthy behaviors like smoking, higher risk of experiencing comorbid health conditions such as metabolic syndrome and cardiovascular disease, and reduced access to health care services and other health‐promoting amenities compared to urban areas. ⁸ , ⁹ , ¹⁰ , ¹¹ These robust and modifiable risk factors are often established early in life and accumulate over time. ¹² , ¹³ , ¹⁴ However, there is still limited evidence on how rural–urban disparities in dementia risk unfold across the lifespan.

Dementia is a slowly progressive syndrome that can be traced to early life, ¹⁵ with the underlying pathology developing many years before observable symptoms arise, underscoring the importance of life‐course analysis in dementia‐related research. The Socioecological Model of Health ¹⁶ and Life Course Theory ¹⁷ together suggest that individuals’ health is impacted by structural and societal contexts (including geographic location) throughout the life course. In this case, at what developmental stages and for how long people live in rural or urban areas may be crucial for dementia incidence. Three primary hypotheses have been put forth regarding the effects of different developmental stages on lifespan health. First, theories about “sensitive periods” suggest that certain developmental stages have a greater impact on later‐life outcomes because there are critical neurobiological and/or socioemotional milestones that must be met. ¹⁸ , ¹⁹ These early ingredients set the stage and can have cascading effects on the rest of the life course. For example, childhood, adolescence, and early adulthood are critical periods for cognitive development, ²⁰ , ²¹ , ²² making rural–urban residence during these periods especially impactful for later‐life cognitive health and dementia risk, above and beyond the effects of later developmental stages. Second, life‐course trajectories may be influenced continually by ongoing interactions between individuals and their contexts. In this case, developmentally‐proximal stages, such as rural or urban residence in adulthood, can have a greater impact on later‐life health outcomes because the effects of early life on later health can be reversed, augmented, or negated with individual and contextual changes with age. ²² Third, and finally, it is possible that lifespan advantages and disadvantages accumulate across the lifespan (aligned with Cumulative Disadvantage Theory), where living in a rural area in childhood and adulthood confers even more risk for later‐life cognitive health than each developmental stage in isolation. ²³ , ²⁴

RESEARCH IN CONTEXT

Systematic review: The authors reviewed the relevant scientific literature regarding theoretical and empirical research on life‐course rural/urban residence and cognitive outcomes. Although there is limited work on this topic, the available research suggests that residing in rural areas during childhood or late life is linked to poorer cognitive outcomes. The relevant articles have been appropriately cited.

Interpretation: Using life‐course rural residence data, our findings suggest that rural residence across the lifespan, especially childhood rural residence, is associated with lower levels of later‐life cognition across multiple domains.

Future directions: It is crucial for future research to understand the mechanisms underlying rural–urban disparities in cognitive function across the lifespan. In addition, conducting cross‐national comparisons will provide valuable insights into these mechanisms, particularly across regions where the impact of rural–urban disparities on cognitive outcomes is more (and less) pronounced.

Understanding how rural–urban disparities in later‐life cognition unfold across the life course can offer valuable insights for when in the life course prevention and intervention policies might be most effective. Although evidence exists on life‐course rural–urban residence and later‐life cognitive decline, ²⁵ research on how rural–urban residence from early life to old age affects different cognitive function domains and dementia risk remains limited. Using nationally representative data from the United States, we aimed to investigate the associations between rural–urban residence in childhood, adulthood, and residential mobility and later‐life cognitive outcomes. Specifically, we had three primary research questions: (RQ1) To what extent is childhood rural–urban residence associated with later‐life cognitive function and status?; (RQ2) To what extent is adulthood rural–urban residence associated with later‐life cognitive function and status? And is this association confounded by childhood rural‐urban residence?; and (RQ3) To what extent is residential mobility (rural‐to‐rural, urban‐to‐urban, rural‐to‐urban, urban‐to‐rural) associated with later‐life cognitive function and status? We evaluated both cognitive function and status/diagnosis as outcomes because cognitive function domain scores provide detailed measures of specific cognitive abilities, which facilitate early detection and intervention of cognitive decline, whereas measures of status/diagnosis are clinically meaningful andfacilitate clinical treatment and management.

2. METHOD

2.1. Study participants and data sources

We used data from the Health and Retirement Study (HRS) 2016 Harmonized Cognitive Assessment Protocol (HCAP) Respondent and Informant Data and Summary Cognitive and Functional Measures Data, which includes 3496 participants 65 years of age and older ²⁶ who underwent cognitive batteries to assess their cognitive function and impairment. Childhood and adulthood rural–urban residence statuses were obtained from the 2016 HRS Core and Harmonized HRS datasets, respectively. ²⁷ , ²⁸ Demographic information (e.g., age, sex, race, ethnicity, and parental education) was extracted from the HRS Research and Development (RAND) dataset. ²⁹ , ³⁰ Additional details about the variables, data sources, and study years can be found in Table 1. We excluded participants who arrived in the United States after the age of 10 (n = 423), which resulted in a final sample size of 3073 participants (weighted n = 45,138,192). In addition, due to missingness on key analytic variables and/or covariates, the sample sizes for each research question and cognitive outcome vary slightly across research questions (see Figure 1 for more detail). The respondents were also asked to nominate a friend or family member for an informant interview to aid with the assessment of the cognition of the respondents. Of all HCAP cases, 87% have both informant and respondent interviews. ³¹ The present study was not pre‐registered, but all analytic code and materials to reproduce the analyses can be found on the Open Science Framework at: https://osf.io/ceryx/. We did not have any a priori hypotheses for this study. Data can be downloaded from the HRS website.

TABLE 1.

Data sources for variables of interest.

Variable	Data source	Study year	Notes
Childhood rural–urban residence	Health and Retirement Study (HRS) Core	2016	Based on the interview question "Were you living in a rural area most of the time when you were (in grade school/in high school/about age 10)?"; imputed based on waves at years 1996–2016
Adulthood rural–urban residence	Harmonized HRS	2016	Based on residential addresses
Cognitive function	HRS Harmonized Cognitive Assessment Protocol (HCAP)	2016
Demographic characteristics	HRS RAND	2016	Missing ages were imputed based on waves at years 1992–2016
Respondent/informant	HRS‐HCAP	2016

FIGURE 1.

A flowchart of participants included in the analyses. ^a n = 2492 for diagnosis outcomes; n = 2398 for language fluency, visuospatial functioning, orientation time and place domain scores (after excluding 94 records from informants [which had no domain scores]); n = 2396 for the memory domain score (after excluding 94 records from informants and 2 records with a missing score); and n = 2394 for the executive function domain score (after excluding 94 records from informants and 4 records with a missing score). ^b n = 2497 for diagnosis outcomes; n = 2387 for language fluency, visuospatial functioning, orientation time and place domain scores (after excluding 110 records from informants); n = 2385 for the memory domain score (after excluding 110 records from informants and 2 records with a missing score); and n = 2383 for the executive function domain score (after excluding 110 records from informants and 4 records with a missing score). ^c n = 2463 for diagnosis outcomes; n = 2369 for language fluency, visuospatial functioning, orientation time and place domain scores (after excluding 94 records from informants); n = 2367 for the memory domain score (after excluding 94 records from informants and 2 records with a missing score); and n = 2365 for the executive function domain score (after excluding 94 records from informants and 4 records with a missing score).

2.2. Measures

2.2.1. Exposure: Childhood and adulthood rural–urban residential statuses and mobility

As part of the HRS Core interview, participants reported on their childhood rural–urban residential status using a dichotomous item that asked: “Were you living in a rural area most of the time when you were (in grade school/in high school/about age 10)?” Adulthood rural–urban residential status was classified using Rural–Urban Continuum Codes in 2016, which are codes that represent whether the address in a given county is considered rural or urban in relation to its population count and (non‐) adjacency to a metro area. These characteristics are represented on a 1–9 point scale and were further dichotomized by HRS into rural–urban status for public use. Codes 1–3 are considered “urban” (or 0) and 4–9 are considered “rural” (or 1). We then defined residential mobility as four categories based on childhood to adulthood rural–urban status: urban‐to‐rural, rural‐to‐urban, rural‐to‐rural, and urban‐to‐urban.

2.2.2. Outcomes: Cognitive function and status

Cognitive function and diagnosis were assessed using the HCAP, which was developed to enable cross‐national comparisons of cognitive aging and has been tested in prior research. ³¹ In the present study, we used pre‐constructed measures of five domains of cognitive functions (memory, executive function, language fluency, visuospatial functioning, and orientation) as well as cognitive status (normal, mild cognitive impairment [MCI], and dementia). ³² Items for cognitive function domains come from various well‐established sources, including the Mini‐Mental State Examination (MMSE), Telephone Interview for Cognitive Status (TICS), and Consortium to Establish a Registry for Alzheimer's Disease (CERAD), among others (see Jones et al. ³³ for a full list for each domain). Memory, executive function, language fluency, and visuospatial functioning scores ranged from 0 to 11; orientation time and place domain scores ranged from 0 to 5. The memory, executive function, and language fluency scores were standardized to a mean of 50 and an SD of 10 in the HCAP sample. Visuospatial, orientation place, and orientation time cognitive function domains had highly skewed distributions, where most participants achieved the maximum possible scores of 11, 5, and 5 for visuospatial, orientation place, and orientation time domains, respectively (Figures S1 and S2). As a result, these three domains were dichotomized to distinguish between participants who obtained the full score and those who did not. For all subsequent analyses, these three domains are treated as binary outcomes (1 = full score, 0 = not full score).

For cognitive status, we used three different variants of the outcome in relation to normal cognitive function: diagnosis of dementia, diagnosis of MCI, and diagnosis of dementia or MCI. As developed and reported in prior research, dementia and MCI in the HCAP sample were classified algorithmically using standard diagnostic criteria and comparing test performance to a normative sample (see Manly et al. ³⁴ for more details).

2.2.3. Covariates

We depicted the relationships between childhood rural residence, adulthood rural residence, and late‐life cognition in conceptual causal diagrams in Figure 2. These conceptual diagrams were developed prior to conducting analysis to serve as a framework for guiding the selection of covariates in the models, with nodes representing key variables and directed arrows indicating temporally ordered effects. Specifically, we adjusted for confounders but avoided adjusting for mediators, which are part of the pathways of interest.

FIGURE 2.

Conceptual diagram for this study with confounders for (A) associations between childhood rural–urban and adulthood rural–urban and cognitive function/status; (B) associations between residential mobility and cognitive function/status. SES, socioeconomic status (reflects childhood social, financial, and human capital). Dashed arrows indicate uncertain directions between nodes.

For RQ1 (childhood rural–urban residential status), we considered the following confounders (Figure 2A): age (continuous), sex (female; male), race (white; Black/African American; other race), ethnicity (not Hispanic/Latino; Hispanic/Latino), and parental education (maximum years of maternal and/or paternal education, where responses could range from none to 17+ years; continuous). Childhood socioeconomic status (SES) was pre‐constructed as a continuous index by Vable et al., ³⁵ combining childhood social capital (maternal investment and family structure), childhood financial capital (financial resources, financial instability), and childhood human capital (parental education years). Because the temporal order between childhood rural–urban status and childhood SES was unknown, childhood SES could be either a confounder or a mediator. To eliminate potential bias, we excluded childhood SES from the main analyses and evaluated it in the sensitivity analyses.

For RQ2 (adulthood rural–urban residential status), we controlled for potential confounding from age, sex, race, ethnicity, parent education, adulthood self‐reported education (participants reported on their years of education; responses could range from none to 17+ years; continuous), adulthood wealth quintiles (participants reported on their net non‐housing financial wealth; responses ranged from –1800k to 16,150k U.S. dollars in 2016; categorical), and childhood SES (Figure 2A). Similarly, to account for the uncertain temporal ordering of marital status and adulthood rural–urban residential status, we included them as additional covariates in the sensitivity analyses.

For RQ3 (residential mobility from childhood to adulthood), we considered the potential confounders as age, sex, race, ethnicity, and parental education. In addition, we conducted a separate set of models that adjust for covariates with unknown temporal orders, including childhood SES, adulthood education, adulthood wealth, and marital status, to assess their impact on RQ3 associations (Figure 2B).

2.3. Statistical analysis

We conducted all statistical analyses using R (version 4.0.2 ³⁶ ) and the “survey” ³⁷ package. We interpreted p‐values < .05 as statistically significant. We fit logistic regression and linear regression models for dichotomous outcomes and continuous outcomes, respectively. For the three dichotomous domain scores, coefficients represent the odds of obtaining full scores versus not. For the three dichotomous diagnoses, the odds of each diagnosis (relative to the “normal” category) were modeled. Specifically, to better understand how rural–urban disparities affect disease severity, we use three comparisons: normal versus MCI; normal versus dementia; and normal versus MCI or dementia. Participants diagnosed with MCI were excluded for the normal versus dementia comparison. All the models were weighted by HRS‐HCAP sampling weights to account for differential sampling and participation rates by demographic factors to represent U.S. adults 65 years of age and older. ³⁸

To test RQ1, we regressed the cognitive outcomes on childhood rural–urban status and known confounders (e.g., age, sex, race, ethnicity, and parental education) for each outcome separately. To test RQ2, we regressed the cognitive outcomes on adulthood rural–urban status and known confounders (e.g., age, sex, race, ethnicity, and parental education) for each outcome separately. Furthermore, to assess the extent to which childhood rural–urban status confounds the relationship between adulthood rural–urban status and later‐life cognitive function, we compared the estimates from models excluding versus including childhood rural–urban status.

Finally, to test RQ3 regarding residential mobility from childhood to adulthood, we regressed the cognitive outcomes on residential mobility categories and known confounders for each outcome separately. Due to the long timespan that the residential mobility variable captures, the roles of some of the covariates (i.e., childhood SES, adulthood education, adulthood wealth quintiles, and marital status) as confounders versus mediators were unknown, and we thus conducted additional follow‐up analyses to adjust for these covariates with unknown temporal orders.

We also conducted several sets of sensitivity analyses to better understand the boundary conditions of the findings. For RQ1, we conducted a sensitivity analysis to include childhood SES as a confounder. For RQ2, we evaluated marital status (dichotomous; married/partnered vs else [married, spouse absent; separated; divorced; widowed; never married]) as a confounder. For all RQs, we conducted sensitivity analyses to compare the results when we consider visuospatial functioning and orientation as continuous variables rather than transformed binary variables (due to skewness). For all RQs, we conducted sensitivity analyses that excluded observations where we had only an informant interview (but no respondent interview) for the analytic sample.

3. RESULTS

Of the sample, 44% lived in rural areas, 50% in urban areas, and 6% had missing values on rural–urban residence during childhood. In adulthood, 29% lived in rural areas, 69% in urban areas, and 1% had missing values (Table 2). In addition, 10% moved from urban to rural areas, 25% from rural to urban, and 18% and 39% stayed in rural and urban areas, respectively; 8% had missing values. In terms of demographics, the final sample had a median age of 76 (range = 65–103). Approximately 59.9% of participants were female, and 77.0% were non‐Hispanic White, 16.7% non‐Hispanic Black, and 4.8% Hispanic/Latino. The participants had a mean education of 13.0 years (SD = 2.7) and a mean non‐housing financial wealth of $157,470 (SD = 478,418). We excluded participants with missing values for each RQ separately (see Figure 1). Among participants diagnosed with dementia or MCI, the proportion of rural residence during either childhood or adulthood was slightly higher than that in the entire sample. Distributions of rural–urban status and residential mobility were similar in the weighted population representing U.S. adults 65 years of age and older (Table S1).

TABLE 2.

Descriptive statistics based on unweighted analytic sample.

	N (%) or mean (SD)
Total	3073
Categorical characteristics, N (%)
Childhood rural–urban status
Urban	1525 (49.6)
Rural	1349 (43.9)
Missing	199 (6.5)
Adulthood rural–urban status
Urban	2127 (69.2)
Rural	904 (29.4)
Missing	42 (1.4)
Residential mobility
Urban‐to‐rural	302 (9.8)
Rural‐to‐urban	776 (25.3)
Rural‐to‐rural	547 (17.8)
Urban‐to‐urban	1210 (39.4)
Missing	238 (7.7)
Diagnosis
Normal	2070 (67.4)
Mild cognitive impairment	678 (22.1)
Dementia	325 (10.6)
Sex
Male	1233 (40.1)
Female	1840 (59.9)
Race
White	2471 (80.4)
Black	518 (16.9)
Other	83 (2.7)
Missing	1 (0.0)
Ethnicity
Non‐Hispanic	2923 (95.1)
Hispanic	149 (4.8)
Missing	1 (0.0)
Marital status
Married, spouse absent/separated/divorced/widowed/never married	1373 (44.7)
Married/partnered	1698 (55.3)
Missing	2 (0.1)
Respondent/informant
Respondent	2946 (95.9)
Informant	127 (4.1)
Continuous characteristics, mean (standard deviation)
Age, years	76.5 (7.6)
Education, years	13.0 (2.7)
Parental education, years	9.6 (3.8)
Wealth, thousand dollars	157.5 (478.4)
Childhood socioeconomic status index	0.2 (0.9)

3.1. Total associations with childhood rural‐urban conditions (RQ1)

People who lived in rural areas as a child had significantly lower memory (standardized coefficient [beta] = –1.68 [95% confidence interval (CI): –2.39, –0.97]), executive function (beta = –1.20 [95% CI: –1.84, –0.55]), and language fluency scores (beta = –0.86 [95% CI: –1.64, –0.08]) (Figure 3 and Table S2). There were no statistically significant effects of childhood rural–urban status on visuospatial, orientation place, and orientation time scores.

FIGURE 3.

Forest plot of effect sizes for the associations between childhood rural–urban residence and adulthood rural–urban residence (with and without adjustment for childhood rural–urban residence) and cognitive function/status for (A) continuous cognitive function scores; (B) binary cognitive function scores; and (C) cognitive status (mild cognitive impairment [MCI], dementia, and MCI or dementia). Effects are coded in the direction of rural; reference groups are childhood and adulthood urban residences. For domain scores on a binary scale, the odds of obtaining full scores were modeled.

Regarding cognitive status, compared with participants who lived in urban areas during childhood, participants who lived in rural areas had a higher odds of being diagnosed with MCI (OR = 1.31 [95% CI: 1.02, 1.69]), or either dementia or MCI (OR = 1.30 [95% CI: 1.04, 1.62]) during later life (Figure 3 and Table S2). Childhood rural–urban residence was not associated with dementia‐only diagnosis.

3.2. Associations with adulthood rural–urban conditions (RQ2)

The associations between adulthood rural–urban status and cognitive functions and status were similar with and without adjustment for childhood rural–urban status in the model (Figure 3 and Table S3), indicating no influential confounding effects of childhood rural–urban status in this study. Considering that childhood residence is a known confounder based on mechanisms (Figure 2), we interpret the effects from the adulthood models that control for childhood residence. There were no significant associations with adulthood rural–urban status for either cognitive function domain scores or diagnosis of dementia or MCI (Figure 3 and Table S3).

3.3. Associations with rural–urban mobility from childhood to adulthood (RQ3)

Based on the results for childhood and adulthood rural–urban residence, we considered urban‐to‐urban as the reference group for the residential mobility variable.

3.3.1. Total associations with residential mobility

Compared to individuals who stayed in urban areas, individuals moving from childhood urban to adulthood rural had lower levels of executive function (beta = –1.73 [95% CI: –2.78, –0.68]) and lower odds of visuospatial functioning (OR = 0.46 [95% CI: 0.30, 0.70]) (Figure 4 and Table S4A).

FIGURE 4.

Forest plot of effect sizes for the associations between residential mobility (with and without adjustment for confounders with unknown temporal orders) and cognitive function/status for (A) continuous cognitive function scores; (B) binary cognitive function scores; and (C) cognitive status (mild cognitive impairment [MCI], dementia, and MCI or dementia). The reference group for all comparisons is urban‐to‐urban. For domain scores on a binary scale, the odds of obtaining full scores were modeled. SES, socioeconomic status.

Compared to individuals who stayed in urban areas, individuals moving from childhood rural to adulthood urban had lower memory scores (beta = –1.87 [95% CI: –2.74, –1.00]), executive function (beta = –1.29 [95% CI: –2.08, –0.50]), and language fluency (beta = –0.98 [95% CI: –1.93, –0.03]), and a lower odds of full orientation time scores (OR = 0.76 [95% CI: 0.58, 0.99]). Individuals moving from rural to urban also had a higher odds of being diagnosed with dementia (OR = 1.56 [95% CI: 1.05, 2.32]), MCI (OR = 1.40 [95% CI: 1.03, 1.92]), or either dementia or MCI (OR = 1.44 [95% CI: 1.10, 1.89]).

Compared to individuals who stayed in urban areas, individuals who stayed in rural areas had lower memory scores (beta = –1.78 [95% CI: –2.77, –0.78]), executive function (beta = –2.05 [95% CI: –2.95, –1.14]), and language fluency (beta = –1.18 [95% CI: –2.26, –0.10]). Individuals who stayed in rural areas also had higher risk for MCI (OR = 1.45 [95% CI: 1.03, 2.05]).

3.3.2. Roles of covariates with unknown temporal orders

Generally, when we adjusted for covariates with unknown temporal orders (i.e., adulthood education, adulthood wealth, marital status, and childhood SES), we observed similar estimates, in terms of effect size magnitude and statistical significance, for associations between residential mobility and the cognitive status/diagnosis outcomes. However, we observed some inconsistencies between residential mobility and cognitive function domains when we accounted for covariates with unknown temporal orders (Figure 4 and Table S4B). Specifically, the effects of urban‐to‐rural on visuospatial function and rural‐to‐urban on memory scores remained statistically significant. However, the effects of urban‐to‐rural on executive function; effects of rural‐to‐urban on executive function, verbal fluency, visuospatial function, and orientation time scores; and the effects of rural‐to‐rural on memory, executive function, and verbal fluency all attenuated toward the null and became non‐significant, but did not change in direction. These results suggest that there are likely complex and temporally‐ordered associations between residential mobility, childhood SES, adulthood education, wealth, marital status, and cognitive function domains across the life course. The effects of residential mobility on cognitive status/diagnoses were unaffected.

3.4. Sensitivity analysis

The estimates from the primary models did not differ considerably in magnitude, direction, or statistical significance in all of the sensitivity analyses (Figures S3–S7).

4. DISCUSSION

Overall, our results suggest that childhood rural residence is associated with lower levels of cognitive function and a higher risk of cognitive impairment in later life, and that these effects of childhood rural residence on later‐life cognitive outcomes persist even for individuals who moved to urban areas in adulthood. The associations between adulthood rural residence and cognitive functions were null. In terms of the association between rural–urban mobility and late‐life cognitive outcomes, individuals who lived in rural areas at any time in the life course (including those maintaining rural residence, moving from urban to rural, and moving from rural to urban from childhood to adulthood) experienced worse cognitive outcomes compared to those who lived in urban areas throughout the life course.

To date, very few studies have applied a life‐course approach to understand how the impact of rural–urban residence unfolds from early life to old age to affect later‐life cognitive health, including cognitive functions and dementia risk. Generally, the present study results are consistent with Socioecological Models of Health ¹⁶ and Life Course Theory, ¹⁷ which suggests that later‐life cognitive health is impacted by individuals’ geographic contexts throughout the lifespan. More specifically, the present findings align most closely with hypotheses posited in the sensitive period and cumulative disadvantage models of human development, ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² such that living in a rural area as a child, or the cumulative disadvantage of living in a rural area at any point in the lifespan, confers risk for worse cognitive health later in life. This is potentially because rural areas have fewer resources (e.g., libraries, schools) to support healthy cognitive development in early life, which can then lead to lifelong disparities in cognitive health. Likewise, rural areas also have less access to health care services and other healthy amenities, which can increase risk for cognitive health problems if experienced at any point in the lifespan.

The present results also align with some prior research in the U.S. context, showing that individuals who resided in rural environments during their early life tend to have worse cognitive outcomes in later life. ²⁵ , ³⁹ A study examining the association between life‐course rural residence and later‐life cognitive function in a Northern California community sample of medically‐insured older adults found that after adjusting for time‐varying socioeconomic status, childhood rural residence was associated with worse later‐life executive function and verbal episodic memory. ²⁵ Another study using the Wisconsin Longitudinal Study sample found that the effect of early‐life rural residence on later‐life cognitive health persists after adjusting for socioeconomic and educational factors. ³⁹ The magnitude of the associations found in the present study are in close agreement with prior work after accounting for the difference in SD units.

Furthermore, our study findings are consistent with prior research in different contexts around the globe, which is vital for understanding the generalizability of rurality as a determinant of cognitive decline and dementia. ⁴⁰ Although urbanization has been slowing in the United States for a long time, rural residency is consistently shown to be associated with poorer cognitive function in countries experiencing faster urbanization. For example, a study in India found that rural residence, especially during childhood, is associated with cognitive impairment, and that individuals who relocated to urban areas during adulthood or late life demonstrated a lower likelihood of developing cognitive impairment, compared with those maintaining rural residence. ⁴¹ Research in Mexico indicates that early‐life and late‐life urban residence can result in cognitive advantages for older adults. ⁴² A Chinese study found similar patterns and further demonstrated that rural–urban disparities in cognitive performance from 2008 to 2018 are narrowing due to rapid economic development and reform of the rural health system. ⁴³

Taken together, the existing evidence underscores the robustness of rurality as a risk factor for worse cognitive health. It is possible that rurality confers risk for worse cognitive health because of a lack of resources and healthy amenities in the environment, variations in social networks, environmental pollutants due to agriculture, changes in wealth, levels and quality of education attainment, and access to health care.

4.1. Strengths and implications

Our study is among the first to examine the association of life‐course rural–urban residence and residence mobility with late‐life cognitive functions in a nationally‐representative sample of older adults in the United States. We focus on various domains of both cognitive function and impairment, including memory, executive function, visuospatial ability, orientation, and MCI and dementia diagnosis. The assessment of cognitive function involved tasks and tests that do not necessitate literacy or mathematical skills, thereby minimizing bias associated with education in measuring cognitive function.

Our study has important implications. Understanding rural–urban disparities in a slowly progressive disease like cognitive impairment and dementia can inform when in the lifespan it is important to intervene to prevent or slow the likelihood of cognitive health problems. The present study suggests that the prevention/intervention strategies may be most effective in early life and for anyone who lives in a rural community across the lifespan. For example, it is important to increase access to health‐promoting resources such as educational and occupational opportunities, as well as health care access in rural areas. Furthermore, health care providers should advocate for early screening for cognitive impairment and other health conditions such as depressive disorders, stroke, and diabetes among older adults residing in rural areas. These health conditions are not only common but also act as risk factors for cognitive decline and dementia. The earlier timing of prevention/intervention strategies may be able to narrow the gap in cognitive health disparities between rural and urban areas, addressing the multifaceted challenges associated with cognitive health in diverse communities.

4.2. Limitations

Our study has several limitations. First, although we used nationally‐representative data of older adults with a cohort study design, we were only able to obtain early‐life measures retrospectively, which may result in potential recall bias. Second, there may be some degree of measurement error for self‐reported rural residence in childhood because rural contexts are not homogeneous, and people may have different conceptualizations of what is considered rural versus urban. These differences may also vary across birth cohorts and the degree of urbanization of the community and surrounding communities. Third, the data on rural–urban residence is limited to two points—one in childhood and one in adulthood—rather than a more detailed trajectory across multiple time points in the lifespan. Consistent with prior work, ⁴⁴ , ⁴⁵ , ⁴⁶ we chose to use only one timepoint of rural–urban residential data in adulthood because most adults in the HRS sample either do not move, or if they do move, they do not move to notably different residential contexts. For example, 92.3% of the present analytic sample did not change in their residential contexts across the five waves (8 years) leading up to the study endpoint (2016). Given that this sample comprises primarily middle and older adults, it is not surprising to see such residential stability, as geographic mobility often occurs during the young adulthood years (if at all), which we were not able to capture in this study. However, this limitation is unlikely to drastically impact our results because prior studies indicate that, within each developmental stage, people in the United States rarely transition between rural and urban areas, ⁴⁴ and HRS participants tend to have stable rural–urban residence in mid‐life and late‐life. ⁴⁵ , ⁴⁶ Fourth, using dichotomous rural–urban residence status may not fully describe the rural–urban continuum. Administrative data and geocoding may be helpful in gaining a more comprehensive understanding of rurality and urbanicity. As data linkage becomes more common, such as census linkage with the HRS, it is possible to obtain more detailed data on the features of rural and urban residential experiences. Fifth, although HRS was intended to be a representative sample of U.S. adults 55+ years of age, and we used sampling weights to account for differential sampling and participation rates by demographic factors, the present study does notably underrepresent certain demographic subgroups who live in rural areas, such as Hispanic/Latino and Native American populations. As such, the present effects may be underestimated due to limits on the populations represented in the data. Finally, although our study effectively managed to control for potential confounders, the results may still be affected by unmeasured variables and residual confounding due to information availability. In addition, future research would benefit from identifying and modeling the mediating mechanisms, such as health care access, that link rural–urban disparities and cognitive health.

4.3. Conclusion

Our findings indicate that rural residence at any point in life, especially during childhood, is associated with worse cognitive outcomes in later life among older adults in the United States. Public health policies and interventions aimed at reducing the risk of cognitive decline and dementia should consider the unique challenges faced by individuals who grew up in rural environments, such as economic and educational disadvantages, as well as behavioral and health risks. Further research is needed to delve into the specific aspects of rural residence, and the downstream socioeconomic factors and health condition factors that contribute to cognitive decline, which could guide comprehensive, lifespan prevention and intervention strategies.

CONFLICT OF INTEREST STATEMENT

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

CONSENT STATEMENT

All participants provided consent to participate in the Health and Retirement Study.

DEI STATEMENT

This study uses data from a large, nationally‐representative panel study of adults 55+ years of age in the United States. Although the goal of the Health and Retirement Study (HRS) was to obtain a diverse and inclusive sample of participants from a wide range of gender, racial, ethnic, and socioeconomic backgrounds that is representative of the U.S. population, the sampling may be biased by the study design (e.g., four‐stage area probability sampling to recruit participants), execution (e.g., participants from marginalized backgrounds are more likely to drop out; study measures were selected without input from participants), and interpretation (e.g., the study authors of this paper do not share all demographic backgrounds of the participants in this study, which may affect how we interpret the results). In addition, the HRS sample notably underrepresents certain demographic subgroups (e.g., Hispanic/Latino; Native American; people who are not heterosexual). To ensure that diversity, equity, and inclusivity are acknowledged in the conduct of this research, we: (1) use sampling weights to account for differential sampling and participation rates by demographic factors and (2) consider the limits on generalizability of the present findings in the Discussion section.

Supporting information

Supporting Information

Supporting Information

Chu L, Wu Y, Karjalainen H, Atherton OE. Lifespan exposures to rural–urban conditions and later‐life cognitive function. Alzheimer's Dement. 2025;21:e70267. 10.1002/alz.70267