Participants and Procedures
The target population in this study was older adults aged ≥60 years living in Taipei city, which is the capital of Taiwan. The participants were recruited using convenience sampling from four districts (Nangang, Wanhua, Daan, and Wenshan) of Taipei, from April (spring) to September (autumn) 2018. We used neighborhood broadcasts and local advertisements to recruit potential participants. Interested individuals contacted the study recruiters or neighborhood representatives. The inclusion criteria were older adults aged 60 years or above, able to walk independently (those who were with assistive walking devices were excluded), and community-dwelling (those who were living in institution were excluded). A total of 148 community-dwelling older adults were recruited and receive a baseline assessment. However, according to the condition of the valid data on accelerometer wear time (described later), 126 participants who remained in this study at baseline. At the baseline survey, each participant was administered a structured questionnaire by trained interviewers. Each participant was asked wear an accelerometer for seven consecutive days and completed the lower-extremity performance tests at the first day.
After one year, a follow-up lower-extremity performance test was conducted. A total of 98 of the older adults attended the on-site follow-up examination (follow-up rate: 77.8%). Finally, those with incomplete or missing data in follow-up survey were subsequently excluded (n = 9). Ultimately, a total of 89 participants who provided complete data for the study variables were included in the analysis. The details of the screening procedure of included participants are presented in Figure 1. We obtained written informed consent from each participant at baseline survey. Each participant was received the convenience store voucher worth $15 USD if they completed the baseline and follow-up tests. Ethical approval was received from the Research Ethics Committee of the National Taiwan Normal University (REC number: 201711HM003). The study was conducted in accordance with the ethical guidelines of the 1975 Declaration of Helsinki and all its revisions.
Accelerometer-assessed Daily Steps
Accelerometers (ActiGraph, Pensacola, FL, USA) were used to objectively assess each participants’ daily step counts. The validity and reliability of such triaxial accelerometer have been widely confirmed [15-20]. For each participant, the accelerometer was used to record movement for seven consecutive days (five weekdays and two weekend days). By following the standard methods [20], we used 60-second epochs for all data analyses [21]. We asked each participant to wear the accelerometer on the right side of his/her waist at all times except for water-based activities. Non-wear time was defined as the periods of not less than 60 consecutive minutes of zero counts per minute (cpm), with an allowance of up to 2 minutes of between 0–99 cpm [17]. Participants with at least three valid days (a valid day was defined as at least 600 minutes of accelerometer wear time), including at least 1 weekend day, were included in this study. We utilized ActiLife software 6.0 (Pensacola, FL, USA) to analyze the accelerometer data. In accordance with the aforementioned step-based recommendation for older adults, we categorized the daily steps into “not taking at least 7,000 steps/day” and “taking at least 7,000 steps/day” [12].
Lower-extremity Performance Measures
The five-times-sit-to-stand test is used to evaluate lower-extremity performance [22]. In taking the test, the participants were instructed to rise from a chair (which was 46 centimeters high and armless) to a full standing position and then return to a seated position as quickly as possible for five repetitions. Each participant performed the test two times [23, 24]. The best performance in terms of the total time taken for all five repetitions (that is, the shortest time) was used for our analysis. For the cross-sectional analysis, due to the data of baseline lower-extremity performance was skewness, we used a sex-specific median for dichotomizing the baseline lower-extremity performance of the participants into “better” and “worse” categories. For the prospective analyses, we categorized the lower-extremity performance of the participants into “maintained or improved” and “declined” by calculating the differences in the lower-extremity performance between the follow-up and baseline for each participant.
Covariates
Self-reported demographic characteristics, health-related behaviors, and the presence/absence of chronic diseases were assessed via interviewer-administered questionnaires. The covariates were sex, age group (60-74 or ≥ 75 years), educational level (university and higher or up to high school), marital status (married or not married), job status (with or without a full time job), living status (alone or with others), self-reported health (good or poor), current smoking status, alcohol consumption, balanced diet, hypertension status, blood lipid levels, diabetes status, depression status, and body mass index. Body mass index (BMI) was calculated using self-reported weight and height (categories: non-overweight and overweight) based on the cut-off points for the Asian population (24 kg/m2) [25]. Moreover, accelerometer-measured sedentary time (< 100 counts/minute) and accelerometer wear time were included as covariates as they could confound the relationship between physical activity and health outcomes [26, 27].
Statistical Analyses
Complete data for all the studied variables from 89 older adults were analyzed. The binary logistic regression models were examined (a) the cross-sectional associations of taking at least 7,000 steps/day at the baseline with baseline lower-extremity performance (binary categories: “better” and “worse” based on a sex-specific median); and (b) the prospective association of at least 7,000 steps/day at the baseline and the difference in lower-extremity performance 1-year follow-up (binary categories: “maintained or improved” and “declined”). Three different logistic regression models were conducted to investigate before and after adjusting for other covariates. The first model showed unadjusted analyses (Model 1). In the adjusted regression models, the analyses were first adjusted for sociodemographic characteristics (Model 2), and then further adjusted for health-related behaviors, chronic diseases, and accelerometer wear time (Model 3). Odds ratios (ORs) and their 95% confidence intervals (CIs) were estimated using the binary logistic regression models. We also conducted a sensitivity analysis using the change in lower-extremity performance between baseline and 1-year follow-up to examine the robustness of the prospective associations. Linear regression models were used to estimate the coefficient and their CIs across the three models in the sensitivity analysis. Statistical analyses were conducted using SPSS 23.0 (IBM Inc., Armonk, NY, USA).