Study population
The Japan Gerontological Evaluation Study 2013 (JAGES 2013) is one of the waves of JAGES, an ongoing population-based panel study aimed at investigating the association between social and behavioral factors and health-related outcomes in the elderly population.[21] JAGES 2013 was conducted from October to December 2013 recruiting 195,290 community-dwelling residents aged ≥65 years who were ineligible for the long-term care certification from 30 municipalities in 14 of 47 Japanese prefectures. In the Hokkaido prefecture, JAGES at Taisetsu community Hokkaido in 2014 (JAGES ATTACH 2014) was developed based on JAGES 2013. JAGES ATTACH 2014 further limited the number of participants compared with the JAGES 2013 by eliminating those living in Taisetsu community Hokkaido (Higashikawa, Higashikagura, and Biei town) aged 70–74 years who were ineligible for the long-term care certification at the beginning of the study, thereby finally enrolling 1,127 participants. Overall, 824 participants responded to the questionnaire by mail (response rate: 73.1%), and a home visit survey was conducted to measure weight, height, and other indicators after agreeing to participate in the study. The follow-up survey was conducted in August 2017. Follow-up data on frailty were obtained by mailing the Kaigo-Yobo Checklist, and data related to total mortality or the number of participants who moved out of their community were obtained from the local basic resident registry. Responding to the questionnaire was considered informed consent. This study was approved by the Ethics Review Committee of Hokkaido University Graduate School of Medicine (NO. 14-024).
Physical activity assessment
PA was evaluated using the short-term International Physical Activity Questionnaire (IPAQ), a useful tool for evaluating PA in young and older Japanese adults [22]. Participants were asked about the number of days they engaged in vigorous PA, moderate PA, or walking during the past 7 days; the options included “none” and frequencies of the given PA. If the answer was not “none,” they were then asked to report the time spent on the given PA. Vigorous PA (heavy lifting, digging, and jogging) requires extensive physical effort, making the participants breathe heavier than normal. Moderate PA requires moderate physical effort and makes the participants breathe somewhat heavier than normal (carrying light loads, gardening, or table tennis, but not walking). Daily walking should last at least 10 minutes at a time, such as walking at work or at home, walking to move from one place to another, and any other walking that one does solely for recreation, sport, exercise, or leisure. The PA volume for each PA intensity (Metabolic Equivalent of Task [MET]-minutes/week) is as follows: vigorous PA volume = 8.0 × vigorous-intensity activity minutes × vigorous-intensity days; moderate PA volume = 4.0 × moderate-intensity activity minutes × moderate days; walking volume = 3.3 × walking minutes × walking days. The total PA volume was defined as the sum of the volumes of vigorous PA, moderate PA, and walking and was further categorized into low, moderate, and high levels according to the guidelines for data processing analysis of the IPAQ. Daily walking time was defined as the weekly average of walking time (walking minutes × walking days) and then classified into 4 categories: daily walking ≤0.5 h, daily walking 0.5–1 h, daily walking 1–2 h, and daily walking >2 h.
Frailty assessment
Frailty at baseline and follow-up was assessed using the Kaigo-Yobo Checklist, a brief questionnaire for screening frailty in older Japanese adults [23]. It consists of 15 items and includes 3 subscales as follows: home boundness, falling, and lower nutrition. Each item scored 1 for those with a risk of frailty (e.g., low ability, has difficulty, and has no friends); otherwise, a score of 0 was given. The total scores ranged from 0–15, and those who had scores of ≥4 were classified to have frailty. Considering the missing data on 15 items, frailty was further defined as follows: (i) participants whose total scores were ≥4 were defined as frail regardless of missing data; (ii) participants whose total scores were 3 without missing data were defined as non-frail; (iii) participants whose total scores were 2 with ≤1 missing data point on any of the items were defined as non-frail; (iv) participants whose total scores were 1 with ≤2 missing data points on any of the items were defined as non-frail; (v) participants whose total scores were 0 with ≤3 missing data points on any of the items were defined as non-frail, and (vi) participants who did not meet any of the previously described criteria were deemed “missing” and excluded from the analysis.
Covariate assessment
The covariates were self-reported through the questionnaires or measured during the home visiting survey. These included demographic factors (age and sex), lifestyle factors (smoking status, drinking status, work status, and living arrangement), and health status (body mass index [BMI], depressive symptoms, and comorbidity index). Smoking status included “never,” “former,” or “current smoker.” Drinking status included “never,” “occasional,” and “heavy drinker.” Occasional drinking was defined as the daily consumption of Japanese sake, beer, Japanese spirits, whisky, and wine (<23 g/d); for heavy drinkers, the daily consumption amounted to ≥23 g/d. The information related to alcohol consumption was obtained from the Brief-Type Self-Administered Diet History Questionnaire, through which alcohol consumption could be estimated by the amount of each alcoholic drink consumed at a time, with its frequency being recorded [24]. Working status included “currently employed,” “retired,” or “never employed.” Living status included “living alone” or “living with others.” BMI (kg∙m-2) was calculated using height and weight measured during the home visiting survey and categorized into 3 groups: <18.5, 18.5–25, and ≥25. If data on BMI were missing, the data input was self-reported BMI calculated from self-reported height and weight in the questionnaires. The short-term Geriatric Depression Scale was used for screening depressive symptoms, and participants with scores of ≥6 were considered to have depressive tendencies. The comorbidity index was defined as the number of historical incidences of hypertension, stroke, heart disease, diabetes mellitus, hyperlipidemia, musculoskeletal disorders, injuries (e.g., fall or fracture), cancer, and depression.
Statistical analysis
Baseline descriptive statistics of the participants are shown as number (percentage) for categorical variables and as the mean (standard deviation) or median (range) for continuous variables according to PA volume and daily walking time; Chi-squared test and Kruskal–Wallis test were used to test between-group differences. Multiple logistic regression was applied to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for subsequent frailty based on PA volume or daily walking time. There were two multi-adjusted models: Model 1 adjusted for age and sex and Model 2 adjusted for age, sex, BMI, smoking status, drinking status, working status, living arrangement, depressive symptoms, and comorbidity index. The presence of a linear trend was tested by placing the continuous variable of PA volume or daily walking time into the unadjusted or adjusted model. A quadratic trend was tested by placing a second-order polynomial of PA volume or daily walking time into the corresponding models. Sensitivity analysis was conducted by excluding participants with a history of cancer (15 participants), stroke (8 participants), or heart diseases (37 participants) because these participants were susceptible to frailty and these diseases might impact their PA practices.
Statistical analysis was conducted using the SAS statistical software package version 9.4 for Microsoft Windows (SAS Institute Inc., Cary, NC, USA). Two-tailed tests were used for statistical analysis, and P-values <0.05 were considered statistically significant.