Cognition Mediates the Relationship Between Sensory Function and Gait Speed in Older Adults: Evidence from the English Longitudinal Study of Ageing

Abstract

Background/Objective:

Gait speed is an important indicator for assessing overall health status. Previous studies have reported the important role of sensory function in gait speed; however, the underlying mechanism is still unclear. This study aimed to examine whether cognition mediates the association of sensory function with gait speed among English older adults.

Methods:

Gait speed was assessed by “timed walking test”. Hearing was measured by using a hearing screening device. Vision was self-reported. Cognition was assessed by questionnaire. Baron and Kenny’s causal steps method and Sobel test were used to examine the mediating effect.

Results:

Among 4,197 participants aged 60 years and older, 13.5% had poor hearing and 12.6% had poor vision, 2.6% had both poor hearing and poor vision. Multiple linear regression models suggested that poor hearing (β= – 1.905, p < 0.001), poor vision (β= – 1.309, p = 0.004), and poor dual sensory function (β= – 2.442, p = 0.013) was associated with worse cognition. Cognition was correlated with gait speed (β= 0.004, p < 0.001). Poor hearing (β= – 0.072, p < 0.001), poor vision (β= – 0.031, p = 0.029), and poor dual sensory function (β= – 0.081, p = 0.011) was associated with slower gait speed. After introducing cognition into the models, regression coefficients between sensory function and gait speed decreased (β= – 0.066, p < 0.001 for hearing; β= – 0.027, p = 0.054 for vision; β= – 0.073, p = 0.020 for combined hearing and vision). Sobel test identified the significant mediating effect of cognition on the association between sensory function and gait speed.

Conclusion:

Cognition partially mediates the association between sensory function and gait speed. Efforts to maintain mobility performance in older adults should consider protecting both sensory function and cognition.

Keywords

Cognition gait speed hearing mediating effect vision

INTRODUCTION

Walking speed, also termed gait speed, is recognized as a practical and informative indicator for assessing overall health status [1]. Decline in gait speed is common in aging and has the potential to predict adverse outcomes, including but not limited to cardiovascular disease, falls, disability, and mortality [2 –5]. Walking is a complex functional activity, relating to the musculoskeletal condition, sensory and perceptual function, endurance and habitual activity level, cognitive status, mental health, as well as the characteristics of the environment in which one walks [1].

Evidence is accumulating that age-related sensory changes such as hearing loss or poor vision contribute to decline in cognition and gait speed [6 –10]. Meanwhile, several cognitive processes including executive function and episodic memory have been consistently linked to gait performance and risk of falls in older adults [11 –14]. The mechanisms underlying these associations are complicated. Sensory systems influence higher order neurological processes and cognition, which are required for planning movements, divided attention and responding to changes within the environment [15]. There is a hypothesis, suggesting that under the conditions of sensory impairment, greater cognitive resources are dedicated to sensory information processing at the expense of other cognitive processes such as attention which are critical for maintaining postural control and balance, and thus the gait speed is declined [16 –18]. Another hypothesis indicates that long term lack of neural stimulation of the cerebral cortex by deficient hearing and visual infomation cause the decline in executive function and memory [19], and in turn changes the movement performance and the respond to environment [15]. However, few observational studies have examined the role of cognition in mediating sensory impairment and gait speed. And it is still not clear to what extent cognition mediates this association.

Here we conducted a cross-sectional study by using a large, population-based sample derived from English Longitudinal Study of Ageing (ELSA 2014-2015) to assess whether cognitive function may mediate the associations of hearing and vision with gait speed in older adults.

METHODS

Study sample

In current study, we used data from English Longitudinal Study of Ageing (ELSA), Wave 7 (2014-2015). ELSA is a prospective and nationally representative sample of adults aged 50 or older in England. The baseline (Wave 1) survey was conducted in 2002-2003, and a total of 12,099 people have taken this survey. The sample was followed up every 2 years with health examinations taking place every 4 years. Details about the sampling procedure, and study methodology can be found in the original article [20]. Ethical approval was obtained from the National Research Ethics Committee and participants have given informed consent.

At the end of Wave 7 in 2015, data were collected on 9,666 participants aged 50 or over. We selected the older adults aged 60 years and older in this study (n = 7,409). Those who had related diseases and conditions including cancer (n = 485), Parkinson’s disease (n = 67), dementia (n = 132), Alzheimer’s disease (n = 25), stroke (n = 361), fracture (n = 48), congestive heart failure (n = 44), blind (n = 31), and ear infection or cochlear implant (n = 165) were excluded. Meanwhile, 1,854 people who shared the same household with primary respondents were also excluded so that all the observations in the sample were independent. After filtering, 4,197 individuals were included in this study.

Gait speed

In ELSA, the walking test involved timing how long it took to walk a distance of 2.44 meters by using a stopwatch. Respondents began with both feet together at the beginning of the course. The interviewer started timing as soon as the respondents placed either foot down on the floor across the start line. They were asked to walk to the other end of the course at their usual speed, just as if they were walking down the street to the shops, and to walk all the way past the other end of the tape before stopping. Timing was stopped when either foot was placed on the floor across the finish line. Respondents were then asked to repeat the test by lining up their feet and walking back along the course, all the way past the other end. Gait speed for each test was calculated as the distance in meters (2.44) divided by the time in seconds. In this study, we used the mean gait speed of the two tests and handled it as a continuous variable [21].

Objective hearing acuity test

A hearing test was conducted on participants by using a hearing screening device (HearCheck Screener, Siemens, Germany). Participants were asked to remove their hearing aid. Three pure high frequency (3 kHz: 75, 55, and 35 dB HL) tones and three mid-frequency (1 kHz: 55, 35, and 20 dB HL) tones were tested for each ear. The hearing acuity was classified as normal hearing (heard all 6 tones), moderate hearing difficulty (heard 3– 5 tones), and severe hearing loss (heard 0– 2 tones) in ELSA. In this study, the hearing test results of both ears were combined. Participants who had normal hearing in both ears were divided into normal hearing group. Participants who had severe hearing loss in either of two ears were divided into poor hearing group. The rest participants were divided into moderate hearing group [22].

Self-reported vision

Participants were asked to rate their eyesight using the question: ‘Is your eyesight (using glasses or corrective lens as usual) excellent, very good, good, fair, poor, or registered blind?’. Blind was excluded [6, 23]. Answer of ‘excellent’ or ‘very good’ was defined as optimal vision. Answer of ‘fair’ or ‘poor’ was defined as poor vision. Thus vision was classified as optimal, good, and poor vision [24].

Combined hearing and vision function

Participants with both normal hearing and optimal vision, as defined above, were classified as having good dual sensory function. Participants with both poor hearing and poor vision were classified as having poor dual sensory function. The rest participants were classified as having moderate dual sensory function.

Cognition

Cognition included 3 domains, i.e., memory (immediate and delayed), verbal fluency, and time orientation. Memory performance was assessed using the 10-word list-learning test from the Health and Retirement Study. Participants were provided with 10 words and asked to recall them immediately (immediate recall) and again five minutes later (delayed recall). Memory score was the number of correct words, ranging from 0 to 20. The immediate recall was the ability for participants to learn or store new information, whereas delayed recall show the ability to recall the information after a period of distraction from the information. Verbal fluency was assessed by asking the participant name as many animals as possible in one minute. The overall score in this sample ranged from 0 to 60. It was a test for executive functioning and language ability [25], evaluating self-initiated activity, organization, and abstraction/mental flexibility [26]. Time orientation was assessed by standard questions on the date (day, month, and year) and the day of the week in this study, with the overall score ranging from 0 to 4. As a complex cognitive task, time orientation was an assessment of memory and executive function for human brain [27, 28]. Global cognition was the sum of scores for the 3 domains, and higher score indicates better cognitive performance.

Other variables

Covariates in this study were age, sex, educational level, cigarette smoking, alcohol drinking, physical activity, depressive symptoms, hearing aid, heart attack, hypertension, diabetes, arthritis, and osteoporosis. Educational level was categorized as high, intermediate, and low level [29]. Cigarette smoking was categorized as never, former, and current. Alcohol drinking was grouped as daily, weekly/monthly, and rarely/never. Physical activity was assessed using the physical activity questionnaire, and was further classified into three groups as low active, moderate active, and high active [30 –32]. Depressive symptoms were measured using the eight-item Centre for Epidemiological Studies Depression Scale 8 (CES-D8) [33]. Participants were asked to whether ever wears a hearing aid. Heart attack, hypertension, diabetes, arthritis, and osteoporosis were dichotomized as “no” or “yes” based on answers to the question of “Has a doctor ever diagnosed you with ‘this problem’?”

Statistical analysis

Participants’ characteristics were described according to sensory function groups and comparison across groups was conducted by using one-way ANOVA for quantitative variable, Kruskal-Wallis test for ordinal variable and chi-square test for qualitative variable. Baron and Kenny’s causal steps method [34] was used to explore the relationships between sensory function, cognition and gait speed. In this method, X indicates independent variable (sensory function); Y indicates dependent variable (gait speed); M indicates mediator (cognition). Four multiple linear regression models were fitted to assess if variables meet the following conditions. Firstly, X is associated with the Y (Fig. 1, Path c). Secondly, X is associated with the M (Fig. 1, Path a). Thirdly, M is associated with the Y after adjusting for X (Fig. 1, Path b). Finally, X significantly decreases its effect on the Y when M is included in the models as a covariate (Fig. 1, Path c’), if so, partial mediation is considered to have occurred. In addition, we used Sobel test [35] to examine the mediating effect of cognition on the association of sensory function with gait speed. This test decomposes the total effect of variable into direct effect and indirect effect and calculates the percentage of the main association explained by the mediator.

Fig.1

Results of mediation analysis for hearing, cognition, and gait speed (A for moderate hearing versus normal hearing; B for poor hearing versus normal hearing). X, independent variable (cause); Y, dependent variable (outcome); M, mediator. Path a: the relationship between X and M. Path b: the relationship between M and Y, with X included in the model. Path c: the relationship between X and Y. Path c’: the relationship between X and Y, with M included in the model. β, regression coefficient. Models were basically adjusted for the potential confounders: age, sex, educational level, cigarette smoking, alcohol drinking, physical activity, hearing aid, vision, depressive symptoms, heart attack, hypertension, diabetes, arthritis, and osteoporosis. ^*p < 0.05; ^**p < 0.01; ^***p < 0.001.

All models were basically adjusted for the potential confounders: age, sex, educational level, cigarette smoking, alcohol drinking, physical activity, hearing aid, depressive symptoms, heart attack, hypertension, diabetes, arthritis, and osteoporosis. Hearing and vision, as the main independent variables were adjusted mutually in the separate analysis. Data were analyzed using Stata version 15.0 (Stata Corp LP, College Station, TX). p values were two-sided with p < 0.05 considered statistically significant.

RESULTS

A total of 4,197 individuals aged 60 years or over were included this study. Descriptive statistics for participants’ characteristics grouped by sensory function were listed in Table 1. Overall, 13.5% of participants had poor hearing and 12.6% of participants reported poor vision. Univariate analyses showed that both cognition and gait speed were significantly different across the sensory function groups (p < 0.001). The participants with poor hearing or poor vision had the lowest level of cognition and slowest gait speed.

Table 1

Characteristics of participants at Wave7 in ELSA (n = 4,197)

Characteristic	Hearing				Vision
	Normal	Moderate	Poor	p	Optimal	Good	Poor	p
	n = 1,543	n = 1,809	n = 524		n = 2,036	n = 1,630	n = 527
Age, y, (mean±SD)	67.64±5.91	72.42±7.96	78.44±9.64	<0.001	70.21±7.76	71.84±8.56	74.75±9.75	<0.001
Female, %	48.8	44.8	45.8	0.067	41.4	48.8	52.8	<0.001
Educational level, % -Low	16.8	28.2	42.4	<0.001	19.2	28.4	46.6	<0.001
-Intermediate	34.8	35.2	30.0		34.2	36.2	27.0
-high	48.4	36.6	27.5		46.6	35.4	26.4
Cigarette smoking, % -Never	33.3	35.2	35.5	0.573	36.2	33.4	29.5	<0.001
-Former	56.5	54.5	55.9		54.6	56.7	55.4
-Current	10.2	10.3	8.6		9.2	9.9	15.1
Alcohol drinking, % -Rarely	17.5	23.6	30.0	<0.001	17.5	23.7	35.4	<0.001
-Weekly	59.5	54.2	50.5		58.0	56.2	45.9
-Daily	23.0	22.2	19.5		24.5	20.1	18.7
Physical activity, % -Low	21.3	32.9	49.2	<0.001	23.6	33.1	57.5	<0.001
-Moderate	54.4	49.6	38.2		52.3	50.4	32.8
-High	24.3	17.4	12.6		24.1	16.5	9.7
Hearing aid, %	1.0	13.8	57.8	<0.001	13.7	15.9	20	0.002
Depressive symptoms,	2.81±1.18	2.99±1.31	3.24±1.48	<0.001	2.74±1.13	3.05±1.33	3.54±1.57	<0.001
(mean±SD)
Heart attack, %	3.8	5.8	8.8	<0.001	4.4	5.8	8.6	<0.001
Hypertension, %	35.7	45.1	50.4	<0.001	38.5	44.2	49.4	<0.001
Diabetes, %	10.5	13.7	17.7	<0.001	11.0	13.0	21.5	<0.001
Osteoporosis, %	6.4	8.2	11.6	<0.001	6.3	9.0	11.2	<0.001
Arthritis, %	37.5	43.0	51.1	<0.001	35.6	43.2	54.6	<0.001
Cognition, (mean±SD)	37.62±8.69	33.97±9.22	29.65±9.72	<0.001	36.38±9.38	33.71±9.50	31.00±10.07	<0.001
Gait speed, m/s, (mean±SD)	0.96±0.26	0.86±0.27	0.75±0.27	<0.001	0.93±0.26	0.87±0.28	0.75±0.30	<0.001

Bold for p≤0.05 in multi-sample comparison; One-way ANOVA for quantitative variable; Kruskal-Wallis test for ordinal variable; Chi-square test for qualitative variable.

Figures 1 and 2 displayed the results of multiple linear regression models for hearing and vision based on Baron and Kenny’s causal steps method. For Path a, worse hearing or vision was associated with lower level of cognition (β= – 0.960 95% CI: – 1.561∼ – 0.360, p = 0.002 for moderate hearing and β= – 1.905 95% CI: – 2.962∼ – 0.848, p < 0.001 for poor hearing; β= – 0.992 95% CI: – 1.563∼ – 0.421, p = 0.001 for good vision; and β= – 1.309 95% CI: – 2.211∼ – 0.407, p = 0.004 for poor vision). For Path b, after adjusting for all the covariates as well as hearing and vision, the cognition was positively correlated with gait speed (β= 0.004 95% CI: 0.003∼ 0.005, p < 0.001). For Path c, worse hearing or vision was associated with slower gait speed (β= – 0.034 95% CI: – 0.052∼ – 0.015, p < 0.001 for moderate hearing and β= – 0.072 95% CI: – 0.105∼ – 0.040, p < 0.001 for poor hearing; β= – 0.016 95% CI: – 0.033∼ 0.001, p = 0.070 for good vision; and β= – 0.031 95% CI: – 0.059∼ – 0.003, p = 0.029 for poor vision). After introducing cognition into the models, regression coefficients between sensory function and gait speed (Path c’) decreased (β= – 0.030 95% CI: – 0.049∼ – 0.012, p = 0.001 for moderate hearing and β= – 0.066 95% CI: – 0.099∼ – 0.034, p < 0.001 for poor hearing; β= – 0.013 95% CI: – 0.030∼ 0.005, p = 0.148 for good vision; and β= – 0.027 95% CI: – 0.055∼ 0.001, p = 0.054 for poor vision), indicating the potential mediating effect of cognition on the association between sensory function and gait speed.

Fig.2

Results of mediation analysis for vision, cognition, and gait speed (A for good vision versus optimal vision; B for poor vision versus optimal vision). X, independent variable (cause); Y, dependent variable (outcome); M, mediator. Path a: the relationship between X and M. Path b: the relationship between M and Y, with X included in the model. Path c: the relationship between X and Y. Path c’: the relationship between X and Y, with M included in the model. β, regression coefficient. All models were basically adjusted for the potential confounders: age, sex, educational level, cigarette smoking, alcohol drinking, physical activity, hearing aid, hearing, depressive symptoms, heart attack, hypertension, diabetes, arthritis, and osteoporosis. ^*p < 0.05; ^**p < 0.01; ^***p < 0.001.

Figure 3 showed the results of multiple linear regression models for combined hearing and vision. 2.6% of participants reported exposure to both poor hearing and poor vision, and 21.9% of participants had good dual sensory function. Compared with participants with good dual sensory function, participants with poor dual sensory function had lower level of cognition (Path a: β= – 2.442 95% CI: – 4.366∼ – 0.518, p = 0.013) and slower gait speed (Path c: β= – 0.081 95% CI: – 0.143∼ – 0.019, p = 0.011). After introducing cognition into the models, regression coefficients between dual sensory function and gait speed decreased (Path c’: β= – 0.073 95% CI: – 0.135∼ – 0.012, p = 0.020).

Fig.3

Results of mediation analysis for combined hearing and vision, cognition, and gait speed (A for moderate dual sensory function versus good dual sensory function; B for poor dual sensory function versus good dual sensory function). X, independent variable (cause); Y, dependent variable (outcome); M, mediator. Path a: the relationship between X and M. Path b: the relationship between M and Y, with X included in the model. Path c: the relationship between X and Y. Path c’: the relationship between X and Y, with M included in the model. β, regression coefficient. All models were basically adjusted for the potential confounders: age, sex, educational level, cigarette smoking, alcohol drinking, physical activity, hearing aid, depressive symptoms, heart attack, hypertension, diabetes, arthritis, and osteoporosis. ^*p < 0.05; ^**p < 0.01; ^***p < 0.001.

Sobel test was further conducted to assess the mediating effect. In Table 2, we reported unadjusted and all adjusted results of Sobel test. In full model, the total and direct effects of worse hearing on gait speed were – 0.036 (95% CI: – 0.050∼ – 0.022, p < 0.001) and – 0.032 (95% CI: – 0.046∼ – 0.018, p < 0.001); the indirect effect of cognition was – 0.003 (95% CI: – 0.005∼ – 0.001, p = 0.002), with 9.25% of the total effect being mediated. The total and direct effects of worse vision on gait speed were – 0.016 (95% CI: – 0.028∼ – 0.004, p = 0.012) and – 0.013 (95% CI: – 0.025∼ – 0.001, p = 0.034); the indirect effect of cognition was – 0.003 (95% CI: – 0.005∼ – 0.001, p = 0.004), with 16.32% of the total effect being mediated. The total and direct effects of worse dual sensory function on gait speed were – 0.035 (95% CI: – 0.055∼ – 0.015, p < 0.001) and – 0.030 (95% CI: – 0.050∼ – 0.010, p = 0.002); the indirect effect of cognition was – 0.005 (95% CI: – 0.007∼ – 0.003, p < 0.001), with 14.46% of the total effect being mediated.

Table 2

Sobel test of the mediating effects of cognition for the association of sensory function with gait speed

Effect	Association with gait speed
	Model 1				Model 2
	β (95% CI)	Sobel-Z	p	Mediation (%)	β (95% CI)	Sobel-Z	p	Mediation (%)
Hearing
Total	– 0.105 (– 0.119, – 0.091)	– 15.562	<0.001		– 0.036 (– 0.050, – 0.022)	– 4.773	<0.001
Direct	– 0.072 (– 0.086, – 0.058)	– 10.805	<0.001		– 0.032 (– 0.046, – 0.018)	– 4.360	<0.001
Indirect	– 0.033 (– 0.039, – 0.027)	– 12.289	<0.001	31.70	– 0.003 (– 0.005, – 0.001)	– 3.084	0.002	9.25
Vision
Total	– 0.087 (– 0.101, – 0.073)	– 13.000	<0.001		– 0.016 (– 0.028, – 0.004)	– 2.516	0.012
Direct	– 0.062 (– 0.074, – 0.050)	– 9.691	<0.001		– 0.013 (– 0.025, – 0.001)	– 2.120	0.034
Indirect	– 0.025 (– 0.031, – 0.019)	– 9.830	<0.001	28.57	– 0.003 (– 0.005, – 0.001)	– 2.902	0.004	16.32
Combined hearing and vision
Total	– 0.135 (– 0.155, – 0.115)	– 13.470	<0.001		– 0.035 (– 0.055, – 0.015)	– 3.634	<0.001
Direct	– 0.091 (– 0.111, – 0.071)	– 9.367	<0.001		– 0.030 (– 0.050, – 0.010)	– 3.128	0.002
Indirect	– 0.044 (– 0.052, – 0.036)	– 11.264	<0.001	32.47	– 0.005 (– 0.007, – 0.003)	– 3.500	<0.001	14.46

Model 1 unadjusted covariate. Model 2 adjusted for age, sex, educational level, cigarette smoking, alcohol drinking, physical activity, hearing aid, depressive symptoms, heart attack, hypertension, diabetes, arthritis, and osteoporosis. Hearing and vision, as the main independent variables were adjusted mutually in the separate analysis.

DISCUSSION

In this population of community-dwelling older adults, 13.5% had poor hearing and 12.6% had poor vision, while 2.6% reported exposure to both poor hearing and poor vision. In multiple linear analyses, worse hearing or/and vision is independently associated with poorer cognition and slower gait speed, while cognition is positively related to gait speed. According to Baron and Kenny’s causal steps method and Sobel test, cognition partially mediates the relationship between sensory function and gait speed.

Several previous studies have reported the risk effect of impaired hearing or/and impaired vision on gait performance in older community population. A prospective study conducted in older Finnish women showed that impaired hearing correlated cross-sectionally with slower maximal gait speed and lower walking endurance, and that impaired hearing at baseline could predict new self-reported walking difficulties after 3 years [36]. Another longitudinal study conducted in United States indicated that impaired visual acuity, contrast sensitivity, and stereoacuity were associated with greater risk of walking limitations during 5 years of follow-up [10]. In line with them, our results suggested the risk effect of poor hearing and vision on gait speed in older English. One commonly accepted explanation was that cognition such as memory and executive function might mediate the relationship between sensory function and gait performance.

As with our finding, a cross-sectional study by Tien et al. found that persons with vision impairment or hearing loss had a lower mean Mini-Mental State Examination score than those with normal vision or hearing in non-institutionalized residents aged 50+ years [7]. Based on three large-scaled cohorts of aging, Maharani et al. performed growth curve analysis and reported that both single and dual sensory impairments in older adults were independently associated with accelerated rates of decline in cognitive abilities [6]. Sensory impairment causes cognitive decline that is either potentially remediable (the ‘information-degradation’ hypothesis) or permanent (the ‘sensory-deprivation’ hypothesis) [14]. According to the ‘information-degradation’ hypothesis [37], declines in cognitive performance manifest as a consequence of compensating for impaired sensory input. Individuals with sensory impairment may compensate for sensory deficits via increased reliance on cognitive resources, thereby reducing the cognitive resources available for other cognitive tasks. The ‘sensory-deprivation’ hypothesis emphasizes that chronic reallocation of cognitive resources may produce permanent changes in cognitive performance over time [38].

Meanwhile, recent converging evidence from epidemiological, structural imaging, functional imaging, and genetic studies suggests that high-order cognitive control mechanisms influence gait [12]. Of the cognitive functions examined, attention and executive function are empirically considered as having important relationships with gait speed. Interestingly, Holtzer et al. successfully demonstrated that memory and verbal intelligence quotient might be reliable predictors of gait speed as well [11]. Our study also showed the positive correlation of global cognition including memory, verbal fluency, and time orientation with gait speed in older adults, suggesting that the cognitive correlates of gait speed are multifaceted. Shared brain substrates that underlie complex and distinct cognitive motor circuits explain, in part, the association between executive function and episodic memory and gait. Hippocampal volume and metabolism that are critical for memory function are also related to gait. Furthermore, volumetric and functional neuroimaging studies showed that the prefrontal cortex subserves executive function and gait [12].

To our knowledge, this is the first observational study to characterize the relationship between sensory function, cognition and gait speed, by focusing on the role of cognition in mediating sensory function and gait speed. As we expected, when cognition was introduced in the linear models as a covariate, sensory function decrease their effects on the gait speed, suggesting the partial mediation of cognition. And further Sobel test proved the statistic significance of mediating effect. Individuals with hearing or visual impairment have to allocate more attention resources to processing sensory information which lead to the reduction of resources for walking task. Additionally, cognitive decline caused by long-termed exposure to sensory impairment could change the gait performance. An experimental study suggested that external auditory cues were useful for patients with Parkinson’s disease to reduce interference and to maintain gait performance during more complicated functional activities, because it was less attention-demanding than performing the task with no cue, which requires the use of cognitive processes to efficiently divide attention between tasks [39]. Another experimental study proved that for elderly, the task of reintegrating sensory information perturbs balance and requires additional attentional demand [40]. This study provides the epidemiological evidence that cognition mediates the association of sensory function with gait speed among community-dwelling older adults.

A major strength of this study is its large sample size, which provides sufficient power of test to detect the mediating effects that are present. Besides, hearing was measured in an objective and valid method, which could minimized misclassification bias. This study had several limitations. Firstly, this cross-sectional study cannot make conclusions about the longitudinal nature of the relationships between sensory function, cognition and gait speed. Furthermore, the current design prevents us from making causal attributions. The selection of variables as “cause” and “effect” is purely theoretical. Future research should be directed at clarifying the longitudinal nature of these relationships and experimentally manipulating the variables through interventions. Secondly, there were 7.6% of eligible participants excluded from the hearing analyses, because they have not performed the hearing measurement. Their inclusion, otherwise, could have modified the results as found in the present analyses. Thirdly, ELSA did not take objective measures for vision; self-reported evaluation is prone to reporting bias. Fourthly, the cognitive tests in this study were primarily administered verbally, which possibly biased against those with severe hearing and vision impairment. However, those who had ear infection or cochlear implant or blindness were excluded from analyses. Besides, in ELSA, if a participant normally uses reading glasses or hearing aids, these must be used during testing. This suggests that participants with poor sensory function can use ancillary tools to help them understand problems from cognitive test. Finally, despite controlling for many potential covariates, residual confounding such as effect of proprioception may influence our observed associations between sensory function, cognition, and gait speed.

Gait speed is a robust indicator of health status and a strong predictor of several adverse health outcomes in older adult population. Thus, identifying modifiable risk factors of slower gait speed and their underlying mechanism is a public health priority. The evidence that we provide supports attempts to maintain hearing and vision and protect cognitive ability, as they are proximal influences on gait. Furthermore, our findings on the role of cognition in mediating sensory function and gait speed would deepen our understanding on the etiology and development of Alzheimer’s disease and dementias.

Footnotes

ACKNOWLEDGMENTS

This study uses data from the English Longitudinal Study of Ageing (ELSA). The authors appreciate the ELSA consortium for making the survey and making it available online freely, and all the participants for providing these data.

Authors’ disclosures available online ().

References

Fritz

, Lusardi

(2009) White paper: “Walking speed: The sixth vital sign”. J Geriatr Phys Ther 32, 46–49.

Abellan van Kan

, Rolland

, Andrieu

, Bauer

, Beauchet

, Bonnefoy

, Cesari

, Donini

, Gillette Guyonnet

, Inzitari

, Nourhashemi

, Onder

, Ritz

, Salva

, Visser

, Vellas

(2009) Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people an International Academy on Nutrition and Aging (IANA) Task Force. J Nutr Health Aging 13, 881–889.

Artaud

, Singh-Manoux

, Dugravot

, Tzourio

, Elbaz

(2015) Decline in fast gait speed as a predictor of disability in older adults. J Am Geriatr Soc 63, 1129–1136.

White

, Neogi

, Nevitt

, Peloquin

, Zhu

, Boudreau

, Cauley

, Ferrucci

, Harris

, Satterfield

, Simonsick

, Strotmeyer

, Zhang

(2013) Trajectories of gait speed predict mortality in well-functioning older adults: The Health, Aging and Body Composition Study. J Gerontol A Biol Sci Med Sci 68, 456–464.

Sabia

, Dumurgier

, Tavernier

, Head

, Tzourio

, Elbaz

(2014) Change in fast walking speed preceding death: Results from a prospective longitudinal cohort study. J Gerontol A Biol Sci Med Sci 69, 354–362.

Maharani

, Dawes

, Nazroo

, Tampubolon

, Pendleton

; Sense-Cog WP1 group (2018) Visual and hearing impairments are associated with cognitive decline in older people. Age Ageing 47, 575–581.

Tay

, Wang

, Kifley

, Lindley

, Newall

, Mitchell

(2006) Sensory and cognitive association in older persons: Findings from an older Australian population. Gerontology 52, 386–394.

Zheng

, Swenor

, Christ

, West

, Lam

, Lee

(2018) Longitudinal associations between visual impairment and cognitive functioning: The Salisbury Eye Evaluation Study. JAMA Ophthalmol 136, 989–995.

, Simonsick

, Ferrucci

, Lin

(2013) Hearing loss and gait speed among older adults in the United States. Gait Posture 38, 25–29.

10.

Swenor

, Simonsick

, Ferrucci

, Newman

, Rubin

, Wilson

(2015) Visual impairment and incident mobility limitations: The health, aging and body composition study. J Am Geriatr Soc 63, 46–54.

11.

Holtzer

, Verghese

, Xue

, Lipton

(2006) Cognitive processes related to gait velocity: Results from the Einstein Aging Study. Neuropsychology 20, 215–223.

12.

Holtzer

, Wang

, Lipton

, Verghese

(2012) The protective effects of executive functions and episodic memory on gait speed decline in aging defined in the context of cognitive reserve. J Am Geriatr Soc 60, 2093–2098.

13.

Watson

, Rosano

, Boudreau

, Simonsick

, Ferrucci

, Sutton-Tyrrell

, Hardy

, Atkinson

, Yaffe

, Satterfield

, Harris

, Newman

(2010) Executive function, memory, and gait speed decline in well-functioning older adults. J Gerontol A Biol Sci Med Sci 65, 1093–1100.

14.

Lin

, Yaffe

, Xia

, Xue

, Harris

, Purchase-Helzner

, Satterfield

, Ayonayon

, Ferrucci

, Simonsick

, Health ABC Study Group (2013) Hearing loss and cognitive decline in older adults. JAMA Intern Med 173, 293–299.

15.

Muir

, Gopaul

, Montero Odasso

(2012) The role of cognitive impairment in fall risk among older adults: A systematic review and meta-analysis. Age Ageing 41, 299–308.

16.

, Simonsick

, Ferrucci

, Lin

(2013) Hearing loss and gait speed among older adults in the United States. Gait Posture 38, 25–29.

17.

Shumway-Cook

, Woollacott

(2000) Attentional demands and postural control: The effect of sensory context. J Gerontol A Biol Sci Med Sci 55, M10–M16.

18.

Monge

, Madden

(2016) Linking cognitive and visual perceptual decline in healthy aging: The information degradation hypothesis. Neurosci Biobehav Rev 69, 166–173.

19.

Wayne

, Johnsrude

(2015) A review of causal mechanisms underlying the link between age-related hearing loss and cognitive decline. Ageing Res Rev 23, 154–166.

20.

Steptoe

, Breeze

, Banks

, Nazroo

(2013) Cohort profile: The English longitudinal study of ageing. Int J Epidemiol 42, 1640–1648.

21.

Zaninotto

, Sacker

, Head

(2013) Relationship between wealth and age trajectories of walking speed among older adults: Evidence from the English Longitudinal Study of Ageing. J Gerontol A Biol Sci Med Sci 68, 1525–1531.

22.

Ray

, Popli

, Fell

(2018) Association of cognition and age-related hearing impairment in the English Longitudinal Study of Ageing. JAMA Otolaryngol Head Neck Surg 144, 876–882.

23.

de la Fuente

, Hjelmborg

, Wod

, de la Torre-Luque

, Caballero

, Christensen

, Ayuso-Mateos

(2018) Longitudinal associations of sensory and cognitive functioning: A structural equation modelling approach. J Gerontol B Psychol Sci Soc Sci. doi: 10.1093/geronb/gby147.

24.

Smith

, Timmis

, Pardhan

, Latham

, Johnstone

, Hamer

(2017) Physical inactivity in relation to self-rated eyesight: Cross-sectional analysis from the English Longitudinal Study of Ageing. BMJ Open Ophthalmol 1, e000046.

25.

Clark

, Gatz

, Zheng

, Chen

, McCleary

, Mack

(2009) Longitudinal verbal fluency in normal aging, preclinical, and prevalent Alzheimer’s disease. Am J Alzheimers Dis Other Demen 24, 461–468.

26.

Smith

, Neal

, Peryer

, Sheehan

, Tan

, Myint

(2018) Orientation and verbal fluency in the English Longitudinal Study of Ageing: Modifiable risk factors for falls? Int Psychogeriatr. doi: 10.1017/S1041610218002065.

27.

Valax

, Tremblay

, Sarocchi

(1996) What month is it? The process of temporal orientation on a unit of the year scale. Acta Psychol (Amst) 94, 309–317.

28.

Koriat

, Fischhoff

(1974) What day is today? An inquiry into the process of time orientation. Mem Cognit 2, 201–205.

29.

Olaya

, Bobak

, Haro

, Demakakos

(2017) Trajectories of verbal episodic memory in middle-aged and older adults: Evidence from the English Longitudinal Study of Ageing. J Am Geriatr Soc 65, 1274–1281.

30.

Hamer

, de Oliveira

, Demakakos

(2014) Non-exercise physical activity and survival: English longitudinal study of ageing. Am J Prev Med 47, 452–460.

31.

Demakakos

, Hamer

, Stamatakis

, Steptoe

(2010) Low-intensity physical activity is associated with reduced risk of incident type 2 diabetes in older adults: Evidence from the English Longitudinal Study of Ageing. Diabetologia 53, 1877–1885.

32.

Dhalwani

, O’Donovan

, Zaccardi

, Hamer

, Yates

, Davies

, Khunti

(2016) Long terms trends of multimorbidity and association with physical activity in older English population. Int J Behav Nutr Phys Act 13, 8.

33.

Turvey

, Wallace

, Herzog

(1999) A revised CES-D measure of depressive symptoms and a DSM-based measure of major depressive episodes in the elderly. Int Psychogeriatr 11, 139–148.

34.

Baron

, Kenny

(1986) The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc Psychol 51, 1173–1182.

35.

Mackinnon

, Warsi

, Dwyer

(1995) A simulation study of mediated effect measures. Multivariate Behav Res 30, 41.

36.

Viljanen

, Kaprio

, Pyykko

, Sorri

, Koskenvuo

, Rantanen

(2009) Hearing acuity as a predictor of walking difficulties in older women. J Am Geriatr Soc 57, 2282–2286.

37.

Pichora-Fuller

(2003) Cognitive aging and auditory information processing. Int J Audiol 42(Suppl 2), 2S26–2S32.

38.

Humes

, Busey

, Craig

, Kewley-Port

(2013) Are age-related changes in cognitive function driven by age-related changes in sensory processing? Atten Percept Psychophys 75, 508–524.

39.

Rochester

, Hetherington

, Jones

, Nieuwboer

, Willems

, Kwakkel

, Van Wegen

(2005) The effect of external rhythmic cues (auditory and visual) on walking during a functional task in homes of people with Parkinson’s disease. Arch Phys Med Rehabil 86, 999–1006.

40.

Teasdale

, Simoneau

(2001) Attentional demands for postural control: The effects of aging and sensory reintegration. Gait Posture 14, 203–210.