Executive Function Predicts Decline in Mobility after a Fall: The MYHAT Study

Abstract

Background:

Evidence suggests that better cognitive functioning is associated with better mobility in older age. It is unknown whether older adults with better cognitive function are more resilient to mobility decline after a fall.

Methods:

Participants from the Monongahela Youghiogheny Healthy Aging Team (MYHAT) study were followed annually for up to 9 years for incident falls. We examined one-year (mean 1.0 year, SD 0.1) change in mobility pre- to post-fall using the Timed Up and Go (TUG) in relation to pre-fall cognition (executive function, attention, memory, and visuospatial function) among incident fallers (n = 598, mean age 79.1, SD = 7.0). Linear regression models tested the association of cognition with change in TUG. Interaction terms were tested to explore if age, sex, body mass index, physical activity, depressive symptoms, or visual acuity modified the associations of cognition and mobility among fallers. The association between cognition and one-year change in TUG was also tested in a comparison sample of non-fallers (n = 442, mean age 76.3, SD = 7.2).

Results:

Overall, mobility decline was greater in fallers compared to non-fallers. In fully-adjusted models, higher executive function, but not attention, memory, or visuospatial function, was associated with less decline in mobility among incident fallers. The effect was significantly stronger for those who were older, sedentary, and had lower body mass index. Higher scores in memory tests, but not in other domains, was associated with less mobility decline among non-fallers.

Conclusions:

Higher executive function may offer resilience to mobility decline after a fall, especially among older adults with other risk factors for mobility decline. Future studies should assess whether executive function may be a helpful risk index of fall-related physical functional decline in geriatric settings.

INTRODUCTION

It is estimated that more than one in every four older adults fall each year ¹. Falls can lead to a myriad of negative effects on health, function, and overall well-being, even in the absence of injury ². Declines in physical function are a common result of a fall ³, and the impetus for disability and loss of independence. Fall prevention efforts tend to focus on the avoidance of falls. However, given the high prevalence of falls among older adults, it is also of interest to understand whether certain characteristics predict fall-related detrimental effects on mobility.

There is growing evidence of associations between cognition and motor function supported by observational studies ^4,5, cognitive training interventions ⁶, and work demonstrating shared brain networks responsible for cognition and mobility ^7,8. Since not all domains are affected similarly by the aging process⁹, the associations between cognition and mobility may depend on the cognitive domain measured. Mobility has been shown to be more closely related to attention and executive function¹⁰, however, work also suggests that memory and other cognitive tasks also contribute to changes in mobility, specifically gait speed¹¹. Whether better cognition provides resilience to the impact of a fall on mobility, and to what extent different cognitive domains contribute, has yet to be explored.

In the present study, we sought to quantify the association between function in multiple cognitive domains before an incident fall and change in mobility during the same annual follow-up period when the fall occurred. We compared these results to a sample of non-fallers to account for age-related decline in mobility over a similar time period. Since the impact of a fall on mobility is likely determined by a combination of cognition and other locomotor risk factors, we also explored whether the association of pre-fall cognition with change in mobility depends on other pre-fall characteristics. These findings have important clinical implications for identifying older patients whose cognitive function along with other characteristics may make them more or less resilient to the effects of a fall on mobility, and for targeting older adults who may benefit from cognitive interventions.

METHODS

Participants

The sample was drawn from the Monongahela-Youghiogheny Healthy Aging Team (MYHAT) study, an ongoing population-based cohort study of mild cognitive impairment, based in a region of southwestern Pennsylvania¹². From 2006 to 2008, an age-stratified random sample of participants aged 65 years or older was recruited from publicly available voter registration lists. Recruitment criteria included being age 65 or older, living within the selected area, and not living in a long-term care institution. Participants were excluded if they were too ill to participate, had severe vision or hearing impairment, or were decisionally incapacitated. Of 2036 original participants, 54 were excluded from detailed assessment due to substantial cognitive impairment (age-education corrected Mini-Mental State Examination of less than 21 out of 30 ¹³). The remaining 1982 participants underwent the full baseline assessment, and were invited to participate in subsequent annual in-person assessments for a maximum of 9 years, or 10 total assessments, at the time of this report. The University of Pittsburgh Institutional Review Board approved all study procedures. All participants provided written informed consent.

Falls

At each MYHAT annual assessment, participants are asked to self-report if they have fallen in the preceding year, and if yes, how many times they had fallen. Details about the fall(s), such as when it occurred in the preceding year or its severity, were not assessed. Participants who reported a fall at baseline were considered “prevalent fallers” and were excluded from these analyses. We defined a participant as an “incident faller” if they reported a fall for the first time after the baseline assessment, and as a “non-faller” if they did not report a fall at baseline or any follow-up assessments.

Study sample selection

To select the sample of incident fallers from the initial 1982 participants, we excluded 531 (26.8%) prevalent fallers and 711 non-fallers. From the 740 remaining incident fallers, if the participant reported incident falls at multiple assessments, we used the only earliest reported incident fall. We excluded 66 who either reported having a stroke or a transient ischemic attack (TIA) (n=50) or did not answer this question (n=16) at the pre- or post-fall assessments¹⁴. We then excluded 5 who were missing all four cognitive predictors of interest (see below) at the pre-fall assessment, and 71 who were missing the Timed Up and Go test (see below) at either the pre- or post-fall assessment, resulting in an overall sample size of 598 incident fallers.

For non-fallers, we considered change in mobility between the first and second assessments. Of the 711 (35.9%) non-fallers, we excluded 182 who had no follow-up after baseline and 4 who had further follow-up but a completely missing second assessment. Of the remaining 525, we excluded 53 who reported stroke or TIA (n=51) or did not answer this question (n=2) at the first or second assessments. We excluded 2 missing all four cognitive predictors of interest at the pre-fall assessment, and 28 who were missing the Timed Up and Go test at either first or second assessment, resulting in an overall sample size of 442 non-fallers.

Mobility

Mobility was assessed using the Timed Up and Go (TUG) test ¹⁵. The TUG is a measure of balance and walking ability based on the time in seconds to stand up from a chair, walk 20 feet at usual pace (10 feet forward and 10 feet back), and return to the seated position, using an assistive device as needed. The main outcome variable for the present analyses was change in the time (seconds) to complete the TUG between the assessment when the first fall was reported (i.e. post-fall) and the preceding annual visit (i.e., pre-fall). For example, if a participant reported falling for the first time at the third follow-up assessment, we know that they fell between follow-up two and three, and therefore change in mobility was based on change between the second follow-up (i.e., pre-fall) and the third follow-up (i.e., post-fall) assessments. A positive value indicates slowing, i.e., decline in mobility, over time. For non-fallers, we used change in TUG between the first and second assessments.

Cognition

Pre-fall visit cognition was quantified using neuropsychological test composite scores in the domains of executive function (Trail Making Test B ¹⁶, Clock Drawing ¹⁷, Verbal Fluency Letters (P&S)¹⁸), attention (Trail Making Test A ¹⁶, Digit Span Forward ¹⁹), visuospatial skill (WAIS-III Block Design ²⁰), and memory (WMS-R Logical Memory (immediate and delayed recall)¹⁹, WMS-R Visual Reproduction (immediate and delayed recall)¹⁹, Fuld Object Memory Test with Semantic Interference²¹. Each test score was standardized using the baseline mean and standard deviation for the entire MYHAT baseline sample of 1982. For participants with at least one non-missing test, standardized test scores were averaged within each domain to form composite z-scores. We further standardized the composite z-scores using means and standard deviations of the current study sample of 598 incident fallers, and likewise for the sample of 442 non-fallers, so their regression coefficients can be interpreted in units of the sample standard deviation of those respective groups.

Other Covariates

In addition to age, we examined other characteristics that may influence age-related mobility decline and could act as potential effect modifiers of the association between cognition and change in mobility. These characteristics included sex, body mass index (BMI, kg/m²), being physically active through everyday activities such as walking, climbing stairs, housework, gardening, or childcare (self-reported moderate or more intense, 0=no, 1=yes), depressive symptoms (modified Center for Epidemiological Studies Depression Scale (mCES-D ^22,23), scored as integers from 0-20), and visual acuity (0=least impaired eye 20/40 or better, 1=least impaired eye worse than 20/40). We also included other characteristics that could confound the main associations in multivariable models: health status as indicated by number of prescription medications, self-reported subjective health (poor or fair vs. good vs. very good or excellent), and number of incident falls reported during the period that change in mobility was measured.

Statistical Methods

We first computed descriptive summary statistics for the demographic characteristics and other characteristics of the entire sample, for incident fallers (pre-fall assessment), and for non-fallers (first assessment), separately. Then, to examine the association of the cognitive predictors (executive function, attention, visuospatial skill, and memory) at the pre-fall assessment (or first assessment for non-fallers) with change in mobility, we fit three sets of linear regression models: (1) unadjusted model, (2) adjusting for each covariate separately (age, sex, BMI, physical activity, depressive symptoms, visual acuity, number of medications, subjective health, number of falls), and (3) adjusting for all covariates together. Age, BMI, depressive symptom score, number of medications, and reported number of falls were centered at their means for each group. Finally, we tested models including the interaction between each cognitive domain significantly associated with change in TUG among fallers and each pre-fall risk factor (i.e., age, sex, BMI, physical activity, depressive symptoms, and visual acuity), adjusting for main effects, to explore if the effect depended on other pre-fall characteristics. All statistical analyses were done using R version 3.4.0 ²⁴. Statistical significance was set at p < 0.05.

RESULTS

Participant characteristics are reported in Table 1. Incident fallers were on average significantly older, had higher BMI, reported more medications, worse subjective health, and had slightly lower attention scores, but higher memory scores, at the first (pre-fall) visit than non-fallers (Table 1). Fallers also had significantly slower TUG test at both pre- and post-fall visits, and significantly greater slowing between visits, than non-fallers. The average number of falls for fallers was 1.3 (SD=0.9) during the one-year follow-up time (mean 1.0 year, SD 0.1). Among the 598 incident fallers, 269 (45.5%) had slower performance on the TUG, with an average decline of 19.3%, 129 (21.6%) remained the same, and 200 (33.4%) had faster performance on the TUG, with an average improvement of 12.7%.

Table 1:

Pre-fall characteristics of the study sample (Mean (SD) unless otherwise noted.)

	Overall (n = 1040)	Fallers (n = 598)	Non-fallers (n = 442)	p-value
Demographics
Age	77.9 (7.2)	79.1 (7.0)	76.3 (7.2)	<0.001
Female N (%)	622 (59.8)	371 (62.0)	251 (56.8)	0.096
Education group N (%)				0.538
Less than high school	118 (11.3)	72 (12.0)	46 (10.4)
High school	458 (44.0)	267 (44.6)	191 (43.2)
More than high school	464 (44.6)	259 (43.3)	205 (46.4)
Covariates
BMI (kg/m2)	27.7 (5.4)	27.9 (5.3)	27.3 (5.4)	0.011
Physically active N (%)	710 (66.3)	411 (68.7)	299 (67.8)	0.787
Depressive symptoms (mCES-D)	0.6 (1.7)	0.7 (1.9)	0.4 (1.3)	0.092
Impaired visual acuity N (%)	261 (25.7)	145 (25.0)	116 (26.7)	0.562
Number of medications	4.1 (3.0)	4.5 (3.1)	3.6 (2.8)	<0.001
Subjective health N (%)				0.011
Poor/Fair	122 (11.7)	80 (13.3)	42 (9.5)
Good	480 (46.2)	288 (48.2)	192 (43.4)
Very good/Excellent	438 (42.1)	230 (38.5)	208 (47.1)
Cognitive predictors
Attention z-score	0.1 (0.8)	0.0 (0.8)	0.1 (0.7)	0.033
Executive function z-score	0.1 (0.7)	0.1 (0.8)	0.1 (0.7)	0.364
Memory z-score	0.2 (0.8)	0.2 (0.9)	0.1 (0.8)	0.044
Visuospatial skill z-score	0.1 (1.0)	0.1 (1.0)	0.1 (1.0)	0.353
Timed Up and Go
Pre-fall or first assessment (s)	12.5 (3.3)	12.8 (3.9)	12.0 (3.1)	<0.001
Post-fall or second assessment (s)	12.7 (3.6)	13.3 (3.9)	12.0 (3.1)	<0.001
Change in TUG	0.3 (2.3)	0.5 (2.5)	0.0 (2.1)	0.031

Notes: Percentages exclude participants with missing data. P-values compare fallers and non-faller distributions using Wilcoxon rank sum test for quantitative variables (age, BMI, depressive symptoms, number of medications, cognitive predictors) and Fisher’s exact test for categorical variables.

Missing data for fallers: 9 BMI, 18 visual acuity, 2 number medications, 3 memory z-score, 70 visuospatial skill z-score

Missing data for non-fallers: 7 BMI, 1 physical activity, 7 visual acuity, 1 memory z-score, 18 visuospatial skill z-score

BMI: body mass index, mCES-D: modified Center for Epidemiological Studies Depression Scale, TUG: Timed Up and Go

For incident fallers, executive function (β_exec = −0.43, SE = 0.10, p < .001), memory (β_mem = −0.25, SE = 0.10, p = .011), and visuospatial skill (β_visuo = −0.43, SE = 0.10, p = .029) were significantly associated with change in TUG in unadjusted models (Figure 1). Executive function was the only cognitive domain significantly associated with change in TUG in fully-adjusted model; one standard deviation higher pre-fall executive function scores was associated with 0.31 seconds less change in TUG (i.e. less gait slowing; β_exec = −0.31, SE = 0.12, p =.007). Age was the only pre-fall risk factor that substantially attenuated the association of executive function with TUG (Figure 1). The interaction models revealed that the relationship between higher executive function score and decreased post-fall gait slowing was significantly stronger for those who were older (β_exec×age = −0.05, SE = 0.02, p < .001), sedentary (β_exec×PA = 0.54, SE = 0.20, p = .008), and had lower BMI (β_exec×BMI = 0.06, SE = 0.02, p = .003) (Figure 2).

Figure 1. Associations of pre-fall or first assessment cognitive scores with change in TUG in incident fallers and non-fallers.

Plots show estimated coefficients and 95% confidence intervals for the effect of pre-fall (for incident fallers) or first assessment (for non-fallers) cognitive scores on change in time to complete the Timed Up and Go (TUG) as measured at the following annual assessment. More negative coefficients indicate that higher cognitive scores associate with less change in TUG (i.e. less gait slowing) at follow up.

Figure 2. Significant interactions between pre-fall executive function and pre-fall risk factors for change in TUG outcome for incident fallers.

The association between pre-fall executive function and change in time to complete the Timed Up and Go (TUG) significantly differed depending on the level of body mass index (BMI, left panel), age (middle panel) or physical activity (PA, right panel).

For non-fallers, memory was the only domain significantly associated with change in TUG in unadjusted models (β_mem −0.21, SE 0.10, p = .032) and fully-adjusted models β_mem −0.25, SE 0.12, p = .036) (Figure 1).

Supplementary Tables 1-4 show the unadjusted, adjusted for each covariate separately, and fully adjusted model statistics for executive function, memory, visuospatial skill, and attention for fallers and non-fallers (Tables A and B, respectively).

DISCUSSION

In this population-based sample of 598 older adults who reported a first-time fall, we examined whether cognitive functioning before a fall predicted change in mobility, and further explored whether or not the effect depended on other pre-fall locomotor risk factors. Our results suggest that higher executive functioning, but not other cognitive domains, predicts less mobility decline, and that this effect is fall-related since this association was not found among non-fallers. Although this effect is robust to and independent of other known locomotor risk factors, it appears stronger for vulnerable subgroups, including those who were older, sedentary, and had lower BMI. These findings support the link between cognition, particularly executive function, and mobility previously reported in observational studies^4,5, but now in a sample of fallers.

The impact of a fall on gait slowing was quite remarkable in this cohort. The average change in time to complete the TUG was significantly greater among fallers compared to non-fallers. Further, the gait slowing among fallers was also larger than what was previously observed in the full MYHAT cohort (0.5 and 0.18 seconds, respectively) ²⁵, or in other cohorts of similarly aged older adults ²⁶. This is consistent with previous reports that falls are predictive of physical function decline and other mobility impairments^2,3, regardless of the severity or cause for the fall.

The effect size of the association between executive function and annual decline in mobility among incident fallers is noteworthy. In the present study, we found that one standard deviation of executive function score was significantly associated with 0.31 seconds less annual decline in the TUG in the fully-adjusted model. Albeit seemingly small, this is similar to the decline in TUG performance in the MYHAT sample overall²⁵. In addition, other common locomotor risk factors were not strongly associated with gait slowing independently of executive function; highlighting the potential importance of executive function. One could speculate that promoting executive function could possibly reduce the annual mobility decline observed among fallers and make it more similar to that observed in the general community of older adults, and that examining executive functioning in older adults could help predict the impact that a future fall would have on physical function.

We did not find that attention, visuospatial function, or memory were associated with change in mobility in our sample of incident fallers. Associations between these domains and physical function are less consistent than the domain of executive function⁴. The association between attention and physical function outcomes, for example, has primarily been found when attention is measured with a divided attention task ⁴. Of note, there were more missing data for visuospatial function, which may have reduced the power to detect smaller associations. Executive function more closely aligns with performance on the TUG test compared to cognitive processes underlying attention, visuospatial function, and memory. We did find, however, that better memory performance at the first assessment predicted less decline in gait speed over the one-year follow-up period among non-fallers. Memory has been shown to be associated with gait speed decline in the Health ABC Cohort of well-functioning older adults.¹¹ However, these findings suggest that memory may not provide resilience to decline among older adults who have fallen in the past year.

Recent work from neuroimaging studies provides possible explanations as to why pre-fall executive function, but not other cognitive domains, predicts the extent of decline in mobility after a fall. A consistent and complementary finding that emerges across different imaging modalities is that networks important for executive function, including basal ganglia, cerebellum, and frontal and parietal cortices are involved in mobility control^27,28. Moreover, reduced volume in multiple regions of gray and white matter along with increased white matter hyperintensity burden are associated with poor mobility outcomes, particularly gait, in aging ²⁹. The association of frontal and parietal cortical regions aligns with the behavioral finding of cognitive processes, particularly executive function, being associated with mobility outcomes among older adults. Therefore, executive functioning may be a proxy for underlying integrity of selected brain networks important for mobility.

Better executive functioning may be particularly important for counteracting the negative impact of a fall on mobility among the oldest-old, those who are sedentary, and those who have lower BMI. Interaction models revealed that higher executive function scores predicted slower decline in mobility after a fall to a greater extent among participants with these characteristics. The oldest-old, those who are sedentary, and those with lower BMI are more likely to be in the spiral of decline associated with a reduced physiological reserves, including the brain, that leads to decline in mobility and increased risk for fall, and then circling back to decline in physiological reserves³⁰. Our findings suggest that mobility resilience after a fall may be partially explained by preserved brain integrity, as indicated by higher executive function, even among those with a physically frail phenotype.

The results of this study should be interpreted within the context of the following strengths and limitations. We were able to measure change in mobility from when the fall was reported relative to the previous annual visit; allowing examination of the effect of cognition on physical function in the context of a new fall(s) in the past year. Details about the fall, including when it occurred between the assessments and whether an injury or fracture occurred, was not measured, limiting inferences about the extent that better cognition contributes to less decline in, or faster recovery of, mobility directly related to a fall. We did adjust for number of self-reported falls in the fully-adjusted models, but this is likely underestimated. Body mass index was based on self-reported weight, which may have limited accuracy. Our sample is derived from a well-characterized population-based cohort of older adults, which is more representative of the population than clinical samples of fallers, and was relatively large despite considering only participants who reported an incident fall. Since population-based studies of older adults are more heterogeneous and allow for subgroups to be identified, interaction models allowed for a more nuanced understanding of how pre-fall cognition relates to change in mobility among fallers. Lastly, cognition was measured with a battery of tests that allowed the differential influence of four separate cognitive domain composites to be examined in relation to change in mobility, which has clinical implications for more personalized prevention or treatment recommendations.

This study has potential implications for geriatric clinical assessment in identifying those more or less vulnerable to the negative effects of a fall on mobility. Falls in late adulthood may trigger a harmful cycle leading to decline in physical function, and greater fall risk, with loss of independence and poor quality of life. Tests of executive function, if administered routinely in geriatric settings, can provide useful information on the likelihood of mobility decline after a fall, and could become part of a risk index of fall-related mobility decline. Our results also have implications in the context of novel intervention strategies to reduce mobility decline. Initial evidence from clinical trials in older adults suggests that improved executive function may be an important protective factor against age-related declines in mobility^31,32. In sum, future studies should assess the relevance of measuring executive function in geriatric settings, both as an index of risk of future fall-related mobility decline, and as a potential target of interventions to minimize or reduce fall-related mobility decline.