Risk factors for decline in gait speed during walking while talking in older adults

Abstract

Background & aims:

Slow gait speed during Walking While Talking (walking while reciting alternate letters of the alphabet; WWT) is associated with an increased risk of developing dementia and falls. The aim of this study was to examine longitudinal changes in WWT-speed and to identify risk factors that may modify the rate of change in WWT-speed.

Methods:

A total of 431 older participants (55.7% female; M Age=76.8 ± 6.4 years; mean follow up 2.1 ± 1.8 years) enrolled in the Central Control of Mobility in Aging study were examined. WWT-speed (cm/s) was measured with a computerized walkway. The following baseline measures were examined as risk factors: demographics [age, sex, education], medical illnesses [hypertension, diabetes, cardiac arrhythmias, history of stroke, Parkinson’s disease, kidney disease, arthritis], cognitive functions [global cognition, executive function, processing speed], physical and sensory functions [unipedal stance time, gait speed during single task walking, visual acuity], psychological variables [depression, anxiety] and falls. Linear mixed effect models were used to examine 1) change in WWT-speed over time, and 2) risk factors associated with change in WWT-speed over time.

Results:

WWT-speed declined in an accelerating non-linear fashion over time after adjusting for baseline age, sex and education. The rate of decline in WWT-speed was modified by older age (b −0.16 95%CI −0.22, −0.09), poorer balance (b −1.73 95%CI −2.57, −0.90), and faster gait speed during single task walking (b −0.06 95%CI −0.08, −0.04).

Significance:

This study identified fixed and modifiable risk factors of faster decline in WWT-speed over time in community-residing older adults.

1. Introduction

Walking while talking (WWT), performing a challenging cognitive task while walking, is a dual-task that exposes early cognitive and motor impairments in aging [1]. Slow gait speed during WWT is associated with mild cognitive impairment and risk of incident dementia, over and above gait speed assessed without an additional task (referred to as single-task walking, STW) [2,3]. Slow WWT-speed also represents a poorer health outcome that in turn puts older adults at an increased risk for falls [4]. Hence, preventing accelerated slowing of WWT-speed offers potential means to prevent incidence of falls. To do this, there is a need to understand how WWT-speed changes over time in older age and the factors associated with faster changes in WWT-speed.

Only a few studies have examined WWT-speed longitudinally. In community-dwelling older adults (n = 85) gait variability during WWT increased over two years, without decline in WWT-speed [5]. By contrast, in those with mild Alzheimer’s disease (n = 25), WWT-speed declined over two years [6]. These studies are limited due to small samples and short follow-up periods [5,6]. Longitudinal studies from large population-based samples with longer follow-ups provide a better understanding of changes in WWT-speed in the general older population. Furthermore, risk factors for WWT-speed have only been examined at cross-section. In community-dwelling older adults without dementia (n = 186–1037), poorer cognition (i.e. executive function, attention, memory [7]), functional dependence [8] and fear of falling [9] were associated with slow WWT-speed. Other mobility-related measures (i.e. Timed Up and Go test time [8]) were associated with greater gait variability or gait cost (the percent decline in speed between STW and WWT), but not with WWT-speed. Longitudinal studies are needed to identify risk factors associated with faster decline in WWT-speed.

Therefore, the aims of this study were, in a community-dwelling sample of older adults without dementia, to (1) establish longitudinal changes in WWT-speed, and (2) determine the demographic, medical, cognitive, physical and psychological factors that modify the rate of change in WWT-speed. We hypothesized that WWT-speed would decline over time, and specific risk factors are associated with faster decline in WWT-speed.

2. Methods

2.1. Study participants

The Central Control of Mobility and Aging Study (CCMA) is a population-based longitudinal study of adults aged 65 and older, recruited from population lists of lower Westchester County, New York. Those with dementia (physician diagnosis or meeting criteria at consensus case conferences [10]), inability to ambulate independently, active medical conditions interfering with assessments, major loss of vision (acuity worse than 20/100) and/or hearing (worse than 40 dB HL at 2000 Hz) were excluded. Annual follow-ups were conducted from 2011 to 2018. Ethical clearance was obtained from the Institutional Review Board of the Albert Einstein College of Medicine. Written informed consent was obtained from all participants.

2.2. Single-task cognition

Single-task cognitive performance was assessed while participants recited alternate letters of the alphabet, starting with “A” for 10 s in standing. This task was chosen as it requires a range of cognitive functions (attention switching, working memory, response inhibition, processing speed, semantic memory of the alphabet) and therefore is challenging for older adults [11]. Correct and incorrect answers were recorded to create a corrected response rate (CRR, percentage of correct answers × the number of correct answers per second) [12].

2.3. Gait assessments

Gait was assessed at each visit using a 8.5 m GAITRite walkway [13], in a quiet well-lit room. For STW, participants completed one trial at comfortable walking speed. For WWT, participants completed one trial while reciting alternate letters of the alphabet starting from “A”. They were instructed to pay equal attention to both walking and the cognitive task to control for task prioritization [14]. Answers were recorded during WWT to calculate the CRR. Participants who used walking aids were only assessed if they were able to complete walks without the device. Gait speed (centimeters/seconds [cm/s]) was obtained from the GAITRite software. The cost of gait speed and cognitive performance during WWT were calculated as (single task performance – WWT performance)/single task performance× 100% [11].

2.4. Assessment of baseline risk factors

We selected the following factors based on prior cross-sectional associations with WWT-speed [8,15] and longitudinal associations with STW-speed [16].

2.5. Demographics & medical conditions

Age, sex and years of education were recorded with self-report. Body mass index (BMI) was calculated using participant’s height and weight. History of medical illnesses: (a) cardiovascular risk factors (hypertension, diabetes, cardiac arrhythmia) (b) neurological conditions (stroke, Parkinson’s disease), (c) musculoskeletal conditions (osteoarthritis, rheumatoid arthritis) and (d) chronic kidney disease, was recorded with self-report and, corroborated with informants or medical records when available.

2.6. Cognitive functions

Global cognition with the total score on the Repeatable Battery of the Assessment of Neuropsychological Status (RBANS) [17], executive function with Trail Making Test interference (TMT) B - TMT-A) [18] and digit span of the Wechsler Adults Intelligence Scale-III (WAIS-III) [19], processing speed with digit symbol substitution test of WAIS-III and TMT-A and episodic memory with total recall of Free and Cued Selective Reminding Test, were assessed.

2.7. Physical and sensory measures

Balance was assessed using unipedal stance time (seconds). Participants stood on their dominant leg (self-reported) for a maximum of 30 s. Based on a prior study of the association between unipedal stance time and falls risk [20], we used a cut-off score of ≤ 10 s to distinguish those with poor versus good balance. Maximum dominant hand (self-reported) grip strength (kilograms) was assessed with a Jamar Dynamometer. Visual acuity was measured with a Snellen’s chart (<20/80 = poor vision) [21].

2.8. Psychological factors

Depression was assessed with the 30-item Geriatric Depression Scale [22].

2.9. Falls

The number of falls in the past year was assessed with a standardized questionnaire at the baseline.

2.10. Data analysis

STATA (StataCorp LLC, College Station, TX, United States) version 16.1 was used in the analyses.

2.11. Aim 1: to examine longitudinal changes in WWT-speed over time

Linear mixed effect models were used to examine the associations between WWT-speed (dependent variable) and time (independent variable). Both linear and non-linear models were examined. As non-linear models had a better goodness of fit (when assessed with likelihood ratio test), a quadratic term of time, baseline age, sex and education were included as the fixed effects. Subject study ID was included as the random effect.

2.12. Aim 2: to determine the risk factors associated with change in WWT-speed

To identify the factors that modify the rate of change in WWT-speed, first, each factor was tested as an interaction term with time in separate models. The example shows the unadjusted model for effects of hypertension on WWT-speed over time:

$$WWTspeed=beta1\left(time\times time\right)+beta2hypertension+beta3\left(time\times hypertension\right)$$

If beta3 was significant, the interpretation is that individuals with hypertension at baseline may have a faster (or slower) rate of change in WWT-speed, compared to those without hypertension. The models were adjusted for baseline age, sex, and education.

Next, to build a final model that includes all factors that modify the rate of change in WWT-speed, those identified in separate models were entered into a single model in a stepwise fashion and, retained only if they remained significant. The factors examined were carefully chosen to represent multiple demographic, medical, cognitive and sensorimotor factors based on a priori knowledge [8,15,16,23]. We considered the presence or absence of these factors at baseline as we were interested using baseline values to predict the change in WWT-speed. Similar models were built to examine longitudinal changes in WWT-cost in speed and associated factors of change in WWT-cost in speed. In a sensitivity analysis, baseline WWT-speed also was examined as a modifier for change in WWT-speed over time.

We performed two secondary analyses; (1) for the factors that did not modify the rate of change in WWT-speed or WWT-cost in speed, we examined if risk factors have an overall effect on WWT-speed or cost. These models were adjusted for age, sex, education and any risk factor that modifies the rate of change in WWT-speed (2) although our primary outcome was WWT-speed (poor gait during WWT is a marker of adverse health outcomes), changes in cognitive performance during standing and WWT, cost in cognitive performance were also examined using models as described above.

3. Results

Of 588 CCMA participants, 431 participants with at least two followup data (who can contribute to change in WWT-speed) were included. Excluded participants had higher prevalence of hypertension, poorer global cognition, executive function and processing speed. The mean follow-up time was 2.1 ± 1.8 years (range 1–8 follow-ups). Table 1 summarizes participants’ characteristics at baseline. The mean age of participants was 76.8 ± 6.4 years and 55.7% (n = 240) were females. Supplementary Table 1 shows sample size and mean age of participants at different follow-ups.

Table 1.

Participants’ characteristics at baseline.

	n = 431
Age (years), mean, SD	76.8	6.4
Female, n, %	240	55.7
Education(years), mean, SD	14.7	2.9
Medical conditions
Hypertension, n, %	259	60.9
Diabetes, n, %	84	19.7
Cardiac arrhythmia, n, %	76	18.0
Stroke, n, %	24	5.6
Parkinson’s disease, n, %	2	0.5
Osteoarthritis, n, %	224	53.9
Rheumatoid arthritis, n, %	20	4.8
Kidney disease, n, %	35	8.2
BMI (kg/m2), mean, SD	29.20	6.9
Cognitive tests
Total RBANS Score (40–160), mean, SD	91.8	12.1
TMTA (time to complete), mean, SD	50.7	27.5
TMT interference (time to complete), mean, SD	86.8	76.5
Digit Symbol (0–133), mean, SD	52.6	15.0
Digit Span (0–16), mean, SD	10.2	2.6
Free recall (0–48), mean, SD	29.5	7.1
Sensorimotor functions
Unipedal stance ≤10 s, n, %	191	48.9
Maximum grip strength, mean, SD	26.4	9.3
Poor vision, n, %	12	2.8
Psychological factors
Geriatric Depression Score, mean, SD	4.5	3.8
Age-associated syndromes
Falls, n, %	80	18.8
Gait speed
Single task walking (cm/s), mean, SD	98.6	23.0
WWT (cm/s), mean, SD	69.7	24.3
Reciting alternate letters of the alphabet
CRR in standing, mean, SD	0.7	0.2
CRR in walking, mean, SD	0.8	0.4
Cognitive cost,%	−45.4	276.9

SD, standard deviation; BMI, body mass index, TMT, Trail Making Test; WWT, walking while talking; CRR, corrected response rate.

3.1. Longitudinal changes in WWT-speed over time

WWT-speed declined in a non-linear fashion over time (Fig. 1). The rates of change in WWT-speed at 1, 2, 3.2 and 5.3 years of follow-up, corresponding to the 25th, 50th, 75th and 95th percentiles of the time variable in the cohort (due to non-linear changes the rates of WWT- speed decline are different at each follow-up) are summarized in Table 2. Corresponding to change over time, Supplementary Fig. 1 shows decline in WWT-speed at different ages: slower decline until mid-70 s followed by a faster decline. WWT-speed decline was significant in both men and women, and in those < 75 and ≥ 75 years at baseline (Supplementary Table 2).

Fig. 1.

Decline in gait speed during Walking While Talking during the study follow-up.

Table 2.

Rate of decline in Walking While Talking (WWT)-speed at different time points (n = 431).

	Rate of decline in WWT-speed
	b	(95% CI)	p
Time= 1	0.60	(−0.12,1.32)	0.10
Time= 2	−0.51	(−0.94,0.08)	0.02
Time= 3	−1.83	(−2.30, −1.36)	<0.001
Time= 5.3	−4.03	(−5.19, −2,87)	<0.001

^aThe time points of 1, 2, 3.2 and 5.3 years of follow ups were selected corresponding to the 25th, 50th, 75th and 95th quartiles of the time variable in the cohort.

3.2. Risk factors associated with decline in WWT-speed over time

Table 3 summarizes the modifying effects of baseline risk factors on the time coefficient. In individual models, interaction terms of baseline age, cardiac arrhythmia, TMT-A, unipedal stance time, baseline STW-speed and depression with time were significant. When introduced into one model, only baseline age, unipedal stance time and baseline STW-speed remained significant (Fig. 2). Older baseline age was associated with a faster decline in WWT-speed, indicating an additional decline by 0.16 cm/s (95% CI-0.22, −0.09) per year with each additional year of baseline age (eg. for each 5-year increase in age, decline is 0.8 cm/s faster per year). Those with poorer balance showed faster decline in WWT-speed by 1.73 cm/s (95% −2.57, −0.90) per each year compared to people with good balance. Those with faster baseline STW-speed showed faster decline in WWT-speed by 0.06 cm/s (95% −0.08, −0.04) per year with each additional 1 cm/s of baseline STW-speed. In the sensitivity analysis, baseline WWT-speed did not modify decline in WWT-speed.

Table 3.

Factors that modified decline in WWT-speed over time (n = 431).

	b	95%CI	p-value
Demographic factors
Baseline age×time	−0.13	−0.19, −0.08	0.00
Gender×time	−0.14	−0.90, 0.62	0.72
Education×time	0.10	−0.04, 0.23	0.16
Medical conditions
Hypertension×time	−0.15	−0.91, 0.62	0.71
Diabetes×time	−0.68	−1.66, 0.31	0.18
Cardiac Arrhythmia×time	−1.34	−2.38, −0.29	0.01
Stroke×time	−1.32	−2.86, 0.23	0.10
Parkinson’s disease×time	−6.81	−13.74, 0.12	0.05
Osteoarthritis×time	−0.29	−1.05, 0.48	0.46
Rheumatoid arthritis×time	−0.92	−2.57, 0.73	0.27
Kidney disease×time	−0.33	−1.98, 1.31	0.69
BMI×time	−0.02	−0.08, 0.04	0.45
Cognitive functions b
Global cognition×time	−0.02	−0.05, 0.02	0.37
\TMT-A×time	−0.02	−0.04, −0.004	0.01
TMT interference×time	−0.01	−0.01, 0.00	0.06
Digit Symbol×time	0.02	−0.01, 0.04	0.19
Digit Span×time	−0.06	−0.21, 0.09	0.44
Free recall×time	0.02	−0.03, 0.07	0.44
Physical & sensory measures
Unipedal stance time×time	−1.60	−2.40, −0.81	<0.001
STW-speed×time	−0.03	−0.04, −0.01	0.002
Vision×time	−1.44	−3.63, 0.76	0.20
Psychological factors
Depression×time	−0.11	−0.22, −0.004	0.04
Age-associated syndromes
Falls×time	−0.07	−1.07, 0.93	0.89

^aAll models were adjusted for baseline age, sex and education appropriately. BMI, body mass index; TMT, Trail making Test. STW-speed, gait speed during single task walking.

^bFor all cognitive tests except for TMT- A and interference higher scores indicate better function.

Fig. 2.

Effect modification of gait speed during Walking While Talking by baseline age (a), unipedal stance time (balance) (b) and baseline gait speed during single task walking (c). Difference in change in gait speed during WWT are shown in those with highest and lowest quartiles of baseline age, baseline gait speed during single task walking and unipedal stance, respectively.

3.3. Longitudinal changes in WWT-cost in speed and associated risk factors

WWT-cost in speed changed in a non-linear fashion: decreased in early follow-ups (b −2.04, 95%CI −2.84, −1.25 at Time=1; b −1.32, 95%CI −1.79, −0.85 at Time=2) and increased in later follow-ups (b 0.97, 95%CI −0.31, 2.25 at Time 5.3; b 1.80, 95%CI 0.02, 3.58 at Time=6.5) (Supplementary Fig. 2a). Greater TMT interference was associated with greater WWT-cost, indicating a modification of cost by 0.01 (95% CI 0.004, 0.02).

3.4. Factors associated with overall slow WWT-speed and greater WWT- cost in speed over time

Chronic kidney disease (b −6.35, 95% CI −12.16, −0.55), global cognition (b 0.23, 95% CI 0.07, 0.38) and processing speed assessed with digit symbol substitution test (b 0.18, 95% CI 0.05, 0.31) were associated with slow WWT-speed. No risk factors were associated with overall greater WWT-cost in speed.

3.5. Change in cognitive performance

Single-task cognitive performance improved over time linearly (b 0.007, 95%CI 0.002, 0.012). WWT cognitive performance changed in a non-linear fashion. Decline was pronounced in later follow-ups (b 0.031, 95%CI 0.012, 0.049 at Time=1; b 0.009, 95% −0.002, 0.020 at Time=2; b= −0.016, 95%CI −0.028, −0.005 at Time=3.2; b= −0.061, 95% CI −0.091, −0.031 at Time=−5.3). Cognitive cost during WWT did not change significantly over time (b 2.711, 95% −2.034, 7.466).

When gait and cognitive performance during WWT were compared, participants showed better cognitive performance (Supplementary Fig. 2d).

4. Discussion

Slow WWT-speed is a simple low-cost marker of falls. We found that WWT-speed continuously declined in older age in a non-linear fashion, as seen in one of our prior studies [24]. Older age, poorer balance, and faster STW-speed were associated with faster decline in WWT-speed.

4.1. Risk factors of faster decline in WWT-speed over time

Decline in WWT-speed started in the earlier follow-ups but was not significant (see Table 2). Corresponding to this, in terms of participants’ age, there was a slower decline in mid-60 s (Supplementary Fig. 1), followed by a sharper decline after mid-70 s. This suggests that older adults retain sufficient cognitive and physical capacity until mid-70 s to perform complex walking conditions such as WWT. This agrees with the behavior in younger adults during WWT. Younger adults slow down during WWT, but the cost in gait speed is smaller compared to older adults due to their greater capacity [25]. People with poorer balance showed a faster decline in WWT-speed. Current evidence on the effect of balance on WWT-speed is cross-sectional and inconsistent. In community dwelling older adults, worse performance on the Berg Balance Scale was not associated with WWT-speed [8]. Whereas in the CCMA study, people with lower postural reserve had a smaller decrease in speed from STW to WWT, compared to those with higher reserve [26]. Direct comparisons between these studies are difficult due to different balance measures (Berg scale is a direct measure of balance whereas postural reserve was examined using clinical gait abnormalities- a proxy for sensorimotor and balance deficits). Longitudinally, people with poorer balance had faster WWT-speed decline. Potentially this is because those with poorer balance already had lower cognitive and sensorimotor resources and more impairments than those with good balance. This finding also highlights an interesting underlying mechanism of WWT test. Currently, WWT is identified as a test of interplay between gait and cognition, but our findings suggest that WWT could be a marker of balance as well. To confirm this, the associations between balance and gait during WWT should be examined in future studies. Unexpectedly, people with faster baseline STW-speed showed faster rate of decline in WWT-speed. People with faster STW-speed were potentially healthier than those with slow STW-speed and may have had a relatively larger WWT-speed range to decline over time and thereby showed a faster decline. Therefore, in this case having faster STW-speed may not necessarily a risk factor for WWT-speed decline.

Cardiac arrhythmia, higher TMT-A time and depression were also modifiers in individual models. These measures represent proxies for brain pathology [27,28], suggesting that altered brain structure and function is associated with faster decline in WWT-speed. Collectively, our findings highlight that improving balance and executive function, decreasing cardiovascular risk factor burden and preventing depression may offer means to reduce faster decline in WWT-speed and thereby associated adverse health outcomes such as falls. However, it should be noted that our study, by examining the effects of demographic, medical, sensorimotor factors at baseline, offer temporal relationships, not causal inferences.

4.2. Longitudinal changes in WWT-cost in speed over time

Unexpectedly, in earlier follow-ups WWT-gait cost in speed decreased over time. This is explained by the faster decline in STW-speed in earlier follow-ups when decline in WWT-speed was slower. This finding agrees with a relatively faster decline in STW-speed that was previously observed [24]. In later follow-ups WWT-gait cost increased over time. In line with prior cross-sectional evidence, poor executive function was associated with greater WWT-gait cost [12]. Compared to WWT-gait cost, WWT-cognitive cost reduced over time (not significant) as cognitive performance during standing improved with time (this may represent a practice effect due to repeated exposure, but the interval between assessments and exposure to single trail at a visit make it less likely). Additionally, we observed, despite our instructions on equal priority participants have prioritized performance on the cognitive task during WWT (Supplementary Fig. 2d).

4.3. Factors associated with overall slower WWT-speed over time

From a clinical viewpoint, risk factors associated with overall slower WWT-speed are less influential as they do not result in accelerated WWT-speed decline. But people with chronic kidney disease, poorer global cognition and slower processing speed appear to have slower overall WWT-speed. This is in line with prior cross-sectional associations between chronic kidney disease [29] and poorer cognition with slow WWT-speed [7].

4.4. Clinical implications

Decline in WWT-speed in aging cannot be attributed to one identifiable cause, rather is associated with multiple medical, cognitive and sensorimotor factors. The WWT task we examined is highly representative of everyday activity. Worsening gait performance during this task indicates it’s complexity in high functioning older adults and was associated with falls, frailty and disability [1,4]. Understanding the risk factors associated with faster WWT-speed decline is a first step to design and test interventions to prevent falls and disability. Interestingly, some underlying factors identified herein are modifiable. Multifactorial interventions incorporating balance training, reducing cardiovascular risk factor burden and depression, and maintaining cognition into older age may reduce decline in WWT-speed and thereby prevent associated adverse outcomes.

4.5. Strengths and limitations

This is one of the few longitudinal studies that examined decline in WWT-speed. We examined a large sample of community-dwelling older adults with a longer follow-up time. A range of risk factors were assessed using standardized procedures. These were based on the strength of previously associations, but not meant to be an exhaustive list and some covariates (i.e. history of falls) were self-reported. Participants who developed severe walking difficulty and could not complete a walk without assistive devices were excluded. Although theoretically this could raise concern, in our study this only corresponded to 0.9% of the overall sample, thereby reducing any bias in results. WWT is a test that approximates everyday activity, but the need to standardize the test to ensure uniformity across participants reduces its ecological validity. We used linear mixed effect models that provide unbiased estimates of longitudinal changes of outcomes in the case of attrition. However, we cannot rule out the possibility of models being not fully specified (capturing all the relationships between covariates and their covariance). Also, there is the possibility of unmeasured or residual confounding. We establish the temporal sequence of relationship of risk factors for WWT-decline. Causality should be tested in future interventions.

5. Conclusions

WWT-speed declined in an accelerating non-linear fashion. People with older age, poorer balance and faster STW-speed had a faster rate of decline in WWT-speed. Effects of modifiable factors to prevent WWT- speed decline need to be examined in future intervention studies.