Bidirectional Association Between Lifestyle and Cognitive Function in Aging Adults

Abstract

Lifestyle research on cognitive health has been largely unidirectional and cross-sectional, with limited attention to multidimensional and bidirectional relationships; this study examines longitudinal associations in middle- and older adults. Data from the Korean Longitudinal Study of Aging (2006–2020) included physiological and psychosocial lifestyle factors and cognition assessed by the Korean version of the Mini-Mental State Examination (K-MMSE). Analyses used Mplus 8.9 with cross-lagged panel model (CLPM) and random intercept CLPM (RI-CLPM), adjusting for sex, age, smoking, drinking, and disability. CLPM showed significant lagged effects of lifestyle on cognition across waves (β = .010 to .013, p < .01), with intermittent reverse effects (β = .013 to .015, p < .05). The RI-CLPM confirmed a stable effect of lifestyle on cognition (β = .013 to .017, p < .05), whereas effects from cognition to lifestyle were mixed and nonsignificant (β = −.054 to .034, p > .05). A multidimensional lifestyle was associated with cognitive maintenance, whereas cognitive decline shows variable influences on lifestyle.

Plain Language Summary

The Role of Lifestyle in Cognitive Function Among Older Adults Over Time

As people age, changes in memory and thinking become more common. Understanding what helps maintain brain health is important for supporting independence and quality of life. Many previous studies have focused on single lifestyle factors—such as exercise or diet—and often looked at them at only one point in time. This study examined how a combination of everyday habits relates to thinking ability over a long period. We used data from over 5,101 middle-aged and older adults in Korea who were followed for 14 years. We looked at several aspects of lifestyle, including physical activity, eating habits, social relationships, and participation in work or community activities. We then examined how these factors were related to changes in cognitive function over time. The findings showed that people with healthier and more active lifestyles tended to maintain better cognitive function over time. In contrast, declines in cognitive function did not consistently lead to changes in lifestyle. This suggests that daily habits may play a more important role in supporting brain health than changes in cognition do in shaping lifestyle. Overall, the results highlight the importance of maintaining a balanced and active lifestyle that includes both physical and social activities. Encouraging older adults to stay physically active, socially connected, and engaged in meaningful daily activities may help support their cognitive health. These findings are especially relevant for health care professionals, including occupational therapists, who work to promote healthy aging and independence in later life.

Keywords

older adults lifestyle cognitive function longitudinal study cross-lagged panel model random intercept cross-lagged panel model

Introduction

Population aging is accelerating worldwide, highlighting the importance of maintaining health and quality of life in later adulthood (World Health Organization [WHO], 2021). Age-related cognitive decline compromises daily functioning and heightens dementia risk, thereby threatening older adults’ independence and well-being (Campbell et al., 2025). In 2020, an estimated 57 million people were living with dementia, with nearly 10 million new cases annually, placing heavy burdens on individuals, families, and societies (Nandi et al., 2024; WHO, 2025). Because cognitive function declines gradually across the life course, identifying factors linked to cognitive decline and dementia is critical for prevention (Hou et al., 2024).

Lifestyle, defined as habitual patterns of daily behavior, is a key determinant of health (Brivio et al., 2023). Health research has traditionally focused on physiological factors such as physical activity, sleep, diet, and substance use (Lippman et al., 2024; Liu et al., 2022). More recently, holistic perspectives have emphasized psychosocial elements, broadening the concept of lifestyle to include social and contextual dimensions (Burke & Dunne, 2022). Park et al. (2023) described a healthy lifestyle as a multidimensional construct encompassing physical activity, nutrition, social relationships, and occupational participation, which together shape health and quality of life. Such multidimensional lifestyles are strongly associated with cognitive preservation, underscoring the need for longitudinal examination of their relationships (Ngandu et al., 2015).

Cognitive function, the capacity to process environmental stimuli, is shaped by repeated behaviors and social interactions (Hsu & Bai, 2022). Cognitive reserve theory suggests that education, physical activity, occupational tasks, and social participation enhance brain plasticity and delay decline (Sumowski, 2015). Regular activity, balanced nutrition, and social engagement thus provide cognitive stimulation and protect brain health (Wen et al., 2025). Examining causal pathways between lifestyle and cognition is therefore crucial to identify modifiable targets for preventing or managing decline (Livingston et al., 2024).

Lifestyle and cognition can be understood within broader theoretical frameworks. Ecological Systems Theory (Bronfenbrenner, 1979) highlights that lifestyle behaviors are shaped by physiological factors and multilayered contextual interactions, supporting the multidimensional approach of this study. Cognitive Reserve Theory proposes that regular lifestyle engagement enhances neuroplasticity and delays decline, while recognizing that deterioration can in turn undermine lifestyle behaviors over time (Jauny et al., 2022; Stern, 2002). Activity Theory further suggests that sustained participation in personal and social activities promotes cognitive health (Havighurst, 1961), consistent with the bidirectional link between decline and reduced social or occupational engagement.

Healthy lifestyle factors such as physical activity, social participation, and diet are associated with reduced risk of cognitive decline (Ngandu et al., 2015). Conversely, cognitive impairment may lower physical activity, increase social isolation, and promote unhealthy eating, thereby worsening cognitive health (Sutton et al., 2022). Yet causal inference between lifestyle and cognition remains limited. Many longitudinal studies emphasize average trends rather than the dynamic, bidirectional interplay between variables (Ngandu et al., 2022). Short follow-up periods and limited time points further constrain interpretation, conflating between-person differences with within-person changes (Borgeest et al., 2020; Kimura et al., 2019). Treating associations as static or cross-sectional may thus underestimate the complexity of lifestyle–cognition links (Heneghan et al., 2023). Addressing these gaps requires methods that capture within-person change and bidirectional interactions across time (Bloomberg et al., 2024).

The present study investigates the longitudinal association between lifestyle and cognitive function in middle- and older-aged adults from a bidirectional perspective. Specifically, we applied cross-lagged panel models (CLPMs) and random intercept cross-lagged panel models (RI-CLPMs) to examine temporal pathways between lifestyle—encompassing physical activity, nutrition, social relationships, and occupational participation—and cognitive function, while controlling for individual-level covariates.

Method

Study Design

This study employed a longitudinal quantitative design using secondary data from the Korean Longitudinal Study of Aging (KLoSA). The analysis examined long-term associations between lifestyle factors and cognitive function among older adults. This research was approved by the Institutional Review Board of Yonsei University (approval number: 1041849-202508-SB-169-01).

Data Source and Sample

The data for this study were drawn from the KLoSA, a nationally representative panel survey conducted by the Korea Employment Information Service to provide systematic statistics on population aging. KLoSA has followed household members aged 45 years and older biennially, collecting extensive information on demographic characteristics, family composition, health status, employment, occupation, income and consumption, subjective expectations, and quality of life, thereby providing a comprehensive dataset for socioeconomic and health policy research.

The baseline sample consisted of 10,254 respondents in the first wave. By the eighth wave, the retained sample included 5,717 individuals, with a panel retention rate of 77.1% (Korea Employment Information Service, 2022). This study utilized data from eight waves (time points) spanning 2006 (Wave 1) to 2020 (Wave 8). The dataset was merged across waves by respondent ID to construct repeated measures, yielding 11,174 person-wave observations. We excluded 1,202 participants who were newly enrolled or had incomplete responses, as well as 4,871 cases without integrated sampling weights.

The final analytic sample comprised 5,101 respondents. Missing data on the Mini-Mental State Examination (MMSE), used as the measure of cognitive function, ranged from 3.9% to 5.0% across waves. To address missingness while minimizing bias and preserving statistical power, we employed full information maximum likelihood (FIML), which uses all available information to estimate parameters under the assumption of missing at random.

Measurements

Observed Variables-Lifestyle

Lifestyle factors influencing health were defined based on the theoretical framework of Lifestyle-DEPER (Decision, Execution, Personal Factor, Environment, Resources) (Park et al., 2023). Health-related lifestyle indicators were operationalized into four domains: physical activity, nutrition, social relationships, and occupational participation. For this study, each domain was recoded into categorical variables.

Physical activity was assessed using two items: frequency of weekly exercise and average duration per session. Weekly total activity time (minutes) was calculated by multiplying these two variables. Following WHO (2020) guidelines, less than 75 minutes per week was coded as “1 = low activity,” 75–150 minutes as “2 = moderate activity,” and 150 minutes or more as “3 = sufficient activity.”

Nutrition was assessed using meal consumption patterns across two consecutive days (yesterday and the day before) (Cho, 2022). Skipping two or more meals per day (out of breakfast, lunch, dinner) was coded as a “skipped day.” Respondents who skipped meals on both days were coded as “1 = persistent skipping,” one skipped day as “2 = irregular intake,” and no skipped days as “3 = regular intake” (Yi et al., 2015).

Social relationships were measured using household size, frequency of contact with close friends, and contact with non-cohabiting parents and offspring (Choi & Jeon, 2021). Household status was dichotomized as “0 = living alone” and “1 = living with others,” and contact variables were dichotomized as less than once per week “0 = low connectivity” and once per week or more “1 = high connectivity.” The four indicators were summed to create a composite score (range: 0–4), which was categorized as “1 = disconnected” (0), “2 = limited” (1), and “3 = active” (≥2).

Occupational participation was measured using seven items on engagement in productive activities and participation in organized groups (religious, social, leisure, cultural, and sports) (Lim & Park, 2023). Productive activity was recoded as a binary variable: “0 = not engaged” and “1 = engaged,” and group participation was recoded as “0 = less than once per week” and “1 = regular participation” (≥ once per week). The indicators were summed to create a composite score (range: 0–7), which was categorized as “1 = no participation” (0), “2 = limited participation” (1), and “3 = diverse participation” (≥2).

The composite lifestyle index was constructed as a continuous variable by summing the recoded scores across the four domains.

Observed Variables—Cognitive Function

Cognitive function was assessed using the Korean version of the Mini-Mental State Examination (K-MMSE), a widely used screening tool for older adults. The K-MMSE evaluates five domains: orientation, memory, attention, language, and visuospatial ability (Kwon, 1989; Park, 1989). In this study, the K-MMSE total score was treated as a continuous variable, with higher scores indicating better cognitive function.

Sociodemographic Factors and Covariates

The sociodemographic variables included in this study were gender, age, educational level, residential area, smoking, alcohol consumption, and disability status. Gender was coded as male (0) and female (1), and age was recategorized as ≤64 years (0) and ≥65 years (1). Educational level was classified as no formal education (0), elementary school graduate (1), middle school graduate (2), high school graduate (3), and college graduate or higher (4). The residential area was categorized into metropolitan (1), small- to medium-sized cities (2), and rural areas (3). Marital status, smoking, alcohol consumption, and disability status were coded as absent (0) or present (1).

To estimate the temporal effects between lifestyle and cognitive function, gender, age, smoking, alcohol consumption, and disability status were included as covariates. These covariates were incorporated into the observed variables at each wave to control for the influence of individual background characteristics.

Data Analysis

To examine the longitudinal relationship between lifestyle and cognitive function, we applied both the CLPM and the RI-CLPM. All analyses were conducted using Mplus version 8.9, employing the Robust Maximum Likelihood Estimation (MLR) method. Missing values were handled using FIML. CLPM allows for the estimation of autoregressive and cross-lagged effects over time, but it does not control for stable between-person heterogeneity. To address this limitation, RI-CLPM was employed, which separates stable between-person differences through random intercepts and thereby minimizes bias in cross-lagged effects, allowing for the analysis of within-person changes (Hamaker et al., 2015; Mulder & Hamaker, 2021).

Descriptive statistics were used to examine the sociodemographic characteristics of participants, and means, standard deviations, skewness, kurtosis, and Pearson correlations were computed for key study variables. Based on CLPM and RI-CLPM, causal associations between lifestyle and cognitive function were analyzed at both the between-person and within-person levels. Six constrained models were compared to evaluate model fit: (1) an unconstrained model, (2–3) equality constraints on autoregressive effects, (4–5) equality constraints on cross-lagged effects, and (6) equality constraints on error covariances (Supplemental 1).

Model fit was assessed using multiple indices: the Comparative Fit Index (CFI) and Tucker–Lewis Index (TLI), with values ≥ 0.90 indicating acceptable fit; the Root Mean Square Error of Approximation (RMSEA), with values ≤ 0.08 considered acceptable and ≤ 0.05 considered good; and the Standardized Root Mean Square Residual (SRMR), with values ≤ 0.08 indicating acceptable fit (Kline, 2016). For the final selected model, covariates including sex, age, smoking, drinking, and disability status were included to control for potential confounding effects.

Results

Demographic Characteristics of the Sample

Among the study participants, women accounted for 2,981 individuals (58.4%), a higher proportion than men. By age group, 3,667 participants (71.9%) were aged 64 years or younger, while 1,434 participants (28.1%) were aged 65 years or older. Regarding education level, 635 (12.4%) had no formal education, 1,420 (27.8%) had completed elementary school, 953 (18.7%) had completed middle school, 1,536 (30.1%) had completed high school, and 557 (10.9%) had attained college or higher education. In terms of residential area, the largest group resided in metropolitan areas (n = 2,176, 42.7%), followed by small-to-medium cities (n = 1,618, 31.7%) and rural areas (n = 1,307, 25.6%). Marital status showed that 4,363 participants (85.5%) had a spouse. For health-related behaviors, 4,200 (82.3%) were non-smokers and 3,098 (60.7%) were non-drinkers. In addition, 4,837 (94.8%) reported having no disabilities (Table 1).

Table 1.

Sociodemographic Characteristics (N = 5,101).

Demographic information	n	%
Sex
Male	2,120	41.6
Female	2,981	58.4
Age ranges
≤64	3,667	71.9
≥65	1,434	28.1
Educational attainment
No formal education	635	12.4
Elementary school	1,420	27.8
Middle school	953	18.7
High school	1,536	30.1
College or higher	557	10.9
Region of residence
Metropolitan area	2,176	42.7
Medium-to-small city	1,618	31.7
Rural area	1,307	25.6
Presence of spouse
Yes	4,363	85.5
No	738	14.5
Smoking
Yes	901	17.7
No	4,200	82.3
Drinking
Yes	2,003	39.3
No	3,098	60.7
Disability status
With a disability	264	5.2
Without a disability	4,837	94.8

Descriptive Statistics and Correlations of Key Variables

The correlations of key variables across measurement waves are presented in Supplemental 2. None of the correlation coefficients exceeded .90, indicating no multicollinearity concerns. Lifestyle variables at each wave were positively and significantly correlated (r = .203 to .616, p < .01). Adjacent waves of lifestyle showed relatively high correlations, confirming temporal stability across repeated measures. Cognitive function variables were also significantly and positively correlated across waves (r = .509 to .741, p < .01), with stronger autocorrelations observed in later waves.

The correlations between lifestyle and cognitive function were statistically significant and positive at all measurement points, ranging from weak (r = .137, p < .01) to moderate levels (r = .422, p < .01). These associations tended to become stronger in the later waves, suggesting that the relationship between lifestyle and cognitive function increased over time.

Normality assumptions for the distribution of each variable across waves were assessed using skewness and kurtosis. All absolute values were within the recommended thresholds (skewness < 2, kurtosis < 7), satisfying the criteria for normality (Kline, 2016).

Model Fit Analysis of Lifestyle and Cognitive Function

The model fit indices for the equality-constrained models of the CLPM and RI-CLPM are presented in Table 2. For the CLPM, the unconstrained model (Model 1) demonstrated the best overall fit (CFI = .993, TLI = .989, RMSEA = .021, SRMR = .015) and the lowest Akaike information criterion (AIC; 326,129.463). However, when considering statistical parsimony, the constrained model with equal cross-lagged effects of lifestyle (Model 4) showed comparable fit indices (CFI = .993, TLI = .989, RMSEA = .021, SRMR = .017) and yielded the lowest Bayesian information criterion (BIC; 326,636.878), suggesting that this model was more parsimonious. Similarly, for the RI-CLPM, Model 1 exhibited acceptable fit (CFI = .989, TLI = .981, RMSEA = .028, SRMR = .034), while Model 4 provided nearly equivalent fit (CFI = .988, TLI = .981, RMSEA = .027, SRMR = .035) but demonstrated the lowest BIC (326,802.075), indicating greater parsimony.

Table 2.

Model Fit Indices for CLPM and RI-CLPM.

Model	Model fit index
Model	χ²	df	CFI	TLI	RMSEA	SRMR	AIC	BIC
CLPM
1	234.929	72	.993	.989	.021	.015	326,129.463	326,652.438
2	1054.173	78	.960	.939	.050	.134	327,277.542	327,761.294
3	874.348	78	.967	.950	.045	.162	327,077.960	327,561.712
4	260.745	78	.993	.989	.021	.017	326,153.126	326,636.878
5	287.118	78	.991	.987	.023	.023	326,187.576	326,671.328
6	363.109	79	.988	.983	.027	.035	326,295.479	326,772.694
RI-CLPM
1	339.985	69	.989	.981	.028	.034	326,278.111	326,820.698
2	535.508	75	.981	.970	.035	.053	326,547.617	327,050.981
3	550.816	75	.981	.969	.035	.069	326,589.702	327,093.066
4	363.835	75	.988	.981	.027	.035	326,298.711	326,802.075
5	368.492	75	.988	.981	.028	.036	326,304.567	326,807.931
6	456.939	76	.985	.976	.031	.036	326,373.455	326,870.281
CLPM (with covariates)
4^C	219.406	78	.995	.987	.019	.011	323,846.603	324,853.331
RI-CLPM (with covariates)
4^C	308.247	75	.992	.978	.025	.019	323,969.338	324,995.677

Note. Model 4^C includes covariates (sex, age, smoking, drinking, and disability status). Bold values indicate the selected final model based on overall model fit indices. χ² = chi-square; df = degrees of freedom; CFI = Comparative Fit Index; TLI = Tucker–Lewis Index; RMSEA = Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Square Residual; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; CLPM = Cross-Lagged Panel Model; RI-CLPM = Random Intercept Cross-Lagged Panel Model.

When covariates (age, sex, smoking, drinking, disability status) were incorporated into both the CLPM and RI-CLPM Model 4, the fit indices improved compared to the baseline models. Therefore, the constrained models with covariates were deemed the most appropriate analytical models, as they offered both parsimony and interpretability (Little, 2024).

Main Findings of CLPM and RI-CLPM

The standardized path coefficients for CLPM Model 4 and RI-CLPM Model 4 with covariates are presented in Table 3. Both models included autoregressive effects, cross-lagged effects, and synchronous correlations, while RI-CLPM additionally estimated between-person effects.

Table 3.

Standardized Path Coefficients of CLPM and RI-CLPM.

Parameter	CLPM				RI-CLPM
	Model 4		Model 4^C		Model 4		Model 4^C
	β	SE	β	SE	β	SE	β	SE
Between-person	–	–	–	–	.798***	.091	.626***	.100
Autoregressive effects
LS1 → LS2	.401***	.015	.386***	.016	.391***	.017	.372***	.018
LS2 → LS3	.821***	.040	.817***	.042	.780***	.051	.763***	.056
LS3 → LS4	1.016***	.032	1.003***	.034	.947***	.046	.925***	.057
LS4 → LS5	.652***	.022	.635***	.023	.605***	.034	.580***	.040
LS5 → LS6	.954***	.029	.945***	.032	.852***	.060	.815***	.072
LS6 → LS7	.799***	.027	.794***	.028	.709***	.054	.681***	.067
LS7 → LS8	.860***	.031	.836***	.033	.742***	.063	.698***	.073
CF1 → CF2	.643***	.014	.559***	.017	.209***	.022	.268***	.028
CF2 → CF3	.909***	.020	.884***	.028	.228***	.030	.314***	.045
CF3 → CF4	.982***	.021	.971***	.027	.336***	.029	.447***	.039
CF4 → CF5	.921***	.019	.872***	.022	.354***	.026	.417***	.035
CF5 → CF6	.975***	.015	.949***	.021	.360***	.026	.414***	.038
CF6 → CF7	.936***	.014	.928***	.021	.410***	.025	.473***	.037
CF7 → CF8	.944***	.015	.891***	.021	.436***	.023	.457***	.033
Cross-lagged effects
LS1 → CF2	.018***	.004	.013**	.004	.020**	.007	.017*	.007
LS2 → CF3	.015***	.003	.012**	.004	.017**	.006	.015*	.006
LS3 → CF4	.015***	.003	.012**	.004	.017**	.006	.015*	.006
LS4 → CF5	.015***	.003	.011**	.003	.017**	.006	.015*	.006
LS5 → CF6	.015***	.003	.011**	.003	.016**	.006	.014*	.006
LS6 → CF7	.013***	.003	.010**	.003	.015**	.005	.013*	.005
LS7 → CF8	.014***	.003	.010**	.003	.015**	.005	.013*	.005
CF1 → LS2	.127***	.013	.107***	.014	.018	.015	.034*	.016
CF2 → LS3	.053**	.016	.035*	.018	–.074***	.019	–.054**	.019
CF3 → LS4	.112***	.014	.091***	.015	.016	.016	.029	.017
CF4 → LS5	.126***	.013	.100***	.014	.021	.017	.027	.017
CF5 → LS6	.037*	.016	.017	.016	–.049*	.023	–.037	.023
CF6 → LS7	.038*	.016	.017	.016	–.046*	.022	–.041	.022
CF7 → LS8	.065***	.017	.038*	.017	.011	.024	.007	.023
Synchronous correlation
LS1 with CF1	.190***	.014	.136***	.015	.107***	.017	.095***	.017
LS2 with CF2	.075***	.013	.079***	.013	.085***	.016	.083***	.016
LS3 with CF3	.024*	.011	.027*	.011	.033*	.016	.026	.016
LS4 with CF4	.038***	.010	.037***	.010	–.002	.015	.002	.015
LS5 with CF5	.067***	.011	.067***	.011	.056**	.017	.056**	.017
LS6 with CF6	.036***	.010	.033**	.010	.069***	.017	.065***	.017
LS7 with CF7	.075***	.013	.075***	.013	.098***	.016	.096***	.016
LS8 with CF8	.119***	.014	.114***	.014	.155***	.017	.148***	.017

Note. ^CCovariates. LS = Lifestyle; CF = Cognitive Function; SE = Standard Error; CLPM = Cross-Lagged Panel Model; RI-CLPM = Random Intercept Cross-Lagged Panel Model.

p < .05. **p < .01. ***p < .001.

For both CLPM and RI-CLPM, the autoregressive paths of lifestyle and cognitive function in Model 4 were statistically significant (β = .209 to 1.016, p < .001). Cross-lagged effects were significant in most cases, except for the paths from cognition to lifestyle in RI-CLPM (β = .011 to .021, p > .05). Given that model fit indices improved with the inclusion of covariates, the interpretation primarily focused on the constrained models with covariates (Model 4^C).

In the CLPM (Model 4^C), as illustrated in Figure 1, the autoregressive paths of lifestyle and cognitive function were significant across all waves (β = .386 to 1.003, p < .001). Cross-lagged effects from lifestyle to subsequent cognitive function were consistently significant across waves (β = .010 to .013, p < .01), while the reverse paths from cognitive function to lifestyle were significant only at some time points (β = .013 to .015, p < .05). Synchronous correlations were significant and positive at all waves (β = .027 to .136, p < .05).

Figure 1.

CLPM of lifestyle (LS) and cognitive function (CF) across T1–T8.

In the RI-CLPM (Model 4^C), as illustrated in Figure 2, the between-person effect was statistically significant (β = .626, p < .001). At the within-person level, autoregressive effects of both lifestyle and cognitive function remained stable across all waves (β = .268 to .925, p < .001). Cross-lagged effects from lifestyle to subsequent cognitive function were consistently significant (β = .013 to .017, p < .05). By contrast, the reverse effects from cognitive function to lifestyle included both negative and positive coefficients but were only partially significant (β = −.054 to .034, p < .05). Synchronous correlations were generally significant, except for LS3 ↔ CF T3 and T4, where they did not reach significance (β = .056 to .148, p < .01).

Figure 2.

RI-CLPM of lifestyle (LS) and cognitive function (CF) across T1–T8.

Discussion

The multidimensional lifestyle of older adults is a protective factor that strengthens cognitive reserve and lowers the risk of cognitive decline and dementia (Kivipelto et al., 2020; Lourida et al., 2019). This study used the CLPM and RI-CLPM to examine the bidirectional relationship between lifestyle and cognitive function in middle-aged and older adults. CLPM results showed that lifestyle consistently predicted cognitive function across all time points. Although the effect sizes were modest, these findings suggest that multidimensional lifestyle is associated with cognitive function over time. In contrast, RI-CLPM revealed stable effects of lifestyle on cognition but largely nonsignificant reverse effects, suggesting a potential role of lifestyle in sustaining cognitive health.

Previous studies consistently show that lifestyle factors support cognitive function and help prevent decline (Dhana et al., 2020; Jia et al., 2023). However, many emphasized physiological factors such as physical activity, sleep, smoking, and alcohol use or relied on cross-sectional designs that could not capture temporal changes or reciprocal effects. Few have tested the reverse pathway by which cognitive decline influences lifestyle behaviors (Wen et al., 2025). Using CLPM and RI-CLPM, this study extends prior research by incorporating multidimensional lifestyle factors—including physiological and psychosocial elements—and statistically estimating the reciprocal relationship between lifestyle and cognition over time while distinguishing within-person from between-person processes (Hamagami & McArdle, 2001; Mund & Nestler, 2019).

The CLPM results underscored the consistent but modest longitudinal association between lifestyle and cognition, aligning with prior findings that multidimensional lifestyle factors strengthen cognitive reserve by supporting neural efficiency and adaptive mechanisms (Lourida et al., 2019; Stern, 2002). Nevertheless, CLPM is constrained by its focus on average population trends, without disentangling stable between-person differences from within-person fluctuations (Mund & Nestler, 2019). Thus, while affirming the role of lifestyle as a primary pathway for maintaining cognitive health, CLPM results must be interpreted with caution.

The RI-CLPM provided more nuanced insights by separating between-person and within-person effects (Mulder & Hamaker, 2021). Results showed sustained effects of lifestyle on cognition but nonsignificant reverse effects from cognition to lifestyle. This suggests that accumulated lifestyle patterns are relatively stable and not easily altered by short-term decline. These findings align with Activity Theory, which emphasizes the stabilizing role of ongoing activities (Havighurst, 1961), and Cognitive Reserve Theory, which highlights the protective impact of cumulative lifestyle engagement (Stern, 2002). The nonsignificant reverse effects may also reflect individual or contextual variability: some individuals reduce activities sharply after decline, while others maintain engagement through social or familial support (Weng et al., 2018). Overall, the lifestyle–cognition relationship is not uniform but may be moderated by personal and contextual factors, underscoring the need for future research on moderators such as social support, health status, and education.

The present study contributes to the literature by analyzing the reciprocal associations between lifestyle and cognition using longitudinal data. Unlike prior studies that focused narrowly on physiological variables, this study incorporated psychosocial elements such as social relationships and occupational participation, thereby providing a broader interpretation of how multidimensional lifestyle factors interact with cognition. By applying both the CLPM and RI-CLPM, the study distinguished between- and within-person processes and revealed temporal dynamics that are often overlooked in unidirectional approaches. These findings underscore that lifestyle may contribute to cognitive health over time, whereas the influence of cognitive decline on lifestyle appears to be is more context-dependent. From a clinical perspective, these findings support the integration of socially embedded roles and meaningful occupational participation with key physiological lifestyle factors such as physical activity, diet, sleep, and substance use in the promotion of cognitive health. Consequently, a multidimensional lifestyle that includes psychosocial factors is associated with cognitive function over time and may support cognitive health, providing empirical evidence for occupational therapy interventions that promote social and occupational engagement in older adults.

Several limitations must also be acknowledged. The longitudinal data were derived from a selective cohort, which may restrict generalizability. The fixed 2-year intervals limited the ability to detect short-term fluctuations. In addition, nutrition was assessed based on meal consumption, which does not capture detailed dietary intake and may limit measurement precision. Cognitive function was assessed using the MMSE, a screening tool that may be subject to ceiling effects, limited sensitivity to subtle changes, and insufficient coverage of executive function. While RI-CLPM enabled the separation of between-person and within-person effects, it could not fully account for heterogeneity in within-person pathways. In addition, the use of constrained models to ensure statistical parsimony may have reduced analytic flexibility, potentially overlooking complex interactions or heterogeneous patterns. Future research should employ more flexible modeling strategies, sensitive assessment tools, and objective longitudinal tracking methods to capture nuanced temporal processes and individual variability more accurately.

Conclusion

This study longitudinally examined the relationship between lifestyle and cognitive function in middle-aged and older adults using both the CLPM and RI-CLPM, showing that lifestyle is modestly associated with cognitive health over time. Although the effect sizes were relatively small, the findings suggest that a multidimensional lifestyle including physiological factors such as physical activity and nutrition and psychosocial elements such as social relationships and occupational participation may contribute to the maintenance of cognitive function. In contrast, reverse effects of cognition on lifestyle were limited, suggesting that accumulated lifestyle patterns are relatively stable and not easily altered by short-term decline. These results highlight the potential importance of multidimensional lifestyle as a key intervention for promoting cognitive health and provide practical evidence for designing management programs for older adults at both policy and clinical levels. Furthermore, the study supports the need for tailored health strategies that reflect the specific characteristics of older adults, emphasizing preventive lifestyle management as an important approach for sustaining cognitive health in aging societies.

Supplemental Material

sj-docx-1-otj-10.1177_15394492261452169 – Supplemental material for Bidirectional Association Between Lifestyle and Cognitive Function in Aging Adults

Supplemental material, sj-docx-1-otj-10.1177_15394492261452169 for Bidirectional Association Between Lifestyle and Cognitive Function in Aging Adults by Young-Myoung Lim and Ji-Hyuk Park in OTJR: Occupational Therapy Journal of Research

Footnotes

ORCID iDs

Young-Myoung Lim

Ji-Hyuk Park

Ethical Considerations

The study protocol was reviewed and approved by the Institutional Review Board of Yonsei University (approval number: 1041849-202508-SB-169-01).

Consent to Participate

The requirement for written informed consent was waived by the board because the study involved the secondary analysis of de-identified, publicly available data and posed minimal risk to participants.

Author Contributions

Young-Myoung Lim was responsible for conceptualization, methodology design, data analysis, and preparation of the original draft, including visualization of the results. Ji-Hyuk Park contributed to the conceptualization of the study, provided supervision throughout the research process, and participated in reviewing and editing the manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data used in this study were obtained from the Korean Longitudinal Study of Aging (KLoSA), a publicly accessible dataset provided by the Korea Employment Information Service.

Patient and Public Involvement Statement

No patient or public involvement. This study was based on secondary analysis of anonymized, publicly available data.

Supplemental Material

Supplemental material for this article is available online.

References

Bloomberg

Muniz-Terrera

Brocklebank

Steptoe

(2024). Healthy lifestyle and cognitive decline in middle-aged and older adults residing in 14 European countries. Nature Communications, 15(1), 5003. https://doi.org/10.1038/s41467-024-49262-5

Borgeest

G. S.

Henson

R. N.

Shafto

Samu

Cam-CAN, & Kievit

R. A.

(2020). Greater lifestyle engagement is associated with better age-adjusted cognitive abilities. PLOS ONE, 15(5), Article e0230077. https://doi.org/10.1371/journal.pone.0230077

Brivio

Viganò

Paterna

Palena

Greco

(2023). Narrative review and analysis of the use of “Lifestyle” in health psychology. International Journal of Environmental Research and Public Health, 20(5), 4427.

Bronfenbrenner

(1979). The ecology of human development: Experiments by nature and design. Harvard university press.

Burke

Dunne

P. J.

(2022). Lifestyle medicine pillars as predictors of psychological flourishing. Frontiers in Psychology, 13, Article 963806.

Campbell

Lavoie

Farraia

Huelin

Zhang

Tahami Monfared

A. A.

(2025). Quality of life in mild cognitive impairment and mild dementia associated with Alzheimer’s disease: A systematic review. Neurology and Therapy, 14(1), 7–26. https://doi.org/10.1007/s40120-024-00676-9

Cho

Y. G.

(2022). Meal frequency, metabolic health, and social norms. Korean Journal of Family Medicine, 43(5), 275. https://doi.org/10.4082/kjfm.43.5E

Choi

K. W.

Jeon

G. S.

(2021). Social network types and depressive symptoms among older Korean men and women. International Journal of Environmental Research and Public Health, 18(21), 11175. https://doi.org/10.3390/ijerph182111175

Dhana

Evans

D. A.

Rajan

K. B.

Bennett

D. A.

Morris

M. C.

(2020). Healthy lifestyle and the risk of Alzheimer dementia: Findings from 2 longitudinal studies. Neurology, 95(4), e374–e383. https://doi.org/10.1212/WNL.0000000000009816

10.

Hamagami

McArdle

J. J.

(2001). Advanced studies of individual differences linear dynamic models for longitudinal data analysis. In New developments and techniques in structural equation modeling. Psychology Press.

11.

Hamaker

E. L.

Kuiper

R. M.

Grasman

R. P. P. P

. (2015). A critique of the cross-lagged panel model. Psychological Methods, 20(1), 102–116. https://doi.org/10.1037/a0038889

12.

Havighurst

R. J.

(1961). Successful aging. The Gerontologist, 1, 8–13. https://doi.org/10.1093/geront/1.1.8

13.

Heneghan

Deng

Wells

Ritchie

Muniz-Terrera

Ritchie

C. W.

. . .Naci

(2023). Modifiable lifestyle activities affect cognition in cognitively healthy middle-aged individuals at risk for late-life Alzheimer’s Disease. Journal of Alzheimer’s Disease, 91(2), 833–846.

14.

Hou

Sun

Liu

Sun

Wang

(2024). A longitudinal study of factors associated with cognitive frailty in middle-aged and elderly population based on the health ecology model. Journal of Affective Disorders, 352, 410–418. https://doi.org/10.1016/j.jad.2024.02.014

15.

Hsu

H. C.

Bai

C. H.

(2022). Individual and environmental factors associated with cognitive function in older people: A longitudinal multilevel analysis. BMC Geriatrics, 22(1), Article 243. https://doi.org/10.1186/s12877-022-02940-9

16.

Jauny

Eustache

Hinault

T. T.

(2022). M/EEG dynamics underlying reserve, resilience, and maintenance in aging: A review. Frontiers in Psychology, 13, Article 861973.

17.

Jia

Zhao

Liu

Liang

. . .Cummings

(2023). Association between healthy lifestyle and memory decline in older adults: 10 year, population based, prospective cohort study. BMJ, 380, Article e072691. https://doi.org/10.1136/bmj-2022-072691

18.

Kline

R. B.

(2016). Principles and practice of structural equation modeling (4th ed.). Guilford Press.

19.

Kimura

Aso

Yabuuchi

Ishibashi

Hori

Sasaki

. . .Matsubara

(2019). Modifiable lifestyle factors and cognitive function in older people: A cross-sectional observational study. Frontiers in Neurology, 10, Article 401.

20.

Kivipelto

Mangialasche

Snyder

H. M.

Allegri

Andrieu

Arai

. . .Carrillo

M. C.

(2020). World-Wide FINGERS network: A global approach to risk reduction and prevention of dementia. Alzheimer’s & Dementia, 16(7), 1078–1094. https://doi.org/10.1002/alz.12123

21.

Korea Employment Information Service. (2022). User guide for Korean Longitudinal Study of Ageing (KLoSA).

22.

Kwon

Y. C.

(1989). Korean version of mini-mental state examination (MMSE-K) Part I: Development of the test for the elderly. Journal of the Korean Neuro-Psychiatric Association, 28, 125.

23.

Lim

Y. M.

Park

J. H.

(2023). Comparative analysis of cluster types according to activity and social participation of oldest-old. Journal of the Korea Contents Association, 23(11), 348–357. https://doi.org/10.5392/JKCA.2023.23.11.348

24.

Lippman

Stump

Veazey

Guimarães

S. T.

Rosenfeld

Kelly

J. H.

. . .Katz

D. L.

(2024). Foundations of lifestyle medicine and its evolution. Mayo Clinic Proceedings: Innovations, Quality & Outcomes, 8(1), 97–111. https://doi.org/10.1016/j.mayocpiqo.2023.11.004

25.

Little

T. D.

(2024). Longitudinal structural equation modeling. Guilford Publications.

26.

Liu

Z. Y.

Wang

Zhang

Y. J.

Zhu

H. L.

(2022). Combined lifestyle, mental health, and mortality in US cancer survivors: A national cohort study. Journal of Translational Medicine, 20(1), 376. https://doi.org/10.1186/s12967-022-03584-4

27.

Livingston

Huntley

Liu

K. Y.

Costafreda

S. G.

Selbæk

Alladi

. . .Mukadam

(2024). Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. The Lancet, 404(10452), 572–628.

28.

Lourida

Hannon

Littlejohns

T. J.

Langa

K. M.

Hyppönen

Kuźma

Llewellyn

D. J.

(2019). Association of lifestyle and genetic risk with incidence of dementia. Jama, 322(5), 430–437. https://doi.org/10.1001/jama.2019.9879

29.

Mulder

J. D.

Hamaker

E. L.

(2021). Three extensions of the random intercept cross-lagged panel model. Structural Equation Modeling: A Multidisciplinary Journal, 28(4), 638–648. https://doi.org/10.1080/10705511.2020.1784738

30.

Mund

Nestler

(2019). Beyond the cross-lagged panel model: Next-generation statistical tools for analyzing interdependencies across the life course. Advances in Life Course Research, 41, 100249. https://doi.org/10.1016/j.alcr.2018.10.002

31.

Nandi

Counts

Bröker

Malik

Chen

Han

. . .Bloom

D. E.

(2024). Cost of care for Alzheimer’s disease and related dementias in the United States: 2016 to 2060. npj Aging, 10(1), 13. https://doi.org/10.1038/s41514-024-00136-6

32.

Ngandu

Lehtisalo

Korkki

Solomon

Coley

Antikainen

. . .Kivipelto

(2022). The effect of adherence on cognition in a multidomain lifestyle intervention (FINGER). Alzheimer’s & Dementia, 18(7), 1325–1334.

33.

Ngandu

Lehtisalo

Solomon

Levälahti

Ahtiluoto

Antikainen

. . .Kivipelto

(2015). A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): A randomised controlled trial. The Lancet, 385(9984), 2255–2263.

34.

Park

J. H.

(1989). Standardization of Korean version of the Mini-Mental State Examination (MMSE-K) for use in the elderly. Part II. Diagnostic validity. Journal of Korean Neuropsychiatric Association, 28, 125–135.

35.

Park

J. H.

Park

H. Y.

Hong

Han

D. S.

Lim

Y. M.

Kim

A. R.

Nam

Park

K. H.

Lim

Bae

Jin

(2023). Conceptual model of establishing lifestyle (Lifestyle-DEPER [Decision, execution, personal factor, environment, resources]) and lifestyle intervention strategies. Therapeutic Science for Rehabilitation, 12(4), 9–22. https://doi.org/10.22683/tsnr.2023.12.4.009

36.

Stern

(2002). What is cognitive reserve? Theory and research application of the reserve concept. Journal of the International Neuropsychological Society, 8(3), 448–460.

37.

Sumowski

J. F.

(2015). Cognitive reserve as a useful concept for early intervention research in multiple sclerosis. Frontiers in Neurology, 6, Article 176. https://doi.org/10.3389/fneur.2015.00176

38.

Sutton

Catling

Segaert

Veldhuijzen van Zanten

(2022). Cognitive health worries, reduced physical activity and fewer social interactions negatively impact psychological wellbeing in older adults during the COVID-19 pandemic. Frontiers in Psychology, 13, Article 823089.

39.

Wen

Zheng

Chen

J. H.

Fong

T. C.

(2025). Contribution of social and lifestyle factors to cognitive status and 5-year change among middle-aged and older Americans. Humanities and Social Sciences Communications, 12(1), 1–10. https://doi.org/10.1057/s41599-025-04521-8

40.

Weng

P. H.

Chen

J. H.

Chiou

J. M.

Y. K.

Chen

T. F.

Chiu

M. J.

. . .Chen

Y. C.

(2018). The effect of lifestyle on late-life cognitive change under different socioeconomic status. PLOS ONE, 13(6), Article e0197676. https://doi.org/10.1371/journal.pone.0197676

41.

World Health Organization. (2020). WHO guidelines on physical activity and sedentary behaviour. https://www.who.int/publications/i/item/9789240015128

42.

World Health Organization. (2021). Decade of healthy ageing: Baseline report. https://www.who.int/publications/i/item/9789240017900

43.

World Health Organization. (2025, March 31). Dementia. https://www.who.int/news-room/fact-sheets/detail/dementia

44.

Y. H.

Kim

Y. J.

Lee

S. Y.

Lee

J. G.

Jeong

D. W.

Cho

Y. H.

. . .Lee

S. H.

(2015). The correlation of meal frequency and nutrition with mental health status in women aged 20-39 years: The 5th Korea National Health and Nutrition Examination Survey, 2010-2012. The Korean Journal of Obesity, 24(2), 101–107. https://doi.org/10.7570/kjo.2015.24.2.101

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.02 MB

0.00 MB