BACKGROUND
Educational attainment has been extensively examined as a potential protective factor against late-life cognitive decline. In most studies of middle-aged and older adults, educational attainment is commonly assessed at a single time point (eg, study enrollment), assuming that education is generally completed by early adulthood (eg, before age 25) and is stable throughout later life. However, adult learners who follow “non-traditional” educational trajectories (eg, delayed enrollment in postsecondary education or those who did not receive a standard high school diploma) are becoming increasingly common in the United States. A cohort study of individuals who graduated high school in 1982 found that 35% of individuals with a high school diploma at age 28 attained higher educational credentials between the ages of 28 and 50 years. Another study showed that attaining at least a bachelor's degree after age 25 years (vs those who had a post-high school certificate or lower) was associated with beneficial health outcomes, such as fewer depressive symptoms and better self-rated health at midlife. These findings broadly suggest that educational attainment after the so-called traditional age may benefit later-life health.
However, higher educational attainment later in life may be less beneficial for health compared to those following a “traditional” educational trajectory (ie, enroll in college full-time immediately after high school graduation), as there are fewer years in the labor market to recoup investments in education (ie, less time to accumulate economic and social benefits, such as personal control, cognitive ability, and social support). Those who pursue non-traditional educational trajectories also tend to come from historically marginalized populations (eg, racial and ethnic minorities and women) and socioeconomically disadvantaged backgrounds, who are generally underrepresented in studies of late-life cognition. Importantly, few studies have examined the relationship between changes in educational attainment during adulthood and late-life cognition, which may have implications for cognitive health and related disparities.
To address this research gap, we analyzed data from the Kaiser Healthy Aging and Diverse Life Experiences (KHANDLE) Study, a racially and ethnically diverse cohort of older adults. KHANDLE presents a unique opportunity to examine the timing of educational attainment on cognitive change, as participants had completed a survey (Kaiser Permanente Northern California Multiphasic Health Checkups; KPNC MHC) that included information on educational attainment between 1964 and 1985 and again reported their educational attainment during KHANDLE interviews starting in 2017. Some KHANDLE participants were in early adulthood between 1964 and 1985, enabling us to identify individuals with non-traditional educational trajectories. We hypothesize that among those with lower levels of educational attainment earlier in life, those who pursued and attained higher educational levels later in life would have higher overall cognition and a slower rate of decline than those with lower education in later life, but worse cognitive outcomes than those who attained higher educational attainment earlier in life.
RESEARCH IN CONTEXT
Systematic review: We reviewed the literature on non-traditional educational attainment, adulthood education, and late-life cognition. Few studies have assessed the timing of educational attainment, which may have implications for the education–cognition relationship. Relevant work is discussed and cited.
Interpretation: This study expands on prior research by demonstrating how the level and timing of educational attainment across two time points are associated with domain-specific cognitive outcomes in a diverse cohort. Individuals with high educational attainment had higher executive function and verbal episodic memory, versus individuals with low educational attainment. Those with later-stage high educational attainment also had higher executive function versus those with low educational attainment, but it was significantly lower than that of those with high educational attainment earlier in adulthood.
Future directions: Findings should be replicated in larger diverse populations to better understand how the level and timing of educational attainment are associated with late-life cognitive function longitudinally.
METHODS
Study population
We used data from KHANDLE, a cohort of community-dwelling older adults residing in the San Francisco Bay Area and Sacramento, California. KHANDLE aims to examine how race, ethnicity, and life-course health and sociocultural factors influence late-life brain health and cognitive decline. Eligible participants were long-term members of KPNC, an integrated healthcare delivery system; aged 65 years and older on January 1, 2017; participants in the MHC during the period 1964 to 1985; and were English or Spanish speakers. Starting in 1964, KPNC began offering voluntary MHCs at Kaiser Permanente Oakland and San Francisco Medical Centers, which collected information on demographics, health-related activities, and health indicators. KHANDLE recruited 1712 community-dwelling individuals aged 65 years and older, with approximately equal proportions of Asian, Black, Latino, and White participants using stratified random sampling by race and ethnicity and educational attainment. Exclusion criteria at KHANDLE baseline included electronic medical diagnosis of dementia or other neurodegenerative diseases or presence of health conditions that would impede participation in study interviews (defined as hospice activity, history of end stage renal disease, or dialysis in the past 12 months or severe chronic obstructive pulmonary disease or congestive heart failure hospitalizations in the past 6 months). KHANDLE participants completed three interview cycles from March 2017 to June 2021, consisting of questionnaires assessing exposure to risk and protective factors of dementia across the life course and cognitive assessments. The first two interview cycles (ie, waves 1 and 2) were conducted in person, and the third interview cycle (ie, wave 3) was conducted by phone due to the COVID-19 pandemic.
The analytic sample comprises KHANDLE participants who, in addition to reporting educational attainment during KHANDLE interviews, provided self-reported educational attainment in the MHC between the ages of 24 and 30 years (N = 624). We excluded individuals who reported higher educational attainment at MHC than KHANDLE (n = 67) and those missing all cognitive measures (n = 3) for a final analytical sample of 554 individuals (). This study was approved by the KPNC Institutional Review Board and all KHANDLE participants provided written informed consent.
Exposure: Timing and level of educational attainment
We harmonized self-reported educational attainment from the MHC and KHANDLE wave 1 as high school graduate or less/general education development (GED), technical or trade education, some college, and college graduate or above. We defined later-life educational attainment as an increase in educational attainment between MHC and KHANDLE educational attainment. We categorized participants into educational attainment change groups based on timing and level of educational attainment as follows: (a) low education (ie, high school graduate or less/GED or technical or trade education at both time points); (b) high education (ie, some college or college graduate or more at both time points); and (c) later-life high education (ie, MHC education reported as high school graduate or less/GED or technical or trade education and KHANDLE education reported as some college or college graduate or above), detailed in Table .
Cognitive outcomes
We assessed the cognitive domains of executive function and verbal episodic memory using the Spanish and English Neuropsychological Assessment Scales (SENAS) during KHANDLE waves 1 to 3. SENAS is a battery of cognitive tests that has undergone extensive development using item response theory (IRT) methodology for valid comparisons of cognitive aging across diverse racial, ethnic, and linguistic groups. The SENAS measures for these two domains are psychometrically matched to have similar reliability of measurement across the ability continuum across diverse older adults. Visual-spatial ability and semantic memory were not routinely measured in KHANDLE. Executive function composite scores were obtained using component tasks of category fluency (animals, supermarket test), phonemic (letter) fluency, and working memory (digit span backward, visual span backward, and list sorting). Verbal episodic memory composite scores were derived from a multitrial word-list learning test. Further details on the administrative procedures, measure development, and psychometric characteristics have been described in detail elsewhere. Each cognitive domain was z-standardized using the mean and standard deviation from the full sample at baseline. Cognitive assessments during wave 3 were assessed by phone due to the COVID-19 pandemic.
Covariates
Sociodemographic information included baseline age (in years) retrieved from KPNC electronic records, sex (male or female), and self-reported race and ethnicity (categorized as Asian, Black, Latino, White) during KHANDLE wave 1. We assessed childhood socioeconomic conditions using both maternal and paternal educational attainment (each categorized as less than ≤12 years vs > 12 years) and childhood financial hardship. Childhood financial hardship was defined as reporting ever (ie, rarely, sometimes, often, or very often) skipping a meal or going hungry for financial reasons. Participants were also categorized as experiencing childhood financial hardship if they responded “rent” or “have other living arrangements” to the question: “During your childhood, did your family mainly: rent, pay a mortgage or own a home, have other living arrangements?”
Statistical analysis
We first examined the distribution of participant characteristics overall and by educational attainment change group. Missingness overall was around 22%. We imputed missing covariates using multiple imputation by chained equations, and the main analyses were applied to 30 imputed datasets. Main analyses included a series of linear mixed-effects models examining the associations between educational attainment change groups and domain-specific (ie, executive function or verbal episodic memory) annual rate of cognitive change. Linear mixed-effects models are fitted with maximum-likelihood estimation, a method that estimates the parameters of an assumed probability distribution based on the observed data. This method allows for obtaining parameter estimates, even in the presence of missing data, by using each individual's available data to compute the maximum likelihood. Model 1 included educational attainment change group (low education as reference), baseline age, time since baseline (in years), sex, race and ethnicity, and an indicator for interview mode. Model 2 additionally adjusted for childhood socioeconomic conditions (maternal education, paternal education, and childhood financial hardship). All models allowed for random intercepts and accounted for practice effects using model constraints based on prior analyses in this cohort. An interaction term between the educational attainment change group and time was included to determine whether the rate of cognitive change differed according to the educational attainment change group. We used the fitted models to predict mean cognition over time for each educational attainment change group and plotted the corresponding trajectories averaged across all other baseline covariates. Adjusted post hoc analyses were additionally performed to estimate mean differences and 95% confidence intervals (CIs) in domain-specific cognition for all pairwise combinations of educational attainment change groups, with the p values adjusted for multiple comparison (k = 3) using the Tukey method.
We conducted a sensitivity analysis to assess the robustness of our study findings. We applied a more stringent definition of non-traditional adult educational attainment by excluding individuals who reported technical or trade education, which may have been completed as an alternative to high school education or after high school completion. All analyses were conducted using R statistical software, version 4.1.1 (R Project for Statistical Computing).
RESULTS
The final analytic sample (n = 554) consisted of 23% Asian, 30% Black, 19% Latino, and 28% White participants (Table ). Descriptive statistics prior to multiple imputation are provided in Table . The average age of participants at KHANDLE wave 1 was 73 years, with a mean follow-up of 2.4 years (range: 0 to 3.9 years). Sixty-three percent were female, and 45% reported childhood financial hardship. Fourteen percent were categorized as low education, 66% high education, and 20% later-life high education. Compared to the overall KHANDLE cohort, individuals included in these analyses tended to be younger (73 vs 76 years) and were more likely to self-identify as Black (30% vs 26%, Table ).
TABLE 1 Descriptive statistics for KHANDLE analytical sample, overall and stratified by educational attainment change group.
Overall | Educational attainment change group | |||
Sample | Low Education | High Education | Later-life high | |
Individuals, n (%) | 554 | 75 (13.5) | 366 (66.1) | 113 (20.4) |
Sociodemographic characteristics | ||||
Baseline age (in years), mean (SD) | 73.2 (4.6) | 74.2 (4.7) | 72.7 (4.6) | 74.0 (4.5) |
Race, n (%) | ||||
Asian | 128 (23.1) | 8 (10.7) | 109 (29.8) | 11 (9.7) |
Black | 168 (30.3) | 27 (36.0) | 95 (26.0) | 46 (40.7) |
Latino | 104 (18.8) | 24 (32.0) | 54 (14.8) | 26 (23.0) |
White | 154 (27.8) | 16 (21.3) | 108 (29.5) | 30 (26.5) |
Female, n (%) | 347 (62.6) | 49 (65.3) | 215 (58.7) | 83 (73.5) |
Male, n (%) | 207 (37.4) | 26 (34.7) | 151 (41.3) | 30 (26.5) |
Childhood financial hardship (yes/no), % | ||||
Yes | 45.4 | 57.2 | 40 | 54.8 |
No | 54.6 | 42.8 | 60 | 45.2 |
Maternal education level, % | ||||
≤ 12 Years | 74.2 | 89.6 | 70.1 | 77.4 |
> 12 Years | 25.8 | 10.4 | 29.9 | 22.6 |
Paternal education level, % | ||||
≤ 12 years | 69.7 | 92.2 | 61.6 | 80.9 |
> 12 years | 30.3 | 7.8 | 38.4 | 19.1 |
Number of cognitive assessments during follow-up | ||||
Executive function, mean (SD) | 2.7 (0.7) | 2.4 (0.8) | 2.7 (0.6) | 2.7 (0.6) |
Verbal episodic memory, mean (SD) | 2.7 (0.7) | 2.4 (0.8) | 2.7 (0.6) | 2.7 (0.6) |
Compared to those in the high education group, individuals in the later-life high education group were least likely to identify as Asian (30% vs 10%) and most likely to identify as Black (26% vs 41%). Later-life high education (vs high education) was also more common among women than men and associated with more childhood financial hardship and low parental education.
Executive function
Adjusting for baseline age, time since baseline, sex, race and ethnicity, interview mode, and practice effects (Model 1), compared to individuals with low educational attainment, individuals with high educational attainment (β = 0.64 SD units; 95% CI: 0.45, 0.83) and later-life high educational attainment (β = 0.24 SD units; 95% CI: 0.02, 0.46) had higher executive function (Table ). After additionally controlling for childhood socioeconomic conditions (Model 2), the association of high educational attainment (β = 0.59 SD units; 95% CI: 0.39, 0.79) and later-life high educational attainment (β = 0.22 SD units; 95% CI: 0.00, 0.44) with higher executive function persisted, albeit minimally attenuated (Figure , Table ). There was no evidence that the annual rate of decline in executive function differed by educational attainment change group (Table ). Pairwise comparisons revealed that individuals with high educational attainment had significantly higher cognition than those with later-life educational attainment (adjusted mean difference: 0.37 SD units; 95% CI: 0.17, 0.57; Tukey-adjusted p value: < .0001).
[IMAGE OMITTED. SEE PDF]
Verbal episodic memory
In Model 1, compared to individuals with low educational attainment, individuals with high educational attainment (β = 0.27 SD units; 95% CI: 0.07, 0.47) had higher verbal episodic memory (Table ). Later-life high educational attainment (β = 0.01 SD units; 95% CI: −0.22, 0.25) was not associated with verbal episodic memory. After additionally controlling for childhood socioeconomic conditions (Model 2), the association of high educational attainment (β = 0.27 SD units; 95% CI: 0.06, 0.48) with higher verbal episodic memory persisted (Figure , Table ). There was no evidence that the annual rate of decline in verbal episodic memory differed by educational attainment change group (Table ). Pairwise comparisons revealed that individuals with high educational attainment had significantly higher verbal episodic memory than those with later-life educational attainment (adjusted mean difference: 0.25 SD units; 95% CI: 0.04, 0.46; Tukey-adjusted p value: 0.015).
Sensitivity analysis
Using a more restrictive definition for educational attainment change groups yielded results similar to those of the main analyses (Table ), although the associations were stronger for educational attainment change groups and executive function.
DISCUSSION
In this study of a racially and ethnically diverse cohort of older adults, our findings suggest that both timing and level of education may influence late-life cognitive function and vary by cognitive domain. Individuals with high educational attainment (ie, any college education by age 30) and later-life high educational attainment had higher levels of late-life executive function compared to those with low educational attainment. Those with high educational attainment also had significantly higher executive function compared to those with later-life high educational attainment. Only those with high educational attainment had significantly higher verbal episodic memory compared to those with low educational attainment. However, the rate of cognitive decline did not differ across educational attainment change groups.
Our findings are consistent with prior literature suggesting that higher educational levels are associated with higher cognitive performance, but not with cognitive decline. In particular, our effect estimates for the association between educational attainment and cognitive performance fall within the range of effect sizes presented by a large meta-analysis, where an individual with a university degree would be expected to have a mean difference of 0.2 to 0.4 SD units higher in cognitive performance versus a comparable individual with less education (eg, a high school diploma). To our knowledge, this study is the first to examine how both timing and level of educational attainment may be associated with domain-specific late-life cognitive function and change. As hypothesized, compared to those who maintained lower educational attainment levels over time, individuals with later-life high educational attainment had higher late-life cognitive function, but effect estimates were significantly smaller compared to those who achieved higher educational attainment earlier in life. Importantly, these results highlight the need for repeated assessments of educational attainment throughout adulthood to better characterize protective associations with late-life cognition.
Several factors have been proposed to explain why higher educational attainment may benefit cognitive function. Education in early life may directly enhance cognitive abilities by exposing individuals to more cognitive stimulation and opportunities to acquire knowledge and skills that may persist into later life. Higher educational attainment may also be indirectly associated with higher cognitive function later in the life course through higher income, more cognitively stimulating environments, and greater access to resources and the ability to invest in one's health. Additional work is needed to assess potential mechanisms specifically linking higher educational attainment in later adulthood to domain-specific cognition.
Some limitations of our studies must be discussed. We cannot rule out the possibility that reporting error, rather than delayed educational attainment, accounts for changes in self-reported educational attainment over time. We used self-reported educational attainment from two time points (ages 24 to 30 and 65 years and older) and do not have information on the specific timing when higher educational attainment was obtained. For example, individuals who obtain higher educational attainment in their 30s may have different cognitive outcomes compared to individuals who obtain higher educational attainment in their 40s and 50s. Given that we found a stronger relationship for the high educational attainment group and executive function compared to the later-life educational attainment, it may be possible that the protective effects of higher educational attainment decrease with later-life timing of such educational attainment. Further work examining the granularity in the timing of later-life higher educational attainment on cognitive outcomes is needed to understand nuances in possible effect modification by different life stages and to examine midlife circumstances that could serve as potential confounders or downstream mediators.
Early- or later-life high educational attainment is likely to be influenced by childhood socioeconomic factors (ie, parental educational attainment) that provide more opportunities and advantages earlier in life and continue throughout an individual's later life. However, controlling for parental educational attainment and childhood financial hardship minimally changed our effect estimates. Additionally, individuals who participate in non-traditional educational pathways are likely to have dependents other than a spouse and full- or part-time occupations. These sociocontextual factors may reflect personality, resilience, or spousal and familial support that could confound the association of later-life educational attainment and late-life cognitive performance. Although we were unable to disentangle these relationships in this study, the interplay of these factors as potential modifiers or mediators should be further explored. Due to the limited sample size, we were underpowered to assess differences in the effects of educational attainment change groups by sex or across race and ethnicity, and there may be some residual confounding. Finally, our sample consists of long-term members of Kaiser Permanente Northern California, most of whom are highly educated, and has limited generalizability to the broader U.S. population, those who have lower levels of education, or to younger birth cohorts.
Nevertheless, our study has several strengths. We had self-reported educational attainment from two time points at least 35 years apart and were able to examine whether higher educational attainment in later adulthood was associated with late-life cognitive function and cognitive change in a unique, ethnically and racially diverse cohort of older adults in Northern California. Our findings also highlight the importance of examining domain-specific cognitive outcomes, as noted by the heterogeneity in our results with educational attainment in later life.
In conclusion, we present preliminary evidence on how timing and level of educational attainment are both associated with cognitive health at a later age among a diverse cohort of older adults. Higher education in early life is beneficial for cognition, but obtaining higher educational levels later in life may also be protective. More large-scale studies are needed to examine how these associations may vary by sex or across race and ethnicity.
ACKNOWLEDGMENTS
We are thankful to all individuals participating in the KHANDLE study. This research was supported by the National Institute on Aging (NIA) (grants P30AG072972, R01AG066132, R01AG052132, R00AG073457, K00AG068431). Funders had no role in study design, data collection, analysis, interpretation of data, manuscript drafting, or decision to submit for publication.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest. Author disclosures are available in the .
CONSENT STATEMENT
This study was approved by the KPNC Institutional Review Board. All KHANDLE participants provided written informed consent.
Beydoun MA, Beydoun HA, Gamaldo AA, Teel A, Zonderman AB, Wang Y. Epidemiologic studies of modifiable factors associated with cognition and dementia: systematic review and meta‐analysis. BMC Public Health. 2014;14(1):643. doi: [DOI: https://dx.doi.org/10.1186/1471-2458-14-643]
Choy S. Nontraditional undergraduates: Findings from the condition of education. U.S. Department of Education, National Center for Education Statistics; 2002. NCES 2002‐012. 2002.
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2024. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the "License"). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
INTRODUCTION
The timing of educational attainment may modify its effects on late‐life cognition, yet most studies evaluate education only at a single time point.
METHODS
Kaiser Healthy Aging and Diverse Life Experiences (KHANDLE) Study cohort participants (N = 554) reported educational attainment (dichotomized at any college education) at two time points, and we classified them as having low, high, or later‐life high educational attainment. Linear mixed‐effects models estimated associations between educational attainment change groups and domain‐specific cognitive outcomes (z‐standardized).
RESULTS
Compared to low educational attainment, high (β= 0.59 SD units; 95% confidence interval [CI]: 0.39, 0.79) and later‐life high educational attainment (β = 0.22; 95% CI: 0.00, 0.44) were associated with higher executive function. Only high educational attainment was associated with higher verbal episodic memory (β = 0.27; 95% CI: 0.06, 0.48).
DISCUSSION
Level and timing of educational attainment are both associated with domain‐specific cognition. A single assessment for educational attainment may inadequately characterize protective associations with late‐life cognition.
Highlights
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
Details

1 Kaiser Permanente Division of Research, Oakland, California, USA
2 Department of Public Health Sciences, University of California, Davis, California, USA
3 Department of Epidemiology, University of California, Los Angeles Fielding School of Public Health, Los Angeles, California, USA
4 Department of Epidemiology and Biostatistics, University of California, San Francisco, California, USA
5 Department of Epidemiology and Population Health, Stanford University, Stanford, California, USA
6 School of Public and Community Health Sciences, University of Montana, Missoula, Montana, USA
7 Department of Neurology, School of Medicine, University of California, Davis, California, USA