Association Between Cognitive Trajectories and Subsequent Health Status, Depressive Symptoms, and Mortality Among Older Adults in the United States: Findings From a Nationally Representative Study

Abstract

Background

Cognitive decline may be an early indicator of major health issues in older adults, though research using population-based data is lacking. Researchers objective was to assess the relationships between distinct cognitive trajectories and subsequent health outcomes, including health status, depressive symptoms, and mortality, using a nationally representative cohort.

Methods

Data were drawn from the National Health and Aging Trends Study. Global cognition was assessed annually between 2011 and 2018. The health status of 4 413 people, depressive symptoms in 4 342 individuals, and deaths among 5 955 living respondents were measured in 2019. Distinct cognitive trajectory groups were identified using an innovative Bayesian group-based trajectory model. Ordinal logistic, Poisson, and logistic regression models were used to examine the associations between cognitive trajectories and subsequent health outcomes.

Results

Researchers identified five cognitive trajectory groups with distinct baseline values and subsequent changes in cognitive function. Compared with the group with stably high cognitive function, worse cognitive trajectories (ie, lower baseline values and sharper declines) were associated with higher risks of poor health status, depressive symptoms, and mortality, even after adjusting for relevant covariates.

Conclusions

Among older adults, worse cognitive trajectories are strongly associated with subsequent poor health status, high depressive symptoms, and high mortality risks. Regular screening of cognitive function may help to facilitate early identification and interventions for older adults susceptible to adverse health outcomes.

It is estimated that one in nine persons aged 65 or older in the United States (US) were living with dementia in 2022, and this number is expected to increase in the coming years due to population aging (1). Dementia is generally characterized by progressive cognitive decline that interferes with the ability to function independently (2). Understanding the patterns and implications of cognitive decline more broadly can help facilitate deeper insights into the development and progression of dementia and its associated adverse health outcomes. Although cognitive decline often becomes increasingly apparent with advancing age (3), the actual pace of decline and the pivotal moments when cognitive deterioration accelerates can vary considerably among individuals (4).

Long-term repeated cognitive assessments provide opportunities to capture heterogeneity of cognitive trajectories in a population of older persons. Some persons may manifest steep cognitive declines, although others may remain stable over time (5–8). For example, one study, which utilized data from seven waves of the Health and Retirement Study (HRS), identified five distinct cognitive trajectory groups (9), including stable high (15.4%), stable moderate (52.6%), stable low (3.6%), high-to-moderate (20.1%), and moderate-to-low (8.3%). However, the majority of studies examining cognitive trajectories have primarily employed a variable-centered approach with mixed-effects models (10–12) or growth curve models (13,14). This approach assumes that cognitive trajectories represent variations of an average trajectory observed within a population. To address potential heterogeneity in population trajectories, most studies using this variable-centered approach have simply compared cognitive trajectories between subgroups according to a prespecified characteristic, such as sex or race (15,16). Such a strategy, however, does not provide a full picture of population heterogeneity.

Long-term decline in cognition, especially a rapid decline, may lead to subsequent adverse health outcomes besides dementia. A better understanding of how distinct cognitive trajectories affect a broad range of health outcomes among older adults may provide novel insights into the consequences of cognitive decline over time. Such insights can empower policymakers to more effectively evaluate the overall health impact of cognitive aging. Previous research, based largely on cross-sectional or longitudinal data with less than three time points, has suggested that cognitive decline is associated with outcomes such as disability and poor psychological well-being (17–19). A small number of studies have investigated whether long-term cognitive trajectories are associated with future dementia, disability, and mortality, and have found that a stable trajectory of high cognitive performance can be crucial for both a low incidence of dementia and healthy aging (20–22). Although health status and depressive symptoms are crucial components of physical and mental health among older adults (23,24), less is known about whether long-term cognitive trajectories are associated with these two health outcomes.

Furthermore, to avoid sample selection bias, it is important to use nationally representative data to examine the relationship between cognitive trajectories and subsequent health outcomes (25,26). For example, an earlier study that examined the associations between distinct cognitive trajectories and death used data from a randomized placebo-controlled trial that excluded persons who developed dementia or disability during the study period (22). As such, the sample was healthier than the general older population, and the heterogeneity of cognitive aging may have been underestimated.

The current study used nationally representative data with long-term follow-up to investigate the association between distinct cognitive trajectories and subsequent health outcomes, including health status, depressive symptoms, and death. To identify these trajectories, and to better capture the shapes of trajectories, and identify distinct trajectory groups, researchers used an innovative Bayesian group-based trajectory model (GBTM) that is more efficient in model selection and more flexible in functional forms than a traditional GBTM model (27).

Method

Data and Study Sample

We used 2011–2019 data from the National Health and Aging Trends Study (NHATS), a population-based longitudinal study of adults aged 65 and older in the United States (28,29). Participants were assessed annually starting in 2011 (Round 1). If a participant could not respond due to severe cognitive or physical impairment, a proxy respondent, usually a family member or primary caregiver, was interviewed. During Round 5, a refreshment sample was added to maintain the national representativeness of the study population.

In latent class modeling approaches for longitudinal data, it is generally recommended to have at least three measurement time points for proper estimation (30,31). Therefore, participants included those with at least three rounds of valid cognitive scores. After estimating the cognitive trajectory group memberships, the final samples included 4 413 participants for health status, 4 342 for depressive symptoms, and 5 955 for mortality. Details about the sample selection procedures can be found in Appendix Figure 1. Appendix Table 1 shows the sample size, the number of decedents and the number of individuals lost to follow up in each NHATS round.

Measures

Cognitive function

At each round of 2011–2018, if a participant could not respond, the assistance of a proxy was enlisted. Specifically, the proxy was asked if the participant could participate in cognitive activities. If the proxy responded “no,” a reason was obtained (eg, the participant has dementia/Alzheimer’s disease and/or is unable to speak or hear), and the participant’s cognitive score was coded as “missing.” If the proxy responded “yes,” three domains of cognition were assessed: (1) memory (an immediate and delayed 10-word recall), (2) orientation (querying the date, month, year, and day of the week and naming the President and Vice President of the United States), and (3) executive function (a clock drawing test with a subscale). The composite cognitive score sums across the three subscales (range: 0–33), with higher scores indicating better cognitive function, following Quinones’ approach (16). The distributions of cognitive scores in each Round are described in Appendix Figure 2. If the participant couldn’t answer a question, refused or did not attempt to draw clock, a score of 0 was given (16).

Health outcomes

We evaluated three health outcomes: health status, depressive symptoms, and death. The distributions for the three health indicators at baseline (ie, Round 1 for the initial cohort and Round 5 for the replenishment sample) are described in Appendix Figure 3.

In 2019, participants were asked, “Would you say that, in general, your health is excellent, very good, good, fair, or poor?” and to rate their health status from 1 (excellent) to 5 (poor). As a robustness check, researchers also dichotomized health status, indicating poor or fair health (yes or no). Depressive symptoms were measured using the Patient Health Questionnaire-4 (PHQ-4) in 2019, which includes four questions: “Over the last month, how often have you (1) had little interest or pleasure in doing things; (2) felt down, depressed, or hopeless; (3) felt nervous, anxious, or on edge; and (4) been unable to stop or control worrying?” Each question was scored from 0 (not at all) to 3 (nearly every day), with a total score ranging from 0 to 12. As a robustness check, researchers also dichotomized depressive symptoms using a cutoff score of greater or equal to 6, as recommended by previous studies (32).

NHATS recorded participants’ death status in each Round. Those who dropped out before Round 9 were excluded from the analysis. Appendix Table 2 shows the characteristics of the drop-out participants. Hispanics, non-Hispanic others, and those older were more likely to drop out than other groups; those with above high school education were less likely to drop out. There were no significant differences for other sociodemographic and health variables.

Covariates

All covariates were measured at baseline when the participants enrolled in the study (ie, 2011 or 2015). Following previous studies (22,24,33–36) on the relationship between cognitive decline and health outcomes, researchers examined the following sociodemographic characteristics: age and its quadratic form (continuous), race/ethnicity (non-Hispanic White, non-Hispanic Black, non-Hispanic other, and Hispanic), sex (female, male), educational attainment (less than high school, high school, or beyond high school), and marital status (never married, married or living with a partner and separated, divorced, or widowed). Researchers have also adjusted for the baseline values of health status (for the health status and mortality analyses) and of depressive symptoms (for the depressive symptoms and mortality analyses), whether the health status or depressive symptoms are proxy-reported (vs self-reported), smoking (yes or no), and number of comorbidities (0, 1–3, 4+), which includes eight self-reported, physician-diagnosed chronic conditions: heart attack, heart diseases, high blood pressure, arthritis, osteoporosis, diabetes, lung disease, and cancer. Additionally, researchers considered the following characteristics: number of children in the household (0, 1–3, 4+), number of siblings (0, 1–3, 4+), and whether the participant was covered by Medicare Part D (yes or no), Medicaid (yes or no), and Tricare (yes or no). Following previous studies (24), researchers categorized the number of children and the number of siblings in the household as previously, to capture potential nonlinearity in the relationship between the number of children and health. The category of “unknown” in these variables is coded as missing. The percentage of missingness for each variable is below 5%. Descriptive statistics of these variables are shown in Table 1.

Table 1.

Descriptive Statistics, by Health Status (HS), Depressive Symptom (DS), and Mortality

	Number of individuals			Mean (SD) or Percentage
	HS	DS	mortality	HS	DS	Mortality
Female	4 413	4 342	5 955	58.6	58.5	58.1
Age	4 413	4 342	5 955	74.7	74.7	76.5
				(6.8)	(6.8)	(7.6)
Race/Ethnicity	4 413	4 342	5 955
White				70.1	70.2	70.5
Black				20.3	20.3	20.5
Hispanic				5.6	5.6	5.4
Other				3.9	3.9	3.7
Education attainment	4 354	4 285	5 885
Below High school				19.2	19.0	22.0
High school				25.6	25.5	26.2
Beyond High school				55.3	55.5	51.8
Number of siblings	4 398	4 329	5 934
0				17.7	17.6	20.9
1–3				58.2	58.2	56.7
4+				24.2	24.2	22.4
Number of children	4 413	4 342	5 955
0				8.2	8.1	8.7
1–3				60.3	60.5	59.2
4+				31.5	31.4	32.1
Medicare Part D	4 229	4 159	5 682	63.3	63.2	64.1
Medicaid	4 300	4 230	5 794	12.9	12.8	14.2
Tricare	4 338	4 267	5 852	5.8	5.9	5.6
Comorbidity	4 363	4 296	5 880
0				10.3	10.4	9.1
1–3				71.9	71.9	70.1
4+				17.9	17.7	20.8
Marital status	4 410	4 339	5 951
Never married				4.0	3.9	3.8
Married/live with a partner				54.9	55.1	51.3
Separated, divorced, widowed				41.1	41.0	44.8
Smoking regularly	4 410	4 340	5 951	48.4	48.4	49.9
Outcome was proxy reported	4 413	4 342	5 955	1.3	1.2	2.2

Notes: Mean (standard deviation) for continuous variables and percentage for categorical variables; SD represents standard deviation.

Statistical Analyses

We applied a Bayesian GBTM approach to identify distinct cognitive trajectory groups using cognitive scores before 2019. As with the frequentist GBTM approach (27), the Bayesian GBTM approach assumes that there are multiple latent subgroups with distinct trajectories among the study population. However, the Bayesian GBTM has several methodological advantages compared with the frequentist GBTM. Most importantly, it employs a Bayesian model-averaging procedure to select functional forms of trajectories. This approach accounts for uncertainty in model selection by incorporating all possible models although efficiently reducing the number of models needed for model fit comparisons (27). For example, when conducting model selection, two models may have very similar fit statistics such as Bayesian information criterion (BIC) and Akaike information criterion (AIC), although exhibiting different trajectory patterns. The frequentist GBTM typically chooses one single model with a particular functional form. The choice can be arbitrary when the two models cannot be distinguished from one another based on fit statistics.

We used the years since the first observation as the time variable (24). Researchers constructed models with varying numbers of latent groups and permitted the functional forms to include combinations of linear, quadratic, and cubic terms. The ultimate functional form of trajectories, for a specific number of latent groups, is determined as a weighted average of linear, quadratic, and cubic terms, tailored to achieve the best fit for the data. Following existing literature (27), the selection of the number of latent groups was based on the Bayesian information criterion (BIC) to identify the optimal model, considering that each group should be sufficiently represented (ie, ensuring each group comprises at least 5% of the participants). Each participant was assigned to the latent group with the highest posterior probability (the group membership is time constant). Trajectories for all groups were plotted with 95% credible intervals (CIs).

Using the identified trajectory group memberships, researchers applied ordinal logistic regression and Poisson regression models to evaluate associations between cognitive trajectories and health status and depressive symptoms, respectively, and logistic regression models to evaluate associations between cognitive trajectories and death, with and without adjusting for covariates. Researchers performed multiple imputations by chained equations to handle the missingness in covariates using Stata package ICE and R package MICE (37,38). The regression coefficients from 10 imputed data sets were combined based on the Rubin’s rule (39). As a sensitivity analysis, researchers performed analyses without multiple imputations. Use of NHATS data for this study was approved by the Yale University institutional review board (IRB Protocol ID: 2 000 029 030). All analyses were performed using R version 4.1.0 and Stata version 17.0.

Results

We identified five cognitive trajectory groups. Their corresponding trajectories are shown in Figure 1. The estimates of model parameters and their 95% CIs are shown in Appendix Table 3. The results of model diagnostics are shown in Appendix Table 4.

Figure 1.

Cognitive trajectories with 95% Credible Intervals. Each line represents the cognitive trajectories for each of the five groups. 95% Credible Intervals are shown in gray areas. The labels assigned to each group were devised in a relative sense to depict the characteristics of each group. This was done because there is no clinical guidance available to precisely define cutoffs for cognitive scores as “high” or “medium high,” etc., or to categorize trajectory changes as “stable” or “sharp decline,” etc.

The five groups can be distinguished by their baseline values because there is no overlap in their estimated trajectories. Groups 1 (“high start, stable”), 2 (“high-medium start, stable”), and 3 (“medium start, slight decline”) started with high or medium baseline levels of cognitive function (23.8, 20.7, and 17.4, respectively), comprising 12.8%, 30.9%, and 28.7% of the participants, respectively. The cognitive trajectories for these three groups remained relatively stable or experienced only a slight decline over time. Approximately 30% of cognitively healthy older adults experienced cognitive decline during our study period. Group 4 (“low-medium start, sharp decline”) started with a baseline value of 14.2, representing 19.4% of the participants, and experienced a relatively sharp cognitive decline over time. Group 5 (“low start, sharp decline”) started with a baseline value of 8.5, accounting for 8.3% of the participants, and experienced a sharp cognitive decline over time. Baseline sample characteristics by trajectory groups are shown in Appendix.

Results examining the associations between cognitive trajectory group memberships and health status are shown in Table 2 (the full table is shown in Appendix Table 6). Results without multiple imputations are shown in Appendix Table 6 and were generally consistent with the results with multiple imputations. The first group was the reference because it had the best trajectory (ie, stably high cognitive function over time). The groups are ordered according to progressively worsening trajectories. Group 5 was considered the worst group because the cognitive scores were always lower than those of other groups. In the unadjusted model, all four groups exhibited significantly poorer health status compared with Group 1 (“high start, stable”). The estimated odds ratios were 1.9 (95% CI: 1.6–2.2), 3.3 (95% CI: 2.8–3.9), 5.6 (95% CI: 4.6–6.9), and 8.7 (95% CI: 6.2–12.3) for Groups 2–5, respectively. To illustrate, for participants in Group 5 (“low start, sharp decline”), the odds of experiencing worse health status were 8.7 times that of participants in Group 1 (“high start, stable”). The increasing magnitudes indicate that participants’ health status worsened as the cognitive trajectories worsened. After adjusting for covariates, the magnitudes of odd ratios became smaller (1.3, 1.6, 1.6, and 2.7 for Groups 2–5, respectively), but the pattern of worsening health status with worse cognitive trajectory persisted. The results did not change substantively when employing logistic regression to examine the binary indicator of health status (see Appendix Table 7).

Table 2.

Associations Between Cognitive Trajectory Group Membership and Health Status

	Odds Ratios and 95% Confidence Intervals
	Without covariates	With covariates
Group membership (Ref. “High start, stable”)
High-medium start, stable	1.9	1.3
	(1.6,2.2)	(1.1, 1.6)
Medium start, slight decline	3.3	1.6
	(2.8,3.9)	(1.3, 1.9)
Low-medium start, sharp decline	5.6	1.6
	(4.6,6.9)	(1.2, 2.0)
Low start, sharp decline	8.7	2.7
	(6.2,12.3)	(1.8, 4.0)

Notes: Estimates are from ordinal logistic regressions. 95% Confidence Intervals were reported in parentheses. N = 4 413. Covariates include age and its quadratic form, race/ethnicity, sex, educational attainment, marital status, whether the health status are proxy-reported, smoking, number of comorbidities, number of children in the household, number of siblings, whether the respondent participant was covered by Medicare Part D, Medicaid, and Tricare, and baseline health status.

Results examining the associations between cognitive trajectory group memberships and depressive symptoms are shown in Table 3 (the full table is shown in Appendix Table 8). In the unadjusted analysis, all four groups showed higher risks of depressive symptoms compared with Group 1 (“high start, stable”). The estimated incidence rate ratios were 1.3 (95% CI: 1.2–1.4), 1.7 (95% CI: 1.6–1.9), 2.5 (95% CI: 2.3–2.7), and 2.8 (95% CI: 2.5, 3.1) for Groups 2–5, respectively. For instance, the incidence rate of depressive symptoms for participants in Group 5 (“low start, sharp decline”) is 2.8 times higher than those in Group 1 (“high start, stable”). After adjusting for covariates, the magnitudes of incidence rate ratios became smaller (1.2, 1.3, 1.6, and 1.7 for Groups 2–5, respectively), but the pattern of increasing incidence of depressive symptoms with worse cognitive trajectory persisted. The results remained consistent when employing logistic regression to examine the binary indicator of depressive symptoms (see Appendix Table 9).

Table 3.

Associations Between Cognitive Trajectory Group Membership and Depressive Symptom

	Incidence Rate Ratios and 95% Confidence Intervals
	Without covariates	With covariates
Group membership (Ref. “High start, stable”)
High-medium start, stable	1.3	1.2
	(1.2, 1.4)	(1.1, 1.3)
Medium start, slight decline	1.7	1.3
	(1.6, 1.9)	(1.2, 1.5)
Low-medium start, sharp decline	2.5	1.6
	(2.3, 2.7)	(1.5, 1.8)
Low start, sharp decline	2.8	1.7
	(2.5, 3.1)	(1.5, 2.0)

Notes: Estimates are from Poisson regressions. 95% Confidence Intervals were reported in parentheses. N = 4 342. Covariates include age and its quadratic form, race/ethnicity, sex, educational attainment, marital status, whether the depressive symptoms are proxy-reported, smoking, number of comorbidities, number of children in the household, number of siblings, whether the respondent participant was covered by Medicare Part D, Medicaid, and Tricare, and baseline depressive symptoms.

Results examining the associations between cognitive trajectory group memberships and mortality are presented in Table 4 (the full table is shown in Appendix Table 10). Results without multiple imputations are shown in Appendix Table 10. In the unadjusted analysis, all four groups showed significantly higher odds of death compared with Group 1 (“high start, stable”). The estimated odds ratios were 2.0 (95% CI: 1.5–2.8), 4.8 (95% CI: 3.5–6.6), 12.1 (95% CI: 8.9–16.7), and 33.4 (95% CI: 23.8–47.7) for Groups 2–5, respectively. For instance, the odds of deaths were 33.4 times higher among participants in Group 5 (“low start, sharp decline”) compared with those in Group 1 (“high start, stable”). The magnitudes of the associations decreased when adjusting for covariates, with odds ratios of 1.4, 2.6, 5.9, and 15.3 for Groups 2–5, respectively.

Table 4.

Associations Between Cognitive Trajectory Group Membership and Mortality

	Odds Ratios and 95% Confidence Intervals
	Without covariates	With covariates
Group membership (Ref. “High start, stable”)
High-medium start, stable	2.0	1.4
	(1.5, 2.8)	(1.0, 2.0)
Medium start, slight decline	4.8	2.6
	(3.5, 6.6)	(1.9 ,3.7)
Low-medium start, sharp decline	12.1	5.9
	(8.9, 16.7)	(4.1, 8.3)
Low start, sharp decline	33.4	15.3
	(23.8, 47.7)	(10.2, 22.9)

Notes: Estimates are from logistic regressions. 95% Confidence Intervals were reported in parentheses. N = 5 955. Covariates include age and its quadratic form, race/ethnicity, sex, educational attainment, marital status, whether the health status or depressive symptoms are proxy-reported, smoking, number of comorbidities, number of children in the household, number of siblings, whether the respondent participant was covered by Medicare Part D, Medicaid, and Tricare, baseline health status, and baseline depressive symptoms.

Discussion

This study examined whether cognitive trajectories are associated with three health outcomes in older US adults: health status, depressive symptoms, and mortality. Researchers identified five distinct trajectory groups for cognitive functioning. Approximately 30% of cognitively healthy older adults experienced cognitive decline during our study period. Researchers found that, in general, compared with the group with stably high cognitive function, worse cognitive trajectories (ie, lower baseline values and sharper declines) were independently associated with significant decreases in health status, increases in depressive symptoms and higher mortality risks, even after adjusting for a comprehensive set of covariates. For health status and depressive symptoms, even after controlling for baseline health in our analyses, their relationship with cognitive functioning can be bidirectional.

Our estimated cognitive trajectories are consistent with those estimated in previous studies using nonpopulation-based data in the United States (4,22), potentially giving us more confidence in the robustness of the findings. For example, using data from 754 community-living older adults aged 70 years or older in greater New Haven, Connecticut, United States (US), and applying the traditional GBTM, one study identified five distinct cognitive trajectories with groups starting high and also having relatively stable trajectories and groups starting low and having relatively sharp declines over time (4). Our results on the relationship between cognitive trajectories and mortality are also consistent with findings from previous studies in other countries and using nonpopulation-based samples in the United States: worse cognitive trajectories (ie, low start and sharp declines over time) are associated with higher mortality risks (22,40).

The current study contributes important new information to the literature by examining population heterogeneity of cognitive trajectories among US older adults. Previous studies have demonstrated that, among US older adults with cognitive impairment, trajectories of health status, physical functioning, and depressive symptoms cannot be determined solely from the baseline values, as two trajectory groups may have similar baseline values (24,36,41). Moreover, trajectory groups that begin with the poorest health may not necessarily experience the greatest decline over time. In contrast, the findings from the current study indicate that, among US general older adults, cognitive trajectory groups exhibit distinct baseline values, and those with the lowest initial values tend to experience the greatest declines over time.

Another unique contribution of our study is the examination of associations between cognitive trajectories and non-mortality outcomes. Although one prior study examined the relationship between trajectories of individual cognitive domains and trajectories of depressive symptoms (42), studies have not examined the relationship between global cognitive trajectories and depressive symptoms in the United States. Therefore, our findings add to the literature by examining this relationship. When comparing Groups 1 and 5, an incidence rate ratio of 2.8 among those in Group 5 suggests a nonnegligible increase in participants’ depressive symptoms. To our knowledge, no prior study has evaluated the relationship between cognitive trajectories and health status. For health status, an odds ratio of 8.7 represents a considerable deterioration when comparing the two groups with the best (ie, Group 1) and the worst (ie, Group 5) trajectories.

This study has several limitations. First, researcher’s sample excluded participants who dropped out of NHATS in the following rounds. Researchers investigation reveals that dropouts were more likely to be Hispanics and non-Hispanic others, older individuals, and less likely to have above high school education (see Appendix Table 2). If anything, based on researcher’s results in Tables 2 and 3, those who have above high school education tended to have better health compared with others. Second, the study examined associations rather than causal relationships between cognitive trajectories and subsequent health outcomes. There could be unobserved confounders such as certain environmental factors that may have contributed to both declining cognitive trajectories and worse health outcomes. Although researcher's data set provides only limited information on environmental factors, an analysis that also accounted for neighborhood inclusiveness (such as strong social connections, willingness to assist, and trust), neighborhood physical conditions (including litter, broken glass, or debris on sidewalks and streets), and whether the sampled individual resides in a metropolitan or non-metropolitan area yielded similar findings, as demonstrated in Appendix Table 11. Finally, it was not possible for us to analyze cognitive trajectories for participants who were unable to complete the cognitive test. Although AD8 scores were administered to all proxies when the sample person couldn’t complete the cognitive test, they were measured on different scales compared with our cognitive scores, making direct comparisons in trajectory modeling impractical.

Despite these limitations, our findings provide insights into the impact of cognitive trajectories on important health outcomes. An older person’s cognitive trajectory contributes important prognostic information beyond a single assessment of cognition. Consequently, it becomes imperative to gain a deeper understanding of the factors that influence an individual’s cognitive trajectory. By doing so, researchers can enhance our comprehension of the intricate relationship between cognitive health and other aspects of health, thereby paving the way for more effective interventions and care strategies. In addition, researchers found that approximately 30% of cognitively healthy older adults at baseline experienced severe cognitive decline over the 7-year follow-up period. This statistic can be valuable for policymakers in estimating the cognitive aging burden within the US older population.

The cognitive trajectories identified in this study have the potential to serve as cognitive aging phenotypes for predicting the risk of poor health status, depressive symptoms, and mortality. The observed associations between cognitive trajectories and health status underscore the potential utility of cognitive assessments as predictive indicators for health outcomes among older adults. Although researcher's findings do not directly imply interventions to improve health status, they highlight the importance of incorporating cognitive assessments into routine clinical care. Early identification of individuals with declining cognitive trajectories may enable health care providers to implement targeted surveillance and preventive interventions, thereby mitigating associated health risks and optimizing overall well-being.

The correlations between cognitive trajectories and depressive symptoms emphasize the interconnectedness of cognitive and mental health domains and underscore the importance of holistic assessments in identifying individuals at heightened risk for depressive symptoms. Clinicians can leverage cognitive assessments alongside traditional depression screening tools to improve mental well-being in older adult populations.

It is striking that the risks for mortality increase much more dramatically with worsening cognitive trajectories compared with poor health status and depressive symptoms. Mortality selection, wherein individuals with worse cognitive trajectories who survive into older age may represent a relatively healthier subgroup, can have mitigated the associations between cognitive trajectories and health status and depressive symptoms. Recognizing cognitive patterns associated with increased mortality can guide end-of-life care planning, with cognitive status considered when developing individualized care plans for older patients nearing the end of life.

Although it appears that baseline cognitive function distinguishes cognitive trajectories among older adults in the United States, future studies should evaluate cognitive trajectories for specific subpopulations, such as individuals with cognitive impairment. For health outcomes such as depressive symptoms and health status, individuals with cognitive impairment often exhibit greater heterogeneity in their trajectories compared with the broader category of older US adults, and baseline values are often insufficient to distinguish subsequent trajectories (24,36,41). Additional research is needed to examine whether this is also the case for cognitive trajectories. Future research should also evaluate how trajectories of health status or depressive symptoms change with cognitive trajectories over time. This approach holds the potential to yield valuable insights into the interplay between cognitive health and other facets of well-being.

Funding

The authors received support from the National Institute on Aging (R21AG074238-01), the National Institute on Minority Health and Health Disparities (R01MD017298; P50MD017356), the Claude D. Pepper Older Americans Independence Center at Yale School of Medicine (P30AG021342), and the Institution for Social and Policy Studies at Yale University.

Conflict of Interest

T.M.G. serves on the Editorial Board of the Journal of Gerontology: Medical Sciences.

Author Contributions

E.Z. and Y.W. designed and conceptualized the study. E.Z. and Y.Z. prepared the data. Y.Z. conducted the data analyses. E.Z., Y.Z., and Y.W. drafted the manuscript. All authors contributed to finalizing the manuscript. The author(s) read and approved the final manuscript.