Relationship between Cognitive Decline and Daily Life Gait among Elderly People Living in the Community: A Preliminary Report

Dementia is a rapidly growing public health problem, affecting around 50 million people around the world. Management of modifiable risk factors can delay or slow onset or progression of the disease [1]. Community-based studies reported reversion rates from mild cognitive impairment (MCI) to normal cognition of approximately 31% [2]. Early diagnosis and intervention for MCI are urgently needed. The standard diagnostic procedure for MCI requires an interview by a physician, a comprehensive neuropsychological assessment by a psychologist that explores various cognitive domains, and brain imaging. However, these procedures are time-consuming, expensive, and restricted in availability. Thus, screening methods, such as neuropsychological tests, such as Montreal Cognitive Assessment [3], have been devised; however, the use of such tests is not widespread in the community as it strains the subject.

Physical and cognitive decline, which increase the risk of developing dementia, are getting a lot of attention [4]. Gait parameters have been recognized as a potentially easy screening tool for MCI. A previous meta-analysis of cross-sectional studies reported that MCI had a slower gait velocity (normal, fast pace, dual-task) than cognitive healthy individuals (CHI). These gait velocities were functional objective parameters, which best discriminated between CHI and MCI [5]. Furthermore, cohort studies suggested that gait variability could predict MCI [6]. Moreover, a previous meta-analysis showed a significant, small effect [5]. However, the gait velocity used in these studies was laboratory-based and was measured in an environment that was different from that of daily life gait. A study comparing laboratory-based gait with daily life gait also reported that daily life gait velocity may be useful to detect the decline in the functional abilities of older adults [7]. Smart home sensor technologies involving continuous real-time gait velocity measurement have shown promising results for the early detection of MCI [8]; however, such studies have limited gait distance, and the installation and management of sensors is difficult. Recently, smartphones with built-in accelerometers have become widely used to measure the amount of physical activity [9]. Thus, measurement of gait velocity in daily life using a smartphone accelerometer could be a self-screening tool of MCI.

In this study, we compared daily life gait measured with smart device accelerometers and laboratory-based gait in CHI and cognitive decline individuals (CDI). We aimed to clarify the relationship between cognitive decline and daily life gait and use it for the development of MCI screening methods.

One hundred fifty-eight community-dwelling elderly people (65 years of age or above) who participated in a cognitive function measurement event hosted by the city office and consented to participate in the study were selected. These events were held in 2017 and 2018 in a total 14 locations (two in May 2017, four in September 2017, four in May 2018, four in September 2018). The inclusion criteria comprised of being able to gait independently and to follow assessment instructions. The exclusion criteria were being diagnosed with dementia and requiring long-term care. In addition, subjects were excluded if they had symptoms that affected walking (neurological conditions, such as stroke or Parkinson’s disease; musculoskeletal disease, such as arthrosis impairing posture or gait; acute disease; or surgery). Excluding three (two CHI, one CDI) subjects who did not have daily life gait data, 155 subjects were included in the analysis.

Cognitive function measurement event was held for 2 days at public halls. On the first day, we collected subject’s characteristics of age, sex, body height and weight, years of education, family arrangement, self-rated health, presence or absence of chronic disease (such as hypertension or orthopedic disease), and pain (such as low back or knee), depression, subjective memory impairment (SMI), quality of life (QOL), instrumental activity of daily living (IADL), and life space using a questionnaire. After that, we conducted a group cognitive function test (5-Cog) and the participants watched a lecture on dementia. At the end, subjects were given an iPod-touch for daily gait measurement and were asked to wear the device all day (except for sleeping and bathing) for a week. The device was placed in a case and could not be seen or operated by the subjects. The 2nd day, we collected the iPod-touch and performed laboratory-based gait measurements. We gave the subjects feedback about the results of the 5-Cog and instructed them to perform exercises to prevent dementia.

Cognitive function was assessed with the 5-Cog test, which enables measurement of the cognitive level of health of older adults in the community and individuals with slight cognitive disabilities. This test consists of five items: attention, memory, visuospatial function, language, and reasoning [10]. Attention was evaluated using the Japanese version of the set dependency activity [11], which assesses alternating attention. In this test, there are three rows on the page (top, middle, and bottom) with three Chinese characters that mean “top,” “middle,” and “bottom.” Some of the characters were placed in the incorrect rows. The subjects were required to choose the characters that were placed in the correct row. In order to assess the memory ability, we used a Category Cued Recall test [12]. A Clock Drawing test, which requires the subjects to draw clock hands showing the time at “ten after eleven” [13], was used to assess visuospatial function. We examined language ability using a category fluency test [14]. The subjects were asked to generate as many examples as possible in 2 min from the semantic category “animals.” To assess abstract reasoning ability, we employed the similarity subset of the Wechsler Adult Intelligence Scale-Revised [15]. The reliability and validity of the 5-Cog test in a Japanese population have been confirmed [16].

Using a DVD, cognitive assessment was conducted in a group setting (maximum 100 participants) by an examiner with use of a projector. All subjects were asked to record their answers on an answer sheet. Before the cognitive assessment, the subjects were asked to perform a motor function test of the upper extremities, circling as many numbers 1–80 as possible in 15 s, to confirm that there was no difficulty in answering. Mean duration of this test was 35 min. We conducted the test with a small group of 10–20 subjects and with three staff so that the test could be performed accurately. For subjects who had difficulty understanding the tasks or had impaired hearing or vision, the staff assisted individually.

The score of each test was adjusted by age, educational years, and sex, indicated by the deviation value (mean: 50, 1 standard deviation (SD): 10). If the score of the test was within 1 SD from the average score, it was rank 3; if it was more than 1 to 1.5 SD lower, it was rank 2; and if it was lower than 1.5 SD, it was rank 1. Therefore, if it was within 1 SD from the average value in five tests, the overall rank would be 15 points (CHI). 14 points or less indicated that one or more tests were lower than 1 SD of the average. In this study, 14 points or less was defined as CDI.

Daily life gait data were collected by a health care application (app) preinstalled on iPod-touch (iOS10.3.1; Apple Japan, Tokyo, Japan). It is commercially available, and it has a 3-axis accelerometer and an air pressure gauge. This app automatically recorded start and stop time of gait, moving distance, and number of steps in one continuous gait. The algorithms for determining walking initiation in each app programming interface are not published. Reliability and validity of measurements have been examined in previous studies [17, 18]. In our preliminary study of healthy adults (n = 12, 21.3 ± 0.6 years old), the number of steps taken with the iPod-touch was highly correlated with the pedometer (Kenz Lifecoder EX, Suzuken Co. Ltd., Nagoya, Japan) (r = 0.928, p < 0.001) results. Furthermore, daily life gait velocity with the iPod-touch was moderately correlated with laboratory-based 10-m gait velocity (normal pace) (ρ = 0.517, p = 0.085) (unpublished data). Six days of data were used for analysis, excluding the date the device was handed over and the date it was collected. The records of gait velocity faster or slower than 2 SD were excluded from the average sex and age [19]. We calculated mean, maximum, and minimum velocity (distance/time; cm/s); velocity variability (coefficient of variation; CoV: SD/mean×100); mean, maximum, and minimum step length (distance/number of steps; cm); step length variability (CoV); and cadence (number of steps/times; steps/min).

Laboratory-based gait was measured using an electronic portable walkway with embedded pressure sensors connected to a computer (Walk Way MW-1000, AMIMA Co. Ltd., Tokyo, Japan: length 4.8 m width 0.6 m, scanning frequency 100 Hz). This gait analyzer is approved as a medical device in Japan. Start and endpoint were marked on the floor 3 m from either mat end to avoid recording acceleration/deceleration. Subjects were asked to walk 11 m as fast as possible (but without running) in two trails, wearing comfortable footwear, in a well-lit and quiet room. Based on the footfalls recorded on the walkway, the software automatically computed gait parameters as the mean of two trials. The following four gait variables were used: velocity (cm/s), stride length (cm), stride length variability (CoV), and cadence (steps/min).

The following evaluations were carried out as factors related to gait and cognitive function. The self-rated health was rated on a 4-point rating scale: “very healthy,” “moderately healthy,” “not very healthy,” and “not healthy.” “Very healthy” and “moderately healthy” were combined as “healthy,” and “not very healthy” and “not healthy” were combined as “unhealthy.” SMI was evaluated by four questions about memory complaints [20]. If one question applied, the patient was judged to have SMI. Depression was evaluated by the use of the 15-item Geriatric Depression Scale (GDS; score range 0–15; a score ≥5 indicated the presence of depressive symptom) [21]. QOL was evaluated by the 8-Item Short-Form Health Survey [22]. Furthermore, we calculated physical component summary and mental component summary (higher score indicates higher QOL). IADL was measured using the five-cog IADL questionnaire. The 15 items asked about abilities of using the phone, planning and managing events as a leader, organizing a meeting, using public transportation, taking a trip, medication management, money management, shopping for commodities, paying bills, handling bank accounts, document preparation, preparing meals, cleaning, organizing clothing and tableware, and writing letters and sentences (able or unable). The number of items answered with “able” was used as the index of IADL (score range, 0–15; a higher score indicates higher functional capacity). Life space was evaluated using the life-space assessment (LSA) [23]. LSA was developed to evaluate mobility status by measuring the life space in community-dwelling elderly individuals (score range, 0–120; a higher score indicates larger life space).

Mean ± SD or median (25th–75th percentiles) were calculated for continuous variables and frequency (%) for categorical variables. To compare differences between CHI and CDI, t test and Mann-Whitney U test were used to analyze the continuous variables, and the χ² test and Fisher exact test were used to analyze the categorical variables. Pearson’s and Spearman’s rank correlation was used to compare daily life and laboratory-based gait parameters. All statistical analysis was performed using IBM SPSS Statistics version 26 (IBM Co., Ltd, Tokyo, Japan); the level of significance was set at p < 0.05.

The characteristics of subjects age were 75.5 ± 5.4 years, 106 (68.4%) were female, years of education was 12.0 (12.0–14.0), number of chronic diseases was 1.0 (0.0–2.0), score of GDS was 3.0 (1.0–5.0), 123 (79.4%) were SMI, and score of 5-Cog’ IADL questionnaire was 14.0 (13.0–15.0). More detailed IADL results are shown in the online supplementary Table (for all online suppl. material, see www.karger.com/doi/10.1159/000528507).

The subjects consisted of 98 CHI (63.2%) and 57 CDI (36.8%). The demographic, health and cognitive functioning of CHI and CDI are presented in Table 1. CDI had a higher prevalence of low back pain than CHI (27 (47.4%) versus 30 (30.9%), respectively; p = 0.037). There was no difference in other demographic and health variables.

In cognitive function, all five tests scored significantly lower in CDI compared to CHI. In the CDI, attention (n = 15, 26.3%), memory (n = 14, 24.6%), visuospatial function (n = 11, 19.3%), language (n = 25, 43.9%), and reasoning (n = 23, 40.4%) were lower than 1 SD. The type of cognitive decline included: decline in memory only (n = 4, 7.0%), decline in memory and other domains (n = 10, 17.5%), single domain decline other than memory (n = 33, 57.9%), and multiple domain decline other than memory (n = 10, 17.5%). The degree of cognitive decline was lower than 1 to 1.5 SD (n = 34, 59.6%) and lower than 1.5 SD (n = 23, 40.4%) in one or more domains.

The number of measurement days was 6.0 (5.5–6.0), daily measurement time was 11.2 (7.6–13.6) hours, total moving distance was 11.5 (7.3–17.9) km, and exclusion rate taken into analysis was 48.0% (37.2–68.2%). No significant difference in measurement status was found between CHI and CDI.

The gait velocity in CDI (78.0 ± 16.2 cm/s) was slower than that in CHI (82.0 ± 16.3 cm/s), but there was no significant difference. On the other hand, the maximum gait velocity in CDI (113.7 [97.0–128.5] cm/s) was significant slower than in CHI (121.2 [105.8–134.3] cm/s) (p < 0.032, Table 2). Other items did not differ significantly between CHI and CDI.

The gait velocity in CDI (193.2 ± 31.3 cm/s) was slower that in CHI (199.3 ± 31.5 cm/s), but there was no significant difference. On the other hand, stride length variability in CDI (2.6 [1.8–4.1]) was significant larger than in CHI (1.8 [1.2–2.7]) (p < 0.001, Table 3). Stride length and cadence did not differ significantly between CHI and CDI.

The mean and maximum gait velocity and cadence in daily life were significantly and weakly correlated with velocity (mean: r = 0.313; Fig. 1, maximum: ρ = 0.293, cadence: ρ = 0.256) and stride length (mean: ρ = 0.254, maximum: ρ = 0.243, cadence: ρ = 0.270) in laboratory-based gait. The stride length variability in laboratory-based gait was significantly and weakly correlated with maximum gait velocity in daily life gait (ρ = −0.260, p = 0.001, Fig. 2).

CDIs were aware of forgetfulness and showed significant cognitive decline, but they were independent in their IADL and were considered to have the diagnostic criteria for MCI [24]. No significant differences were found in demographic and health variables other than prevalence of low back pain. In laboratory-based gait, as in a previous study [25], CDI significantly increased stride variability but did not have slower gait velocity. In this study, cued recall test was used for the memory test. Cued recall is more likely to be retained in MCI compared to free recall [26]. Therefore, it is possible that we may have underestimated memory decline, which may have influenced the difference in gait outcomes.

We focused on the difference in daily life gait parameters between CHI and CDI. CDI could be measured without forgetting as well as in CHI. Mean gait velocity of all subjects was 80.5 ± 16.3 cm/s. In addition, it was weakly correlated with laboratory-based gait velocity (fast pace), which was similar to the findings of a previous study [7]. The result of previous studies of daily life gait velocity in CHI were 61 ± 17 cm/s (n = 76, mean age = 86) with smart home [19], 123 ± 9 cm/s (n = 178, age range = 75–79) with a smartphone global positioning system app [27], and 114 ± 8 cm/s (n = 30, mean age = 68.8 ± 4.4) with smart shoes [28]. Gait velocity in the present study included both indoor and outdoor, and we considered the results to be valid because they were intermediate between the smart home results measured indoors and the smartphone global positioning system and smart shoe results measured outdoors. The exclusion rate was high (48%), most of which was slower than 2 SD from the sex and age mean. In daily life gait, we found repeated short-period of gait (60% of all gait bouts lasted just 30 s or less and 75% of all gait bouts were less than 40 steps) and many short-periods of rest (half of all rest periods lasted 20 s or less) [29]. Our one continuous gait data often recorded dozens of steps in increments of a few minutes. Therefore, we considered the possibility that the device failed to recognize a short-period of rest and included the rest time in the gait data, resulting in more data with slower gait velocity.

Daily life mean and maximum gait velocity were slower in CDI. Furthermore, stride length variability, which was significantly increased by CDI in laboratory-based gait (fast pace), was weakly correlated with daily life maximum gait velocity. Smart home studies have shown that slower gait velocity was associated with lower global cognitive score [19] and that subjects transitioning to MCI did not walk slower but had fewer opportunities to walk faster [30]. In addition, the smart shoes study showed a significant decrease in gait velocity and stride length, and an increase in stride time variability in amnestic MCI compared to CHI [28]. We considered the lower gait velocity in CDI to represent cognitive decline. However, the prevalence of low back pain was higher in CDI than in CHI, which may have influenced the slower gait velocity. Since we have not been able to investigate the details of the frequency and severity of low back pain, we are limited in our ability to consider the impact of the pain on daily life gait. Nevertheless, we considered the effect to be small because no one complained of pain during the laboratory-based gait measurement (fast pace) and we found there was no significant difference between CHI and CDI in the laboratory-based gait velocity.

Community-dwelling elderly showed a possible association between cognitive decline and slower daily life gait velocity. The limitations of the present study include that it was a cross-sectional study and the duration of the daily life gait measurement was only 6 days. Gait changes characteristic of MCI appear when laboratory-based gait results are followed for more than 1 year [31]. Therefore, it is necessary to examine whether changes in daily life gait in MCI appear earlier than in laboratory-based gait in longitudinal studies. In such cases, frequent app updates may alter the data, which may affect the continuation of the study. In addition, this study used the 5-Cog test, which can be administered in groups at community sites. However, it may not be sufficient to rigorously detect MCI. Moreover, it is necessary to examine whether indoor, outdoor, or both indoor and outdoor gait measures are effective in predicting MCI. In addition, changes in gait by type of cognitive decline should be considered.

We thank Akira Iizuka (Gunma University Research Planning Office) for the valuable support and contribution to our research.

Written consent was obtained from each subject after informing them of the full information regarding the purpose of the study, risk and benefits, confidentiality, anonymity, and freedom of participation. This study was conducted in accordance with the Declaration of Helsinki and approved by Gunma University Ethical Review Board for Medical Research Involving Human Subjects (approval number: HS2017-009).

The authors have no conflicts of interest to declare.

This work was supported by JSPS KAKENHI Grant Number 18K10500, the Nakatomi Foundation, and by the Ai-no-bokin.

All the authors contributed significantly to the research and manuscript, and all the authors are in agreement with the content of the manuscript. Tetsuya Yamagami contributed to study concept and design, analysis and interpretation of data, and preparation of the manuscript. Motoi Yagi contributed to the study concept and analysis/interpretation of data. Shigeya Tanaka, Saori Anzai, Takuya Ueda, Yoshitsugu Omori, and Chika Tanaka were involved in data collection, recruitment, and evaluation of the subjects. Yoshitaka Shiba conceived the study and data collection, recruitment, and evaluation of the subjects.

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further inquiries can be directed to the corresponding author.