Exploring the Relationship Between Engagement in Enriching Early-Life Activities During Adolescence and Cognition in Later-Life: Results from the Health and Retirement Study

Abstract

The cognitive benefits associated with mid- to late-life engagement have been demonstrated in several studies. However, the link between engagement in enriching early-life activities (EELAs) during adolescence and later-life cognition has been relatively unexplored in major epidemiological studies. We examined the EELA-cognition relationship in a nationally representative sample of adults aged 50+. A subset of Health and Retirement Study respondents (n=3482) completed cognitive tests and returned a retrospective early-life activity inventory. Linear regression models analyzed the EELA-cognition relationship, and multiple imputation addressed missingness. Each additional EELA was associated with a 0.36 point higher cognitive score (95% CI: 0.24, 0.47). This relationship remained significant after adjusting for potential confounders (B=0.16; 95% CI: [0.06, 0.26]). EELA engagement was associated with better later-life cognitive performance. This study is understood to be the first to examine the EELA-cognition relationship using a large, nationally representative dataset. The findings highlight the importance of early-life engagement during an important developmental period (e.g., adolescence).

As the population continues to age and older adults are soon expected to outnumber children for the first time in United States history (Vespa et al., 2020), identifying modifiable risk factors that may delay or prevent the onset of cognitive decline has become an urgent public health priority. The engagement hypothesis posits that participating in physical, social and intellectually enriching activities acts as a buffer against cognitive decline (Bielak, 2010; Hultsch et al., 1999). This hypothesis aligns with the cognitive reserve theory that attempts to explain inter-individual variability between Alzheimer’s disease and related dementias (ADRD) pathology and the clinical symptoms expressed from these brain changes (Stern, 2012). The cognitive reserve theory proposes that the clinical expression of pathology varies due to one’s experiential resources across the lifespan, such as environmental exposures, educational attainment, or occupation (Stern, 2009, 2012; Tucker & Stern, 2011).

The engagement hypothesis is supported by positive associations between engagement in enriching lifestyle activities (e.g., social interactions, singing or playing music, physical activity, drawing, and crossword puzzles) across mid- to late-life and better cognitive outcomes (Arbuckle et al., 1998; Carlson et al., 2012; Craik et al., 2010; Erickson et al., 2019; James et al., 2011; Kensinger & Gutchess, 2017; Parisi et al., 2012; Stern, 2006; Vemuri et al., 2014; Wilson et al., 2012). These associations are intriguing but incomplete as they do not explore the potential impact of enriching early-life activities (EELAs) during adolescence and subsequent cognition in later-life.

Previous studies that explored the EELA-cognition relationship characterized early-life engagement by deriving a composite measure based on (1) the number of EELAs endorsed prior to age 13 (Chan et al., 2019; Moored et al., 2018; Morris, Ai, et al., 2021; Morris, Chaddock-Heyman, et al., 2021) or (2) various indicators of early-life enrichment such as childhood socioeconomic status, access to reading materials in the home at age 12, years of foreign language instruction, and frequency of involvement in engaging activities early in life (Oveisgharan et al., 2020; Wilson et al., 2005).

Older adults who engaged in more EELAs during childhood had greater educational attainment (Chan et al., 2019; Morris, Ai, et al., 2021), slower rates of cognitive decline (Oveisgharan et al., 2020), better late-life cognitive function (e.g., faster processing speeds, executive functioning) (Chan et al., 2019; Greenfield et al., 2021; Wilson et al., 2005). significantly larger amygdala volumes and hippocampal volumes in men (Moored et al., 2018), and greater functional connectivity (Morris, Chaddock-Heyman, et al., 2021). However, these conclusions are drawn from studies with limited sample sizes (n < 100) or limited sample generalizability (e.g., highly educated, limited racial and ethnic diversity). Additionally, none of the retrospective activity inventories mentioned above focus on activities carried out in high school (e.g., middle adolescence). Adolescence is a time of emerging independence, development, and exploration (Larsen & Luna, 2018). Yet, the relationship between EELAs during adolescence and cognition has been relatively unexplored. Reserve has been theorized to be cumulative across the life-span, suggesting that engagement in early-life activities should also be captured as this may play a key role in boosting or maintaining reserve (Scarmeas & Stern, 2003; Schreiber et al., 2016; Stern, 2006; Wang et al., 2017).

Using the engagement hypothesis as our guiding framework, we took a life course approach and assessed whether engagement in enriching lifestyle activities during middle adolescence, a period of development relatively unexplored in the EELA-cognition literature, is associated with higher levels of cognition in later-life. Data were analyzed from a randomly sampled subset of participants from the Health and Retirement Study (HRS) to examine the EELA-cognition relationship. We hypothesized that better cognitive performance in later-life would be associated with greater EELA engagement during adolescence. If EELAs are positively related to cognitive performance in later-life, they may represent a modifiable lifestyle factor that could be targeted in interventions and incorporated into public policy to promote cognitive health and potentially delay the onset of, reduce adverse outcomes from, or mitigate the costs associated with ADRD in later-life (Alzheimer’s Association, 2021; Sperling et al., 2011).

Research Design

Study Sample

The HRS is sponsored by the National Institute on Aging and directed by the University of Michigan’s Institute for Social Research. Starting in 1992, the HRS has collected data on multiple health, retirement, and aging components in the United States (Juster & Suzman, 1995). The study longitudinally tracks a nationally representative sample of approximately 20,000 Americans over 50 years of age. Evaluations involving age-eligible respondents and their spouses occur every 2 years to assess several aspects of aging, including but not limited to socioeconomic factors (e.g., income and health insurance), physical function, and cognitive performance.

Demographic and cognitive data were obtained from the RAND HRS Longitudinal File 1992–2018, Version 1 (Health and Retirement Study, 2021; RAND HRS Longitudinal File 1992–2018 Version 1, 2021). The RAND HRS Longitudinal File, funded by the National Institute on Aging and the Social Security Administration, is an easy-to-use dataset based on the HRS core data. Given the data were de-identified and publicly available, the Institutional Review Board waived human subjects research approval.

In the spring and fall of 2017, the HRS distributed 10,352 Life History Mail Surveys (LHMS) to a random sample of HRS respondents; 5288 questionnaires were returned (Health and Retirement Study, 2020). The LHMS consisted of 63 questions regarding participants’ childhood experiences (e.g., residential history, education, employment, and medical history). In particular, the survey included a 10-item retrospective inventory assessing whether respondents engaged in enriching lifestyle activities during adolescence (e.g., singing, debate club, and crafting).

Certain criteria had to be met for inclusion into the analytic sample. First, individuals had to have cognitive summary scores available during Wave 10 (2010) of the HRS and had to return the LHMS (Figure 1). Wave 10 (2010) was selected because it contained the largest sample size of LHMS respondents aged 50+ with cognitive data present (n=3482). Longitudinal analysis incorporating all waves of data collection is currently underway.

Figure 1.

Analytic sample flowchart

Measures

Enriching Early-Life Activities (EELAs)

The 2017 LHMS incorporated 10 retrospective questions asking whether respondents participated in cognitively enriching tasks during adolescence (Table 1). The EELA inventory targeted activities often performed outside of the home. The specific wording of the questionnaire was as follows: “In high school, did you take classes or spend time to do the following: (1). learn to play a musical instrument, (2). take singing lessons or sing in a chorus or choir, (3). learn woodwork or carpentry, (4). learn a craft (e.g., knitting, quilting, or embroidery), (5). learn ballet or dance, (6). learn to paint or draw or other art, (7). participate in math or science club, (8). learn drafting or technical drawing, (9). take vocational or trade classes (e.g., auto repair, HVAC), and (10). participate in theater, drama, or debate club.” The available answer choices were “Yes” and “No.” The EELA indicator was coded, similar to previous studies, as a sum of the EELAs endorsed during adolescence. Values ranged from 0 to 10, with a higher score indicating greater engagement in EELAs during adolescence (Chan et al., 2019; Moored et al., 2018; Morris, Ai, et al., 2021). Unlike previous studies, EELA engagement was also categorized as a dichotomous variable, where 1 = performed any EELA and 0 = did not perform any EELAs.

Table 1.

Survey-Weighted Demographic and EELA Characteristics of Analytic Sample

Characteristic	Mean (SE) or %	% Missing
Age	64.6 (0.4)	0
Total Wealth		0
≤ $17,040	20.4
$17,041 - $132,000	23.2
$132,001 - $418,500	26.5
> $418,500	29.9
Gender		0
Male	44.5
Female	55.5
Father’s Education	10.0 (0.2)	15.1
Mother’s Education	10.3 (0.1)	8.4
Education		0
Less than High School	14.9
GED	4.2
High School Graduate	28.1
Some College	25.9
College and Above	27.0
Race/Ethnicity		0.2
Non-Hispanic White	75.0
Non-Hispanic Black	9.7
Hispanic	12.6
Other	2.7
Cohort by Birth Year		0
AHEAD	1.5
CODA	6.7
HRS	30.7
War Babies	18.0
Early Baby Boomers	5.6
Mid Baby Boomers	37.6
Smoking Status		0.4
Never Smoker	44.7
Former Smoker	42.4
Current Smoker	13.0
Diabetes Status		0.03
No	80.5
Yes	19.5
Hypertension		0.1
No	47.7
Yes	52.3
Alcohol Intake (# of drinks/day)	0.4 (0.03)	0.3
Body Mass Index (kg/m2)	28.6 (0.1)	2.3
Depressive Symptoms	1.2 (0.04)	0
Region Born		0.03
South	28.5
Northeast	20.5
Midwest	27.5
West	7.7
Other	15.8
Number of EELAs Endorsed (# out of 10)	2.0 (0.04)
Percentage Endorsing ≥ 1 EELA	79.1
Percentage of Specific EELAs Endorsed
Learn to play a musical instrument	26.9	12.9
Take singing lessons or sing in a chorus or choir	30.9	13.3
Learn woodwork or carpentry	24.3	13.6
Learn a craft (e.g., knitting, quilting, embroidery)	21.2	13.5
Learn ballet or dance	7.5	13.7
Learn to paint or draw or other art	29.1	13.5
Participate in math or science club	8.9	13.7
Learn drafting or technical drawing	16.4	13.9
Take vocational or trade classes (e.g., auto repair, HVAC)	14.3	13.6
Participate in theatre, drama, or debate club	21.4	13.6

Note. n=3482; AHEAD, Asset and Health Dynamics Among the Oldest Old; CODA, Children of the Depression; HRS, Health and Retirement Study; EELAs, Enriching Early-Life Activities; HVAC, Heating Ventilation and Air Conditioning.

Cognition score from Wave 10 (2010)

The HRS cognitive battery was an adapted version of the Telephone Interview for Cognitive Status (TICS) (Brandt et al., 1988). Scores ranged from 0–35 and consisted of several tests across different cognitive domains. Immediate and delayed (5 minutes) recall of a word list assessed memory and contributed up to 20 points to the cognitive score. The Serial 7’s test asked participants to subtract seven from 100 in a serial fashion across five trials (i.e., 100, 93, 86, 79, and 72) to assess working memory; scores ranged from 0–5 points. The backward count test evaluated attention and processing speed. Participants received two points if they successfully counted backwards for 10 consecutive numbers correctly on their first attempt; one point was awarded if they answered correctly on the second attempt, and zero points otherwise.

On the naming task to assess language, participants identified two objects. For the orientation task, participants recalled the current President, Vice President, month, day, year, and day of the week. Each correct answer received one point leading to a maximum of eight possible points. The orientation and naming questions were only asked of newly-enrolled participants or individuals 65 and older. Said differently, respondents underwent the full cognitive battery at enrollment and each subsequent visit upon reaching 65 years of age (Ofstedal et al., 2005).

Covariates

This study adjusted for demographic, socioeconomic, and behavioral factors thought to be associated with EELA engagement and cognition. Covariates included age, race/ethnicity (re-coded as non-Hispanic white, non-Hispanic Black, Hispanic, and non-Hispanic other), gender (male or female), cohort by birth year (see Supplemental Material), and self-reported father’s and mother’s education (in years). Parental education was incorporated to characterize opportunities for engagement and estimate the participant’s early-life socioeconomic environment (Dubow et al., 2009). We also adjusted for total wealth which represents the accumulation of economic resources throughout the life course (Semyonov et al., 2013), region of birth (South, Northeast, Midwest, West, and Other – U.S. Territories or International), smoking status (never, former, current), diabetes status (yes, no), hypertension (yes, no), body mass index (BMI; kg/m²), alcohol intake (# of drinks/day), and the number of self-reported depressive symptoms as measured by the eight-item version of the Center for Epidemiological Studies - Depression scale (CESD-8) (Karim et al., 2015; Turvey et al., 1999).

Statistical Analysis

To assess the relationship between EELAs and later-life cognition, three linear regression models were generated using the 2010 HRS cognitive score as the outcome of interest. We implemented a staged analysis approach, starting with the simplest (unadjusted) model and incorporating additional covariates into each subsequent model. Model 1 consisted of a simple linear regression with the EELA indicator (count and dichotomous) as the predictor variable. Model 2 further adjusted for age (mean centered at 65), gender, race/ethnicity, father’s and mother’s education (mean centered), and cohort by birth year. These covariates have been used previously to account for potential confounding of the EELA-cognition relationship (Chan et al., 2019; Moored et al., 2018; Wilson et al., 2005).

Additional covariates of total wealth, smoking status, diabetes status, BMI, alcohol intake, hypertension status, depressive symptoms, and geographic region born were incorporated into the fully adjusted model to further address potential confounding (Model 3). Educational attainment was not fully complete when participants performed the EELAs; therefore, education was excluded from the model, given its role as a potential mediator along the EELA-cognition pathway (Morris, Ai, et al., 2021). Model assumptions were examined with residual versus fitted, Q-Q, scale-location, and residual versus leverage plots.

Due to the complex survey design and oversampling of certain demographic groups to accurately represent immigrants and racial/ethnic minorities (Health and Retirement Study, 2008), respondent-level weights were applied to allow for population-level inferences (Health and Retirement Study, 2019) using the survey package in StataMP 14 (StataCorp., 2011, 2015). Survey-weighted demographic characteristics of the analytic sample were then compared to the values from the 2010 United States Census to ensure the randomly sampled subset of HRS respondents remained nationally representative (US Census Bureau, 2020). A robust standard error estimator addressed the complex survey design. The mi impute command performed multiple imputation with chained equations (MICE) to address EELA item non-response and missing covariates (Schafer & Olsen, 1998). The outcome of interest was not imputed (Von Hippel, 2007). All analytic variables were incorporated into the imputation model to produce 30 imputed datasets.

Results

Individuals who engaged in EELAs were less likely to report a diagnosis of hypertension; they were also more likely to be non-Hispanic white and reported fewer depressive symptoms, greater wealth, higher educational attainment, and greater parental education compared to individuals who did not engage in any EELAs. After applying respondent-level weights, the weighted sample was 56% female and 75% non-Hispanic white (Table 1). Prior to imputation, the variable with the highest percentage of missingness (15.1%) was self-reported father’s education (Table 1). The mean age was 64.6 (SE = 0.37) years. Approximately 79% endorsed at least 1 EELA during adolescence and two of 10 early-life activities were reported, on average. Respondents with higher education were more likely to report engaging in EELAs during adolescence and had higher levels of later-life cognition.

When expressing EELA as a count (Table 2), each additional EELA was associated with an increase in cognitive performance of 0.36 points (95% CI: 0.24, 0.47) (Model 1). After adjusting for age, gender, race/ethnicity, father’s and mother’s education, and cohort by birth year, a significant association remained between EELA count and late-life cognition (B = 0.18; 95% CI: [0.08, 0.29]) (Model 2). This association remained significant in the fully adjusted model (B = 0.16; 95% CI: [0.06, 0.26]) (Model 3).

Table 2.

Imputed, Survey-Weighted Models, Cognitive Score vs. EELA Count (Coefficient [95% CI])

	Model 1	Model 2	Model 3
# of EELAs	0.36***[0.24,0.47]	0.18***[0.08,0.29]	0.16**[0.06,0.26]
Age, Centered at 65		−0.19***[−0.25,−0.13]	−0.20***[−0.26,−0.14]
Gender (ref: Male)		0.56**[0.16,0.96]	0.67**[0.26,1.08]
Race/Ethnicity (ref: Non-Hispanic White)
Non-Hispanic Black		−2.85***[−3.44,−2.27]	−2.06***[−2.62,−1.50]
Hispanic		−1.96***[−2.67,−1.25]	−1.30**[−2.13,−0.46]
Other		−2.19***[−3.09,−1.29]	−1.79***[−2.67,−0.92]
Father’s Education, Mean Centered		0.11***[0.05,0.18]	0.09**[0.02,0.15]
Mother’s Education, Mean Centered		0.13***[0.06,0.20]	0.10**[0.04,0.17]
Cohort (ref: HRS)
AHEAD		1.46*[0.01,2.91]	1.38[−0.15,2.90]
CODA		0.20[−0.59,0.98]	0.18[−0.56,0.92]
War Babies		−0.51[−1.24,0.21]	−0.61[−1.31,0.10]
Early Baby Boomers		−2.02***[−2.88,−1.16]	−1.65***[−2.49,−0.81]
Mid Baby Boomers		−2.82***[−4.00,−1.65]	−2.55***[−3.76,−1.35]
Total Wealth (ref: ≤ $17,040)
$17,041-$132,000			0.79**[0.23,1.35]
$132,001-$418,500			1.13***[0.68,1.58]
> $418,500			1.68***[1.21,2.15]
Smoking Status (ref: Never)
Former Smoker			−0.36*[−0.68,−0.04]
Current Smoker			−1.06***[−1.65,−0.48]
Diabetes Status (ref: No)
Yes			−0.45*[−0.80, −0.09]
Body Mass Index (kg/m2)			0.02[−0.02,0.04]
Alcohol Intake (drinks/day)			−0.04[−0.25,0.18]
Hypertension (ref: No)
Yes			−0.19[−0.52,0.13]
Depressive Symptoms			−0.32***[−0.43,−0.20]
Region Born (ref: South)
Northeast			0.56*[0.09,1.02]
Midwest			0.20[−0.20,0.61]
West			0.03[−0.52,0.57]
Other			−0.15[−0.72,0.42]
Constant	22.5***[22.2,22.8]	24.1***[23.5,24.8]	23.2***[21.9,24.5]

Note. n=3482; 30 imputations; CI, Confidence Interval; EELAs, Enriching Early-Life Activities; HRS, Health and Retirement Study; AHEAD, Asset and Health Dynamics Among the Oldest Old; CODA, Children of the Depression.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

When expressing EELA as a dichotomous variable (Table 3), any EELA engagement was associated with significantly higher cognitive scores (B = 1.34; 95% CI: [0.86, 1.81]) compared to no EELA engagement during adolescence. After adjusting for age, gender, race/ethnicity, father’s and mother’s education, and cohort by birth year, a significant association remained between EELA participation and late-life cognition (B = 0.69; 95% CI: [0.28, 1.10]). This association persisted after adjusting for total wealth, smoking status, diabetes status, BMI, alcohol intake, hypertension status, depressive symptoms, and geographic region born (B = 0.56; 95% CI: [0.15, 0.97]). Multiply imputed results were consistent with non-imputed results.

Table 3.

Imputed, Survey-Weighted Models, Cognitive Score vs. Binary EELA (Coefficient [95% CI])

	Model 1	Model 2	Model 3
Performed EELAs (ref: No)	1.34***[0.86,1.81]	0.69**[0.28,1.10]	0.56**[0.15,0.97]
Age, Centered at 65		−0.19***[−0.25,−0.14]	−0.20***[−0.26,−0.14]
Gender (ref: Male)		0.55**[0.15,0.94]	0.66**[0.26,1.06]
Race/Ethnicity (ref: Non-Hispanic White)
Non-Hispanic Black		−2.82***[−3.39,−2.24]	−2.03***[−2.58,−1.47]
Hispanic		−1.91***[−2.62,−1.21]	−1.27**[−2.09,−0.45]
Other		−2.14***[−3.05,−1.22]	−1.75***[−2.63,−0.87]
Father’s Education, Mean Centered		0.11***[0.05,0.18]	0.09**[0.02,0.15]
Mother’s Education, Mean Centered		0.13***[0.06,0.21]	0.11**[0.04,0.18]
Cohort (ref: HRS)
AHEAD		1.53*[0.08,2.99]	1.44[−0.08,2.96]
CODA		0.28[−0.50,1.06]	0.25[−0.48,0.99]
War Babies		−0.58[−1.31,0.15]	−0.66[−1.37,0.04]
Early Baby Boomers		−2.07***[−2.94,−1.21]	−1.69***[−2.53,−0.85]
Mid Baby Boomers		−2.92***[−4.10,−1.75]	−2.64***[−3.84,−1.45]
Total Wealth (ref: ≤ $17,040)
$17,041-$132,000			0.78**[0.23,1.33]
$132,001-$418,500			1.13***[0.68,1.58]
> $418,500			1.68***[1.21,2.14]
Smoking Status (ref: Never)
Former Smoker			−0.37*[−0.68,−0.05]
Current Smoker			−1.06***[−1.65,−0.47]
Diabetes Status (ref: No)
Yes			−0.45*[−0.81,−0.09]
Body Mass Index (kg/m2)			0.01[−0.02,0.04]
Alcohol Intake (drinks/day)			−0.03[−0.24,0.18]
Hypertension (ref: No)
Yes			−0.18[−0.52,0.15]
Depressive Symptoms			−0.31***[−0.43,−0.20]
Region Born (ref: South)
Northeast			0.55*[0.09,1.01]
Midwest			0.21[−0.19,0.61]
West			0.05[−0.47,0.57]
Other			−0.13[−0.69,0.43]
Constant	22.1***[21.7,22.6]	24.0***[23.3,24.7]	23.1***[21.9,24.4]

Note. n=3482; 30 imputations; CI, Confidence Interval; EELAs, Enriching Early-Life Activities; HRS, Health and Retirement Study; AHEAD, Asset and Health Dynamics Among the Oldest Old; CODA, Children of the Depression.

^*p < 0.05

^**p < 0.01

^***p < 0.001

Sensitivity Analyses

To further examine the relationship between EELAs and later-life cognition, we excluded 414 respondents who returned the EELA inventory despite reporting that they did not attend high school in the LHMS questionnaire (Supplemental Material). Results were consistent with the main analysis. We performed an additional sensitivity analysis excluding 1370 individuals younger than 65 to better capture the relationship between EELAs and cognition in older adults. Results were similar to the primary findings (Supplemental Material). Although the main focus of this study was to explore associations between EELAs and global cognition, a third sensitivity analysis was performed to explore which, if any, of the five tests from the cognitive battery were driving the significant associations between EELAs and later-life cognition. Results showed that the EELA-cognition relationship was statistically significant for immediate and delayed recall tasks but not for the serial 7’s, backward counting, or naming/orientation tasks (Supplemental Material).

Discussion

In this study, we sought to explore the relationship between engagement in EELAs during adolescence and cognitive performance in later-life. Engagement in enriching early-life activities during adolescence was associated with better cognitive performance in later-life, independent of various demographic, socioeconomic, and behavioral factors. Our study supports the engagement hypothesis and expands upon recent studies of EELA by (1) focusing on middle adolescence, a life course period relatively unexplored in existing retrospective early-life activity inventories and (2) using a large, nationally representative dataset. The retrospective approach used to capture EELAs is commonly used to explore associations between adverse childhood experiences (ACEs) and health outcomes. (Schickedanz et al., 2021). We propose that investigating both ACEs and EELAs is essential to fully capture the breadth of early-life exposures and how these factors may be related to long-term health outcomes (Schickedanz et al., 2021).

These findings also contribute to the growing literature on potential factors associated with cognitive reserve and resilience. Quantifying cognitive reserve relies primarily on educational attainment as an early-life surrogate measure (Stern, 2002). Our findings suggest that we can extend beyond self-reported years of formal education and measure the benefits of early-life extracurricular activities among older adults to supplement existing markers of cognitive reserve (Chan et al., 2019; Moored et al., 2018; Morris, Ai, et al., 2021). To better understand the EELA-cognition relationship, future nationwide epidemiological surveys can administer brief, early-life retrospective questionnaires as an additional source of enriching behavior to complement measures of formal education.

Many potential mechanisms could explain the positive associations found between EELA and cognition. First, EELAs may enhance existing network efficiency, capacity, or flexibility through features that promote plasticity (e.g., neurogenesis and synaptogenesis) or cognitive reserve (e.g., promote educational attainment and social engagement) in brain regions thought to stimulate networks responsible for emotion regulation and executive functioning (Moored et al., 2018; Morris, Chaddock-Heyman, et al., 2021; Noonan et al., 2018; Stern, 2009). Additional research should be carried out to explore which mechanism more closely aligns with the EELA-cognition relationship, particularly for individuals who did not attend high school but reported extensive engagement in EELAs as these individuals may also experience late-life cognitive benefits from EELAs (Chan et al., 2019; Hopper & Iwasaki, 2017; Moored et al., 2018). Incorporating EELA questionnaires into existing brain imaging studies could also enhance our understanding around the potential mechanisms that link EELAs to later-life cognition.

The current findings shed light on intervention and policy implications of a life-course approach to environmental stimulation for combatting and preventing age-related cognitive declines and dementia. Our results suggest that what is good for aging adults was likely also good for them as emerging, young adults, and engagement in enriching activities during adolescence may strengthen later-life cognitive performance. Delaying the diagnosis of dementia by just a few years can lead to many social and economic benefits (Zissimopoulos et al., 2014). Investing in programs that promote EELAs during adolescence may foster positive youth development and shift the growth, reserve, and resilience trajectories of at-risk youth. These programs may ultimately lead to cognitive benefits that extend into late-life, particularly among socially disadvantaged groups or in low-resource communities where individuals may be at higher risk of cognitive decline (Catalano et al., 2004; Chan et al., 2019; Moored et al., 2018).

Strengths

This study examined the association between EELAs performed during middle adolescence (e.g., ages 14–17), a period of the life course relatively unexplored in the EELA-cognition literature, and subsequent later-life cognitive performance. The use of a nationally representative dataset increased generalizability and allowed for population-level inferences. The robust sample size increased the statistical power and provided substantial precision for evaluating associations and adjusting for many covariates that may not be available in other epidemiological studies.

Multiple imputation techniques of the independent variables provided a larger analytic sample size, and results from the sensitivity analyses were consistent with the primary analysis. The EELA inventory included activities performed in different social settings (e.g., drama, dance, and woodwork) with varying levels of cognitive demand (e.g., music, crafts, and math club). Furthermore, by exploring the impact of EELA engagement in middle adolescence, the results provide implications for the optimal timing and venue of future interventions and prevention research.

Limitations

Some key limitations of this study must be considered. First, the analysis excluded all participants without cognitive data available in 2010, leading to potential selection bias favoring those who were in better health. There is also a potential for survival bias among participants who survived to 2017 and were willing to complete the EELA questionnaire. The 10 EELA questions do not encompass all of the enriching early-life activities that could have been performed during adolescence, and the activities are likely to be offered to participants from socioeconomically advantaged backgrounds who attended resource-rich high schools. In an effort to adjust for this potential confounder, we included geographic region born, mother’s education, and father’s education in the model as proxies for early-life socioeconomic status.

The EELA questionnaire was only administered to participants at one timepoint; thus, the reliability of EELA measures over time cannot currently be assessed. As with any retrospective survey, there may be recall bias for activities performed many decades ago. However, studies show that retrospective recall for early-life childhood experiences has good test-retest reliability (ICC > 0.90), strong concordance rates among siblings, and high accuracy, which suggests that retrospective EELA inventories are likely valid and reliable tools for capturing early-life enrichment (Berney & Blane, 1997; Krieger et al., 1998; Schreiber et al., 2016).

Given the specific wording of the HRS EELA questionnaire (e.g., “In high school, did you take classes or spend time to do the following:”), it is unclear whether individuals who did not attend high school but performed EELAs during adolescence experience positive cognitive outcomes in later-life. Previous studies suggest that EELAs are positively associated with cognitive performance among socio-demographically at-risk older adults (Chan et al., 2019; Moored et al., 2018). Also, the analysis was cross-sectional and thus cannot address whether EELA engagement is associated with slower rates of cognitive decline. Longitudinal analysis is currently underway.

There are additional measurement-oriented limitations. The reported cognitive score was not standardized and heavily weighted by memory components (i.e., 20/35 points of the cognitive score are determined by immediate and delayed recall). We also excluded late-life lifestyle activities given their role as a potential mediator and because the inclusion of this measure would dramatically impact the analytic sample size (e.g., psychosocial data are only collected from a 50% subsample of respondents during each evaluation wave) (Jacqui et al., 2017). Some of the covariates incorporated into the fully adjusted model (e.g., BMI, alcohol intake, and depressive symptoms) may act as partial mediators along the EELA-cognition pathway and attenuate associations between EELAs and later-life cognition. Lastly, brain imaging data were not available to explore whether behavioral changes were associated with structural or functional brain changes in our sample.

Future Directions

Future studies of cognitive aging should supplement existing retrospective lifestyle activity inventories to include EELAs and allow researchers to further investigate the relationship between these enriching activities in early-life and cognition in later-life. Following on later-life activity questionnaires (Carlson et al., 2012; Hultsch et al., 1999), adding frequency and duration questions to each EELA item would allow us to explore whether a higher dosage or intensity of EELA engagement is linked to better late-life cognitive performance. However, this level of detail may be subject to increased participant burden or recall bias. Most importantly, it would be beneficial for future EELA questionnaires to include individuals without a formal high school education to determine whether EELAs are related to late-life cognitive health for those at greater risk for ADRD (Livingston et al., 2020).

Conclusion

The positive associations between lifestyle activity engagement during mid- to late-life and cognition has been well established in the literature. However, little is known about the relationship between EELA engagement in adolescence and later-life cognition. Our findings support the engagement hypothesis and underscore the importance of offering EELAs to adolescents to promote cognitive health and inquiring about engagement in early life through retrospective surveys to further explore this research question. These findings may inform policy and intervention research to understand the public health implications and potentially buffer the effects of age-related cognitive declines and dementia in an aging population.

Biographies

Breanna M. Crane, BS: Ms. Crane is a doctoral student whose key research interests involve cognitive aging and how modifiable risk factors (e.g., community mobility, lifestyle engagement, enriching early-life activities) can potentially delay or prevent the onset of age-related cognitive declines and Alzheimer’s disease and related dementias (ADRD).

Karen Bandeen-Roche, PhD: Dr. Bandeen-Roche is a biostatistician known for her research on aging and aging-related frailty. She is currently the Hurley Dorrier Professor of Biostatistics and Chair of the Biostatistics Department at the Johns Hopkins Bloomberg School of Public Health.

Michelle C. Carlson, PhD: Dr. Carlson is a Professor in the Department of Mental Health at the Johns Hopkins Bloomberg School of Public Health and Core Faculty at the Center on Aging and Health. She examines relationships between modifiable lifestyle factors and risk for age-related cognitive/functional declines and Alzheimer’s disease.

Data Availability Statement

The Health and Retirement Study data used for this study are publicly available (Health and Retirement Study, 2020; RAND HRS Longitudinal File 1992–2018 Version 1, 2021).