Longitudinal Associations Between Psychological Resilience and Cognitive Function: Evidence From the U.S. Health and Retirement Study

Abstract

Objectives

Reductions in psychological resilience and declining cognition are common among older adults. Understanding the longitudinal association between them could be beneficial for interventions that focus on age-related cognitive and psychological health. In this study, we evaluated the longitudinal associations between cognition and psychological resilience over time in a nationally representative sample of U.S. older adults.

Methods

A total of 9,075 respondents aged 65 and above from 2006 to 2020 health and retirement study (HRS) were included in the current study. Cognition was measured through a modified 35-point Telephone Interview Cognitive Screen, and psychological resilience in the HRS was calculated using a previously established simplified resilience score. Bivariate latent growth modeling was used to examine the parallel association between psychological resilience and cognitive function over a period of up to 12 years.

Results

Positive correlations existed between the intercepts (r = 0.20, SE = 0.07, p < .001) as well as the slopes (r = 0.36, SE = 0.03, p < .001) for psychological resilience and cognition. The initial level of cognition positively predicted the slope of psychological resilience $(β=0.16,SE=0.01,p<.001)$⁠, whereas a somewhat less robust effect was found for the slope of cognition and the initial level of psychological resilience $(β=0.10,SE=0.03,p<.001)$⁠, after controlling all other covariates.

Discussion

In a population-based sample of U.S. older adults, cognition and psychological resilience could mutually reinforce one another. Clinicians and policy makers may consider recommending tasks associated with improving cognitive function for interventions to bolster psychological resilience among older adults.

It is projected that the total number of older adults with Alzheimer’s disease and related dementias (ADRD) in the United States will reach 14 million by 2050, which may lead to serious challenges for public health and increases in overall social care burden (Alzheimer’s Association & Centers for Disease Control and Prevention, 2018). In the absence of a cure for ADRD, prevention research remains vital to identify protective factors to delay the onset of, or reduce the risk of, dementia. Recently, researchers have emphasized that psychosocial resources might influence the trajectory of cognitive decline (Gershon et al., 2014). One such emerging factor is psychological resilience, which is thought to be important in late life as a component of successful psychosocial adjustment and has been associated with various positive health outcomes (Taylor & Carr, 2021; Zhang et al., 2024).

Along with the accumulated risks and resources of early and midlife, later life is a stage where older adults increasingly face challenges such as comorbid chronic health conditions, physical and cognitive decline, and a potential loss of social support. Although some older adults experience significant decline in mental and physical health as they age, others exhibit great resilience, that is to say, they experience little to no decline (Pruchno & Carr, 2017). Within this context, psychological resilience may help explain why some individuals recover from adversity and sustain healthy functioning in later life. Emerging research suggests that older adults who maintain resilience and employ strategies to sustain functioning throughout their lives may be able to leverage these abilities when facing challenges to their cognitive function in later life (Taylor & Carr, 2021). Although this work is encouraging in this context, the complex association between psychological resilience and cognition has not been examined relative to other personal resources known to benefit the cognitive health (e.g., perceived mastery, perceived optimism) (Toyama & Hektner, 2023).

The purpose of the current study was to systematically assess the longitudinal association between psychological resilience and cognitive function. We draw on the previous work that established the simplified resilience score (SRS) in the health and retirement study (HRS) (Manning et al., 2014). Using HRS and a parallel latent growth model (LGM) approach, we examined each variable’s own trajectory as well as their mutual influences among a nationally representative sample of US older adults.

Psychological Resilience in Later Life

Historically, the importance of being resilient in the face of adversity was first stressed in childhood and adolescence, as resilience measures were often utilized to explain why some children living with highly adverse circumstances emerged as functional and capable individuals (Lösel & Bliesener, 1990). This concept has only more recently been applied to later life. Traditionally, the aging process has often been seen as being accompanied by frailty, vulnerability, and loss, with more research focused on pathology and poor outcomes rather than favorable ones (Fulop et al., 2010). However, older adults are also endowed with the capacity for positive adaptation in the face of challenges, which helps counteract the typical (physical and psychological) health changes that come with aging. The examination of psychological resilience within the context of aging suggests that psychological resilience supports a holistic view of adaptive aging (Jeste et al., 2013; Pruchno & Carr, 2017).

The definition of psychological resilience varies across disciplines, and systematic reviews have indicated that psychological resilience could be understood as several constructs: as an adaptive attribute shaped by social contexts that may be stable over time, as a dynamic process of adaptation to cope with challenges or stressors, and as a positive outcome derived from dealing with stress and adversity (Masten et al., 2023). Consistent with previous research drawing on data from the HRS (Manning et al., 2014; Taylor et al., 2018), in this study, we focused on psychological resilience as “an individual’s ability to navigate adversity through positive adaptation in a manner that protects health and well-being” (Taylor & Carr, 2021). Previous studies suggest that the cumulative effect of a lifetime of confronting adversities could harm but may also foreshadow positive adaptation and strength in older adults (Phillips et al., 2016) and could explain well-being of older age despite declining physical health (Jeste et al., 2013). This focus was recently supported by emerging evidence suggesting that psychological resilience in later life may be malleable and may change in response to social and environmental influences, and thus might also be modulated through targeted interventions (Edwards et al., 2017; Johnson et al., 2021).

Emerging evidence consistently suggests that psychological resilience may be a critical factor in maintaining cognitive health in later life. For example, a large number of cross-sectional studies have suggested that psychological resilience is positively associated with various cognitive domains, including attention, working memory, associative memory performances, short-term memory, verbal fluency, and global cognition (Franks et al., 2023; Ma et al., 2022; Saez-Sanz et al., 2023). Additionally, a few longitudinal studies, which briefly defined resilience as the ability to successfully adapt the life-stress events, have shown that greater psychological resilience may also be associated with reduced cognitive decline among older adults (Jiang et al., 2024). However, far less is known regarding the direction or causal mechanisms linking cognition and psychological resilience among the older population.

Cognitive enrichment theory proposes that cognitive enrichment might be affected by a body of positive attitudes and beliefs (e.g., control beliefs, self-efficacy, and positive emotion) (Hertzog et al., 2008). More generally, theorists argued that older adults who are more optimistic, open to experience, positively motivated and/or goal-directed are more likely to maintain a high quality of life when facing age-related challenges. Previous research indicated that these positive resources such as mastery, self-esteem or interpersonal control may reflect a more holistic construct, psychological resilience, as resilient individuals often share these characteristics (Taylor & Carr, 2021), which may help them rapidly adapt to the disruptive or stressful life events and maintain well-being. Thus, from a theoretical perspective, psychological resilience may provide a common core to a spectrum of positive resource variables which are substantially correlated with one another, and may also have an independent and positive association with health and well-being (Musich et al., 2021; Nygren et al., 2005; Taylor & Carr, 2021). Similarly, these resources may have shared associations with other enrichment factors (e.g., social support or engagement), which may facilitate the maintenance of cognitive function (García et al., 2022; Zhang et al., 2023).

Theoretical Framework

The selective optimization with compensation (SOC) model (Baltes & Carstensen, 1996) describes three strategies of adaptively responding to everyday demands and functional decline in later life and may provide a comprehensive theoretical framework for understanding the relationship between psychological resilience and cognitive function. This model centers on doing the best one can with what one has and describes three dynamic processes of modifying and adapting one’s behavior across the lifespan: selection, optimization, and compensation (Carpentieri et al., 2017). In line with this framework, successful cognitive aging might be achieved if adaptive strategies are used.

Psychological resilience is associated with the capability with which older adults can identify and use SOC strategies when facing losses and changes, which is important for adaptive aging. It can help older adults prioritize goals (selection) according to their importance for increasing gains (optimization) and adjusting to aging-related losses (compensating), yielding favorable health outcomes (Freund, 2008). The use of SOC could further buffer age-related cognitive declines (Robinson et al., 2016). For example, selection may improve memory by helping older adults set goals and prioritize activities (e.g., maintain memory). Whereas optimization (e.g., employing memory strategies) and compensation (e.g., enrolling in a cognitive training program when previously used memory strategies become ineffective) may help older adults to cope with memory decline and achieve the goal (Scheibner & Leathem, 2012).

Protective factors, often referred to as “assets,” “resource” or “strengths,” are crucial for resilience to be achieved (Richardson, 2002). Protective factors have commonly been identified across two levels of functioning: individual factors (e.g., psychological, cognitive) and familial/community support factors (e.g., family cohesion) (Bergeman & Wallace, 1998). These factors can facilitate the capability that enables individuals to resist adversity and underlie the process of adaptation (Windle, 2011). Older adults with higher levels of cognitive function are likely to use adaptive coping strategies, adjust personal values and preference systems, and adopt positive appraisal to reinterpret and cope with stressful life events, which may benefit psychological resilience (Greve & Staudinger, 2006). Prior longitudinal studies suggested that cognitive function could benefit the development of protective psychological resources such as sense of control, life purpose, and optimism (Oh et al., 2020; Zhang et al., 2020). These studies may imply a protecting role of cognition in the growth of psychological resilience. A recent study suggested that cognition is a significant determinant of psychological resilience (Chen et al., 2023). However, it is also plausible that the relationship between psychological resilience and cognitive function is bidirectional, such that psychological resilience facilitates strategies and behaviors that support better cognitive function over time, and in turn, better cognitive function also increases psychological resilience.

Although a review of theory and empirical evidence suggests a positive relationship between psychological resilience and cognitive function among older adults (Jung et al., 2021), many existing studies rely on samples that have not been drawn at random, making it difficult to generalize the findings to the wider population of older adults (e.g., military veterans). Therefore, more research is needed to understand the relationship between psychological resilience and cognitive function using a nationally representative population-based sample with a longer follow-up period. In addition, the limited set of longitudinal studies available generally do not examine the reciprocal effects of psychological resilience on cognition, which may help tailor strategies to maintain cognitive health. Thus, the current study uses a parallel LGM between psychological resilience and cognition to address the previous questions among older adults in a nationally representative sample.

Method

Data

This study utilizes data from HRS, a biennial, ongoing survey of approximately 20,000 adults aged 51 years or older from the United States. The HRS utilized a mixed-mode design randomly assigning half of the core HRS respondents (subsample A) to receive the Leave-Behind Questionnaire (LBQ) in 2006, and the second half (subsample B) received the same questionnaire in 2008. As a follow-up, subsample A also completed the LBQ in 2010, 2014, 2018, and subsample B participated in the LBQ in 2012, 2016, and 2020. For the present analyses, we merged four waves of data (2006, 2010, 2014, 2018) from subsample A starting with the 2006 wave (n = 7,730). Next, we merged four waves of data (2008, 2012, 2016, 2020) from subsample B based on the 2008 wave (n = 7,073). Finally, we combined these two longitudinal datasets into a single dataset by rescaling time. Specifically, 2006 and 2008 were merged and realigned into Time 1, 2010 and 2012 were merged into Time 2, 2014 and 2016 to Time 3, and 2018 and 2020 to Time 4. Exclusion criteria were as follows: (a) respondents who were aged <65 at baseline (n = 5,322); (b) respondents who were aged ≥65 with diagnosed memory issue in Time 1 (n = 268); (c) respondents who were aged ≥65 with missing values in diagnosed memory issue in Time 1 (n = 138). Therefore, 9,075 older adults from baseline (Time 1) were included in the final analysis. Importantly, we do note that there was some attrition over the course of the study: there were 7,274 respondents in Time 2, 5,410 respondents in Time 3, and 3,328 respondents in Time 4.

Measures

Global cognitive function

Global cognitive function was derived from a modified 35-point Telephone Interview Cognitive Screen (Griffin et al., 2020), which combines seven different tasks: immediate word recall, delayed word recall, serial 7s, backward counting, date, object, and president/vice president naming. A single composite score for global cognition was obtained by summing all tasks with the range from 0 to 35. Thus, a higher composite score indicates better global cognitive function. HRS has imputed cognitive function using multivariate, regression-based procedure based on a combination of relevant demographic, health, and economic variables, as well as prior and current wave cognitive variables. The imputation was calculated only for self-respondents in a given wave, not for proxy respondents or nonparticipants. The imputation process is described in more detail within HRS documentation (McCammon et al., 2023). In the current study, we used the imputed cognitive score to include more observations. Its psychometric property has been first validated by Herzog and Wallace (1997). In order to examine the construct validity in our study, we conducted confirmatory factor analysis at each time points. The results indicated that the factor structure was consistent with what was reported in Herzog and Wallace (see Supplementary Table 1). The Cronbach’s alpha in Time 1, Time 2, Time 3, and Time 4 was 0.70, 0.74, 0.76, and 0.75, respectively.

Psychological resilience

Psychological resilience in HRS was measured using the SRS developed by Manning (2014). Guided by the Wagnild and Young resilience scale (Wagnild & Young, 1993), which conceptualized resilience as an ability to adapt successfully to stressful situations (Perna et al., 2012), the domain of SRS parallels Wagnild and Young Scale: (a) perseverance or the ability to keep going despite major setbacks; (b) equanimity, which describes being able to adjust to change, often with humor; (c) meaningfulness or the realization that life has a purpose; (d) self-reliance or recognition of one’s one inner strengths; and (e) existential loneliness or the realization that some experiences must be faced alone. This measure draws 12 questions from LBQ in which individuals indicate how strongly they agree/disagree with the following statements: (a) I feel it is impossible for me to reach the goals that I would like to strive for; (b) So far, I have gotten the important things I want in life; (c) If something can go wrong for me, it will; (d) I am satisfied with my life; (e) what happens in life is often beyond my control; (f) I can do the things that I want to do; (g) The future seems hopeless to me and I can’t believe that things are changing for the better; (h) When I really want to do something, I usually find a way to succeed at it; (i) In most ways, my life is close to ideal; (j) I can do just about anything I really set my mind to; (k) There is really no way I can solve the problems I have; and (l) I have a sense of direction and purpose in life. Among them, the reversed coded items are 1, 3, 5, 7, and 11. All items were standardized (0–1) because they varied in their response ranges. Detailed descriptions of measured items and coding scheme were given in Supplementary Table 5. Respondents’ total resilience score was obtained by summing up all 12 rescaled items. The reliability index (Cronbach’s alpha) was 0.85 for each time point.

Covariates

All covariates controlled in the models were measured from Time 1, including baseline age in years, gender (male = 0; female = 1), education (total years of education), race (White = 0; Black = 1; other ethnic groups = 2), marital status (married = 1; other status = 0), total wealth, medical conditions, activities of daily living, and depression (Jiang et al., 2024; Lou et al., 2023). Total wealth (in U.S. dollars) was calculated as the sum of all a respondent’s wealth components less all debt. Medical conditions were measured by the count of chronic medical conditions reported previously, which include high blood pressure, diabetes, cancer or a malignant tumor, chronic lung disease, cardiovascular disease, stroke, emotional or psychiatric problems, arthritis, and sleep disorders. Activities of daily living consisted of 6 items asking respondents “if they have any difficulties” in bathing, dressing, eating, getting in/out of bed, walking across a room and toileting. Finally, depression was measured by the Center for Epidemiologic Studies Depression scale, with a range of 0–8. A higher score indicated more depressive symptoms.

Statistical analyses

First, descriptive statistics were conducted to describe respondents’ baseline characteristics, and bivariate correlations were calculated between psychological resilience and cognitive function at each time point. Additionally, in order to examine the predictive validity of the composite score of SRS, a series of bivariate correlations were conducted with total SRS, the individual items making up SRS and cognitive function at each time point. The results indicated that compared with each individual item (0.11–0.26), SRS total scores have stronger associations (0.28–0.32) with cognitive function at each time point (Supplementary Figures 1–4). This suggests that using the composite SRS could have a good predicitive validity in the current study.

Second, two unconditional LGMs were fit separately to explore the growth trajectory in global cognition and psychological resilience. LGM is an analytic approach that allows for modeling not only the average change process, but also the variation among individual growth trajectories (Du, 2021). Third, in order to examine the parallel process of change for both variables, and how the two processes were associated simultaneously, multivariate LGM was assessed between cognition and psychological resilience over four time points (Curran & Hussong, 2003). This multivariate model could examine the unique effects of predictors on the change in cognitive status and psychological resilience, whereas also modeling the reciprocal effects of changes in cognition on changes in psychological resilience. The model fit index used included: the chi-square test, Akaike information criterion (AIC), Bayesian information criterion (BIC), comparative fit index (CFI), standardized root mean square residual (SRMR), and root mean square error of approximation (RMSEA). A minimum CFI value of 0.90 and a value less than 0.05 for RMSEA and 0.08 for SRMR were considered a good model fit for this study (Du et al., 2023).We also employed a robust maximum likelihood to handle non-normality, and the missing values were handled by the full information maximum likelihood (FIML) estimation. Finally, as FIML assumes either missing completely at random (MCAR) and missing at random (MAR), we adopted a pattern mixed model to examine the sensitivity of MCAR and MAR assumption. Analyses were conducted using R and its package “lavaan” (Rosseel, 2012). For missing data analysis, we used Mplus version 8 (Muthen et al., 2017).

Results

Sample characteristics

Descriptive statistics were examined for respondents’ demographic information. Means and standard deviations for continuous variables, as well as frequencies and percentages for categorical variables, are presented in Table 1. Among 9,075 respondents at baseline, 57.76% were women (n = 5,242), and the mean age was 74.51 (SD = 7.10). Approximately, 84.72% of the respondents were Caucasian (n = 7,688), and 58.98% of them were married (n = 5,352) at Time 1. The average years of education was 12.27 (SD = 3.16). Mean and standard deviation of psychological resilience and cognition as well as their correlations were shown in Supplementary Table 2. On average, both cognition and resilience tended to decrease over time. In addition, within each time point, both variables were positively associated, with coefficient $(ρ)$ ranging from 0.30 to 0.32.

Univariate latent growth models

Unconditional LGM were used to separately examine the initial level (i.e., intercept) and growth rate (i.e., slope) parameters of cognition and psychological resilience. As is customary in latent growth modeling, the loadings from the intercept to each repeated variable were fixed to 1 across T1–T4. In addition, the loadings of slope factors were fixed to 0, 1, 2, and 3 for the first, second, third, and fourth measurement occasions, respectively. The unconditional LGM for cognition (⁠$χ2$ = 77.76, df = 5, p < .001, AIC = 134,485.39, BIC = 134,549.26, CFI = 0.99, RMSEA = 0.04, SRMR = 0.05) shows a good fit. The LGM for psychological resilience shows a good fit (⁠$χ2$ = 50.67, df = 3, p < .001, AIC = $187,976.12$⁠, BIC = 188,054.37, CFI = 0.99, RMSEA = 0.04, SRMR = 0.02). Table 2 presents parameter estimates for the unconditional LGM of both cognition and psychological resilience. Both variables showed relatively high scores at baseline and, as mentioned before, declined over time. The significant variance for the slope factor of psychological resilience and cognition indicated that there is an interindividual difference in their decline rates.

Bivariate Latent Growth Models

After modeling the growth trajectory for cognition and psychological resilience separately, it was then possible to examine the developmental relationship by modeling them simultaneously using bivariate latent growth modeling, whereas controlling for other covariates. In the bivariate LGM, we allow the initial level of one construct to predict the subsequent linear growth rate of the other construct. We also calculated covariances between resilience and cognition at baseline as well as their growth rates in the model.

Figure 1 shows the paths that were statistically significant (solid lines), with dotted lines indicating nonsignificant paths. The overall model indicated a good fit: $χ2=2,218.85$⁠, df = 62, p < .001, AIC = 313,302.87, BIC = 313,743.89, CFI = 0.95, RMSEA = 0.06, SRMR = 0.05. Older adults who had higher psychological resilience values at baseline tended to have a faster decline of psychological resilience rate over time (r = −0.19, SE = 0.04, p < .001). In contrast, we did not observe any significant relations between the initial status and linear slope factor for cognition.

In addition, we found that a positive and significant correlation between baseline cognition and psychological resilience (r = 0.20, SE = 0.07, p < .001), indicating a cross-sectional association between cognition and psychological resilience. The linear slope of cognition was also positively associated with the slope rate of psychological resilience (r = 0.36, SE = 0.03, p < .001). Thus, a more rapidly decline rate of cognition was related to a faster decline rate of resilience during the study period. Meanwhile, we found that older adults with a higher level of cognition at baseline exhibited a slower decline in psychological resilience $(β=0.16,SE=0.01,p<.001)$⁠. Similarly, older adults with more psychological resilience at baseline had a slower decline rate in cognition, albeit with a smaller effect $(β=0.10,SE=0.03,p<.001)$⁠. The significant and positive crossed-paths between initial levels and slope factor indicated that cognitive decline was accounted for in part by level of psychological resilience and vice versa.

Table 3 summarized the effect of baseline covariates on each growth factor. Most of these variables were significantly related to the baseline status of cognition and psychological resilience, rather than their growth rates. Based on the standardized coefficients, older adults’ baseline age has the strongest effect, among all covariates, on the growth factors of cognition and psychological resilience. More specifically, at Time 1, increases in age significantly reduced both cognition (β = −0.29, SE = 0.01, p < .001) as well as psychological resilience (β = −0.15, SE = 0.00, p < .001). In addition, those with older ages tended to have a faster decline in both cognition (β = −0.50, SE = 0.01, p < .001) and psychological resilience (β = −0.28, SE = 0.00, p < .001). Education was strongly associated with both cognition and psychological resilience at Time 1 (β = 0.42, SE = 0.02, p < .001; β = 0.18, SE = 0.01, p < .001), indicating that higher education was related to higher levels of cognition as well as psychological resilience. Women tended to have slightly higher cognition at Time 1 than men (β = 0.10, SE = 0.08, p < .001), but their differences in terms of psychological resilience were almost negligible (β = 0.03, SE = 0.04, p = .01). Black older adults in the sample demonstrated lower cognition at Time 1, compared with white older adults (β = −0.24, SE = 0.13, p < .001). Older adults with more functional limitations and depression symptoms tended to have lower cognition at Time 1 (β = −0.15, SE = 0.09, p < .001; β = −0.11, SE = 0.03, p < .001).

Additional Analysis on Missingness

For the variables of interest (cognition and psychological resilience), the missingness occurred mainly for the cognition. We first conducted a logistic regression where the dependent variable was a binary variable indicating whether or not a respondent completed all four waves of cognitive measurement (0 = completer; 1 = non-completer). The logistic regression results (see Supplementary Table 3) demonstrated that, compared with those who completed the tests in four waves, the non-completers in the current study tended to be male, older, unmarried, and Black and have lower educational attainment, more chronic conditions, greater depressive symptoms, and difficulty with activities of daily living. In addition, a pattern mixture model was performed to examine if the missing values on cognition could rely on missing not at random (MNAR) process (Enders, 2022). The missing pattern was classified into three groups: early dropouts who quit prior to Time 3 follow-up, late dropouts who leave prior to Time 4 follow-up, and completers who include respondents with intermittent missing values.

We then fitted unconditional LGMs including two dummy codes for early and late dropouts (EDROP and LDROP, respectively). Based on the Enders’ guidance (2022), we set up the growth rate differences between early and late dropout groups as 0, 0.2, and 0.5, which corresponds to the Cohen’s small to medium effect size benchmark. Table 4 showed the results that compared the same LGM under MAR process and MNAR process. The results indicate that although the decline rate may be flattened (less positive) as the effect size increases, both AIC and BIC supported the MAR process. Therefore, we would like to accept the major missing process depending on the MAR.

Discussion

In this study, we drew from 12 years of data (2006–2020) of the HRS to develop a pooled sample of U.S. older adults among whom we could examine the parallel longitudinal associations between psychological resilience and cognitive function. Our study is the first to provide direct evidence that psychological resilience and cognition may co-develop over time; indeed, positive correlations were observed between their intercept as well as slope factors. We further added to prior research by demonstrating a reciprocal directional relationship between psychological resilience and cognition, where the intercept of psychological resilience could positively predict the slope of cognition, and vice versa. Notably, the standardized coefficients suggested that initial level of cognition may have a substantially larger effect on the subsequent change of psychological resilience than vice versa.

Consistent with prior cross-sectional studies, we found a positive association between psychological resilience and cognition at baseline (Ma et al., 2022). This indicates that older adults starting with a higher psychological resilience may also start with better cognitive ability. In addition, the univariate growth models show that on average, both psychological resilience and cognition tended to decline over time, with significant slope variance indicating the presence of interindividual differences in individual trajectories. These findings are consistent with the mainstream aging research that declining on both variables are inevitable consequence of aging as older adult are associated a greater incidence of dependence, chronic physical illness and poor quality of life (Black & Rush, 2002; Gooding et al., 2012). Importantly, the positive correlation between changes in psychological resilience and changes in cognition suggests that older adults experiencing faster declines in psychological resilience may simultaneously experience a larger decline in cognition. This parallel process implies that psychological resilience and cognitive function were each distinct processes, and they may be closely intertwined over late development.

Although several longitudinal studies have confirmed that psychological resilience is an important factor for maintaining cognitive health in adulthood (Jiang et al., 2024; Lou et al., 2023), less research has examined their bidirectional associations. To the best of our knowledge, our study was among the first to indicate the joint longitudinal relationship between psychological resilience and cognitive decline among US older adults. The finding that older adults starting with higher psychological resilience tended to show slower cognitive decline may support the protective role of psychological resilience in promoting cognitive ability. In addition, this result also highlighted that psychological resilience, as a personal resource, could be positively associated with cognitive enrichment and the maintenance of functioning in late life (Hertzog et al., 2008). Multiple mechanisms (e.g., the brain or cognitive reserve hypothesis, vascular hypothesis, and stress hypothesis) may explain the beneficial effects resulting from enhanced level of psychological resilience (Jiang et al., 2024; Lou et al., 2023). Specifically, older adults with higher psychological resilience may be more likely to engage in healthy lifestyles (e.g., reading, writing, or regular exercise) (Chan et al., 2018; Du et al., 2023), which jointly compensate for the damage caused by age or disease-related neurodegeneration, and thus promote cognitive reserve. For example, engaging in healthy lifestyles, especially physical activities, could promote cerebral blood flow and angiogenesis, reduce the accumulation of free radical oxidizing proteins, and thus lower risk for vascular health (e.g., high blood pressure or BMI), and ultimately improve cognitive function (Franks et al., 2023; Moored et al., 2020). Older adults with psychological resilience are more likely to use SOC related strategies to adapt to aging-related losses in daily life (Lang et al., 2002).

Additionally, the broaden-and-build theory of positive emotions suggests that resilience may influence cognition by facilitating positive emotion regulation and adaptive coping style (Fredrickson, 2004). Positive emotion may broaden the scope of attention and cognition, enable flexible and creative thinking, and build personal resources (e.g., physical, intellectual, and social resources), which allow older adults to manage aging-related threats more effectively (Fredrickson, 2004). Emerging evidence from observational studies suggested that people with high levels of resilience are more optimistic, zestful, and energetic to their life and are curious and open to new experiences (Gloria & Steinhardt, 2016). Resilient older adults may tend to rely on positive emotion, which compensates or mediates the adverse health effects of negative emotion (called “undoing effect”), to achieve their effective coping for cognitive decline (Fredrickson, 2004). One practical implication of broaden-and-build theory relevant to our findings is to cultivate positive emotions, which over time may augment enduring personal resources. For example, this could include finding positive meaning or optimism in health behaviors such as physical exercise, or social engagement or interaction, which are significant factors in building cognitive reserve (Zahodne, 2021).

The finding that a high level of cognitive function at baseline was predictive of a slower decline in psychological resilience suggests that older adults’ cognition may be protective of one’s capacity to cope with adversity. This provided empirical evidence to support Greve and Staudinger’s theoretical model that cognitive ability could benefit the maintenance of psychological resilience among older adults (Greve & Staudinger, 2006). One possible explanation is that older adults with higher cognition may tend to adopt positive cognitive reappraisals to cope with stressful life events (using cognitive reappraisal to frame a situation in a more positive way) (Troy et al., 2010). By viewing adversity (e.g., bereavement or functional decline) as opportunities to grow rather than a threat, older adults may be likely to deploy adaptive coping strategies (e.g., problem-focused coping) and practice more positive emotion and self-perception, which has been theorized to improve psychological resilience (Resnick et al., 2011). Better memory, attention, and executive functioning may enable one to learn, remember, and execute strategies that preserve resilience. These findings also have relevance in the context of ADRD. Although at least one cross-sectional study did not find the moderating effect of psychological resilience in the associations between factors such as stress and cognition or dementia risk (Franks et al., 2023). It is possible that these links may only reveal themselves over time, and that they may operate differently in the context of an actual diagnosis of dementia or mild cognitive impairment. Given that psychological resilience may represent an important antecedent to maintaining healthy lifestyle behaviors in the face of dementia-risk, further research should explore the extent to which psychological resilience may be augmented within populations most at risk for ADRD, and whether interventions that address both psychological resilience and cognitive function in concert may prove more valuable than that only focus on one or the other.

Our results align with a process-based theory which conceptualizes resilience as an outcome of dynamic, complex interplay between multiple personal and contextual dimensions (Ong et al., 2009). Surprisingly, the standardized coefficients showed that the effect of initial cognition on the rate of decline of psychological resilience is almost two times that of initial psychological resilience to decline rate of cognition (0.14 vs 0.07, both p < .05). This indicates that cognition is likely the dominant factor for the dynamic associations between cognition and psychological resilience. Still, future research is needed to uncover the mechanism underlying the connection between psychological resilience and cognitive function among older adults.

Although our findings contribute to the ongoing development of models of psychological resilience in late life, the results should be interpreted in light of several limitations. First, although the sensitivity analysis indicated that missingness may be driven by MAR instead of MNAR, we cannot fully rule out the possibility of MNAR. Given the untestable nature of MNAR, bias may still exist in the current analysis. Second, although Manning’s psychological resilience scale (SRS) was aggregated from multiple sources in HRS, some of the meaningful nuanced domains of resilience corresponding to the well-established Wagnild and Young’s psychological resilience (1993), are still unmeasured. Essentially, the current SRS mainly reflects personal resilience rather than the social aspects of this construct. Therefore, it is plausible that future research on psychological resilience and cognitive function using different measures of psychological resilience may show a different result. Additionally, evidence on the predictive validity of SRS for cognitive function is still limited. Therefore, future studies should examine SRS in the context of cognitive function and other key health outcomes in the HRS. Third, as is true for any observational study, the longitudinal model used here does not guarantee any causal inference between psychological resilience and cognition because there may exist other unobserved confounding, particularly confounders that may vary over time. Finally, as the global cognition (a modified 35-point Telephone Interview) is limited to participants whose ages are over 65 years old in the HRS, our work is focused on older adults, rather than those in later middle age. However, cognitive reserve and developmental resilience likely develop throughout adulthood, thus it is important for future longitudinal work to capture resilience throughout the life course.

Despite the earlier limitations, our findings also have several implications for future research, clinical practice, and policy development. Future research should examine the potential mechanisms through which cognitive function may bolster psychological resilience in later life. For example, whereas we used a global cognition measure in this study, subsequent studies should examine the differential links between psychological resilience and specific sub-domains of cognitive function, such as memory and executive function. Also, given that psychological resilience and cognition may reinforce each other in late life, more attention should be given to comprehensive interventions that address both psychological resilience and cognition to improve one’s abilities to maintain instrumental activities of daily living in later life (Lou et al., 2023). However, given that cognitive function was most predictive of later resilience, rather than the other way around, addressing cognition function as an integral part of resilience enhancement programs may offer particular promise in improving health outcomes among older adults. From a practical perspective, this could mean further reinforcing, with both clinicians and patients, how interventions that may improve cognitive function—including those involving exercise and diet—may also facilitate psychological resilience later in life. These findings may also highlight the importance of considering, and potentially finding ways to remediate, deleterious effects on psychological resilience that may accompany cognitive decline among those who are at risk for ADRD the most. Overall, this work emphasizes the importance of integrating both psychological resilience and cognitive function in research and practice.

Conclusions

This study offers a significant advancement in our understanding of the longitudinal association between psychological resilience and cognition among older adults. The results highlight a co-development trajectory between resilience and cognition in later life. We found that cognition and psychological resilience may mutually influence each other’s trajectory over time, but also that, rather than psychological resilience offering the greatest influence on cognitive trajectory, one’s initial cognitive performance may offer the most robust influence on later changes in psychological resilience. Thus, future interventions and programming may wish to focus on means of preserving this resilience in the context of age-related, and disease-related, cognitive decline.

Funding

None.

Conflict of Interest

None.

Author Contributions

C. Du and X. Ding contribute equally to the study. C. Du had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: C. Du, X. Ding, B. Katz, H. Xu, and M. Li.

Acquisition and interpretation of data: All authors.

Statistical analysis: C. Du.

Drafting of the manuscript: All authors.

References