Spatio-temporal gait variables predicted incident disability

Participants came from the population-based Obu Study of Health Promotion for the Elderly [19], which was a part of the National Center for Geriatrics and Gerontology – Study of Geriatric Syndromes [20]. The current study was designed as a prospective study. The baseline examination was conducted between 2011 and 2012. Inclusion criteria included: ≥65 years old when recruited for the study. A total of 15,974 individuals were found eligible for participation. Prior to recruitment, 1661 people were excluded because they had participated in other similar studies, were hospitalized or in residential care, or were certified at levels 3–5 to require support or care by the Long-Term Care Insurance (LTCI). A letter was sent to 14,313 individuals for the purpose of recruitment. A total of 5104 individuals participated in the baseline examination. The detailed recruiting protocol was previously described [20]. We excluded participants based on the following criteria: having any dependency for basic ADL, being certified at any level by LTCI, having specific medical conditions (stroke, Parkinson’s disease, Alzheimer’s disease), having severe cognitive impairment as assessed by a Mini-Mental State Examination (MMSE) score < 20, censoring due to moving away or death, and having missing values for any of these variables. Baseline assessments for medical conditions were conducted by well-trained nurses via face-to-face interview, and the other variables were assessed by well-trained staffs. Finally, 4121 participants were eligible for this study. During follow-up, we monitored incident disability using the records of LTCI.

The detailed protocol was described in previous studies [19, 21]. Participants were instructed to walk along a 6.4 m straight and flat pathway; five trials were conducted. Gait assessment was measured over 2.4 m, with 2 m allowed for acceleration and deceleration. An electronic measuring device (WalkWay MW− 1000, Anima Co., Tokyo, Japan) was mounted at the mid of pathway (2.4 m). The WalkWay is 800 mm wide, 2400 mm long, and 5 mm thick, and mounted with strain gages placed 10 mm apart (14,000 points). Gait speed was calculated from walking time measured by a stop watch, and the mean gait speed over five trials was expressed in meters per second. The other variables were cadence, stride length, and stride length variability acquired from the electronic measuring device. Based on a previous study that used a similar measuring device, spatio-temporal gait variables could be classified into three statistically independent factors representing distinct gait domains; the study also suggested the use of gait variables identified from the three factors [15]. Following the study, we selected variables from the three factors. Cadence was calculated by numbers of step per minutes and stride length was calculated by mean stride length (m) through trials. Variability was calculated by the coefficient of variation (CV) of stride length: CV [%] = [standard deviation / mean] × 100. When fewer than five strides of data points were acquired over five trials, the data were regarded as missing values.

We monitored disability defined as LTCI certification for all subjects during the follow-up period (mean duration: 49.6 months). The LTCI was introduced in Japan and certifies a person as “Support Level 1 or 2” if needing support for daily activities or “Care Level 1, 2, 3, 4, or 5” if needing continuous care [22]. When certified by the LTCI, persons are eligible for various services, according to their nursing care plan, up to the maximum coverage amount for a municipality. A primary care medical doctor evaluates the need for LTCI and a trained healthcare official evaluates the need for nursing care, using a questionnaire that assesses current physical and mental status resulting from their medical conditions (e.g., treatments received or medication regimes). Subcategories of these items were paralysis and limitation of joint movement, movement and balance, complex movement, conditions requiring special assistance, conditions requiring assistance with activities of daily living/instrumental activities of daily living, communication and cognition, behavioral problems, and use of medical procedures during the previous 14 days. Based on the results, a computer program is used to calculate the applicant’s standardized scores for the seven dimensions of physical and mental status and estimated level of care and assign a care-need level based on the total estimated care-minutes. The Nursing Care Needs Certification Board (physicians, nurses, and other experts in health and social services appointed by a mayor) reviews whether the initial assessment is appropriate for each assessed individual, considering the applicant’s primary care physician’s statements and notes written by the assessor during the home visit. In this study, the outcome of disability was defined as a new certification of LTCI at any level during the follow-up period.

Covariates were selected as confounding factors based on selections in previous studies. A face-to-face interview was conducted by well-trained nurses to obtain participants’ information, including age, sex, height, weight, and presence of medical conditions (heart disease, diabetes mellitus, osteoarthrosis, and knee osteoarthritis). In addition, physical inactivity [19] and MMSE [23] were measured by well-trained staff. Physical inactivity was measured through responses to the following two questions: “Do you engage in moderate levels of physical exercise or sports aimed at health?” and “Do you engage in low levels of physical exercise aimed at health?” When participants answered “no” to both of these questions, they were evaluated as physically inactive.

To examine characteristics associated with disability incidence, student-t-tests or χ² tests were used to compare participant characteristics between individuals with disability and those without. To compare the degree of differences in gait variables between groups, effect size (Cohen’s d) was calculated: mean differences/pooled SD. Cohen d was suggested that threshold of large effect was 0.8, medium effects was 0.5, low effects was 0.2. To examine the association between gait variables, we calculated Pearson’s correlation coefficient among gait variables. Cox proportional hazards regression analysis was used to examine associations between each gait variable as a continuous variable and disability incidence. Analysis was conducted separately for each gait variable. Hazard ratios (HR) and 95% confidential intervals (CI) in each model were calculated. Disability was used as the objective variable and gait variables were separately used as explanatory variables, adjusted for covariates of age, sex, height, weight, medical condition, physical inactivity, and MMSE score. To examine the cumulative effects of low gait function reflected in spatio-temporal parameters, each gait variable was converted to categorical data (low function or high function) based on cut-off values. Cut-off values were calculated by using a receiver operating characteristic curve and the Youden index to determine sensitivity and specificity. Based on the cut-off values, participants were classified into either low function or high function categories for each gait variable. Then, the number of low functions for gait variables (gait speed, stride length, cadence, and stride length variability) were directly sored (0–4). For identifying associations of the number of low function with incident disability, the Cox proportional hazards regression analysis was used and adjusted model were also set in the same manner since gait variables were considered as continuous variables in the analysis. All calculations were performed using IBM SPSS Statistics, ver. 20 (IBM Corp., Armonk, NY, USA). Significance was defined as P < .05 for all analyses.