Changes in rural older adults’ sedentary and physically-active behaviors between a non-snowfall and a snowfall season: compositional analysis from the NEIGE study

We used longitudinal data from the Neuron to Environmental Impact across Generations study (NEIGE study). The methods of this study are described in previous study [18]. Participants were older adults without long-term care living in Tokamachi city, Japan. Tokamachi is a rural city, officially registered as a heavy snowfall region during winter, located in the southernmost region of Niigata Prefecture. Participants were from two areas: Matsunoyama (mountain) area or Tokamachi area (downtown). The mountain area (2.89 m maximum snow depth) had more snow than the downtown area (1.97 maximum snow depth) during winter [18]. The average temperature at Tokamachi city in February 2018 was − 0.9 °C (lowest − 8.3 °C, highest 9.1 °C) [19].

Briefly, in 2017, a total of 1346 residents were recruited from a resident registry, using stratified random sampling. In the non-snowfall season (autumn) of 2017, we conducted a questionnaire survey and health examination of 527 older adults who agreed to enrollment in the NEIGE study; at the same time, they were asked to wear an accelerometer. Of these participants, 381 agreed to also wear an accelerometer during the snowfall season (winter) of 2018.

Habitual time spent in activity behaviors were evaluated by Active style Pro HJA-750C (Omron Healthcare, Kyoto, Japan). Active style Pro is a validated accelerometer [20‐23] and comparable to the devices most commonly used in studies conducted in Western countries [24, 25]. Its measurement algorithm has been explained in detail elsewhere [20, 21]. Participants were instructed to wear an accelerometer over the waist on an elasticated belt for seven consecutive days (except during sleep and water-based activities) during snowfall and non-snowfall season, respectively. In the survey during snowfall season, participants were mailed an accelerometer. No acceleration signal detected for longer than 60 consecutive minutes was defined as “non-wear”. Participants with a wear time corresponding to at least 10 h during waking time per day [26], collected over four or more valid days were included in the analysis [27]. The data were collected in 60-s epochs. Activity behaviors were classified into three intensity categories based on metabolic equivalents (METs): SB (≤1.5 METs), LPA (1.6–2.9 METs), and MVPA (≥3.0 METs) [28, 29].

Participants reported their age, gender, living arrangement (with others/ alone), and self-rated health (very good/ good/ fair/ poor) in autumn. We classified participants between the ages of 65 and 74 years as young-old, and those between ages 75 and 84 years as old-old. Body mass index (BMI) was calculated from height and weight (kg/m²) measured by body composition analyzer MC-780A (TANITA corporation, Tokyo, Japan).

The Chi-Square test or t test was performed to compare participant characteristics. McNemar’s test was used to compare non-snowfall and snowfall season’s proportions of those adhering to physical activity guidelines (i.e. ≥150 min/week of MVPA in bouts of at least 10 min) [1]. A 10 min-bout MVPA was defined as 10 or more consecutive minutes above the moderate intensity threshold, with the allowance of 1–2 min interruption intervals [27].

We described activity behaviors during the non-snowfall and snowfall season using CoDa approach. As all participants spent some time in every behavior there was no need for a method to deal with zeros. Compositional changes [∆SB, ∆LPA, ∆MVPA] were then determined by Aitchison’s perturbation method [30‐32]. The ratios of each component in the composition, such as [SBsnow/SBnon-snow, LPAsnow/LPAnon-snow, MVPAsnow/MVPAnon-snow] for seasonal changes, were calculated and were then divided by the sum of these ratios. An equal composition of these three activities in the non-snowfall and snowfall seasons would result in a compositional change of [1/3, 1/3, 1/3]. Compositional changes were plotted as ternary diagrams, with some significant guide values illustrated.

Of the 381 older adults who agreed to wear an accelerometer during both the non-snowfall and snowfall seasons (response rate: 28.3%), 26 were excluded for; not meeting wearing time criteria (n = 23), bone fracture (n = 1), unreturned accelerometer (n = 1) and malfunction (n = 1). The analytic sample was 355 who had valid accelerometer data.

When compared to participant characteristics between those who had valid accelerometer data in snowfall season and those who did not, a significant difference was found in age groups (young-old: 73.7%, old-old: 59.8%, Chi-Square test). Participants who participated in the survey in the snowfall season were significantly more physically active (approx. 10 more minutes of MVPA per day) than those who did not (t-test).

Table 1 presents the characteristics of the participants. The mean (SD) age was 72.9 (5.4) years (52.4% women) and most of the participants were living with others, < 25.0 kg/m² BMI, and good perceived health. Compared to women, men were more likely to be from the mountain area, be living with others and to have attained ≥150 min/week of MVPA. There were no significant seasonal differences in proportion of those adhering to global physical activity guidelines (non-snowfall season: overall 24.2%, men 33.1%, women 16.1%, snowfall season: overall 21.7%, men 27.2%, women: 16.7%).

Mean accelerometer wear time was 891.7 min/day in non-snowfall season and 886.8 min/day in snowfall season. Older adults spent most of their time on SB and LPA. Table 2 presents the descriptive statistics of time spent in each activity behavior. In men, time spent in each activity (SB, LPA, and MVPA) was 53.9, 40.8, and 5.3%, respectively during the non-snowfall season, and 64.6, 31.6, and 3.8%, respectively during the snowfall season, corresponding to mean seasonal change of 0.445 for SB, 0.287 for LPA, and 0.268 for MVPA (Fig. 1). Since larger distance from 1/3 indicates greater change in time spent in activity, the largest seasonal change was observed in SB, compared to LPA and MVPA. In women, time spent in each activity was 47.9, 47.9, and 4.2%, respectively during non-snowfall season while that was 55.5, 41.0, and 3.5%, respectively during snowfall season, corresponding to mean seasonal change of 0.409 for SB, 0.302 for LPA, and 0.289 for MVPA. Significant unfavorable seasonal changes in activity behaviors were observed in both men and women, but the degree of change was larger in men in every behavior if we contrast the changes (the ratios) between men and women. Similar findings were observed after stratification by residential area, but the degree of change in activity behaviors was larger in mountain area than in downtown, regardless of gender.

Strengths of this study included objective assessment of activity behaviors and a novel statistical approach. Even though there has been increasing research using objective methods [13‐15, 42], no study has treated with consideration to the co-dependence of time spent in activity behaviors. Moreover, we provided results from a seldom-studied elderly population in a rural snowfall area. However, several limitations should be considered to interpret the findings. We may under/overestimate of SB and LPA since Active style Pro cannot provide postural information. Also, some activities (e.g. snow removal and skiing) may be not captured accurately. Second, we may underestimate the decline of activity level since most of the days that participants wore an accelerometer were relatively good weather in spite of heavy snowfall area. It has previously been observed that longer day length is associated with increased device-based physical activity in the older population [42, 43]. Third, as Tokamachi city is a rural area where heavy snowfall is common during winter, it is not necessarily representative of Japanese rural areas with different climates. People in areas with less snowfall may experience a smaller decline of physical activity during snowfall season. More research is needed in the different climate zones from different geographic areas. Finally, selection bias may have occurred. In general, accelerometry respondents have been more active and healthier than non-participating older adults [44]. In this study, those had valid accelerometer data in snowfall season were more physically active than those who did not.

Our findings suggest several implications in terms of both development of interventions to protect against seasonal physical activity decline and physical activity surveillance/ monitoring. Further research regarding how to stay active during winter may be required for health promotion, particularly in regions that experience long winters and with severe weather (e.g. heavy snow). Given that approximately 60% of the land in Japan has snow and a cold winter and that a quarter of the population lives in those areas [45], leading an active lifestyle during winter potentially has a significant public health impact.

Since SB is the most affected by season, it may be better to focus on developing strategies to reduce time spent in sitting. Older adults might be particularly influenced by seasonal outside conditions due to reduced physical functioning and mobility, and spend much of their time indoors. It thus may be effective to provide supplementary resources for indoor activities (e.g. gymnastic exercise programs), sharing of household chores, and making educational/ instrumental supports for safe snow removal [46]. Further approach includes replacing mentally-passive with mentally-active SB may be effective for health, particularly for preventing cognitive decline [47, 48] and depression [49, 50].

As for physical activity surveillance/ monitoring, investigators should be aware of the potential for under- or overestimation of levels of activity especially when the interest is in its between-individual variation, including community-level and country-level comparisons. Seasonality also should be considered when intervention studies are performed.