Skip to main content
NCBI home page
Frontiers in Psychology logoLink to Frontiers in Psychology
. 2026 Mar 19;17:1564721. doi: 10.3389/fpsyg.2026.1564721

Analysis of disparate factors affecting cognitive function among populations with different educational levels: a large-scale longitudinal study

Dezhen Dai 1,†, Bingbing Xiang 2,†, Yunke Dai 3,†, Pingliang Yang 3, Na Zhu 3,*, Shun Wang 3,*
PMCID: PMC13043331  PMID: 41937829

Abstract

Objective

To explore the influencing factors and patterns of cognitive function among populations with different educational levels.

Methods

Using data from the China Health and Retirement Longitudinal Study (2015–2020), we analyzed 29,620 subjects aged 45–85 years who completed cognitive function assessments. Participants were stratified by educational level: illiterate (n = 7,670), primary school (n = 7,897), junior high school (n = 8,904), and high school and above (n = 5,149). Mixed-effects models were used to analyze cognitive function determinants across educational groups, with sensitivity analyses performed to verify result robustness.

Results

Cognitive function scores demonstrated a significant educational gradient, with the highest scores in the high-education group [20.0 (18.0–23.0)] and lowest in the illiterate group [16.0 (13.0–18.0)]. Age negatively correlated with cognitive function, with stronger effects in higher education groups (illiterate: β = −0.665; high-education: β = −1.033, p < 0.001). Gender effects varied by education level: males showed cognitive advantages in the illiterate group (β = 0.716, p < 0.001), but disadvantages in the high-education group (β = −0.739, p < 0.001). Internet use demonstrated enhanced protective effects with increasing education (illiterate: β = 0.254; high-education: β = 0.411, p < 0.001). Urban residence benefits strengthened with education level (illiterate: β = 0.188; high-education: β = 0.439, p < 0.001). Memory-related diseases showed the most significant impact in the high-education group (β = −2.325, p < 0.05).

Conclusion

Educational level appears to act not only as an independent correlate of cognitive function but also as a potential modifier of its associations with gender, Internet use, residential environment, and chronic disease burden. These observations suggest educational background warrants consideration when designing cognitive health strategies for older adults.

Keywords: China Health and Retirement Longitudinal Study (CHARLS), cognitive function, educational attainment, longitudinal study, population aging

Highlights

  • Gender effects on cognition vary by education

  • Internet’s cognitive protection increases with educational level

  • Age-related cognitive decline intensifies at higher educational levels

  • Memory-related diseases show the most significant impact in high-education groups

Introduction

Cognitive impairment has emerged as a major public health challenge amid global population aging, significantly impacting elderly quality of life and healthcare systems (He et al., 2024; Luo et al., 2024). This condition not only increases mortality and disability risk but also creates substantial socioeconomic burden (Tahami Monfared et al., 2022). Current estimates indicate approximately 132 million individuals worldwide have mild cognitive impairment, with projections reaching 152 million by 2050, demonstrating marked demographic and geographic variation (GBD 2019 Dementia Forecasting Collaborators, 2022).

Educational attainment, a key cognitive reserve indicator, fundamentally influences cognitive trajectory (Bratsberg et al., 2024; Singh-Manoux et al., 2011). The cognitive reserve hypothesis posits that higher education enhances neural plasticity and network optimization, potentially moderating cognitive decline (Barba et al., 2021). Education’s impact likely extends beyond direct effects, with emerging evidence suggesting its role as an effect modifier (Jia et al., 2022). Studies indicate education may indirectly affect cognition by shaping health behaviors, social engagement patterns, and self-management capabilities (Jia et al., 2022; Nguyen et al., 2013).

Contemporary societal changes have introduced new cognitive function determinants. Digital technology adoption, particularly Internet use, may serve as a cognitive protective factor (Ihle et al., 2020; Amland et al., 2024). Evolving modes of social participation provide novel cognitive engagement opportunities (Deng et al., 2019; Siette et al., 2020). Additionally, changing chronic disease patterns, especially the rising prevalence of metabolic disorders, present new considerations in cognitive health (Sanchez-Orti et al., 2022; Maksyutynska et al., 2024).

Prior investigations have largely treated education as a confounder rather than exploring its potential effect modification (He et al., 2024; Ghahremani et al., 2024). Most studies employed cross-sectional designs, limiting understanding of temporal relationships (Clare et al., 2017; Sedgwick, 2015). The China Health and Retirement Longitudinal Study (CHARLS), a nationwide prospective cohort study, offers an opportunity to address these limitations (Zhu et al., 2024). Using probability-proportional-to-size sampling, CHARLS collected comprehensive data on cognitive function and its determinants across educational strata.

The present study examined CHARLS data from 2015 to 2020 to address three key questions: (1) whether cognitive function determinants differ systematically across educational levels, (2) how these relationships evolve over time, and (3) the extent and nature of effect modification by educational attainment. This investigation advances current knowledge by analyzing education’s moderating role using longitudinal data, examining temporal trends in cognitive risk factors, and employing multiple sensitivity analyses to ensure robust findings. Results may inform the development of education-specific cognitive intervention strategies.

Methods

Research design and data source

We conducted a prospective cohort study using data from the CHARLS collected in 2015, 2018, and 2020. CHARLS employed multi-stage stratified probability sampling, covering 150 counties across 28 Chinese provinces, targeting residents aged ≥45 years (Zhao et al., 2014). This study was conducted in accordance with the Declaration of Helsinki and was approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052-11015), and all participants provided written informed consent.

Study population

Inclusion criteria comprised: (1) age 45–85 years; (2) completion of cognitive function assessment; and (3) available educational attainment data. We excluded participants with missing key variables or outlier cognitive assessment scores. The final analytic sample included 14,236 participants with complete data across all three waves (29,620 observations). Participants were stratified by educational level into four groups: illiterate (n = 7,670), primary school (n = 7,897), junior high school (n = 8,904), and high school and above (n = 5,149).

Variable measurement

Primary outcome variable

The cognitive function total score (range: 0–31) comprised four domains: Memory (0–20 points): immediate word recall (10 words, 0–10 points) and delayed word recall (0–10 points); Orientation (0–5 points): assessment of temporal orientation (year, month, day, week, time); Calculation (0–5 points): serial subtraction by 7 s; Visuospatial ability (0–1 point): pentagon copying task

Independent variables

Demographic characteristics: age (continuous), sex (male/female), marital status (married/unmarried); Lifestyle factors: smoking and alcohol consumption (current or former/never), sleep duration (<8 h, ≥8 h), napping (yes/no), Internet use (yes/no), social activity participation (yes/no); Environmental factors: residence (urban/rural), life satisfaction (satisfied/unsatisfied); Health status: chronic conditions (14 common diseases), physical pain (yes/no).

Statistical analysis

The dataset comprised predominantly binary and categorical variables; age and cognitive function total score were the only continuous measures. Normality of these variables was assessed using the Shapiro–Wilk test prior to analysis. Neither variable satisfied the normality criterion in any educational subgroup, and continuous variables were therefore summarized as median (interquartile range). Between-group differences were examined using Kruskal–Wallis tests with Bonferroni correction for post-hoc pairwise comparisons. Categorical variables were expressed as frequency (percentage) and compared using chi-square tests, with Fisher’s exact test applied where expected cell counts fell below five. Spearman correlation assessed relationships between cognitive function and potential predictors.

Multiple linear regression models were used to examine factors associated with cognitive function in each educational group, adjusting for demographic, lifestyle, environmental, and health-related variables. Standardized coefficients (β) were reported for all associations.

We constructed four sequential models: Model 1: demographic characteristics only; Model 2: Model 1 + lifestyle factors; Model 3: Model 2 + environmental factors; Model 4: Model 3 + health status factors. All models included year and individual-level random effects. Residual diagnostics including Q-Q plots and density curves were examined for all subgroup models, confirming that the normality assumption for model residuals was adequately satisfied. To formally test whether educational level functions as an effect modifier, we fitted a pooled mixed-effects model incorporating interaction terms between educational level and five key predictors (age, sex, internet use, urban residence, and memory-related disease). All Model 4 covariates were retained as main effects, with an identical random-effects structure. Interaction significance was assessed using likelihood ratio tests and Type III F-tests (Satterthwaite approximation).

We tested result robustness using two alternative educational stratification schemes: Binary classification: low (illiterate + primary school) versus high (junior high school + high school and above). Three-level classification: illiterate, medium (primary + junior high school), and high education. Statistical analyses were performed using R version 4.0.3, with two-sided p < 0.05 considered statistically significant.

Results

Baseline characteristics

Based on CHARLS data from 2015–2020, we included 29,620 participants: 9,145 in 2015, 9,979 in 2018, and 10,496 in 2020. Participants were stratified by educational level into illiterate (n = 7,670), primary school (n = 7,897), junior high school (n = 8,904), and high school and above groups (n = 5,149) (Table 1; Figure 1).

Table 1.

Characteristics of the study population from CHARLS.

Variable Category Overall (N = 29,620) Group Illiterate (N = 7,670) Group Primary (N = 7,897) Group Middle (N = 8,904) Group High (N = 5,149) p
Life satisfaction (n, %) Satisfied 27,206 (91.9%) 6,957 (90.7%)c,d 7,207 (91.3%)c,d 8,222 (92.3%)a,b,d 4,820 (93.6%)a,b,c <0.001
Dissatisfied 2,414 (8.1%) 713 (9.3%)c,d 690 (8.7%)c,d 682 (7.7%)a,b,d 329 (6.4%)a,b,c
Gender (n, %) Female 14,155 (47.8%) 4,396 (57.3%)b,c,d 3,726 (47.2%)a,c,d 3,936 (44.2%)a,b,d 2097 (40.7%)a,b,c <0.001
Male 15,465 (52.2%) 3,274 (42.7%)b,c,d 4,171 (52.8%)a,c,d 4,968 (55.8%)a,b,d 3,052 (59.3%)a,b,c
Marital status (n, %) Unmarried/divorced/widowed/separated 3,562 (12.0%) 965 (12.6%) 969 (12.3%) 1,035 (11.6%) 593 (11.5%) 0.151
Married 26,058 (88.0%) 6,705 (87.4%) 6,928 (87.7%) 7,869 (88.4%) 4,556 (88.5%)
Ever/current smoke (n, %) No 16,555 (55.9%) 4,798 (62.6%)b,c,d 4,389 (55.6%)a,c,d 4,647 (52.2%)ab 2,721 (52.8%)ab <0.001
Yes 13,065 (44.1%) 2,872 (37.4%)b,c,d 3,508 (44.4%)a,c,d 4,257 (47.8%)ab 2,428 (47.2%)ab
Ever/current alcohol (n, %) No 17,525 (59.2%) 4,737 (61.8%)b,c,d 4,731 (59.9%)a,c,d 5,157 (57.9%)ab 2,900 (56.3%)ab <0.001
Yes 12,095 (40.8%) 2,933 (38.2%)b,c,d 3,166 (40.1%)a,c,d 3,747 (42.1%)a,b 2,249 (43.7%)a,b
Age (years) 59.00 (52.00–66.00) 62.00 (54.00–68.00)b,c,d 59.00 (52.00–68.00)a,c,d 56.00 (52.00–63.00)a,b,d 58.00 (53.00–63.00)a,b,c <0.001
Daily sleep time (h) (n, %) < 8 22,072 (74.5%) 5,608 (73.1%)c,d 5,745 (72.7%)c,d 6,652 (74.7%)a,b,d 4,067 (79.0%)a,b,c <0.001
≥ 8 7,548 (25.5%) 2062 (26.9%)c,d 2,152 (27.3%)c,d 2,252 (25.3%)a,b,d 1,082 (21.0%)a,b,c
Nap (n, %) No 11,954 (40.4%) 3,433 (44.8%)b,c,d 3,310 (41.9%)a,c,d 3,469 (39.0%)a,b,d 1742 (33.8%)a,b,c <0.001
Yes 17,666 (59.6%) 4,237 (55.2%)b,c,d 4,587 (58.1%)a,c,d 5,435 (61.0%)a,b,d 3,407 (66.2%)a,b,c
Internet (n, %) No 22,510 (76.0%) 6,352 (82.8%)b,c,d 6,380 (80.8%)a,c,d 6,588 (74.0%)a,b,d 3,190 (62.0%)a,b,c <0.001
Yes 7,110 (24.0%) 1,318 (17.2%)b,c,d 1,517 (19.2%)a,c,d 2,316 (26.0%)a,b,d 1959 (38.0%)a,b,c
Social activities (n, %) No 12,854 (43.4%) 3,792 (49.4%)b,c,d 3,623 (45.9%)a,c,d 3,658 (41.1%)a,b,d 1781 (34.6%)a,b,c <0.001
Yes 16,766 (56.6%) 3,878 (50.6%)b,c,d 4,274 (54.1%)a,c,d 5,246 (58.9%)a,b,d 3,368 (65.4%)a,b,c
Residential area (n, %) Rural 19,327 (65.2%) 5,678 (74.0%)b,c,d 5,588 (70.8%)a,c,d 5,557 (62.4%)a,b,d 2,504 (48.6%)a,b,c <0.001
Urban 10,293 (34.8%) 1992 (26.0%)b,c,d 2,309 (29.2%)a,c,d 3,347 (37.6%)a,b,d 2,645 (51.4%)a,b,c
Troubled with body pain (n, %) No 16,922 (57.1%) 4,347 (56.7%) 4,448 (56.3%) 5,115 (57.4%) 3,012 (58.5%) 0.0711
Yes 12,698 (42.9%) 3,323 (43.3%) 3,449 (43.7%) 3,789 (42.6%) 2,137 (41.5%)
Co-morbidities (n, %)
Hypertension No 23,838 (80.5%) 6,008 (78.3%)b,c,d 6,368 (80.6%)ac 7,305 (82.0%)a,b 4,157 (80.7%)a <0.001
Yes 5,782 (19.5%) 1,662 (21.7%)b,c,d 1,529 (19.4%)ac 1,599 (18.0%)a,b 992 (19.3%)a
Dyslipidaemia No 26,516 (89.5%) 6,955 (90.7%)c,d 7,216 (91.4%)c,d 7,961 (89.4%)a,b,d 4,384 (85.1%)a,b,c <0.001
Yes 3,104 (10.5%) 715 (9.3%)c,d 681 (8.6%)c,d 943 (10.6%)a,b,d 765 (14.9%)a,b,c
Hyperglycaemia No 27,966 (94.4%) 7,252 (94.6%)d 7,470 (94.6%)d 8,432 (94.7%)d 4,812 (93.5%)a,b,c 0.0112
Yes 1,654 (5.6%) 418 (5.4%)d 427 (5.4%)d 472 (5.3%)d 337 (6.5%)a,b,c
Cancer No 29,371 (99.2%) 7,606 (99.2%) 7,808 (98.9%)c 8,849 (99.4%)b 5,108 (99.2%) 0.004
Yes 249 (0.8%) 64 (0.8%) 89 (1.1%)c 55 (0.6%)b 41 (0.8%)
Chronic lung disease No 27,406 (92.5%) 6,993 (91.2%)c,d 7,206 (91.2%)c,d 8,365 (93.9%)a,b 4,842 (94.0%)a,b <0.001
Yes 2,214 (7.5%) 677 (8.8%)c,d 691 (8.8%)c,d 539 (6.1%)a,b 307 (6.0%)a,b
Liver disease No 28,676 (96.8%) 7,438 (97.0%) 7,658 (97.0%) 8,619 (96.8%) 4,961 (96.3%) 0.178
Yes 944 (3.2%) 232 (3.0%) 239 (3.0%) 285 (3.2%) 188 (3.7%)
Heart disease No 26,682 (90.1%) 6,906 (90.0%)c,d 7,091 (89.8%)c 8,103 (91.0%)a,b,d 4,582 (89.0%)a,c 0.001
Yes 2,938 (9.9%) 764 (10.0%)c,d 806 (10.2%)c 801 (9.0%)a,b,d 567 (11.0%)ac
Stroke No 29,068 (98.1%) 7,511 (97.9%) 7,745 (98.1%) 8,758 (98.4%) 5,054 (98.2%) 0.217
Yes 552 (1.9%) 159 (2.1%) 152 (1.9%) 146 (1.6%) 95 (1.8%)
Kidney disease No 28,172 (95.1%) 7,258 (94.6%)c 7,511 (95.1%) 8,515 (95.6%)a 4,888 (94.9%) 0.0246
Yes 1,448 (4.9%) 412 (5.4%)c 386 (4.9%) 389 (4.4%)a 261 (5.1%)
Digestive system disease No 24,380 (82.3%) 6,110 (79.7%)c,d 6,375 (80.7%)c,d 7,493 (84.2%)a,b,d 4,402 (85.5%)a,b,c <0.001
Yes 5,240 (17.7%) 1,560 (20.3%)c,d 1,522 (19.3%)c,d 1,411 (15.8%)a,b,d 747 (14.5%)a,b,c <0.001
Emotional / mental disorders No 29,419 (99.3%) 7,608 (99.2%) 7,840 (99.3%) 8,848 (99.4%) 5,123 (99.5%) 0.190
Yes 201 (0.7%) 62 (0.8%) 57 (0.7%) 56 (0.6%) 26 (0.5%)
Memory-related disease No 29,396 (99.2%) 7,606 (99.2%) 7,842 (99.3%) 8,848 (99.4%) 5,100 (99.0%) 0.136
Yes 224 (0.8%) 64 (0.8%) 55 (0.7%) 56 (0.6%) 49 (1.0%)
Arthritis or rheumatism No 22,707 (76.7%) 5,412 (70.6%)b,c,d 5,933 (75.1%)a,c,d 7,045 (79.1%)a,b,d 4,317 (83.8%)a,b,c <0.001
Yes 6,913 (23.3%) 2,258 (29.4%)b,c,d 1964 (24.9%)a,c,d 1859 (20.9%)a,b,d 832 (16.2%)a,b,c
Asthma No 28,747 (97.1%) 7,406 (96.6%)c,d 7,591 (96.1%)c,d 8,698 (97.7%)a,b 5,052 (98.1%)a,b <0.001
Yes 873 (2.9%) 264 (3.4%)c,d 306 (3.9%)c,d 206 (2.3%)a,b 97 (1.9%)a,b
Cognitive function score
Total cognitive Scores 18.00 (15.00–21.00) 16.00 (13.00–18.00)b,c,d 18.00 (15.00–20.00)a,c,d 19.00 (16.00–21.00)a,b,d 20.00 (18.00–23.00)a,b,c <0.001
Memory 9.00 (7.00–11.00) 8.00 (6.00–10.00)b,c,d 9.00 (7.00–11.00)a,c,d 10.00 (8.00–12.00)a,b,d 11.00 (9.00–13.00)a,b,c <0.001
Orientation 1 393 (1.3%) 246 (3.2%)b,c,d 87 (1.1%)a,c,d 48 (0.5%)a,b,d 12 (0.2%)a,b,c <0.001
2 1,396 (4.7%) 706 (9.2%)b,c,d 337 (4.3%)a,c,d 283 (3.2%)a,b,d 70 (1.4%)a,b,c
3 4,332 (14.6%) 1,631 (21.3%)b,c,d 1,206 (15.3%)a,c,d 1,113 (12.5%)a,b,d 382 (7.4%)a,b,c
4 9,343 (31.5%) 2,530 (33.0%)b,c,d 2,638 (33.4%)a,c,d 2,819 (31.7%)a,b,d 1,356 (26.3%)a,b,c
5 14,156 (47.8%) 2,557 (33.3%)b,c,d 3,629 (46.0%)a,c,d 4,641 (52.1%)a,b,d 3,329 (64.7%)a,b,c
Computation 1 2,413 (8.1%) 1,055 (13.8%)b,c,d 630 (8.0%)a,c,d 525 (5.9%)a,b,d 203 (3.9%)a,b,c <0.001
2 1907 (6.4%) 711 (9.3%)b,c,d 511 (6.5%)a,c,d 479 (5.4%)a,b,d 206 (4.0%)a,b,c
3 3,791 (12.8%) 1,226 (16.0%)b,c,d 1,040 (13.2%)a,c,d 1,061 (11.9%)a,b,d 464 (9.0%)a,b,c
4 7,351 (24.8%) 1936 (25.2%)b,c,d 2071 (26.2%)a,c,d 2,180 (24.5%)a,b,d 1,164 (22.6%)a,b,c
5 14,158 (47.8%) 2,742 (35.7%)b,c,d 3,645 (46.2%)a,c,d 4,659 (52.3%)a,b,d 3,112 (60.4%)a,b,c
Drawing 0 11,267 (38.0%) 4,021 (52.4%)b,c,d 3,096 (39.2%)a,c,d 2,831 (31.8%)a,b,d 1,319 (25.6%)a,b,c <0.001
1 18,353 (62.0%) 3,649 (47.6%)b,c,d 4,801 (60.8%)a,c,d 6,073 (68.2%)a,b,d 3,830 (74.4%)a,b,c
Open in a new tab

Superscript letters indicate significant differences between groups (p < 0.05), a: significant difference vs 1; b: significant difference vs 2; c: significant difference vs 3; d: significant difference vs 4. p < 0.05 denotes statistical significance. Group Illiterate: Illiterate; Group Primary: Primary school; Group Middle: Middle school; Group High: High school/vocational high school + Junior college or above.

Figure 1.

Flowchart diagram of the Charlson database sample selection displaying years 2015, 2018, and 2020, selection criteria, inclusion and exclusion criteria, data processing, and breakdown into four educational groups: Illiterate, Primary, Middle, and High, with corresponding sample sizes.

Open in a new tab

Flowchart of the sample selection process.

Demographic analysis showed that the illiterate group was significantly older [62.0 (54.0–68.0) years] with a higher proportion of females (57.3%), while the high-education group was younger [58.0 (53.0–63.0) years] with more males (59.3%). Marital status was comparable across groups, with married individuals comprising over 87% in each group. The high-education group exhibited more favorable lifestyle characteristics, with significantly higher rates of Internet usage (38.0%) and social activity participation (65.4%) compared to other groups (p < 0.001).

Health status comparison revealed a higher chronic disease burden in the illiterate group, with significantly higher prevalence of hypertension (21.7%) and arthritis (29.4%). Cognitive function scores showed a clear educational gradient: the high-education group achieved the highest scores [20.0 (18.0–23.0)], while the illiterate group scored lowest [16.0 (13.0–18.0)]. This pattern was consistent across all cognitive domains. In calculation ability, 60.4% of the high-education group achieved maximum scores compared to 35.7% in the illiterate group. Pentagon-copying task completion rates were 74.4% and 47.6% for high-education and illiterate groups, respectively (Table 1).

Analysis of influencing factors of cognitive function

Stratified Spearman correlation analysis identified education-moderated association patterns between cognitive function and candidate predictors across all four groups (Figure 2). Age exhibited the strongest and most consistent negative association, with correlation magnitude progressively intensifying across educational strata (illiterate: r = −0.156 to high-education: r = −0.276, all p < 0.001). Lifestyle factors including Internet use and smoking, as well as environmental factors such as urban residence, similarly demonstrated education-dependent association gradients. Health-related variables, including chronic diseases and physical pain, showed predominantly negative associations across all groups. Notably, gender and memory-related diseases displayed differential association patterns across educational strata. These crude association profiles, captured simultaneously across all predictor domains in Figure 2, provided preliminary evidence of educational effect modification and informed the multivariable linear regression and mixed-effects model analyses presented below.

Figure 2.

Three side-by-side heatmaps display standardized coefficients for various health and demographic factors by education level across 2015, 2018, and 2020. Rows list factors like age, chronic diseases, and behaviors; columns represent education levels. Red indicates positive effects, blue indicates negative effects, with significance denoted by asterisks. Age consistently shows strong negative association across all years and education levels. A color bar on the right provides the standardized coefficient scale.

Open in a new tab

Correlation network of cognitive function and associated factors across educational groups. Red bubbles indicate positive correlations; green bubbles indicate negative correlations. Bubble size represents correlation strength. Group illiterate: illiterate; group primary: primary school; group middle: middle school; group high: high school/vocational high school + junior college or above. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001.

Multivariate analysis (Figure 3; Supplementary Table S1) further elucidated education-dependent patterns. Throughout 2015–2020, age consistently showed negative correlations with cognitive function (all p < 0.001), with the effect strengthening at higher educational levels. By 2020, the age effect coefficient in the high-education group (−0.109) substantially exceeded that of the illiterate group (−0.071).

Figure 3.

Four correlation matrices for Illiterate, Primary, Middle, and High educational groups display significant relationships among cognitive, health, and lifestyle variables using colored circles, with green indicating negative and red indicating positive correlations. Circle size reflects strength.

Open in a new tab

Standardized effects of various factors on cognitive function across education levels (2015–2020). Group illiterate: illiterate; Group primary: primary school; Group middle: middle school; Group high: high school/vocational high school + junior college or above. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001.

Gender effects showed temporal and educational variation. In 2020, males demonstrated cognitive advantages in the illiterate group (β = 0.716, p < 0.001) but disadvantages in higher education groups (high-education group: β = −0.739, p < 0.001). Internet use maintained positive associations across all educational levels, particularly evident in 2018 and 2015 (p < 0.001). Smoking behavior showed negative impacts across all educational groups in 2015, most prominently in the high-education group (β = −0.689, p < 0.001).

Sleep duration demonstrated significant negative associations in medium-education and high-education groups by 2020 (medium-education: β = −0.345, p < 0.05; high-education: β = −0.500, p < 0.05). Cancer showed substantial negative impact in the medium-education group (2020: β = −1.964, p < 0.01), while memory-related diseases significantly affected the high-education group in 2015 (β = −2.325, p < 0.05). Life satisfaction demonstrated negative associations across all educational groups in 2020 (p < 0.01), while urban residence showed consistent positive effects across all educational groups in 2018 and 2015.

Mixed-effects model analysis

The stepwise mixed-effects model analysis (Table 2) revealed distinctive patterns of key influencing factors across educational strata. Baseline cognitive function demonstrated a clear educational gradient in the intercept terms, progressively increasing from the illiterate group (15.888) to the high-education group (20.251, p < 0.001), indicating a robust positive association between educational attainment and baseline cognitive performance.

Table 2.

Mixed effects models of cognitive function across educational groups.

Group illiterate Group primary Group middle Group high
Variable Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4
(Intercept) 15.888*** 15.667*** 15.677*** 15.770*** 17.545*** 17.488*** 17.520*** 17.582*** 18.612*** 18.592*** 18.547*** 18.646*** 20.251*** 20.189*** 20.035*** 20.127***
Age −0.661*** −0.668*** −0.676*** −0.665*** −0.697*** −0.664*** −0.683*** −0.666*** −0.922*** −0.869*** −0.889*** −0.878*** −1.106*** −1.024*** −1.052*** −1.033***
Male 0.148 0.127 0.118 0.125 −0.200* −0.034 −0.032 −0.034 −0.357*** −0.173* −0.175* −0.167* −0.676*** −0.449*** −0.453*** −0.450***
Married 0.049 0.006 0.016 0.012 0.110 0.078 0.071 0.083 0.114 0.071 0.079 0.080 0.153 0.079 0.064 0.054
Smoking 0.056 0.068 0.058 −0.418*** −0.414*** −0.394*** −0.460*** −0.439*** −0.432*** −0.668*** −0.614*** −0.607***
Alcohol 0.028 0.029 0.030 0.003 0.008 0.008 0.048 0.045 0.045 −0.015 −0.009 −0.006
Daily sleep time (≥ 8 h) −0.105 −0.111 −0.112 −0.046 −0.054 −0.055 −0.152 −0.168* −0.174* −0.250* −0.241* −0.238*
Nap 0.255** 0.241** 0.234** 0.126 0.112 0.112 0.027 0.025 0.019 0.090 0.070 0.069
Internet use 0.299** 0.259* 0.254* 0.214* 0.176 0.169 0.324*** 0.272** 0.267** 0.519*** 0.416*** 0.411***
Social activities 0.153* 0.144 0.140 0.150* 0.146 0.132 0.188** 0.188** 0.183* 0.185 0.177 0.179
Residential area (Urban) 0.192* 0.188* 0.174* 0.162 0.236** 0.221** 0.440*** 0.439***
Life satisfaction (Dissatisfied) −0.472*** −0.478*** −0.676*** −0.667*** −0.572*** −0.555*** −0.571** −0.552**
Troubled with body pain −0.030 0.002 −0.085 −0.026
Hypertension −0.067 0.016 −0.142 0.008
Dyslipidaemia 0.144 0.415* 0.123 −0.093
Hyperglycaemia −0.367 −0.511* −0.014 −0.143
Cancer 0.173 0.236 −1.518** 0.357
Chronic lung disease −0.318 −0.305 −0.286 −0.471
Liver disease 0.113* 0.083 0.113* 0.041
Heart disease 0.357* 0.380* 0.695*** 0.173
Stroke −0.086 −0.550 −0.940** −0.435
Kidney disease 0.237 −0.259 0.079 0.010
Digestive system disease −0.178 −0.005 −0.230 −0.166
Emotional and mental disorders 0.003 −0.604 −1.248* −0.501
Memory related disease −0.526 −0.000 −1.147* −1.146
Arthritis −0.108 −0.224* −0.045 −0.139
Asthma 0.146 −0.243 −0.145 0.895
Open in a new tab

Values represent standardized coefficients (*p < 0.05, **p < 0.01, ***p < 0.001). Each model builds upon the previous one: The model presented in this table includes random intercepts for time and individual differences, Model 1: Demographics; Model 2: Model 1 + Lifestyle; Model 3: Model 2 + Environment; Model 4: Model 3 + Health Conditions. Group Illiterate: Illiterate; Group Primary: Primary school; Group Middle: Middle school; Group High: High school/vocational high school + Junior college or above.

Age-related cognitive decline was significant across all educational groups (p < 0.001), with the magnitude of decline intensifying at higher educational levels (illiterate: β = −0.665; high-education: β = −1.033). This pattern suggests an accelerated rate of cognitive decline among individuals with higher educational attainment.

Lifestyle factors showed education-dependent effects: smoking’s negative impact on cognition was more pronounced in higher education groups (high-education: β = −0.607, p < 0.001). Internet use demonstrated increasingly positive effects with higher educational levels (illiterate: β = 0.254; high-education: β = 0.411, p < 0.001). Notably, napping showed significant positive effects exclusively in the illiterate group (β = 0.234, p < 0.01).

Environmental factors also exhibited educational gradients. The positive impact of urban residence strengthened with increasing educational levels (illiterate: β = 0.188; high-education: β = 0.439, p < 0.001). Life satisfaction showed consistent negative associations across all educational strata (β range: −0.478 to −0.667, p < 0.001).

Health conditions demonstrated varying impacts across educational levels. Heart disease showed the strongest positive association in the medium-education group (β = 0.695, p < 0.001). Memory-related diseases exhibited significant negative impacts in the medium-education group (β = −1.147, p < 0.05), while cancer showed significant negative associations exclusively in this group (β = −1.518, p < 0.01).

These findings underscore the complex interplay between educational attainment and cognitive function determinants. While the high-education group exhibited accelerated age-related cognitive decline, they also demonstrated enhanced benefits from protective factors such as Internet use and urban residence.

Sensitivity analysis

Alternative educational stratification schemes (Table 3; Supplementary Table S2) consistently confirmed the robust relationship between educational level and cognitive function. Under the three-level stratification scheme (Illiterate, Primary + Middle, High), baseline cognitive scores demonstrated a clear ascending pattern (15.770, 18.100, 20.127, p < 0.001). This gradient remained evident in the two-level stratification (Illiterate + Primary, Middle + High) (16.560, 19.124, p < 0.001).

Table 3.

Mixed effects models of cognitive function across Two-Level educational stratification.

Group Illiterate + Primary Group Middle + High
Variable Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4
(Intercept) 16.646*** 16.453*** 16.464*** 16.560*** 19.182*** 19.116*** 19.011*** 19.124***
Age −0.742*** −0.733*** −0.747*** −0.732*** −0.958*** −0.891*** −0.918*** −0.904***
Male 0.101 0.151* 0.146* 0.149* −0.439*** −0.233*** −0.236*** −0.234***
Married 0.066 0.024 0.029 0.030 0.132 0.066 0.073 0.069
Smoking −0.120 −0.110 −0.105 −0.567*** −0.531*** −0.522***
Alcohol 0.031 0.035 0.037 0.032 0.032 0.032
Daily sleep time (≥ 8 h) −0.076 −0.082 −0.084 −0.224** −0.231*** −0.234***
Nap 0.216*** 0.200*** 0.197*** 0.092 0.084 0.075
Internet use 0.271*** 0.224** 0.219** 0.487*** 0.395*** 0.389***
Social activities 0.181*** 0.173** 0.167** 0.206*** 0.202*** 0.199***
Residential area (Urban) 0.224*** 0.214*** 0.373*** 0.359***
Life satisfaction (Dissatisfied) −0.590*** −0.588*** −0.608*** −0.586***
Troubled with body pain −0.025 −0.064
Hypertension −0.052 −0.096
Dyslipidaemia 0.246* 0.142
Hyperglycaemia −0.381* −0.075
Cancer 0.334 −0.712
Chronic lung disease −0.287* −0.355*
Liver disease 0.098** 0.094**
Heart disease 0.443*** 0.526***
Stroke −0.305 −0.720**
Kidney disease −0.014 0.093
Digestive system disease −0.103 −0.249*
Emotional and mental disorders −0.345 −1.046*
Memory related disease −0.329 −1.073**
Arthritis −0.228** −0.177
Asthma −0.000 0.091
Open in a new tab

Values represent standardized coefficients (*p < 0.05, **p < 0.01, ***p < 0.001). Each model builds upon the previous one: The model presented in this table includes random intercepts for time and individual differences, Model 1: Demographics; Model 2: Model 1 + Lifestyle; Model 3: Model 2 + Environment; Model 4: Model 3 + Health Conditions. Group Illiterate: Illiterate; Group Primary: Primary school; Group Middle: Middle school; Group High: High school/vocational high school + Junior college or above.

Age-related effects further corroborated our primary findings: cognitive decline intensified with increasing educational level. The three-level stratification revealed progressively stronger age effects from illiterate to high-education groups (−0.665, −0.824, −1.033, p < 0.001), with this pattern remaining significant in the two-level classification.

Among lifestyle factors, Internet use maintained its education-dependent protective effect, showing strongest benefits in the high-education group (β = 0.411, p < 0.001). Similarly, smoking’s negative impact demonstrated increased magnitude at higher educational levels. Social and environmental factors maintained consistent patterns across stratification schemes, with urban residence and life satisfaction effects remaining stable across alternative groupings.

Health status effects demonstrated robust patterns: heart disease showed maximal positive association in the Primary + Middle group (β = 0.545, p < 0.001). Mental and emotional disorders and stroke exhibited their strongest negative impacts within this same group (β = −0.941, p < 0.05; β = −0.742, p < 0.01, respectively).

These sensitivity analyses not only validated our primary findings but also reinforced the significance of educational level as a key moderator of cognitive function determinants. The consistency across different stratification approaches strengthens the reliability of our main conclusions regarding education’s role in cognitive function modulation.

Formal interaction testing in the pooled model provided additional statistical support for the stratified findings. Educational level showed statistically significant overall modification of the associations of age [F(3) = 9.86, p < 0.001], sex [F(3) = 16.25, p < 0.001], and internet use [F(3) = 3.40, p = 0.017] with cognitive function, while urban residence and memory-related disease did not reach significance at this level. The direction of all significant interaction contrasts was broadly consistent with the stratified estimates in Table 2, suggesting that the observed between-group differences are likely to reflect genuine effect modification rather than chance findings (Supplementary Table S3).

Discussion

This large-scale longitudinal study, drawing on nationally representative CHARLS data spanning 2015–2020, systematically examined the moderating role of educational attainment in cognitive function determinants among middle-aged and older Chinese adults. Our findings suggest that educational level not only independently associates with cognitive performance but may also modulate its relationships with gender differences, Internet use, residential environment, and chronic disease burden.

Among the observed education-dependent patterns, the reversal of the gender-cognition association across educational strata is particularly striking. In the illiterate group, male sex corresponded with relatively higher cognitive scores, yet this direction appeared to invert among those with middle or higher education. This complexity is not entirely unexpected. Female superiority in verbal memory has been repeatedly documented, persisting even at advanced ages (Arenaza Urquijo et al., 2024; Marseglia et al., 2024), and some evidence points to a stronger association between college-level education and executive function preservation specifically among women (Lam et al., 2024). Within lower-education cohorts, the cognitive demands inherent in domestic responsibilities and community-level social roles may have afforded women a source of ongoing mental engagement—an informal pathway to cognitive maintenance that resonates with documented patterns of female neuropathological resilience (Guo et al., 2023). Conversely, in higher-education groups, longstanding disparities in occupational access and schooling quality between men and women may have resulted in unequal cognitive reserve accumulation despite nominally similar attainment levels (Arenaza Urquijo et al., 2024). These interpretations, however, rest on unmeasured constructs and cross-sectional comparisons within the data, and should be regarded as tentative pending more targeted investigation.

Across all educational subgroups, Internet use carried a positive association with cognitive function, with the magnitude of this association appearing to increase alongside educational attainment. Prior evidence has linked digital engagement to a lower likelihood of mild cognitive impairment (Li et al., 2022), and educational background has emerged as a meaningful correlate of how older adults engage with online environments (Macdonald and Hülür, 2021). The relationship between digital exclusion and cognitive disadvantage also appears socioeconomically patterned (Wang et al., 2024), and individuals with higher educational attainment tend to show both more active internet engagement and more favorable cognitive profiles (Afsar, 2013). A plausible reading of the education-graded association is that it captures variation in the cognitive intensity of online activity rather than differences in connectivity per se—higher-educated users may be more inclined toward information-seeking and learning-oriented tasks. That said, the binary measurement of internet use in this study precludes any examination of frequency or content, and reverse causation—whereby better cognitive function facilitates internet adoption—cannot be ruled out from observational data alone. The pattern nonetheless raises the possibility that digital inclusion initiatives for educationally disadvantaged older adults may need to prioritize skill development alongside access.

A particularly counterintuitive observation concerns memory-related disease, whose negative association with cognitive function was most pronounced in the high-education group. Rather than contradicting cognitive reserve theory, this finding sits comfortably within threshold-based formulations (Ophey et al., 2024; Guzzetti et al., 2019), which hold that compensatory neural processes may sustain function in the face of accumulating pathology—but only up to a point, beyond which decline may proceed more rapidly in those with higher reserve. Whether the pattern here reflects this compensatory ceiling effect or instead mirrors systematic differences in disease presentation or severity across educational groups remains an open question. At minimum, the observation cautions against treating high educational attainment as a reliable indicator of cognitive resilience in the presence of memory-related conditions, and points to the value of vigilant screening in this population.

Urban residence was associated with higher cognitive scores across all educational groups, though the strength of this association appeared to scale with educational level—a finding broadly in line with research linking urban living to more cognitively stimulating environments for older adults (Miura et al., 2023). The capacity to translate the structural advantages of urban settings into health gains has been linked to individual human capital (Yuan et al., 2023), while community-level social infrastructure may partially buffer against the cognitive costs of urban stressors (Tian et al., 2024). Whether the education-dependent amplification of the urban association observed here reflects active resource utilization, residential selection, or residual confounding is difficult to assess within the current observational framework.

The main strengths of this study are as follows: Firstly, based on large-sample and multi-time-point data, it can better control confounding factors and observe time trends. Secondly, the mixed-effects model was used to consider individual differences and time effects. Thirdly, the robustness of the results was verified through sensitivity analysis of multiple grouping schemes.

This study has several methodological considerations that limit the interpretation of these findings. Participant continuity across all three waves was incomplete, introducing the possibility of selective attrition and cohort-related bias. The cognitive instrument, though widely adopted in aging research, covers a restricted set of domains and may lack the sensitivity to detect early or domain-specific changes. The binary operationalization of internet use represents a notable constraint, collapsing variation in frequency, purpose, and proficiency into a single indicator. Beyond these measurement issues, confounding by variables not captured in CHARLS—among them physical activity patterns, nutritional intake, and finer-grained socioeconomic indicators—cannot be adequately controlled. These considerations suggest that the present findings are more appropriately viewed as generating hypotheses than as establishing firm conclusions, and that causal inference would require evidence from experimental or natural experiment designs.

Based on the research findings, we suggest: (1) Formulating differential cognitive function maintenance strategies for populations with different educational levels, with particular attention to the education-dependency of gender differences. (2) Strengthening digital skills training for people with a low educational level and attaching importance to the protective effect of Internet use on cognitive function. (3) Considering differences in educational level in chronic disease management and carrying out targeted health education. (4) Designing social participation forms suitable for different educational levels to improve participation.

Future research suggestions: (1) Conducting longer-term follow-up studies to deeply explore the causal relationships of influencing factors. (2) Using more refined cognitive function assessment tools. (3) Exploring the interaction between educational level and other socioeconomic factors. (4) Conducting intervention studies to verify the practical value of the findings of this study.

Conclusion

Drawing on longitudinal CHARLS data, this study suggests that educational attainment may serve as a meaningful modifier of the factors associated with cognitive function in older Chinese adults. Age-related cognitive decline appeared more pronounced at higher educational levels, the gender-cognition association shifted across strata, and the benefits of Internet use and urban residence both appeared to strengthen alongside educational attainment, while memory-related disease carried a particularly notable negative association in the high-education group. These observations point to the potential value of incorporating educational background into the design of targeted cognitive health initiatives.

Acknowledgments

The authors wish to thank all the participants and staff of the China Health and Retirement Longitudinal Study (CHARLS).

Funding Statement

The author(s) declared that financial support was received for this work and/or its publication. This study was funded by the Chengdu Medical College (Grant No. 24LHLNYX1-10).

Footnotes

Edited by: Eirini Mavritsaki, Birmingham City University, United Kingdom

Reviewed by: Eva Pettemeridou, University of Limassol, Cyprus

Nikola Savic, Singidunum University, Serbia

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement

The studies involving humans were approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052-11015). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

DD: Conceptualization, Data curation, Formal analysis, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. BX: Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing. YD: Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing. PY: Conceptualization, Methodology, Writing – original draft, Writing – review & editing. NZ: Conceptualization, Formal analysis, Methodology, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing. SW: Conceptualization, Data curation, Funding acquisition, Methodology, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg.2026.1564721/full#supplementary-material

table_1.docx (81.2KB, docx)
table_2.docx (30.7KB, docx)
table_3.docx (19.7KB, docx)

References

  1. Afsar B. (2013). The relation between internet and social media use and the demographic and clinical parameters, quality of life, depression, cognitive function and sleep quality in hemodialysis patients. Gen. Hosp. Psychiatry 35, 625–630. doi: 10.1016/j.genhosppsych.2013.05.001, [DOI] [PubMed] [Google Scholar]
  2. Amland T., Grande G., Scherer R., Lervåg A., Melby-Lervåg M. (2024). Cognitive factors underlying mathematical skills: a systematic review and meta-analysis. Psychol. Bull. 151, 88–129. doi: 10.1037/bul0000457 [DOI] [PubMed] [Google Scholar]
  3. Arenaza Urquijo E. M., Boyle R., Casaletto K., Anstey K. J., Vila-Castelar C., Colverson A., et al. (2024). Sex and gender differences in cognitive resilience to aging and Alzheimer's disease. Alzheimers Dement. 20, 5695–5719. doi: 10.1002/alz.13844 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Barba C., Garcia A., Clay O. J., Wadley V. G., Andel R., Dávila A. L., et al. (2021). Quality of education and late-life cognitive function in a population-based sample from Puerto Rico. Innov. Aging 5:igab16. doi: 10.1093/geroni/igab016, [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bratsberg B., Fjell A. M., Rogeberg O. J., Skirbekk V. F., Walhovd K. B. (2024). Differences in cognitive function at 18 y of age explain the association between low education and early dementia risk. Proc. Natl. Acad. Sci. USA 121:e1882950175. doi: 10.1073/pnas.2412017121, [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Clare L., Wu Y., Teale J. C., Wu Y. T., MacLeod C., Matthews F., et al. (2017). Potentially modifiable lifestyle factors, cognitive reserve, and cognitive function in later life: a cross-sectional study. PLoS Med. 14:e1002259. doi: 10.1371/journal.pmed.1002259, [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Deng J., Huang K., Hu X., Deng J.-H., Huang K.-Y., Hu X.-X., et al. (2019). Midlife long-hour working and later-life social engagement are associated with reduced risks of mild cognitive impairment among community-living Singapore elderly. J. Alzheimer's Dis 67, 1067–1077. doi: 10.3233/jad-180605, [DOI] [PubMed] [Google Scholar]
  8. GBD 2019 Dementia Forecasting Collaborators (2022). Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the global burden of disease study 2019. Lancet Public Health 7, e105–e125. doi: 10.1016/s2468-2667(21)00249-8, [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Ghahremani M., Smith E. E., Ismail Z. (2024). Improving dementia prognostication in cognitively normal older adults: conventional versus novel approaches to modelling risk associated with neuropsychiatric symptoms. Br. J. Psychiatry 226, 129–136. doi: 10.1192/bjp.2024.136, [DOI] [PubMed] [Google Scholar]
  10. Guo L., Cao J., Hou J., Li Y., Huang M., Zhu L., et al. (2023). Sex specific molecular networks and key drivers of Alzheimer’s disease. Mol. Neurodegener. 18:39. doi: 10.1186/s13024-023-00624-5, [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Guzzetti S., Mancini F., Caporali A., Manfredi L., Daini R. (2019). The association of cognitive reserve with motor and cognitive functions for different stages of Parkinson's disease. Exp. Gerontol. 115, 79–87. doi: 10.1016/j.exger.2018.11.020, [DOI] [PubMed] [Google Scholar]
  12. He C., Zhou Z., Kan M., He C. Y. Y., Kan M. M. P., Chan D. H. Y., et al. (2024). Modifiable risk factors for mild cognitive impairment among cognitively normal community-dwelling older adults: a systematic review and meta-analysis. Ageing Res. Rev. 99:102350. doi: 10.1016/j.arr.2024.102350, [DOI] [PubMed] [Google Scholar]
  13. Ihle A., Bavelier D., Maurer J., Oris M., Kliegel M. (2020). Internet use in old age predicts smaller cognitive decline only in men. Sci. Rep. 10:8969. doi: 10.1038/s41598-020-65846-9, [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Jia F., Liu H., Xu K., Sun J., Zhu Z., Shan J., et al. (2022). Mediating effects of cognitive reserve on the relationship between frailty and cognition in older people without dementia. Eur Geriatr Med 13, 1317–1325. doi: 10.1007/s41999-022-00703-8, [DOI] [PubMed] [Google Scholar]
  15. Lam J. O., Whitmer R. A., Corrada M. M., Kawas C. H., Vieira K. E., Quesenberry C. P., et al. (2024). Gender differences in the association between education and late-life cognitive function in the LifeAfter90 study: a multiethnic cohort of the oldest–old. Alzheimers Dement. 20, 7547–7555. doi: 10.1002/alz.14217, [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Li W., Yue L., Xiao S. (2022). Association between internet use, cognitive function, and Globus pallidus volumes: a study among the elderly in Chinese communities. Front. Public Health 10:886974. doi: 10.3389/fpubh.2022.886974 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Luo H., Hu H., Zheng Z., Sun C., Yu K. (2024). The impact of living environmental factors on cognitive function and mild cognitive impairment: evidence from the Chinese elderly population. BMC Public Health 24:2814. doi: 10.1186/s12889-024-20197-2, [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Macdonald B., Hülür G. (2021). Internet adoption in older adults: findings from the health and retirement study. Cyberpsychol. Behav. Soc. Netw. 24, 101–107. doi: 10.1089/cyber.2019.0736, [DOI] [PubMed] [Google Scholar]
  19. Maksyutynska K., Stogios N., Prasad F., Gill J., Hamza Z., de R., et al. (2024). Neurocognitive correlates of metabolic dysregulation in individuals with mood disorders: a systematic review and meta-analysis. Psychol. Med. 54, 1245–1271. doi: 10.1017/S0033291724000345, [DOI] [PubMed] [Google Scholar]
  20. Marseglia A., Dartora C., Samuelsson J., Poulakis K., Mohanty R., Shams S., et al. (2024). Biological brain age and resilience in cognitively unimpaired 70-year-old individuals. Alzheimers Dement. 21:e14435. doi: 10.1002/alz.14435 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Miura K. W., Tokunaga S., Sekiguchi T., Sugimoto H., Otake-Matsuura M. (2023). Effect of home-based group conversation intervention using smartphone application on cognitive health and psychological well-being of older adults with subjective cognitive concerns in Japan: a randomized controlled trial protocol. Front. Psychol. 14:1114790. doi: 10.3389/fpsyg.2023.1114790, [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Nguyen H. T., Kirk J. K., Arcury T. A., Ip E. H., Grzywacz J. G., Saldana S. J., et al. (2013). Cognitive function is a risk for health literacy in older adults with diabetes. Diabetes Res. Clin. Pract. 101, 141–147. doi: 10.1016/j.diabres.2013.05.012, [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Ophey A., Wirtz K., Wolfsgruber S., Balzer-Geldsetzer M., Berg D., Hilker-Roggendorf R., et al. (2024). Mid- and late-life lifestyle activities as main drivers of general and domain-specific cognitive reserve in individuals with Parkinson's disease: cross-sectional and longitudinal evidence from the LANDSCAPE study. J. Neurol. 271, 5411–5424. doi: 10.1007/s00415-024-12484-0, [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Sanchez-Orti J. V., Balanza-Martinez V., Correa-Ghisays P., Selva‐Vera G., Vila‐Francés J., Magdalena‐Benedito R., et al. (2022). Specific metabolic syndrome components predict cognition and social functioning in people with type 2 diabetes mellitus and severe mental disorders. Acta Psychiatr. Scand. 146, 215–226. doi: 10.1111/acps.13433, [DOI] [PubMed] [Google Scholar]
  25. Sedgwick P. (2015). Bias in observational study designs: cross sectional studies. BMJ 350:h1286. doi: 10.1136/bmj.h1286, [DOI] [PubMed] [Google Scholar]
  26. Siette J., Georgiou A., Brayne C., Westbrook J. I. (2020). Social networks and cognitive function in older adults receiving home- and community-based aged care. Arch. Gerontol. Geriatr. 89:104083. doi: 10.1016/j.archger.2020.104083, [DOI] [PubMed] [Google Scholar]
  27. Singh-Manoux A., Marmot M. G., Glymour M., Sabia S., Kivimäki M., Dugravot A. (2011). Does cognitive reserve shape cognitive decline? Ann. Neurol. 70, 296–304. doi: 10.1002/ana.22391, [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Tahami Monfared A. A., Byrnes M. J., White L. A., Zhang Q. (2022). The humanistic and economic burden of Alzheimer's disease. Neurol Ther 11, 525–551. doi: 10.1007/s40120-022-00335-x, [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Tian W., Cao K., Kwan M., Kwan M.-P., Chiu M. Y. L., Chen H. (2024). How does urbanization affect the cognitive function among older adults: a geospatial analysis in China. Health Place 88:103259. doi: 10.1016/j.healthplace.2024.103259, [DOI] [PubMed] [Google Scholar]
  30. Wang Y., Wu Z., Duan L., Liu S., Chen R., Sun T., et al. (2024). Digital exclusion and cognitive impairment in older people: findings from five longitudinal studies. BMC Geriatr. 24:406. doi: 10.1186/s12877-024-05026-w [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Yuan Y., Peng C., Burr J. A., Lapane K. L. (2023). Frailty, cognitive impairment, and depressive symptoms in Chinese older adults: an eight-year multi-trajectory analysis. BMC Geriatr. 23:843. doi: 10.1186/s12877-023-04554-1, [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Zhao Y., Hu Y., Smith J. P., Strauss J., Yang G. (2014). Cohort profile: the China health and retirement longitudinal study (CHARLS). Int. J. Epidemiol. 43, 61–68. doi: 10.1093/ije/dys203, [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Zhu Y., Yin H., Zhong X., Zhang Q., Wang L., Lu R., et al. (2024). Exploring the mediating roles of depression and cognitive function in the association between sarcopenia and frailty: a cox survival analysis approach. J. Adv. Res. 76, 605–613. doi: 10.1016/j.jare.2024.12.021 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

table_1.docx (81.2KB, docx)
table_2.docx (30.7KB, docx)
table_3.docx (19.7KB, docx)

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Articles from Frontiers in Psychology are provided here courtesy of Frontiers Media SA

RESOURCES