Depressive Symptom Trajectories and Cognition Among Older American Couples: A Dyadic Perspective

Abstract

Objectives

This study examined whether trajectories of depressive symptoms of one spouse are associated with the other spouse’s memory.

Methods

Longitudinal data from the Health and Retirement Study (2004–2016) were used (N = 5690 heterosexual couples). Latent-class growth analysis and structural equation models examined the actor and partner effects of depressive symptom trajectories on memory.

Results

Four depressive symptom trajectories were identified (i.e., persistently low, increasing, decreasing, and persistently high). Compared to the low trajectory group, the increasing and persistently high trajectories were associated with worse memory for both men and women. While none of the wives’ depressive symptom trajectories was significantly associated with husbands’ memory (p > .05), husbands’ decreasing trajectory was linked to wives’ better memory (β = 0.498, 95% CI = 0.106, 0.890).

Discussion

Older adults with increasing and persistently high depressive symptoms may experience worse memory. Psychosocial interventions targeting depressive symptoms among older men may be beneficial to their spouses’ memory.

Keywords

dyadic data latent class growth analysis older couples depressive symptom trajectory memory

Introduction

Alzheimer’s disease and related dementia (ADRD) have become an increasingly urgent public health issue with substantial socioeconomic ramifications. Extensive research has found depressive symptoms and/or depression as a strong risk factor of cognitive decline and dementia (Baumgart et al., 2015; Diniz et al., 2013; Mourao et al., 2016; Ownby et al., 2006). However, research concerning the relationship between depressive symptoms and cognition primarily has focused on intra-individual characteristics without consideration of potential interpersonal effects from spouses. Leveraging longitudinal dyadic data from older American couples, this study aimed to bridge these gaps by examining how trajectories of depressive symptoms of one spouse are associated with the other spouse’s subsequent cognition, using an innovative dyadic analytic technique.

Depressive Symptom Trajectories and Cognitive Aging

An emerging body of literature has documented the association of individuals’ longitudinal trajectory of depressive symptoms with cognition (Kuchibhatla et al., 2012; Richards, 2011). Compared to those constantly reporting a low level of depressive symptoms, individuals with a persistently high or increasing/worsening trajectory of depressive symptoms over time had worse cognitive performance (Formánek et al., 2020; Zhu et al., 2022) or faster cognitive decline (Formánek et al., 2020). Existing research suggests that employing a trajectory-based approach for assessing heterogeneous depressive symptom trajectories could calibrate risk-stratification to identify older adults at higher risk for cognitive decline, which could thereby help to inform the development of targeted psychosocial interventions for dementia prevention (Kuchibhatla et al., 2012; Richards, 2011).

However, existing studies of the link between depressive symptom trajectories and cognition predominantly have been focused on intra-individual characteristics (Formánek et al., 2020; Zhu et al., 2022). Because older adults are embedded in a broader societal context and interact with other people, a better understanding of how interpersonal factors may influence cognitive aging is needed. In particular, in the marital context, older married couples may influence each other’s psychological and cognitive health (Dufouil & Alpérovitch, 2000; Lu & Shelley, 2019; Townsend et al., 2001). However, a couple-based perspective is very lacking in existing research.

Spousal Interdependence in Health and Gender Differences

According to the interdependence theory and communal coping perspectives (Lewis et al., 2006), couples’ health behaviors may influence each other as they live in a shared environment, co-experience life events, and cope with common problems as a unit. An extensive literature has documented the role of spousal interdependence in various physical and mental health outcomes among older couples, such as frailty (Monin et al., 2016), functional limitations (Hoppmann et al., 2011), health behaviors (Li et al., 2013; Monin et al., 2015), distress (Lu & Shelley, 2019; Townsend et al., 2001), and quality of life (Bourassa et al., 2015). Among these factors, spousal interdependence in mental health, especially depressive symptoms concordance, has been studied most extensively (Meyler et al., 2007). In contrast, research on spousal interdependence in cognitive health is lagging.

More importantly, the extent to which one’s partner’s depressive symptom trajectories relate to one’s cognition within the spousal context remains largely unexplored. Depressive symptoms could be experienced by both spouses, since the psychological responses may occur from their communal coping with external life events and other stressors (Falconier & Kuhn, 2019). Furthermore, one member with depressive symptoms in a couple may withdraw socially, leading to reduced social interaction and cognitive stimulation for one’s partner, the accumulation of which could lead to cognitive decline (Dufouil & Alpérovitch, 2000). Over time, the psychological responses may affect a diversity of health dimensions of their own (Kaup et al., 2016) and their spouses (Monin et al., 2018; Schulz et al., 2009). Despite this potential link, to our knowledge no prior studies have examined the spousal effect of depressive symptom trajectories on cognition among older couples. In a few studies that examined the relationship between depressive symptoms and cognition within couples (Gerstorf, Hoppmann, Kadlec, et al., 2009; Lee et al., 2012), depressive symptoms were measured at a single time point without considering the changes of depressive symptoms over time.

Furthermore, existing studies document gendered patterns in spousal health interdependence. In one study, wives’ depressive symptoms predicted husbands’ memory decline, whereas husbands’ better episodic memory predicted less cognitive decline for their wives (Gerstorf, Hoppmann, Kadlec, et al., 2009). Using longitudinal data from 2684 married couples in Korea, Lee and colleagues (2012) reported that wives’ psychological wellbeing and cognitive function have stronger effects on those of husbands than vice versa. Another study reported that in older couples one’s depressive symptoms predicted one’s partner’s lower cognitive functioning, regardless of gender (Monin et al., 2018). Overall, it has been reported that wives are more inclined to make efforts to enhance their husbands’ health behaviors and emotional wellbeing than vice versa, and women’s health is more influenced by the marital context than men (Lee et al., 2012; Reczek & Umberson, 2012; Thomeer et al., 2015). Based on gendered patterns suggested in prior studies (Ayotte et al., 2010; Lee et al., 2012), we hypothesized that wives’ depressive symptom trajectory influences husbands’ cognition rather than vice versa.

Research Objectives

Taken together, despite well-established evidence concerning the influential role of depressive symptoms on cognitive decline and the spousal interdependence regarding health among older couples, a research gap remains on the association between depressive symptoms trajectory and cognition within the spousal context and potential gender differences. Employing a prospective dyadic design, this study addresses this knowledge gap by examining whether different depressive symptom trajectories among older heterosexual American couples predict one’s own and one’s partner’s memory 2 years later. Specifically, the objectives of this study included (1) identifying latent-class memberships of depressive symptom trajectories among wives and husbands, separately; (2) examining whether one’s depressive symptom trajectory predicts one’s own and one’s spouse’s memory 2 years later; and (3) investigating potential gender differences in the spousal effect. To achieve these objectives, we used latent-class growth analysis and structural equation modeling techniques to analyze longitudinal dyadic data from a nationally representative sample of older American couples. Study findings have the potential to identify subgroups of older adults at higher risk for poor memory and ultimately inform the development of targeted psychosocial intervention strategies.

Methods

Longitudinal data from the Health and Retirement Study (HRS) were employed. HRS is one of the most extensively used health surveys focusing on American older adults’ health, economic, and employment/retirement conditions. Launched in 1992, HRS investigators at the University of Michigan have been conducting follow-up surveys every 2 years. At baseline, respondents were interviewed face-to-face by trained investigators, most often in the respondent’s home (Juster & Suzman, 1995). Follow-up surveys are mostly done via telephone except for those aged 80+ and those who requested a face-to-face interview (Fisher & Ryan, 2018). Refresher cohorts have been added every 6 years (i.e., 1998, 2004, 2010, and 2016) to supplement sample size. At each wave, about 20,000 nationally representative older adult respondents were recruited, with oversamples from Black and Hispanic households and Florida residents (HRS, 2021). HRS respondents are primarily community-dwelling older adults (Sonnega et al., 2014) and their data have been used widely to investigate depressive symptom trajectories among older Americans (Hybels et al., 2016; Liang et al., 2011; Soh et al., 2022; Xiang & Cheng, 2019), providing important implications for clinical psychological intervention targeted at community-dwelling older adults.

This study analyzed longitudinal HRS data provided by the Rand Corporation with harmonized cross-wave variable coding (Bugliari et al., 2021). To maximize sample size, the 2004 HRS survey was used as the baseline because a refresher cohort was added in that year’s wave of data collection. The Rand-HRS data included cognition measures up to the 2016 survey, the outcome period set for purposes of this study’s data analyses. Therefore, in this study, we characterized older couples’ depressive symptom trajectory between 2004 and 2014 (six waves of information), and the outcome variable (i.e., memory) measured in the subsequent wave (i.e., 2 years after the depressive symptom trajectory) to ensure temporal directionality. A total of seven waves of HRS surveys were used in this study.

Among 19,280 HRS respondents participating in the 2004 survey, 12,204 were couple households, meaning both the spouse/partner HRS respondents answered the same set of HRS questionnaires, thus providing information for dyadic research. We further restricted the sample to include only heterosexual couples who were married for at least 5 years (N = 11,461). The restriction of length of current marriage is to rule out the potential influence of short-term marriage and provide time for couple interdependence in health to develop (Koss, 2017; Lu et al., 2021). Finally, at least one member of the couple must have survived to the 2016 survey and provided a memory score. The resulting final analytic sample size was 5690 older couples.

Measures

Depressive Symptom (2004–2014)

Depressive symptoms were measured by the 8-item Center for Epidemiologic Studies Depression (CESD) scale. Respondents and their spouses were asked about the presence (coded as 1) or absence (coded as 0) of certain depressive-related feelings (e.g., depressed, lonely) in the past 1 week. Summing up all items resulted in a total CESD score, which ranged from 0 to 8. A higher CESD score indicates more depressive symptoms. A cutoff score of 4 was used to define an elevated level of depressive symptoms (Kong et al., 2021; Murchland et al., 2021).

Memory Outcome (2016)

HRS included a battery of cognitive tests. Memory was assessed by respondents and their spouses to assess ability to recall a list of immediate or delayed words. The combination of two word list recall tasks is a well-validated and widely used tool to measure memory in HRS (McArdle et al., 2007). Summing all correct items determined the total memory score, which ranged from 0 to 20 with a higher score meaning better memory. Memory items were asked of all-age HRS respondents. However, mental status, another dimension of cognitive function, was obtained only from HRS respondents aged above 65. Considering that many respondents included in this study were below age 65 without the mental status measure, we used the memory measure as the outcome variable to maximize the utility of the valuable dyadic sample.

Individual- and Household-Level Covariates (2004)

Several baseline individual-level measures were included in the current analyses because they were considered as potential confounders to the relationship between depressive symptom trajectory and memory (Gerstorf, Hoppmann, Kadlec, et al., 2009; Zaninotto et al., 2018). Demographic information confounds included age (in years, continuous), ethnicity (Hispanic/non-Hispanic), and education (in years, continuous). The race variable was dichotomous (White/nonwhite) collapsing Black and other races into a “nonwhite” category because other races accounted for only 4%–5% of the respondents and dichotomous variables can facilitate model estimation. As a robustness check, we re-ran the analyses using the 3-level race variable and found our results unchanged. Retirement status recorded whether the individuals were retired (yes/no). Additionally, Limitations of Activities of Daily Living (ADL), which assessed the respondent/spouse’s difficulty in performing daily tasks (e.g., eating, bathing) were also included. ADL limitations range from 0 to 5 with a higher score indicating greater level of dependence. Further, number of chronic conditions was included, measured by counting the number of doctor-diagnosed medical conditions (i.e., heart diseases, diabetes, high blood pressure, cancer, lung disease, stroke, psychiatric problems, arthritis) the couple had, ranging from 0 to 8 with a higher score indicating a larger number of chronic conditions.

In addition, three baseline household-level covariates shared by the couple were entered into the statistical analyses. Length of current marriage was the number of years the couple had been married to each other. We used the length of current marriage reported by HRS respondents, a mix of both wives and husbands. Household size was measured by the number of people (including the couple) living in the couple’s household. Number of children measured the number of living children or step-children that the HRS couple had. Finally, household income was measured by summing up the family members’ income, employer pension, Social Security, and other government transfers and were log-transformed. Household wealth counted the value of all financial assets (e.g., vehicles, stocks, bonds, debts) of the couple and the value of their housing (both primary and secondary residence) and was log-transformed.

Data Analysis

Initial descriptive analysis was conducted to determine the older couples’ characteristics. To examine gender differences, paired t-tests and McNemar’s test were used to assess the distributional differences of continuous and categorical variables, respectively. In addition, Pearson’s product-moment correlation and Cohen’s kappa coefficient were used to test the interdependence of couples’ continuous and categorical variables separately, consistent with prior research (Koss, 2017; Lu et al., 2021).

Next, latent-class growth analysis (LCGA) was used to characterize the pattern of wives’ and husbands’ depressive symptom trajectories. LCGA is a semi-parametric maximum likelihood-based statistical method to group homogeneous individuals with a similar trajectory in the longitudinal data. LCGA model specification is an iterative process and needs to be gradually adjusted to the maximum logical number of subgroups (Nguena Nguefack et al., 2020). In this study, depressive symptom trajectory was estimated by using the CESD sum score as the outcome variable and the linear time term as a covariate.¹ Choosing the optimal number of classes was based on traditional model fitting indices: the Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample size adjusted BIC (SABIC) (Jung & Wickrama, 2008). In addition, the Lo-Mendell-Rubin likelihood ratio test (LMR LRT) was used to compare two models (Lo et al., 2001). In addition, the decision took into consideration model interpretability and research context (Kong et al., 2021). Once the optimal number of classes is decided, LCGA can assign respondents to the subgroup with similar trajectory by estimating the highest posterior group probability of each individual. In other words, each respondent received a new class membership based on their depressive symptom trajectory.

Finally, we examined the association of older couples’ depressive symptom trajectory with their memory by employing structural equation modeling (SEM). SEM is a widely used method to analyze dyadic data and is particularly appropriate to test the Actor-Partner Interdependence Model (APIM) (Koss, 2017; Lu et al., 2021). Essentially, APIM assumes that one’s characteristics and explanatory variables influence both his/her own outcomes (actor affects), and his/her partner’s outcomes (partner effect) (Kenny & Ledermann, 2010). The proposed theoretical APIM model is depicted in Figure 1. We hypothesized that wives’ depressive symptom trajectories between 2004 and 2014 would predict their own memory 2 years later and link to their husbands’ prospective memory score. Similarly, we hypothesized that husbands’ depressive symptom trajectories over time would be predictive not only of their own but also of their wives’ memory 2 years later. All analyses were conducted in R Studio using the packages “psych” (Revelle, 2021), “lcmm” (Proust-Lima et al., 2017; Wardenaar, 2020), “tidyLPA” (Rosenberg et al., 2019), “ggplot2,” and “lavaan” (Rosseel, 2012).

Figure 1.

Actor-partner interdependence conceptual model of couple depressive symptom trajectories and their association with cognition. Note: Model is adjusted for individual-level and couple-level covariates.

Results

Table 1 displays the characteristics of the couples at baseline of the depressive symptom trajectory (i.e., 2004). On average the older couples were aged 62 (wives) and 65 (husbands) and had about 13 years of education. The majority were White, non-Hispanic with no/minimal ADL difficulties, chronic conditions, and depressive symptoms. On average, the couples were married for 35.95 years (SD = 12.42), had about 3 children, and 2.5 people were living in the household. Gender comparisons showed that compared to wives, husbands were about 3 years older and had 0.1 more years of education. Husbands reported more medical conditions and lower memory scores than wives. However, husbands had fewer depressive symptoms than wives. Couple interdependence indicated that age and education of wives and husbands were highly correlated (Pearson produce-moment correlation coefficient, r > 0.6). Race and ethnicity were also highly concordant within the couple (Kappa coefficient, k > 0.8). Couples had statistically significant, but relatively small, interdependence in depressive symptoms (r = 0.25, CI: [0.23, 0.28]) and memory (r = 0.26, CI: [0.23, 0.29]).

Table 1.

Descriptive Analyses of Couple Characteristics in 2004.

	WivesN = 5,690	HusbandsN = 5,690	Gender difference^a test statistic, p-value	Couples’ interdependence^b estimate, 95%CI
Individual-level demographic
Age, range. range 50–94, mean ± SD	61.96 ± 7.42	64.82 ± 7.80	t = −49.89, p < .001	r = 0.84, CI: (0.83, 0.85)
Race—nonwhite, n(%)	790 (13.88)	819 (14.40)	χ² = 3.48, p = .06	k = 0.82, CI: (0.80, 0.85)
Ethnicity—Hispanic, n(%)	549 (9.65)	547 (9.61)	χ² = 0.01, p = .93	k = 0.86, CI: (0.84, 0.88)
Education, range 0–17, mean ± SD	12.81 ± 3.00	12.92 ± 3.43	t = −2.97, p = .003	r = 0.64, CI: (0.63, 0.66)
Retirement status—yes, n(%)	2333 (41.92)	3180 (61.76)	χ² = 579.73, p < .001	k = 0.38, CI: (0.35, 0.40)
Individual-level health
ADL, range 0–5, mean ± SD	0.15 ± 0.60	0.14 ± 0.60	t = 0.70, p = .48	r = 0.09, CI: (0.07, 0.12)
Number of chronic condition, range 0–8, mean ± SD	1.52 ± 1.24	1.60 ± 1.30	t = −4.01, p < .001	r = 0.19, CI: (0.17, 0.22)
Depressive symptoms, range 0–8, mean ± SD	1.22 ± 1.84	0.87 ± 1.51	t = 10.53, p < .001	r = 0.25, CI: (0.23, 0.28)
Memory^c, range 0–20, mean ± SD	9.95 ± 3.41	8.74 ± 3.39	t = 20.12, p < .001	r = 0.26, CI: (0.23, 0.29)
Couple-level characteristic
Length of marriage, range 5–66.20, mean ± SD	35.95 ± 12.42
Household size^d, range 2–15, mean ± SD	2.50 ± 1.00
Number of children, range 0–11, mean ± SD	3.28 ± 1.89
Household income^e, range 0–14.87, mean ± SD	10.90 ± 1.16
Household wealth^f, range 0–17.23, mean ± SD	12.19 ± 2.35

Note: The ranges of continuous variables are observed ranges.

^aA paired t-test was used to assess the gender mean difference of continuous variables. McNemar’s test was used to assess the gender distribution difference of binary variables.

^bPearson’s product-moment correlation was used to test for couple interdependence for continuous variables. Cohen’s kappa coefficient was used to test the couple interdependence of categorical variables.

^cMemory was measured in 2016.

^dHousehold size included the couple.

^eHousehold income was log-transformed. Before transformation, the household income ranged between 0 and 2,876,200, mean = 85,105, SD = 119,827.3, median = 57,043, skewness = 9.86, Q₁ = 32,265, Q₃ = 97,783.

^fHousehold wealth was log-transformed. Before transformation, the household wealth ranged between 0 and 30,440,000, mean = 621,946.5, SD = 1,448,486, median = 283,000, skewness = 10.71, Q₁ = 113,000, Q₃ = 283,000.

The fitting indices of the LCGA model for wives and husbands are presented in Table 2. For both genders, the 4-class model had the highest log-likelihood and smallest information criterion values (ICs) compared to models with fewer or more classes. Entropy of the 4-class model was also high. LMR LRT results suggested the 4-class model performed significantly better than the 3-class model, while the 5-class model did not do better than the 4-class model. All model fitting indices suggested the 4-class model was the best fit for wives and husbands.

Table 2.

Fitting Indices of Latent-Class Growth Analysis Models for Wives and Husbands.

	Log-likelihood	AIC	BIC	SABIC	Entropy	LMR LRT (p-value)
Wives
1-class	−64539.4	129084.7	129104.7	129095.1	1.00
2-class	−58933.1	117878.2	117918.1	117899	0.94	<.001
3-class	−57459.2	114936.3	114996.1	114967.5	0.91	<.001
4-class	−56684.0	113392	113471.6	113433.5	0.91	<.001
5-class	−56684.0	113398	113497.5	113449.9	0.77	>.99
Husbands
1-class	−52423.94	104853.9	104873.7	104864.2	1.00
2-class	−46815.2	93642.4	93682.1	93663.0	0.95	<.001
3-class	−45350.49	90719.0	90778.5	90749.9	0.93	<.001
4-class	−44583.72	89191.4	89270.7	89232.6	0.93	<.001
5-class	−44255.03	88540.1	88639.2	88591.5	0.90	>.99

Note: AIC = Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample size adjusted BIC; LMR LRT = Lo-Mendell-Rubin likelihood ratio test. The indices in bold indicate the 4-class model achieved the best fit in both wives' and husbands' models.

Figure 2 visualizes the 4-class depressive symptom trajectory over time for wives and husbands, separately. The general patterns of 4 trajectories in men and women were similar and thus named in the same manner to facilitate interpretation. The persistently low trajectory, in which the CESD score was minimal over time, encompassed most of the wives (n = 4150, 73.55%) and husbands (n = 4374, 79.77%). In contrast, the persistently high trajectory characterized by a chronically elevated level of depressive symptoms (i.e., CESD score above 4), had the smallest number of wives (n = 279, 4.95%) and husbands (n = 213, 3.88%). In addition, the decreasing trajectory, where CESD stayed at a moderate level but declined over time, consisted of 726 wives (12.87%) and 440 husbands (8.02%). Finally, 487 wives (8.63%) and 456 husbands (8.32%) were in in the increasing trajectory group whose CESD score started at a low level but increased over time.

Figure 2.

Latent classes of depressive symptom trajectories for wives and husbands. Note: The horizontal line (y = 4) indicates the threshold for the elevated level of depressive symptoms. Respondents with a CESD score of 4 and above are defined to have an elevated level of depressive symptoms.

Finally, SEM results testing APIM are presented in Table 3. Model fitting was good (comparative fit index [CFI] = 0.909, root mean squared error of approximation [RMSEA] = 0.052, 95% confidence interval [CI] =0.050, 0.054). There was significant actor effect. Specifically, compared to the low trajectory group, the three trajectories of wives (i.e., increasing, high, decreasing) were associated with their own poorer memory (β < 0, p < .05). Comparison of the coefficient size showed wives increasing trajectory was associated with lowest own memory (β = −0.841, 95% CI = −1.261, −0.421), followed by the high trajectory (β = −0.764, 95%CI = −1.328, −0.199). For husbands, their increasing and high depressive symptom trajectories were also predictive of their own poorer memory compared to the low trajectory group. As for partner effect, husbands’ decreasing trajectory was associated with wives’ better memory (β = 0.498, 95% CI = 0.106, 0.890). However, no partner effect was observed in the husbands’ model, indicating wives’ depressive symptom trajectory was not predictive of their husbands’ memory.

Table 3.

Structural Equation Modeling Results for Testing the Actor-Partner Interdependence Model.

	Wives’ memory		Husbands’ memory
	β	95%CI	β	95%CI
Wives’ trajectory	Actor effect		Partner effect
Low	ref		ref
Increasing	−0.841***	(−1.261, −0.421)	0.066	(−0.347, 0.479)
Decreasing	−0.384*	(−0.684, −0.085)	0.167	(−0.129, 0.465)
High	−0.764**	(−1.328, −0.199)	0.029	(−0.508, 0.566)
Husbands’ trajectory	Partner effect		Actor effect
Low	ref		ref
Increasing	0.059	(−0.327, 0.446)	−0.711***	(−1.096, −0.327)
Decreasing	0.498*	(0.106, 0.890)	−0.047	(−0.434, 0.341)
High	0.097	(−0.531, 0.726)	−1.166***	(−1.798, −0.533)

Note: Chi-square = 2122.382, df = 253, p < .001, CFI = 0.909, TLI = 0.883, RMSEA = 0.052, 95%CI = 0.050, 0.054.

Note: The model was adjusted for individual-level demographic and health covariates and couple-level characteristics. The coefficient estimates of couple-level covariates were constrained to be the same for wives and husbands to ensure the shared household characteristics had the same couple effect. We further correlated some covariates, such as age and education, to improve model fit.

*P<0.05; **P<0.01; ***P<0.001

Discussion

Using 12-year longitudinal dyadic data from older American couples, this study is the first of which we are aware to characterize the trajectories/courses of depressive symptoms over time and to examine whether different depressive symptom trajectories predict prospective cognition within the spousal context. Our findings suggest significant and strong actor effect. That is, one’s own depressive symptom trajectory was associated with one’s own cognition. However, partner effect was observed only among husbands with improving depressive symptoms. Wives’ depressive symptom trajectory was not associated with husbands’ cognition.

First, our results indicate couples’ interdependence in cognition is statistically significant, although the correlation is relatively small in magnitude, indicating that there is substantial variability in cognition concordance among couples. This finding adds to existing evidence that cognition among older couples is interconnected (Gerstorf, Hoppmann, Anstey, et al., 2009; Hoppmann & Gerstorf, 2009). Studies have suggested older couples may perform tasks collaboratively by pooling their cognitive capacities and/or knowledge (Hoppmann & Gerstorf, 2009). Therefore, there is great opportunity for future research to investigate the determinants of the spousal interdependence in cognition to enrich dyadic understanding of cognitive aging.

Most notably, four distinct depressive symptom trajectories were identified for wives and husbands: persistently low, increasing, decreasing, and persistently high. The depressive symptom trajectories identified in the current study are congruent with those reported in prior research (Formánek et al., 2020; Hsu, 2012; Xiang & Cheng, 2019). Consistent with Musliner et al. (2016), a small proportion (less than 5%) of older men and women in the present sample persistently experienced elevated depressive symptoms. However, different depressive symptom trajectories were identified in other studies. Zhu and colleagues (2022) reported five trajectories, including non-depressed, mild-, increasing-, decreasing-, and persistently high depressive symptoms. Another study reported six distinct trajectories (Liang et al., 2011). In contrast to Soh et al. (2022), we did not identify a fluctuating depressive symptom trajectory. These discrepancies could be attributed to differences in study design (e.g., sample size, statistical approach, or study duration) and research contexts.

In addition, gender-based differences in depressive symptom trajectories in late adulthood have been reported in previous research (Hsu, 2012; Montagnier et al., 2014). Findings in the present study indicate that overall older women were more likely to experience worsened depressive symptoms over time (i.e., increasing trajectory) than their male counterparts. It is also worth noting that depressive symptoms of wives with increasing trajectory started at a low level but exceeded the cutoff of elevated depressive symptoms later, whereas depressive symptoms of husbands remained below the threshold during the entire study period. These findings jointly indicate that older women generally experience more fluctuations in the longitudinal course of depressive symptoms than do older men (Hsu, 2012).

Second, depressive symptom trajectories had a primary actor effect on couples’ cognitive function. Specifically, compared to those with minimal depressive symptoms over time, older men and women in increasing and persistently high trajectory groups had poorer memory 2 years later. In line with existing evidence (Kaup et al., 2016; Zhu et al., 2022), these results suggest that worsened and chronic elevated depressive symptoms over time could be detrimental to cognitive health in later life. Further, consistent with the strong evidence of depression as a risk factor for ADRD, these findings indicate that a small percentage of older adults with elevated depressive symptoms throughout the study period were at risk for worse cognition. Study results validate prior cognition research in supporting the benefits of using trajectories of depressive symptoms instead of measuring depressive symptoms at merely one time point (Kaup et al., 2016). These findings collectively indicate that more nuanced understanding of depressive symptoms, particularly longitudinal patterns/changes in symptoms, are necessary to achieve an enhanced comprehension of the depression trajectory-cognition link.

Interestingly, we found that, compared to the low depressive symptom trajectory group, decreasing trajectory predicted poorer memory among wives. This finding contradicts prior findings that decreased depressive symptoms were not associated with cognitive change compared to those with low trajectory (Zhu et al., 2022). It is possible that even though wives experienced decreasing depressive symptoms, they were still experiencing a high level of depressive symptoms, and thereby worse cognition, over time. Indeed, our results demonstrate that older women had higher levels of depressive symptoms than older men at baseline.

Third, all but one of the partner effects of depressive symptom trajectories on cognition were not significant in the current study, offering limited evidence for crossover effects in cognition. Existing evidence on the partner effects of depressive symptoms on cognition remain inconclusive. For instance, a study reported that wives’ depressive symptoms predicted their husbands’ subsequent cognitive decline (Gerstorf, Hoppmann, Kadlec, et al., 2009), whereas another study documented that one’s depressive symptoms did not affect one’s partner’s cognitive function (Lee et al., 2012). While partner effects need to be investigated in greater depth, we postulate our result could be explained by the presence of strong actor effects possibly reducing statistical power to detect significant partner effects. Another potential explanation for the limited partner effect observed in our study is that health interdependence among couples occurs within-domain, as opposed to cross-domains (Lee et al., 2012). That is, one’s cognition may be influenced more by one’s partner’s cognition, than by the partner’s mental health. Nevertheless, because partner influences may have additive effects over time (Monin et al., 2018), the small and significant partner effect found in our study is worthy of attention and validation.

Nevertheless, study results unveil the gendered partner effects of depressive symptom trajectory on cognition. Specifically, there was no partner effect of wives’ depressive symptom trajectory on husbands’ memory. However, husbands’ decreasing depressive symptom trajectory predicted wives’ better memory 2 years later. Put differently, similar to Gerstorf, Hoppmann, Kadlec, et al. (2009), our result indicates that husbands’ improving mood over time could be beneficial to wives’ brain health, but not vice versa. A potential explanation is that, from a caregiving perspective, improved psychological wellbeing among husbands may result in lower role strain and thereby more time for self-care among wives who are generally emotionally committed to their husbands’ health, which ultimately could lead to wives’ improved wellbeing in general, including cognition (Thomeer et al., 2015). Alternatively, husbands’ improved psychological wellbeing may lead to higher perceived spousal support among wives and increased marital satisfaction (Ko & Lewis, 2011; Ryan et al., 2014), which could be beneficial to wives’ cognition in the long term. These findings further reveal that wives’ depressive symptoms over time did not significantly influence husbands’ subsequent cognition, which is largely consistent with previous dyadic research showing that husbands’ wellbeing is generally not affected by wives’ physical and mental health conditions (Ayotte et al., 2010; Hoppmann & Gerstorf, 2009).

Limitations and Strength

Several limitations of the study warrant discussion. First, this study used memory as the cognition indicator due to the unavailability of other cognition measures (e.g., mental status) among HRS respondents aged below 65. Studies measuring various dimensions of older couples’ cognition are needed to validate and extend findings of the current investigation. Second, the study’s memory outcome was measured at only one time point, which limited our ability to explore the pattern and rate of cognitive change over time at the couple level, especially given the documented variations in the rate of cognitive decline across various depressive symptom trajectories (Formánek et al., 2020). Additionally, the single assessment of memory precludes our ability to test alternative pathways through which one’s cognition may shape one’s partner’s mental wellbeing. Furthermore, the depressive symptom trajectories characterized in this study are at the individual level. The patterns of various health trajectories at the couple-level warrant future longitudinal dyadic investigation. Moreover, we restricted the sample to those having at least one member who survived and provided memory data in 2016. Such a restriction omitted many observations due to death. The analytic respondents selected into our study sample may be healthier and thus reported better memory scores.

Third, the study sample was restricted to married and heterosexual couples. Our finding cannot be generalized to unmarried older adults, who are reported to have poorer physical and psychological health than their married counterparts (Wilson & Oswald, 2005). Future studies need to examine health interdependence and spousal effects for health status among other relationship types, such as lesbian, gay, bisexual, and transgender (LGBT) couples and non-marital cohabitating older couples. Additionally, due to limitations of secondary data analysis design, variables such as marital quality and childhood socioeconomic status for both members of the couple could not be included in the present analyses. Moreover, over 80% of the study samples were non-Hispanic Whites. Future studies using more diverse samples are needed to determine potential racial/ethnic and cultural differences.

Despite these limitations, this study expands the literature in two important aspects. First, the findings supplement existing knowledge by generating nuanced understanding of long-term depressive symptom trajectories and their association with prospective cognition among older American couples using advanced dyadic analytic approaches. Findings further illustrated the gendered partner effect. Further, by employing longitudinal data over a period of 12 years, the results augment comprehension of the relationship of the temporal ordering of depressive symptoms and cognition. Moreover, the study’s methodology offers a unique approach by utilizing a dyadic perspective to examine the couple interdependence of the depressive symptoms-cognition linkage.

Implications and Future Research

The findings of this study have important practical and research implications. First, although all but one of the partner effects of depressive symptom trajectories on cognition were not significant in the current study, future studies need to continue investigating the spousal interdependence of cognition and potential mechanisms to inform the development of couple-oriented interventions for brain health of older adults (Martire et al., 2010). Second, the empirical results reveal that older adults with intensifying and persistently high depressive symptoms over time are at higher risk of poorer cognition than those with persistently low depressive symptoms, providing evidence for the need to offer targeted interventions. Considering the diverse progression of depressive symptoms, more nuanced analytic approaches, such as trajectory-based models, are recommended for further research. Moreover, due to limitations of the secondary analysis design, we were not able to examine whether respondents in various depressive symptom trajectories utilized mental health services to modify their mental health status (e.g., taking antidepressants, receiving care from mental health professionals). Future studies including measures of mental health care and associated psychological and cognitive health outcomes are needed. Lastly, the study offers some evidence for conducting timely assessments of depressive symptoms and delivering consummate treatment in response to symptomatic changes, preferably by linking spousal data, which may promote cognitive health among older adults.

Conclusion

The results of this study indicate that persistently high and increasing depressive symptom trajectories were associated with worse memory among older men and women 2 years later and therefore warrant additional clinical and research attention. Furthermore, husbands’ declining depressive symptom trajectory predicted wives’ better prospective memory, implying psychosocial interventions targeting depressive symptoms among older men may be beneficial to their spouses’ cognition. A dyadic perspective is recommended for future cognition research to extend and validate our findings and elucidate potential mechanisms underlying cognitive aging in the spousal context.

Footnotes

Acknowledgment

DK and PL designed the study, interpreted the results, drafted and revised the manuscript. PL performed the data analysis. PS, JW, and MS interpreted the results and revised the manuscript.

Declaration of Conflicting Interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Dexia Kong

Mack Shelley

Note

References

Ayotte

B. J.

Yang

F. M.

Jones

R. N.

(2010). Physical health and depression: A dyadic study of chronic health conditions and depressive symptomatology in older adult couples. Journals of Gerontology - Series B Psychological Sciences and Social Sciences, 65 B(4), 438–448. https://doi.org/10.1093/geronb/gbq033

Baumgart

Snyder

H. M.

Carrillo

M. C.

Fazio

Kim

Johns

(2015). Summary of the evidence on modifiable risk factors for cognitive decline and dementia: A population-based perspective. Alzheimer’s & Dementia, 11(6), 718–726. https://doi.org/10.1016/j.jalz.2015.05.016

Bourassa

K. J.

Memel

Woolverton

Sbarra

D. A.

(2015). A dyadic approach to health, cognition, and quality of life in aging adults. Psychology and Aging, 30(2), 449–461. https://doi.org/10.1037/pag0000025

Bugliari

Carroll

Hayden

Hayes

Hurd

Karabatakis

Main

Marks

McCullough

Meijer

Moldoff

Pantoja

Rohwedder

St Clair

(2021). RAND HRS longitudinal file 2018 (V1) documentation. https://www.rand.org/well-being/social-and-behavioral-policy/centers/aging/dataprod/hrs-data.html

Diniz

B. S.

Butters

M. A.

Albert

S. M.

Dew

M. A.

Reynolds

C. F.

(2013). Late-life depression and risk of vascular dementia and Alzheimer’s disease: Systematic review and meta-analysis of community-based cohort studies. British Journal of Psychiatry, 202(5), 329–335. https://doi.org/10.1192/bjp.bp.112.118307

Dufouil

Alpérovitch

(2000). Couple similarities for cognitive functions and psychological health. Journal of Clinical Epidemiology, 53(6), 589–593. https://doi.org/10.1016/S0895-4356(99)00189-4

Falconier

M. K.

Kuhn

(2019). Dyadic coping in couples: A conceptual integration and a review of the empirical literature. Frontiers in Psychology, 10, Article 571. https://doi.org/10.3389/fpsyg.2019.00571

Fisher

G. G.

Ryan

L. H.

(2018). Overview of the health and retirement study and introduction to the special issue. Work, Aging and Retirement, 4(1), 1–9. https://doi.org/10.1093/workar/wax032

Formánek

Csajbók

Wolfová

Kučera

Tom

Aarsland

Cermakova

(2020). Trajectories of depressive symptoms and associated patterns of cognitive decline. Scientific Reports, 10(1), Article 20888. https://doi.org/10.1038/s41598-020-77866-6

10.

Gerstorf

Hoppmann

C. A.

Anstey

K. J.

Luszcz

M. A.

(2009). Dynamic links of cognitive functioning among married couples: Longitudinal evidence from the Australian longitudinal study of ageing. Psychology and Aging, 24(2), 296–309. https://doi.org/10.1037/a0015069

11.

Gerstorf

Hoppmann

C. A.

Kadlec

K. M.

McArdle

J. J.

(2009). Memory and depressive symptoms are dynamically linked among married couples: Longitudinal evidence from the AHEAD study. Developmental Psychology, 45(6), 1595–1610. https://doi.org/10.1037/a0016346

12.

Hoppmann

Gerstorf

(2009). Spousal interrelations in old age - a mini-review. Gerontology, 55(4), 449–459. https://doi.org/10.1159/000211948

13.

Hoppmann

Gerstorf

Hibbert

(2011). Spousal associations between functional limitation and depressive symptom trajectories: Longitudinal findings from the study of asset and health dynamics among the oldest old (AHEAD). Health Psychology, 30(2), 153–162. https://doi.org/10.1037/a0022094

14.

HRS (2021). HRS survey design and methodology. HRS. https://hrs.isr.umich.edu/documentation/survey-design

15.

Hsu

H. C.

(2012). Group-based trajectories of depressive symptoms and the predictors in the older population. International Journal of Geriatric Psychiatry, 27(8), 854–862. https://doi.org/10.1002/gps.2796

16.

Hybels

C. F.

Bennett

J. M.

Landerman

L. R.

Liang

Plassman

B. L.

(2016). Trajectories of depressive symptoms and oral health outcomes in a community sample of older adults. International Journal of Geriatric Psychiatry, 31(1), 83–91. https://doi.org/10.1002/gps.4292

17.

Jung

Wickrama

K. A. S.

(2008). An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology Compass, 2(1), 302–317. https://doi.org/10.1111/j.1751-9004.2007.00054.x

18.

Juster

F. T.

Suzman

(1995). An overview of the health and retirement study. The Journal of Human Resources, 30(suppl), S7–S56. https://doi.org/10.2307/146277

19.

Kaup

A. R.

Byers

A. L.

Falvey

Simonsick

E. M.

Satterfield

Ayonayon

H. N.

Smagula

S. F.

Rubin

S. M.

Yaffe

(2016). Trajectories of depressive symptoms in older adults and risk of dementia. JAMA Psychiatry, 73(5), 525–531. https://doi.org/10.1001/jamapsychiatry.2016.0004

20.

Kenny

D. A.

Ledermann

(2010). Detecting, measuring, and testing dyadic patterns in the actor-partner interdependence model. Journal of Family Psychology : JFP : Journal of the Division of Family Psychology of the American Psychological Association (Division 43), 24(3), 359–366. https://doi.org/10.1037/a0019651

21.

L. K.

Lewis

M. A.

(2011). The role of giving and receiving emotional support in depressive symptomatology among older couples: An application of the actor-partner interdependence model. Journal of Social and Personal Relationships, 28(1), 83–99. https://doi.org/10.1177/0265407510387888

22.

Kong

Solomon

Shelley

(2021). Gender-based depression trajectories following heart disease onset: Significant predictors and health outcomes. Aging & Mental Health, 0(0), 1–8. https://doi.org/10.1080/13607863.2021.1891202

23.

Koss

C. S.

(2017). Beyond the individual: The interdependence of advance directive completion by older married adults. Journal of the American Geriatrics Society, 65(7), 1615–1620. https://doi.org/10.1111/jgs.14939

24.

Kuchibhatla

M. N.

Fillenbaum

G. G.

Hybels

C. F.

Blazer

D. G.

(2012). Trajectory classes of depressive symptoms in a community sample of older adults. Acta Psychiatrica Scandinavica, 125(6), 492–501. https://doi.org/10.1111/j.1600-0447.2011.01801.x

25.

Lee

Paddock

S. M.

Feeney

(2012). Emotional distress and cognitive functioning of older couples: A dyadic analysis. Journal of Aging and Health, 24(1), 113–140. https://doi.org/10.1177/0898264311423703

26.

Lewis

M. A.

McBride

C. M.

Pollak

K. I.

Puleo

Butterfield

R. M.

Emmons

K. M.

(2006). Understanding health behavior change among couples: An interdependence and communal coping approach. Social Science & Medicine, 62(6), 1369–1380. https://doi.org/10.1016/j.socscimed.2005.08.006

27.

K. K.

Cardinal

B. J.

Acock

A. C.

(2013). Concordance of physical activity trajectories among middle-aged and older married couples: Impact of diseases and functional difficulties. Journals of Gerontology - Series B Psychological Sciences and Social Sciences, 68(5), 794–806. https://doi.org/10.1093/geronb/gbt068

28.

Liang

Quiñones

A. R.

Bennett

J. M.

(2011). Multiple trajectories of depressive symptoms in middle and late life: Racial/ethnic variations. Psychology and Aging, 26(4), 761–777. https://doi.org/10.1037/a0023945

29.

Mendell

N. R.

Rubin

D. B.

(2001). Testing the number of components in a normal mixture. Biometrika, 88(3), 767–778. https://doi.org/10.1093/biomet/88.3.767

30.

Kong

Lee

Shelley

(2021). Advance directive completion among older American couples: A dyadic perspective on the role of cognitive function and other factors. The Journals of Gerontology: Series B, 77(6), 1–10. https://doi.org/10.1093/geronb/gbab077

31.

Shelley

(2019). Why spouses depress each other: A cross-national study to test the shared resource hypothesis in depressive symptom concordance within older adult couples. Asian Social Work and Policy Review, 13(3), 307–319. https://doi.org/10.1111/aswp.12183

32.

Martire

L. M.

Schulz

Helgeson

V. S.

Small

B. J.

Saghafi

E. M.

(2010). Review and meta-analysis of couple-oriented interventions for chronic illness. Annals of Behavioral Medicine, 40(3), 325–342. https://doi.org/10.1007/s12160-010-9216-2

33.

McArdle

J. J.

Fisher

G. G.

Kadlec

K. M.

(2007). Latent variable analyses of age trends of cognition in the health and retirement study, 1992-2004. Psychology and Aging, 22(3), 525–545. https://doi.org/10.1037/0882-7974.22.3.525

34.

Meyler

Stimpson

J. P.

Peek

M. K.

(2007). Health concordance within couples: A systematic review. Social Science and Medicine, 64(11), 2297–2310. https://doi.org/10.1016/j.socscimed.2007.02.007

35.

Monin

Doyle

Levy

Schulz

Fried

Kershaw

(2016). Spousal associations between frailty and depressive symptoms: Longitudinal findings from the cardiovascular health study. Journal of the American Geriatrics Society, 64(4), 824–830. https://doi.org/10.1111/jgs.14023

36.

Monin

Doyle

Van Ness

Schulz

Marottoli

Birditt

Feeney

Kershaw

(2018). Longitudinal associations between cognitive functioning and depressive symptoms among older adult spouses in the cardiovascular health study. The American Journal of Geriatric Psychiatry : Official Journal of the American Association for Geriatric Psychiatry, 26(10), 1036–1046. https://doi.org/10.1016/j.jagp.2018.06.010

37.

Monin

Levy

Chen

Fried

Stahl

Schulz

Doyle

Kershaw

(2015). Husbands’ and wives’ physical activity and depressive symptoms: Longitudinal findings from the cardiovascular health study. Annals of Behavioral Medicine, 49(5), 704–714. https://doi.org/10.1007/s12160-015-9705-4

38.

Montagnier

Dartigues

J. F.

Rouillon

Pérès

Falissard

Onen

(2014). Ageing and trajectories of depressive symptoms in community-dwelling men and women. International Journal of Geriatric Psychiatry, 29(7), 720–729. https://doi.org/10.1002/gps.4054

39.

Mourao

R. J.

Mansur

Malloy-Diniz

L. F.

Castro Costa

Diniz

B. S.

(2016). Depressive symptoms increase the risk of progression to dementia in subjects with mild cognitive impairment: Systematic review and meta-analysis. International Journal of Geriatric Psychiatry, 31(8), 905–911. https://doi.org/10.1002/gps.4406

40.

Murchland

A. R.

Zeki Al Hazzouri

Zhang

Elfassy

Grasset

Riley

A. R.

Wong

Haan

M. N.

Jones

R. N.

Torres

J. M.

Glymour

M. M.

(2021). Estimating the effects of Mexico to U.S. migration on elevated depressive symptoms: Evidence from pooled cross-national cohorts. Annals of Epidemiology, 64, 53–66. https://doi.org/10.1016/j.annepidem.2021.08.014

41.

Musliner

K. L.

Munk-Olsen

Eaton

W. W.

Zandi

P. P.

(2016). Heterogeneity in long-term trajectories of depressive symptoms: Patterns, predictors and outcomes. Journal of Affective Disorders, 192, 199–211. https://doi.org/10.1016/j.jad.2015.12.030

42.

Nguena Nguefack

H. L.

Pagé

M. G.

Katz

Choinière

Vanasse

Dorais

Samb

O. M.

Lacasse

(2020). Trajectory modelling techniques useful to epidemiological research: A comparative narrative review of approaches. Clinical Epidemiology, 12, 1205–1222. https://doi.org/10.2147/CLEP.S265287

43.

Ownby

R. L.

Crocco

Acevedo

John

Loewenstein

(2006). Depression and risk for Alzheimer disease: Systematic review, meta-analysis, and metaregression analysis. Archives of General Psychiatry, 63(5), 530–538. https://doi.org/10.1001/archpsyc.63.5.530

44.

Proust-Lima

Philipps

Liquet

(2017). Estimation of extended mixed models using latent classes and latent processes: The R package lcmm. Journal of Statistical Software, 78(2), 1–56. https://doi.org/10.18637/jss.v078.i02

45.

Reczek

Umberson

(2012). Gender, health behavior, and intimate relationships: Lesbian, gay, and straight contexts. Social Science and Medicine, 74(11), 1783–1790. https://doi.org/10.1016/j.socscimed.2011.11.011

46.

Revelle

(2021). Package “psych . https://cran.rstudio.org/web/packages/psych/psych.pdf

47.

Richards

(2011). Prevalence and clinical course of depression: A review. Clinical Psychology Review, 31(7), 1117–1125. https://doi.org/10.1016/j.cpr.2011.07.004

48.

Rosenberg

J. M.

Beymer

P. N.

Anderson

D. J.

Lissa

C. J. Van

Schmidt

J. A.

(2019). tidyLPA: An R package to easily carry out Latent Profile Analysis (LPA) using open-source or commercial software. Journal of Open Source Software, 3(30), Article 978. https://doi.org/10.21105/joss.00978

49.

Rosseel

(2012). Lavaan: An R package for structural equation modeling and more. Version 0.5–12 (BETA). Journal of Statistical Software, 48(2), 1–36. https://doi.org/10.18637/jss.v048.i02

50.

Ryan

L. H.

Wan

W. H.

Smith

(2014). Spousal social support and strain: Impacts on health in older couples. Journal of Behavioral Medicine, 37(6), 1108–1117. https://doi.org/10.1007/s10865-014-9561-x

51.

Schulz

Beach

Hebert

Martire

Monin

Tompkins

Albert

(2009). Spousal suffering and partner’s depression and cardiovascular disease: The cardiovascular health study. The American Journal of Geriatric Psychiatry, 17(3), 246–254. https://doi.org/10.1097/JGP.0b013e318198775b

52.

Soh

Tiemeier

Kawachi

Berkman

L. F.

Kubzansky

L. D.

(2022). Eight-year depressive symptom trajectories and incident stroke: A 10-year follow-up of the HRS (health and retirement study). Stroke, 53(8), 2569–2576. https://doi.org/10.1161/strokeaha.121.037768

53.

Sonnega

Faul

Ofstedal

M. B.

Langa

Phillips

Weir

(2014). Cohort profile: The health and retirement study (HRS). International Journal of Epidemiology, 43(2), 576–585. https://doi.org/10.1093/ije/dyu067

54.

Thomeer

M. B.

Reczek

Umberson

(2015). Gendered emotion work around physical health problems in mid- and later-life marriages. Journal of Aging Studies, 32, 12–22. https://doi.org/10.1016/j.jaging.2014.12.001

55.

Townsend

A. L.

Miller

Guo

(2001). Depressive symptomatology in middle-aged and older married couples: A dyadic analysis. Journals of Gerontology - Series B Psychological Sciences and Social Sciences, 56(6), S352–S364. https://doi.org/10.1093/geronb/56.6.S352

56.

Wardenaar

K. L.

(2020). Latent class growth analysis and growth mixture modeling using R: A tutorial for two R-packages and a comparison with mplus (pp. 1–19). Psyarxiv.

57.

Wilson

C. M.

Oswald

A. J.

(2005). How does marriage affect physical andpsychological health? A survey of the longitudinal evidence. IZA Discussion paper No. 1619. JEL Classification: 10, 12. https://ftp.iza.org/dp1619.pdf

58.

Xiang

Cheng

(2019). Trajectories of major depression in middle-aged and older adults: A population-based study. International Journal of Geriatric Psychiatry, 34(10), 1506–1514. https://doi.org/10.1002/gps.5161

59.

Zaninotto

Batty

G. D.

Allerhand

Deary

I. J.

(2018). Cognitive function trajectories and their determinants in older people: 8 Years of follow-up in the English longitudinal study of ageing. Journal of Epidemiology and Community Health, 72(8), 685–694. https://doi.org/10.1136/jech-2017-210116

60.

Zhu

Xie

Zhong

Blumenthal

J. A.

(2022). Trajectories of depressive symptoms and subsequent cognitive decline in older adults: A pooled analysis of two longitudinal cohorts. Age and Ageing, 51(1), Article afab191. https://doi.org/10.1093/ageing/afab191