Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jul 5.
Published in final edited form as: J Aging Health. 2022 Oct 4;35(5-6):335–344. doi: 10.1177/08982643221130381

A 13-Year Time-Lagged Description of General Cognitive and Functional Abilities in Older Men: A Cross-Lagged Panel Model

Peter Martin 1, Leonard W Poon 2, Gina Lee 1, Hardeep K Obhi 3, Kalpana J Kallianpur 4,5, Bradley Willcox 4,5, Kamal Masaki 4,5
PMCID: PMC10321552  NIHMSID: NIHMS1907221  PMID: 36194185

Abstract

Objectives:

The purpose of this study was to evaluate a cross-lagged panel model of general cognition and functional abilities over 13 years. The goal was to determine whether general cognitive abilities predict or precede functional decline versus functional abilities predicting cognitive decline.

Methods:

The sample included 3508 men (71–93 years of age at baseline) of the Kuakini Honolulu-Asia Aging Study who were tested repeatedly using a global cognitive test and an assessment of functional capacity. Education and age served as covariates. Cross-lagged models were tested, assessing stationarity of stability and cross-lags.

Results:

The overall model fit the data well. Cognitive scores had better stability than functional abilities and predicted functional abilities more strongly than functional abilities predicted cognitive scores over time. The strength of all cross-lags increased over time.

Discussion:

These longitudinal data show that cognitive scores predicted functional decline in a population-based study of older men.

Keywords: cognitive function, physical function, well being

The literature shows general cognition and everyday functional abilities decline in normal aging among healthy adults (Mendes de Leon et al., 2002; Salthouse, 2012). There is also evidence that general cognition and everyday functions are positively related and changes are not independent (Lee et al., 2019). At one end of the continuum, high cognition and everyday functioning among normal, healthy adults are expected. On the other end of the continuum, older individuals with moderate to severe dementia (e.g., stages 5, 6, or 7, Reisberg et al., 1982) could only carry out limited everyday functional activities (Hill et al., 1995). This paper addresses the question on the relationship between general cognition and functional abilities over time. Does general cognitive ability decline prior to observed functional changes? Do they occur at the same time? Alternatively, does functional change precede observable general cognitive changes? Importantly, are these relationships stable longitudinally?

In the United States, about 14% of older adults (i.e., 70 years and older) have some form of dementia (Larson, Langa, & Yates, 2013; Plassman, Langa, Fisher et al., 2007). An additional 22% show cognitive impairment (Plassman, Langa, & Fisher et al., 2008). Older adults with cognitive impairment have an increased risk of functional dependence, hospitalization, institutionalization, and death (Arrighi et al., 2010; Miller & Weissert, 2000; Yaffe et al., 2010). It is therefore important to understand the interrelationship between cognitive and functional abilities.

Rajan et al. (2012) investigated the association of cognitive function with age-related transition to and progression of disability in activities of daily living (ADLs) in a population-based longitudinal cohort of nondisabled older adults. They reported that cognitive function was associated with the rate of progression of ADL disability. Han et al. (2016) also examined the association between cognitive abilities and functional changes. Community-dwelling persons aged 70 years or older who participated in the Yale Precipitating Events Project were assessed over 108 months. Participants who had not declined in cognitive abilities incurred the lowest incidence rates of disability in ADLs and instrumental activities of daily living (IADLs).

Cognitive and functional abilities (or physical health) are important in later life; the directional relationship between these constructs has been examined by some studies (e.g., Gross et al., 2011; Infurna et al., 2011; Nelson et al., 2020; Okely & Deary, 2020). Infurna et al. (2011) and Nelson et al. (2020) tested the directionality of the association between memory and physical health or functional limitations of older adults with data from the Health and Retirement Study by employing and comparing different sets of latent change score models. Both studies revealed that the effect of memory from the previous time point on physical health of the subsequent wave had stronger effects compared to the opposite direction, which is the effect from physical health on memory. Okely and Deary (2020) also tested the association between physical functioning and cognition over 9 years using the bivariate dual change model (BDCSM), examining multiple cognitive ability domains (i.e., verbal memory, processing speed, and visuospatial ability) and physical functions (i.e., walking speed, grip strength, and lung function). The results of their models indicated that declines in visuospatial ability led to declines in walking speed after controlling for covariates. In general, the study findings of Okley and Dearly highlighted that cognitive function serves as an early marker of declines in physical function. The study of Ghisletta and colleagues (2019), on the other hand, indicated the opposite direction when compared to the findings of Infurna et al. (2011), Nelson et al. (2020), and Okley and Deary (2020). They examined the effect of metabolic risk and fluid intelligence on the subsequent change of each other by using the BDCSM, and the results showed significant effects of metabolic risk on subsequent changes in fluid intelligence. Although not all of the studies examining the directionality of the relationship between cognitive and functional abilities specifically included functional health as their construct, the results of the studies that employed the BDCSM method seem to agree with the notion of cognitive function as a source of the subsequent change in functional/physical health.

Gross et al. (2011) used a different methodological approach to examine changes in cognitive functioning with tests that probed different cognitive domains (i.e., memory, inductive reasoning, processing speed, and global mental status). They employed parallel process latent growth models to analyze trajectory changes in these cognitive domains and ADL functioning among older adults. The authors concluded inductive reasoning best explained the variance of changes in everyday functioning at baseline, and memory best predicted changes of everyday functioning over time. However, the variance of changes in everyday functioning over time explained by cognition was comparatively lower than that of health domains such as physical health. In addition, even though Gross et al. (2011) examined a parallel process latent growth model of cognition and functional abilities, they did not test the predictive pathways from disability in functional status to changes in cognition.

There are other studies that did not test parallel processes or bidirectional influences of these two constructs, but rather tested how cognitive functioning impacts everyday functioning/functional abilities (e.g., Burdick et al., 2005; Carpenter et al., 2006; Farias et al., 2013; Shimada et al., 2016; Yam & Marsiske, 2013). For example, Burdick et al. (2005) defined and examined several predictors of functional impairment of older adults in assisted living facilities; factors included cognitive, behavioral, and health status. Their findings indicated that greater cognitive impairment significantly predicted worse functional abilities. In another study (Shimada et al., 2016), cognitive frailty and physical limitations were predictors of disability in IADLs.

Shimada et al. also uncovered a significant association between IADLs and cognitive impairment and reported that cognitive impairment was the strongest risk factor for instrumental functional deficits (Shimada et al., 2016). Others, including Yam and Marsiske (2013), explored cognitive functioning as a longitudinal predictor of IADL disability among older adults over a 5-year period using multilevel longitudinal analyses. Their results indicated that everyday cognition contributed primarily to the explanation of between-person variance in self-reported IADL, suggesting that individuals who showed less decline in cognitive function at a given occasion also reported less IADL disability at that time. However, everyday cognition did not contribute significantly to the explanation of within-person effects.

The general literature on the influence of cognition on ADLs established that worse cognitive status and greater cognitive impairment are associated with greater functional limitations. One study (Jefferson et al., 2006) included the elements of executive functioning as predictors of instrumental activities of daily living (IADLs) among older adults and demonstrated that frontal lobe executive functions may be associated with changes in IADL disability.

Several studies have shown important associations between cognitive and functional abilities but did not determine whether cognitive and functional abilities have dual influences on each other over time. Furthermore, it is not clear whether basic physical function and instrumental activities are similarly affected by levels of cognitive function. Finally, few studies address the importance of stationarity when testing cross-lagged models, evaluating whether the cross-lagged effects are stable over time or whether they change in strength.

There are different possible mechanisms that could explain why functional ability and cognition influence each other over time. One way to explain this association is by evaluating this as part of resource models (Martin, 2002) or as part of the disablement process (Sun et al., 2020; Verbrugge & Jette, 1994). The disablement process model suggests that cognitive impairment can lead to functional disability. ADL disability is hypothesized as an outcome influenced by physical and cognitive abilities as well as environmental barriers and psychological factors. A resource model suggests that high cognitive levels serve as an individual resource protecting functional ability. In tandem, high levels of functional ability may provide necessary physical resources to maintain higher levels of cognitive function. Each function may serve as a resource for the other (Martin, 2002). However, as older adults move through the later years of their lives, the association between cognition and functional ability may change as they are more likely to become disabled. A decline in one resource (i.e., cognition or functional ability) may have an effect on the other resource and, as a result, the association between cognition and functional ability may increase over time. Some have suggested that at various levels of dementia, the association between cognitive function and functional ability gets stronger (Liuseifert et al., 2014). It is not clear whether a change in this association is noticed in cognitively functioning individuals.

Therefore, it is important to assess the stability or stationarity of cross-lags in the association of both resources. This question has not been sufficiently addressed in the literature. As such, to address these gaps, we explored whether basic physical function and instrumental activities equally predict cognitive changes over time and vice versa.

The current study aimed to test a cross-lagged panel model of cognitive and functional abilities over four time points (Figure 1). By using a global measure of cognitive function, we may predict impending changes in functional abilities performance by conducting cross-lagged analyses. It is unclear which resource is the “lead” component in this relationship, so we viewed the different pathways as competing hypotheses. However, based on age-related decline in cognitive and physical function, we hypothesized lower stationarity for the cross-lags over time.

Figure 1.

Figure 1.

Open in a new tab

Conceptual Model.

Method

Participants

The participants are members of the Kuakini Honolulu Heart Program (HHP), a long-term prospective study of 8006 men, who were 45–68 years at the study inception (Exam 1) in 1965–68. All Japanese men on the island of Oahu, Hawaii with contact information were invited, and study participants were drawn from first- and second-generation Japanese American men (Yano et al., 1984) using the World War II Selective Services Registration file of American men of Japanese ancestry born between 1900 and 1919. The registration file was inspected for birthdate (1900–1919) and apparent Japanese surname and/or listings of Japanese origin. Based on the list, men who lived on outer islands, died prior to 1965, and did not have a current address were excluded; all 12,261 men from the registration file were first contacted by mail and then by follow-up telephone calls or home visits (Worth & Kagan, 1970). A total of 8006 individuals responded and participated during Exam 1.

The Kuakini Honolulu-Asia Aging Study (HAAS) was initiated with the fourth exam cycle of the Kuakini HHP to study cognitive function, dementia, disability, and other conditions of aging. This project was approved by the Kuakini Medical Center IRB, and written informed consent was obtained from the participant and/or a family member at every examination.

We utilized data from four time points (exams), spanning the period from Exam 4 (1991–1993) to Exam 8 (2001–2004). We did this because data collection on cognitive and functional abilities began at Exam 4, when participants were 71–93 years old.

Participants who were diagnosed with dementia at Exam 4 were excluded from our study. Dementia prevalence was determined by repeated screenings for cognitive functioning with the Cognitive Abilities Screening Instrument (CASI, White et al., 1996). CASI scores range from 0 to 100, with high scores indicating better cognitive function. Scores lower than 74 are considered as an indicator of possible dementia. Participants with a score lower than 74 were invited for a return evaluation. If the second CASI score was also lower than 74 or if a score on the informants’ questionnaire on cognitive decline was over a defined cut-off score, another visit was scheduled for a complete dementia assessment. Given our interest in examining cognition in a non-clinical sample, we excluded 223 men from the analyses who had a dementia diagnosis at Exam 4 (Abbott et al., 2004). It is important to note that only a small percentage of the sample were diagnosed with prevalent dementia after the active screening process. Participants who developed dementia at a later time point were not excluded from the analysis.

Exam 7 (1999–2000) was excluded because data on functional abilities were not collected at that exam cycle. However, the distance between all reported time points was about equivalent (i.e., 3 years between Exams 4 and 5, three years between Exams 5 and 6, and four years between Exams 6 and 8). In total, our analytic sample included 3508 men. The participant selection is depicted in Figure 2. After evaluating the fit of our model with the analytic sample, we also computed the analysis with participants who continued the study over all time points.

Figure 2.

Figure 2.

Open in a new tab

Participant Selection.

We compared participants at Exam 4 with those who had dropped out of the study before Exam 4. The continuing participants were younger, had more years of education, and fewer children (all at Exam 1). They were more likely to be married at Exam 1 and less likely to have worked in skilled or professional occupations.

We also compared participants who dropped out of the study after Exam 4 or had incomplete data, those who continued to participate throughout all exam cycles. Continuing participants were significantly younger, significantly more educated and had significantly higher cognitive scores at Exam 4. They also had fewer depressive symptoms, as measured by the CES-D, and were less likely to be limited in their IADLs. No significant differences between continuing participants and non-participants were seen for basic ADLs.

Participant characteristics are presented in Table 1. At Exam 4, participants were on average 78 years old and had a mean education level of 10.6 years. Overall, participants had relatively high levels of cognitive functioning (M = 84.71 out of 100 points) at Exam 4, but the scores declined to 75.34 at Exam 8. Participants had about one or two functional limitations throughout all exam points.

Table 1.

Participant Demographics and Scores on Study Variables.

Exam 4
 Exam 5
 Exam 6
 Exam 8
(1991–1993)
 (1994–1996)
 (1997–1999)
 (2001–2004)
N 3,508 2,592 1,956 1,171
Age (yrs.) 77.50 (4.44) 79.81 (4.10) 82.60 (3.90) 86.37 (3.38)
Education (Yrs.) 10.56 (3.19)
CASI 84.71 (11.92) 82.02 (12.42) 78.35 (16.52) 75.34 (18.51)
BADLs 0.90 (1.61) 0.30 (0.96) 0.50 (1.31) 0.49 (1.24)
IADLs 1.22 (1.68) 0.99 (1.59) 1.42 (1.84) 1.30 (1.78)
ADLs 1.39 (2.10) 1.27 (2.22) 1.92 (2.77) 1.81 (2.74)
Open in a new tab

Note. Mean values (standard deviations in parentheses). CASI = Cognitive Abilities Screening Instrument. BADLs = basic activities of daily living; IADLs = instrumental activities of daily living; ADLs = activities of daily living (i.e., combined BADL and IADL scores). BADLs, IADLs, and ADLs are coded as impairments.

Measures

The Cognitive Abilities Screening Instrument (CASI) provides an assessment of global cognitive functioning across nine cognitive domains: attention, concentration, orientation, short-term memory, long-term memory, language abilities, visual construction, list-generating fluency, and abstraction/judgment (Teng et al., 1994). The assessment consists of 25 items across the nine domains, which can be summed to create a total score ranging from 0 to 100 (highest). Scores lower than 74 compare to a score of 22 on the Mini-Mental State Examination and are indicators of possible dementia (Abbott et al., 2004). Cronbach’s alpha for the CASI in our sample ranged from 0.75 in Exam 4 to 0.90 in Exam 8.

Functional abilities were measured in three different ways. First, basic activities of daily living (BADLs) were assessed with six items (Young et al., 1995) where participants noted difficulties such as walking in their home, getting out of bed, eating, dressing, bathing, and toileting. Functional abilities were also measured with IADLs through five items: having difficulties with light housework, shopping, preparing meals, paying bills, and talking on the phone. We also combined BADL and IADL scores to create a total score of activities of daily living (ADLs) in an effort to have a more global assessment of functional abilities. Internal consistencies for these measures were quite variable, ranging from 0.55 (Exam 5) to 0.90 (Exam 8) for BADLs, from 0.32 (Exam 4) to 0.66 (Exam 8) for IADLs, and from 0.34 (Exam 4) to 0.82 (Exam 8) for ADLs. It is perhaps more appropriate to think of these activities as “count” variables, rather than as consistent summary scores, because participants may continue with some activities while not engaging in others. It is noteworthy that the internal consistencies get stronger over time. All ADL items were recoded so that high scores reflected high activity levels.

Analytic Approach

We conducted analyses using Mplus version 8 (Muthén & Muthén, 1998–2017). First, the fit of the hypothesized model, M1, was evaluated (Figure 1). Second, five additional models testing for stationarity were included. M2 and M3 tested for stationarity of functional and cognitive abilities, respectively, and M4 was a test of stationarity for both functional and cognitive abilities. Finally, models M5–M7 examined stationarity for functional abilities predicting cognitive abilities at later time points, cognitive abilities predicting functional abilities at later time points, and cognitive and functional abilities predicting each other over time, respectively. In addition to these analyses, we conducted a sensitivity analysis by including only participants who had complete data across occasions.

Results

Table 2 summarizes the bivariate correlations of the study variables. Age and education were significantly associated with general cognitive scores and with functional abilities at all four time points.

Table 2.

Bivariate Correlations of Study Variables.

Age Education CASI 4 CASI 5 CASI 6 CASI 8 ADL 4 ADL 5 ADL 6 ADL 8
Age 1.00
Education −0.20** 1.00
CASI 4 −0.36** 0.36** 1.0
CASI 5 −0.30** 0.32** 0.68** 1.0
CASI 6 −0.30** 0.25** 0.59** 0.71**
CASI 8 −0.34** 0.31** 0.52** 0.57** 0.74** 1.0
ADL 4 −0.19** 0.12** 0.39** 0.23** 0.18** 0.05 1.0
ADL 5 −0.18** 0.10** 0.30** 0.47** 0.29** 0.22** 0.42** 1.0
ADL 6 −0.17** 0.06* 0.27** 0.39** 0.60** 0.31** 0.30** 0.44** 1.0
ADL 8 −0.25** 0.13** 0.27** 0.33** 0.50** 0.63** 0.11** 0.34** 0.45** 1.0
Open in a new tab

Note. CASI = Cognitive Abilities Screening Instrument. ADL = activities of daily living.

*

p < .05.

**

p < .01.

To evaluate power for structural equation models, RMSEA fit indices can be used to compare model fit with a “not-close fit” hypothesis (Kline, 2016). We used the Preacher and Coffman (2006) R code calculating the power with α = 0.05, nine degrees of freedom (our hypothesized model) and a sample size N = 3508. Comparing the close-fit hypothesis (RMSEA = 0.024) with the not-so-close hypothesis (RMSEA = 0.10), power exceeds 0.99.

Next, three separate structural equation models were computed for each functional abilities score (i.e., BADL, IADL, and the combined ADL score). Results are summarized in Table 3. All models fit the data very well, χ2 (df = 9) = 30.45, p < 0.001, CFI = 0.998, RMSEA = 0.026 [CI = 0.16–0.037] for the BADL model, χ2 (df = 9) = 23.31, p < .01, CFI = 0.998, RMSEA = 0.021 [CI = 0.011–0.032] for the IADL model, and χ2 (df = 9) = 26.89, p = .002, CFI = 0.998, RMSEA = 0.024 [CI = 0.014–0.034] for the ADL model. To test for sensitivity, the ADL model was computed with participants with full data from Exam 4 to Exam 8 (N = 662). The overall fit was also excellent, χ2 (df = 9) = 9.91, p = 0.36, CFI = 1.0, RMSEA = 0.012 [CI = 0.00–0.047].

Table 3.

Model Comparison for Basic Activities of Daily Living (BADL), Instrumental Activities of Daily Living (IADL), and a Combined Scale (ADL) (N = 3508).

Model 1 (BADL) Model 2 (IADL) Model 3 (ADL)
CASI, Exam 4
 Age −0.30*** −0.30*** −0.30***
 Education 0.30*** 0.30*** 0.30***
CASI, Exam 5
 CASI, Exam 4 0.68*** 0.71*** 0.69***
 ADL, Exam 4 0.09*** 0.07*** 0.09***
CASI, Exam 6
 CASI, Exam 5 0.64*** 0.68*** 0.66***
 ADL, Exam 5 0.13*** 0.06*** 0.09***
CASI, Exam 8
 CASI, Exam 6 0.87*** 0.85*** 0.83***
 ADL, Exam 6 0.03 0.07*** 0.09***
ADL, Exam 4
 Age 0.12*** 0.14*** 0.16***
 Education 0.03 0.10*** 0.09***
ADL, Exam 5
 ADL, Exam 4 0.58*** 0.29*** 0.41***
 CASI, Exam 4 0.14*** 0.23*** 0.23***
ADL, Exam 6
 ADL, Exam 5 0.51*** 0.23*** 0.37***
 CASI, Exam 5 0.24*** 0.33*** 0.32***
ADL, Exam 8
 ADL, Exam 6 0.43*** 0.27*** 0.34***
 CASI, Exam 6 0.45*** 0.53*** 0.53***
χ2, p-value 30.45, p < .001 23.31, p < .01 26.89, p < .01
df 9 9 9
CFI 0.998 0.998 0.998
RMSEA, CI 0.026 [0.016–0.037] 0.021 [0.011–0.032] 0.024 [0.014–0.034]
Open in a new tab

Note. CASI = Cognitive Abilities Screening Instrument. ADL = activities of daily living.

The results for associations were similar between the three separate functional ability scores (BADL, IADL, and ADL) and cognitive scores. All of the functional ability measures at later time points were predicted by cognitive abilities. Higher cognitive ability scores significantly predicted higher functional scores at later time points; higher functional scores also predicted higher cognitive ability scores at later time points, except for the last time point (Exam 8), when BADL did not significantly predict cognitive function. The strongest association was obtained for cognitive scores predicting the IADL and ADL scores at the last exam of the study (β = 0.53).

Because the ADL model is the most comprehensive model, we proceeded using this model to test for stationarity (Kenny, 1975). The results are summarized in Table 4. The hypothesized model (M1), as indicated above, fit the data well, χ2 (df = 9) = 26.89, p = .002, CFI = 998. RMSEA = 0.024 [CI = 0.014–0.034].

Table 4.

Fit Indices for Model Testing.

Model N χ2 df Δχ2 CFI RMSEA
M1 Hypothesized 3,508 26.89 9 — 0.998 0.024
M2 Stationarity for ADL 3,508 28.02 11 1.13 0.998 0.021
M3 Stationarity for CASI 3,508 89.31 11 62.42* 0.990 0.045
M4 Stationarity for ADL and CASI 3,508 397.91 13 371.02* 0.951 0.092
M5 Stationarity for ADL on CASI 3,508 267.21 11 240.32* 0.967 0.081
M6 Stationarity for CASI on ADL 3,508 28.46 11 1.57 0.998 0.021
M7 Stationarity for ADL on CASI and CASI on ADL 3508 301.37 13 27.97* 0.963 0.080
Open in a new tab

Note. All models are compared to Model 1.

*

p < 0.05.

Six additional models were computed to test stationarity. Model 2 (M2) tested for stationarity of functional abilities. To evaluate this model, the stabilities (i.e., the autoregressive effects) for ADLs were constrained to be equal. Because the fit was not significantly worse than its comparison model M1, Δχ2 (df = 2) = 1.13, p = .57, the results suggest that the stabilities across time were equal. However, testing stationarity of cognitive scores (M3) resulted in a significantly decreased fit of the model, Δχ2 (df = 2) = 62.42, p < .001. Furthermore, when setting stabilities for cognitive and functional abilities to be the same (M4), the model comparisons yielded a significant difference, Δχ2 (df = 4) = 371.02, p < .001. When allowing stationarity of cross-lags, (i.e., setting the value to be the same for every cross-lag), the model fit significantly decreased for functional abilities predicting cognitive abilities (M5), Δχ2 (df = 2) = 240.32, p < .0, but not for cognitive scores predicting functional abilities (M6), Δχ2 (df = 2) = 1.57, p = .46. The final model assumed stationarity for both sets of cross-lags, assuming no predictive change over time. This model, M7, also did not fit as well with the data when compared to the hypothesized model M1, Δχ2 (df = 4) = 27.97, p < .001.

Taken together, the results suggest there is stability (i.e., the autoregressive effects) for functional ability over time. Furthermore, the associations between functional and cognitive abilities appear to strengthen for both directional effects. Figure 3 depicts the final model using the combined functional ability scores. The stability of cognitive function was moderately high, ranging from β = .66, p < .001 to β = .83, p < .001. In contrast, the stabilities for functional abilities were considerably lower, ranging from β = 0.34, p < .001 to β = 0.41, p < .001. All cross-lags were significant. The paths from cognitive to functional abilities were consistently stronger than the pathways from functional to cognitive abilities. The strongest cross-lagged association was obtained for cognitive scores at Exam 6 predicting functional abilities at Exam 8 (β = 0.53).

Figure 3.

Figure 3.

Open in a new tab

Cross-Lagged Panel Model Cognition and Functional Abilities (Activities of Daily Living Total score).

We computed a final cross-lagged panel model with the smaller sample of participants who had complete data across all time points to evaluate the sensitivity of results. Results indicated that stabilities (i.e., the autoregressive effects) for cognition (ranging from 0.65 to 0.68) and functional abilities (ranging from 0.26 to 0.29) were lower when compared to the analytic sample. Furthermore, the path from functional ability at Exam 4 to cognition at Exam 5 was not significant (β = 0.03, p > 0.05), and most of the cross-lagged coefficients were smaller in size when compared to the analytic sample.

Discussion

The aim of this study was to add to our understanding and provide clarity on the relationship between general cognition and functional abilities among older adults over a 13-year period. Data from 3508 men were included in the current analyses; those were men aged 71–93 at baseline (in 1991–1993) with 3 follow-up examinations through 2001–04.

We added age and education as covariates in our model. In addition to evaluating cross-lagged effects, we also tested stationarity in stability and cross-lagged coefficients. Three results are noteworthy: 1) the model fit was acceptable, 2) stabilities were higher for general cognition compared to functional abilities, and 3) cross-lags for cognitive scores predicting functional abilities were stronger than for functional abilities predicting cognitive scores.

The model fit very well with the data, suggesting that the data were well represented in the model. It is important to note that the participants were not demented and were functioning well at the first time of testing. General cognition and everyday performances of the men were high and exceptional over the 13-year span. Functional scores were also high for these men during the same timeframe. Global cognition was compromised only very slightly over 13 years and their functional abilities remained fairly constant. Owing to the high functional scores over 13 years, the amount of variance might not be sufficiently high to detect meaningful changes for these men.

This study added to our knowledge of longitudinal changes in cognitive and functional abilities among older adults. The study replicated the close relationship between changes in cognition and function reported in the literature (Carpenter et al., 2006; Farias et al., 2013; Yam & Marsiske, 2013). The major finding of our cross-lagged panel model showed participants with greater levels of cognitive functioning had higher scores in functional abilities at a later time point. The results answer our initial question of whether cognitive performances predict functional abilities, or vice versa. The results are consistent with findings reported by Gross et al. (2011) who noted that cognitive function explained the variance of changes in everyday functioning at baseline and over time. Other studies (Burdick et al., 2005; Han et al., 2016; Rajan et al., 2012; Shimada et al., 2016) also indicated that greater cognitive impairment predicted worse functional abilities.

The results are consistent with other research demonstrating that the influence of cognitive function on physical activity was stronger than the opposite influence (Cheval et al., 2020). Cognitive function may be particularly relevant to ADL function, given the importance of organization and regulation of mental processes such as formulation of plans, monitoring their implementation, and changing to new plans (Mograbi et al., 2014). It is conceivable that cognitive resources are needed to engage in daily activities. It may be more difficult to stay active and to plan and execute daily activities if cognitive functions are compromised. On the other hand, one could more easily stay cognitively active even if basic ADL functioning is compromised. For example, individuals could still engage in reading, playing games, or connecting socially, even if their mobility or household activities are compromised.

Utilizing cognitive abilities to stay active is consistent with resource theories (Hall & Fong, 2007; Martin, 2002) and studies demonstrating that by staying active, more brain activity is generated in the frontal lobe (Cheval et al., 2018). Cognitive function serves as an individual resource to “protect” physical function over time. Likewise, physical function, to a lesser extent, helps to maintain less decline in cognitive function. However, as time goes by, the linkage or spillover effect appears to get stronger, perhaps caused by the disablement process (Verbrugge & Jette, 1994). The results suggest that cognitive abilities should be screened regularly to predict the potential onset of disability with the goal of delaying the disablement process as much as possible (Connolly et al., 2017; Sun et al., 2020)

The sensitivity analysis replicated the results for the survivorship sample. The regression coefficients were weaker for most cross-lags, and the general conclusion indicates that the effect of cognitive function on functional abilities is weaker at each time point when compared to the effect of functional abilities on cognitive function. It is likely that cognitive function is the more dominant variable affecting decline in functional abilities.

We also compared three different versions of activities of daily living in their association with general cognitive abilities. When comparing these measures, the combined version of basic and instrumental activities better predicted changes in cognitive functioning than basic or instrumental activities alone. Likewise, cognitive functioning predicted the combined ADL measure more strongly than basic or instrumental activities of daily living.

As is true for all research, the study has a number of limitations. First, the sample is from a small region of the United States and includes only male participants of Japanese descent. Therefore, the results cannot be generalized to other geographic areas or demographic groups. Owing to the strong performances in cognition and function over the 13-year period, it would be important to replicate the results with other longitudinal data.

Another limitation includes the limited measures of our study. The Kuakini Honolulu-Asia Aging Study did not use a comprehensive functional assessment at all time points, and our cognitive function assessment was limited to the CASI, a broad cognitive abilities assessment. Likewise, the assessment of functional abilities was somewhat narrow in scope, and the ADL measure had somewhat poor internal consistency, particularly in early assessments. This could have underestimated the overall effect of functional abilities on cognitive function. Finally, although the study drew on longitudinal data, functional abilities data were not available at all time points, changing the test interval at the last two measures. We also cannot rule out that other important variables could have predicted our outcomes; as such, future studies should explore other cognitive reserve variables and genetic, as well as physiological makers that could contribute to cognitive and functional abilities. A positive observation from this study was men who were 71–93 years in age and not demented at the beginning of the study were able to maintain a relatively high level of general cognition and functional status 13 years later. Finally, we acknowledge the limitation of cross-lagged panel models that cannot disentangle between- and within-person effects.

In spite of these limitations, the results from 3508 participants over about 13 years in our cross-lagged panel model provide important and consistent information about the interplay between cognitive and functional abilities. Both seem to affect each other in important ways, suggesting precipitous cognitive decline is a good indicator of impending functional loss. Early clinical intervention for cognitive decline could help to preserve functional abilities later. As more and more individuals reach very old age, maintaining and optimizing cognitive and functional abilities should be a priority for older adults, their families and their communities.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute on Aging (R01AG027060, R01AG038707, U01 AG019349), USDA National Institute of Food and Agriculture (IOW04116), Hawaii Community Foundation (Grant 2004-0463), National Heart, Lung, and Blood Institute (N01-HC-05102), National Institute of General Medical Sciences (1P20GM125526), National Institutes of Health (N01-AG-4–2149).

Footnotes

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.

References

  1. Abbott RD, White LR, Ross GW, Masaki KH, Curb JD, & Petrovitch H (2004). Walking and dementia in physically capable elderly men. Journal of the American Medical Association, 292(12), 1447–1453. 10.1001/jama.292.12.1447 [DOI] [PubMed] [Google Scholar]
  2. Arrighi HM, Neumann PJ, Lieberburg IM, & Townsend RJ (2010). Lethality of Alzheimer disease and its impact on nursing home placement. Alzheimer Disorders Association Disorders, 24(1), 90–95. 10.1097/WAD.0b013e31819fe7d1 [DOI] [PubMed] [Google Scholar]
  3. Burdick DJ, Rosenblatt A, Samus QM, Steele C, Baker A, Harper M, Mayer CG, Brandt J, Rabins P, & Lyketsos CG (2005). Predictors of functional impairment in residents of assisted-living facilities: The Maryland Assisted Living Study. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 60(2), 258–264. 10.1093/gerona/60.2.258 [DOI] [PubMed] [Google Scholar]
  4. Carpenter GI, Hastie CL, Morris JN, Fries BE, & Ankri J (2006). Measuring change in activities of daily living in nursing home residents with moderate to severe cognitive impairment. BMC Geriatrics, 6(1), 7. 10.1186/1471-2318-6-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cheval B, Orsholits D, Sieber S, Courvoisier D, Cullati S, & Boisgontier MP (2020). Relationship between decline in cognitive resources and physical activity. Health Psychology, 39(6), 519–528. 10.1037/hea0000857 [DOI] [PubMed] [Google Scholar]
  6. Cheval B, Radel R, Neva JL, Boyd LA, Swinnen SP, Sander D, & Boisgontier MP (2018). Behavioral and neural evidence of the rewarding value of exercise behaviors: A systematic review. Sports Medicine, 48(6), 1389–1404. 10.1007/s40279-018-0898-0 [DOI] [PubMed] [Google Scholar]
  7. Connolly D, Garvey J, & McKee G (2017). Factors associated with ADL/IADL disability in community dwelling older adults in the Irish longitudinal study on ageing (TILDA). Disability Rehabilitation, 39(8), 809–816. 10.3109/09638288.2016.1161848 [DOI] [PubMed] [Google Scholar]
  8. Farias ST, Chou E, Harvey DJ, Mungas D, Reed B, DeCarli C, Park L, & Beckett L (2013). Longitudinal trajectories of everyday function by diagnostic status. Psychology and Aging, 28(4), 1070–1075. 10.1037/a0034069 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Ghisletta P, Mason F, Dahle CL, & Raz N (2019). Metabolic risk affects fluid intelligence changes in healthy adults. Psychology and Aging, 34(7), 912–920. 10.1037/pag0000402 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Gross AL, Rebok GW, Unverzagt FW, Willis SL, & Brandt J (2011). Cognitive predictors of everyday functioning in older adults: Results from the ACTIVE cognitive intervention trial. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 66B(5), 557–566. 10.1093/geronb/gbr033 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Hall PA, & Fong GT (2007). Temporal self-regulation theory: A model for individual health behavior. Health Psychology Review, 1(1), 6–52. 10.1080/17437190701492437 [DOI] [Google Scholar]
  12. Han L, Gill TM, Jones BL, & Allore HG (2016). Cognitive aging trajectories and burdens of disability, hospitalization and nursing home admission among community-living older persons. Journal of Gerontology, Series A: Biological Sciences and Medical Sciences, 71(6), 766–771. 10.1093/gerona/glv159 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hill RD, Bäckman L, & Fratiglioni L (1995). Determinants of functional abilities in dementia. Journal of the American Geriatric Society, 43(10), 1092–1097. 10.1111/j.1532-5415.1995.tb07006.x [DOI] [PubMed] [Google Scholar]
  14. Infurna FJ, Gerstorf D, Ryan LH, & Smith J (2011). Dynamic links between memory and functional limitations in old age: Longitudinal evidence for age-based structural dynamics from the ahead study. Psychology and Aging, 26(3), 546–558. 10.1037/a0023023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jefferson AL, Paul RH, Ozonoff AL, & Cohen RA (2006). Evaluating elements of executive functioning as predictors of instrumental activities of daily living (IADLs). Archives of Clinical Neuropsychology, 21(4), 311–320. 10.1016/j.acn.2006.03.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Kenny DA (1975). Cross-lagged panel correlation: A test for spuriousness. Psychological Bulletin, 82(6), 887–903. 10.1037/0033-2909.82.6.887 [DOI] [Google Scholar]
  17. Kline RB (2016). Principles and practice of structural equation modeling (4th ed.). Guilford. [Google Scholar]
  18. Larson EB, Yaffe KM, & Langa FE (2013). New insights into the dementia epidemic. New England Journal of Medicine, 369(24), 2275–2277. 10.1056/NEJMp1311405 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lee M-T, Jang Y, & Chang W-Y (2019). How do impairments in cognitive functions affect activities of daily living functions in older adults? PLoS One, 14(6), Article e0218112. 10.1371/journal.pone.0218112 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Liu-Seifert H, Siemers E, Sundell K, Price KL, Han B, Selzler KJ, Aisen PS, Cummings JL, Raskin J, & Mohs RC (2014). Cognitive and functional decline and their relationship in patients with mild Alzheimer’s dementia. Journal of Alzheimer’s Disease, 43(3), 949–955. 10.3233/JAD-140792 [DOI] [PubMed] [Google Scholar]
  21. Martin P. (2002). Individual and social resources predicting well-being and functioning in later years: Conceptual models, research, and practice. Ageing International, 27(2), 3–29. 10.1007/s12126-002-1000-6 [DOI] [Google Scholar]
  22. Mendes de Leon CF, Guralnik JM, & Bandeen-Roche K (2002). Short-term change in physical function and disability: The Women’s Health and Aging Study. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 57(6), S355–S365. 10.1093/geronb/57.6.S355 [DOI] [PubMed] [Google Scholar]
  23. Miller EA, & Weissert WG (2000). Predicting elderly people’s risk for nursing home placement, hospitalization, functional impairment, and mortality: A synthesis. Medical Care Research and Review, 57(3), 259–297. 10.1177/107755870005700301 [DOI] [PubMed] [Google Scholar]
  24. Mograbi DC, Faria CA, Fichman HC, Paradela EM, & Lourenço RA (2014). Relationship between activities of daily living and cognitive ability in a sample of older adults with heterogeneous educational level. Annals of the Indian Academy of Neurolology, 7(1), 71–76. 10.4103/0972-2327.128558 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Muthén LK, & Muthén BO (1998–2017). Mplus user’s guide (8th ed.). Muthén & Muthén. [Google Scholar]
  26. Nelson NA, Jacobucci R, Grimm KJ, & Zelinski EM (2020). The bidirectional relationship between physical health and memory. Psychology and Aging, 35(8), 1140–1153. 10.1037/pag0000579 [DOI] [PubMed] [Google Scholar]
  27. Okely JA, & Deary IJ (2020). Associations between declining physical and cognitive functions in the Lothian Birth Cohort 1936. The Journals of Gerontology: Series A, 75(7), 1393–1402. 10.1093/gerona/glaa023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Plassman BL, Langa KM, Fisher GG, Heeringa SG, Weir DR, Ofstedal MB, Burke JR, Hurd MD, Potter GG, Rodgers WL, Steffens DC, McArdle JJ, Willis RJ, & Wallace RB (2008). Prevalence of cognitive impairment without dementia in the United States. Annals of Internal Medicine, 148(6), 427–434. 10.7326/0003-4819-148-6-200803180-00005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Plassman BL, Langa KM, Fisher GG, Heeringa SG, Weir DR, Ofstedal MB, Burke JR, Hurd MD, Potter GG, Rodgers WL, Steffens DC, Willis RJ, & Wallace RB (2007). Prevalence of dementia in the United States: The aging, demographics, and memory study. Neuroepidemiology, 29(1–2), 125–132. 10.1159/000109998 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Preacher KJ, & Coffman DL (2006). Computing power and minimum sample size for RMSEA. [Computer software]. Available from http://quantpsy.org/ [Google Scholar]
  31. Rajan KB, Hebert LE, Scherr P, Dong X, Wilson RS, Evans DA, & Mendes de Leon CF (2012). Cognitive and physical functions as determinants of delayed age at onset and progression of disability. Journal of Gerontology, Series A: Biological Sciences and Medical Sciences, 67(12), 1419–1426. 10.1093/gerona/gls098 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Reisberg B, Ferris SH, de Leon MJ, & Crook T (1982). The global deterioration scale for assessment of primary degenerative dementia. American Journal of Psychiatry, 139(9), 1136–1139. 10.1176/ajp.139.9.1136 [DOI] [PubMed] [Google Scholar]
  33. Salthouse T. (2012). Consequences of age-related cognitive declines. Annual Review of Psychology, 63(1), 201–226. 10.1146/annurev-psych-120710-100328 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Shimada H, Makizako H, Lee S, Doi T, Lee K, Tsutsumimoto K, Harada T, Hotta R , Bae S, Nakakubo S, Harada K, & Suzuki T (2016). Impact of cognitive frailty on daily activities in older persons. The Journal of Nutrition, Health & Aging, 20(7), 729–735. 10.1007/s12603-016-0685-2 [DOI] [PubMed] [Google Scholar]
  35. Sun Q, Jiang N, Lu N, & Lou VWQ (2020). The bidirectional relationship between cognitive function and loss hierarchy of activities of daily living among older adults with disabilities in China: A cross-lagged analysis. Unpublished preprint; 10.21203/rs.3.rs-45857/v1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Teng EL, Hasegawa K, Homma A, Imai Y, Larson E, Graves A, Sugimoto K, Yamaguchi T, Sasaki H, Chiu D, et al. (1994). The cognitive abilities screening instrument (CASI): A practical test for cross-cultural epidemiological studies of dementia. International Psychogeriatrics, 6(1), 45–62. 10.1017/S1041610294001602 [DOI] [PubMed] [Google Scholar]
  37. Verbrugge L, & Jette AM (1994). The disablement process. Social Science and Medicine, 38(1), 1–14. 10.1016/0277-9536(94)90294-1 [DOI] [PubMed] [Google Scholar]
  38. White L, Petrovitch H, Ross GW, Masaki KH, Abbott RD, Teng EL, Rodriguez BL, Blanchette PL, Havlik RJ, Wergowske G, Chiu D, Foley MS, Murdaugh C, & Curb D (1996). Prevalence of dementia in older Japanese-American men in Hawaii: The Honolulu-Asia aging study. Journal of the American Medical Association, 276(12), 955–960. 10.1001/jama.1996.0354012003303010.1001/jama.276.12.955 [DOI] [PubMed] [Google Scholar]
  39. Worth RM, & Kagan A (1970). Ascertainment of men of Japanese ancestry in Hawaii through World War II selective service registration. Journal of Chronic Diseases, 23(5), 389–397. 10.1016/0021-9681(70)90022-6 [DOI] [PubMed] [Google Scholar]
  40. Yaffe K, Lindquist K, Vittinghoff E, Barnes D, Simonsick EM, Newman A, Satterfield S, Rosano C, Rubin SM, Ayonayon HN, & Harris T (2010). The effect of maintaining cognition on risk of disability and death. Journal of the American Geriatric Society, 58(5), 889–894. 10.1111/j.1532-5415.2010.02818.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Yam A, & Marsiske M (2013). Cognitive longitudinal predictors of older adults’ self-reported IADL function. Journal of Aging and Health, 25(8 Suppl), 163S–185S. 10.1177/0898264313495560 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Yano K, Reed DM, & McGee DL (1984). Ten-year incidence of coronary heart disease in the Honolulu Heart Program. Relationship to biologic and lifestyle characteristics. American Journal of Epidemiology, 119(5), 653–666. 10.1093/oxfordjournals.aje.a113787 [DOI] [PubMed] [Google Scholar]
  43. Young DR, Masaki KH, & Curb JD (1995). Associations of physical activity with performance-based and self-reported physical functioning in older men: The Honolulu Heart Program. Journal of the American Geriatrics Society, 43(8), 845–854. 10.1111/j.1532-5415.1995.tb05525.x [DOI] [PubMed] [Google Scholar]

RESOURCES