Article Text

Download PDFPDF

Day length and weather conditions profoundly affect physical activity levels in older functionally impaired people
  1. D Sumukadas1,
  2. M Witham1,
  3. A Struthers2,
  4. M McMurdo1
  1. 1
    Section of Ageing and Health, University of Dundee, Dundee, UK
  2. 2
    Department of Clinical Pharmacology, University of Dundee, Dundee, UK
  1. Dr D Sumukadas, Section of Ageing and Health, Division of Medicine and Therapeutics, Ninewells Hospital and Medical School, University of Dundee, Dundee DD1 9SY, UK; d.sumukadas{at}nhs.net

Abstract

Background: Regular physical activity is vital for maintaining the health and independence of older people. Few objective data exist on the effect of weather on physical activity levels in this group. The objective of this study was to evaluate the effect of weather using an objective measure of physical activity.

Methods: This was a retrospective study of 127 participants, >65 years old, who were enrolled in a previous randomised controlled trial. The main outcome was daily activity counts measured using the RT3 triaxial accelerometer over 1-week periods. These were correlated with local weather data including daily maximum temperature, sunshine, precipitation and wind speed that were obtained from the metrological office.

Results: The mean age of the subjects was 78.6 years; 90/127 were female; and 720 usable daily counts were obtained for the 127 participants. The mean daily counts showed a striking seasonal variation, with maximum activity in June and minimum in February (137 557 vs 65 010 counts per day, p<0.001). Day length, mean maximum temperature and mean daily sunshine were able to explain 72.9% of the monthly variance in daily activity levels. Daily counts showed moderate correlation with day length (r = 0.358, p<0.001), maximum temperature (r = 0.345, p<0.001), duration of sunshine (r = 0.313, p<0.001) and rain (r = −0.098, p = 0.008) but not with wind speed (r = 0.093, p = 0.12). Multivariate analysis showed that day length, sunshine duration and maximum temperature were independent predictors of daily activity (adjusted R2 = 0.16).

Conclusions: Physical activity levels among older people are much higher in summer than in winter. Day length, sunshine duration and maximum temperature have a significant influence on physical activity levels.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Low physical activity levels are a common problem in older adults,1 but, despite extensive research, few interventions have been shown to successfully achieve sustained increases in everyday activity levels in older people.2 The determinants of physical activity are thought to be manifold in older people, encompassing physical, social, psychological and environmental domains.36 Although these broad categories are well established, there is considerably less information on the features within each domain that determine physical activity.

Research on the determinants of physical activity in older adults has often been based on subjective measures of physical activity, particularly self-report such as recall diaries. These are known to be highly inaccurate,79 and do not provide a firm foundation for in-depth exploration of the determinants of activity. Furthermore, such methods lack the temporal resolution to allow examination of changes in activity on a day-by-day basis.

Commonsense suggests that changes in the weather should be associated with changes in activity levels. Although there are a number of studies that have examined the association between weather and activity,10 few have been carried out in older adults, many have used crude measures of activity or weather, examining for example season rather than day-to-day variations in weather, and fewer still have used objective measures of physical activity.11

The purpose of this analysis was therefore to examine the relationship between objectively measured physical activity in older adults and weather on a day-to-day basis.

METHODS

We recently conducted a randomised double-blind placebo-controlled trial examining the effect of the angiotensin-converting enzyme inhibitor perindopril on functional capacity in older people without heart failure.12 Ethics approval was obtained from the Tayside Committee on Medical Research Ethics and the study conformed to the principles of the Helsinki Declaration. We recruited 130 people over the age of 65 years with self-reported limitations in activities of daily living between October 2003 and March 2006. Participants initially attended a screening visit, at which time baseline outcome measures were recorded. If eligible, they were randomised to receive either perindopril or placebo for 20 weeks and the outcomes were re-measured at 10 and 20 weeks. The principal outcome measure of the original study had been a change in 6 min walking distance, a submaximal test of exercise capacity in older people,13 and secondary outcomes were changes in physical activity measured by accelerometry (RT3), the timed-get-up-and-go test;14 the sit-to-stand test;15 self-reported disability using the Nottingham Extended Activities of Daily Living scale;16 and self-reported quality of life using the EuroQol 5-D questionnaire.17

The RT3 Triaxial Research Tracker (Stayhealthy Inc. USA) measures movement along three orthogonal axes. Acceleration is measured periodically and stored as an activity count in memory. It was used in the 7-day mode to record physical activity counts at intervals of 1 min and participants were asked to wear the accelerometer over the anterior aspect of the hip for a week during waking hours. Triaxial accelerometry has been validated for use in an older population,18 and the RT3 accelerometer is able to distinguish between walking and sedentary tasks in older people.19 Of the total number of participants from the randomised controlled trial 127 had RT3 data available and these were used in the present study.

Data from the RT3s were downloaded as Microsoft Excel files. Activity counts were summed over 24 h periods (midnight to midnight) with the first and last days discarded as data were incomplete for these time periods.

Weather data

Weather data were obtained from the Scottish Meteorological Office (Edinburgh, UK) archive. Daily weather data, including temperature and wind speed at 09:00, precipitation in the previous 24 h and maximum and minimum temperature in the previous 24 h, are recorded daily. We used data recorded at the Invergowrie weather station, on the outskirts of Dundee, UK. All participants lived within a 20-mile radius of this weather station.

Statistical analyses

Daily count data were entered into an Excel spreadsheet, along with baseline demographic data and daily weather data for each day on which count data were recorded. Analyses were performed using SPSS V.15 (SPSS, Chicago, USA). Univariate and multivariate regression analyses were performed using the linear regression function (p<0.05 to enter, stepwise). Weather variables affect within-person variations in daily activity as well as explain differences between individuals. Single-level models of the effect of weather variables may therefore give misleading results.20 Multilevel modelling was performed in order to further explore the contribution of variances not only within an individual but also between individuals. We used the SPSS mixed linear function, with the subject number as the second-level domain to estimate between individual variance. Restricted maximum likelihood estimation was used.

Maximum temperature and sun duration showed no evidence of non-linearity in their relationship with daily activity. For analysis of precipitation and wind, we performed a series of tests dichotomising precipitation at 5 mm intervals, and wind at 5-knot intervals, to examine whether a threshold effect on activity levels was evident. No such effect was seen; no dichotomisation point provided superior correlation with physical activity compared with a standard bivariate correlation and we thus included these variables as linear variables.

To test how much of the variation in activity with change in day length was due to changes in duration of activity (ie, being active for a longer proportion of each day) and how much was due to increased intensity of activity, we calculated the number of minutes in which non-zero counts were recorded by the activity monitors (minutes of activity per day). The mean daily count per minute of activity was then calculated by dividing the total daily count by the minutes of activity.

RESULTS

Baseline daily activity records suitable for analysis were obtained from 127 participants of the original study. A total of 720 days of physical activity was recorded by these 127 patients (mean 5.6 days per patient). Demographic details of the 127 patients are given in table 1. Day length varied from 6.6 to 17.4 h; the mean daily maximum temperature was 13.5°C (range −0.1 to 26.4). The mean duration of sunshine was 4.3 h (range 0 to 14.3), mean wind speed at 09:00 was 5.5 knots (range 0 to 28) and the mean precipitation per 24 h was 1.7 mm (range 0 to 30.8).

Table 1 Baseline patient details and correlation with day length at the time of recruitment

Association with month-to-month variation in daily activity

Comparing daily activity counts with the month of acquisition revealed a striking variation (fig 1), with mean activity counts reaching a maximum of 137 578 (SD 72 695) counts/24 h in June and a minimum of 65 011 (SD 22 715) counts/24 h in February. Strong correlations were seen between mean 24 h counts per month and mean day length, mean maximum daily temperature and mean daily sunshine (table 2). Combining these three factors in a stepwise linear multivariate analysis explained 72.9% of the month-to-month variance in daily count levels (adjusted R2 = 0.729).

Figure 1

Mean 24 h activity counts versus month of acquisition.

Table 2 Univariate correlates of daily activity counts

Association with day-to-day variation in daily activity levels

Univariate regression analyses were performed, regressing daily counts against day length, duration of sunshine, amount of rain, wind speed recorded at 09:00 and maximum temperature. Results are shown in table 2. Stepwise linear multivariate regression analysis revealed that day length, maximum temperature and sunshine duration remained in the model as independent associates of daily activity counts (adjusted R2 = 0.16) on this analysis.

Association with within-person variation in daily activity levels

Because day length was closely correlated with both maximum temperature (r = 0.76) and hours of sunshine (r = 0.50), we calculated the within-patient percentage difference in exposure to both sunshine and maximum temperature on a day-by-day basis. As day length differs minimally over the week-long activity recording period, this analysis provides an additional way of separating the effect of weather conditions from that of day length. Univariate correlations are given in table 2. In stepwise multivariate linear regression analysis, variations in both sunshine duration and maximum temperature were independently associated with within-person variation in daily activity counts (adjusted R2 = 0.046).

Multilevel modelling of weather effects

We confirmed our findings by constructing a multilevel model, with daily activity readings as the level 1 domain and different patients as the level 2 domain. Day length and weather variables were then added as covariates. Day length, maximum temperature and duration of sunshine were significantly associated with daily activity in this model. The estimated increase in counts attributable to weather factors were 3207 (95% confidence interval (CI) 703 to 5710, p = 0.012) per hour of day length; 1481 (95% CI 235 to 2727, p = 0.02) per degree Celsius increase in maximum temperature; 1914 (95% CI 1047 to 2782, p<0.001) per hour increase in duration of sunshine; −519 (95% CI −1355 to 316, p = 0.22) per millimetre increase in precipitation and 331 (95% CI −276 to 938, p = 0.29) per knot increase in wind speed.

Contribution of day length to intensity of activity

There was little change in the minutes of activity recorded per day across the year, with a peak average of 672 min of activity per day in September, and a trough average of 578 min of activity per day in February (p<0.001, Student t test). By contrast, counts per minute of activity (activity intensity) varied markedly through the year, with a peak average of 215 counts per minute in July and a trough average of 107 counts per minute in December (p<0.001, Student t test). Intensity of activity, as opposed to minutes of activity, accounted for 85% of the variance in daily activity counts.

Within-person changes over time

Follow-up accelerometry data were available at 5 months for 79/127 (62%) of participants. In the original trial, daily activity counts were not affected by perindopril therapy; data from both arms of the study are therefore examined here. Out of a total of 79 patients, 31 (39%) were enrolled during the summer months (May to August). These patients had a mean of 149 264 counts/24 h at enrolment, falling by a mean of 22 260 counts/24 h at follow-up 5 months later. Patients with counts above the median demonstrated a fall in counts at follow-up compared with those with counts below the median (−45 360 vs 2379, p = 0.007, Student t test). Of the total of 79 patients 23 (29%) were enrolled during the winter months (November to February). These patients had mean daily counts of 77 261/24 h at enrolment, rising by a mean of 7932 counts/24 h. Patients with counts below the median at enrolment showed a similar rise in counts to those with baseline counts above the median (7389 vs 8430, p = 0.93).

Correlation between baseline characteristics and day length

To examine whether our findings were due to changes in the type of patient recruited in summer compared with those recruited in winter, we compared mean day length at the time of the baseline assessment with a range of other baseline variables. Results are shown in table 1.

DISCUSSION

Our results demonstrate a striking difference in older people’s objectively measured daily activity in summer compared with winter, with patients enrolled in summer being twice as active as those enrolled in winter. Day length, duration of sunshine and maximum temperature were all independent associates of variation in daily activity, both at the level of the individual and between individuals. Furthermore, we have shown that the variation in daily activity with day length and weather is not simply due to being active for a longer period of time each day; it is the intensity of activity that rises in summer rather than the duration of activity.

Many factors are likely to influence activity levels, and models to explain variation in activity levels typically include domains such as physical function, environmental variables, psychological variables and social variables.21 In most studies, a large proportion of the variance in daily activity remains unexplained,22 suggesting that other factors, such as weather conditions, may play a part. At the level of the individual, changes in weather make a small (5%) but significant contribution to daily changes in activity. Factors such as changes in daily routine, daily variation in symptoms and daily variation in mood are likely to explain additional intra-individual variation. Between individuals, a somewhat larger proportion (16%) of the differences in activity were attributable to season and weather; other factors such as physical and psychological function, environment and social habits are likely to explain additional variance at this level. When considering changes in activity at group level through the year, day length and weather explained the majority of the variance (73%).

There have been few studies of the influence of weather and day length on activity in older people, and many of these studies have focused on large time periods, for example the difference between winter and summer.23

Could our results be due to fitter people being recruited in the summer than those recruited in the winter? It would seem more likely that wintertime would be the time when fitter people would be recruited; one would expect less fit people to be more prone to illness in winter and thus less likely to wish to enter trials. When followed up, the changes in physical activity over time are of a smaller magnitude than would be expected from baseline cross-sectional data. Differences in patient characteristics between summer and winter could partly explain this. Baseline measures of physical and psychological function demonstrate changes with season, but other variables (eg, age, creatinine clearance, medication number) that can be viewed as surrogates for illness burden did not vary with season. It is not possible from our dataset to ascertain whether the variations in physical and psychological function are occurring because of changes in daily activity (leading, for example, to deconditioning in winter) or whether they represent changes in the patient populations recruited at different time points.

The strengths of our study are that we used an objective measure of physical activity which has been validated in older people and that we obtained data on both weather and activity on a day-by-day basis, giving far better temporal resolution than has been obtained in most other studies of the effect of weather or season on activity. Furthermore, data were collected over more than one cycle of seasons. Our study population was older than in most studies to date, with prospective definition of baseline physical and psychological functions.

There are several limitations to our study. We studied patients recruited to a randomised controlled trial. Such patients may not necessarily be representative of older patients in the wider community, although the trial was designed to recruit patients with functional impairment. The study was not designed to examine changes longitudinally; ideally, monitoring activity levels over at least a 12-month period would give more accurate data as to how much of the variation we observed was due to season and weather effects on the individual. We cannot tell from our data what activities people undertook during summer that they did not undertake during winter; further work examining this topic is necessary, as is work examining how weather conditions interact with other physical and psychological variables to mediate changes in activity levels. Our dataset did not include enough days with thunder or snow to test their effects; changes in humidity were also not sufficient to merit examination of this variable. The RT3 accelerometer is known to show inter- and intra-monitor variability that increases with an increase in intensity of activity.24 However, as many older people function at lower intensities of activity, this may not affect activity counts to a major extent. Although accelerometry correlates reasonably well with gold standard measures of physical activity such as doubly labelled water measurement, it is known to underestimate energy expenditure for some types of activity, for example activity using the arms or climbing stairs and gradients.25

We have shown a potentially important link between seasonal and weather effects and daily activity in older people. Further work is needed to examine the effects of weather on individual activity patterns over a longer time period, but our results suggest that researchers and clinicians should adjust for weather and day length when measuring changes in activity levels in response to interventions in older people. It is not possible to change the weather conditions, but finding ways to ameliorate the impact of adverse weather conditions may be a fruitful way of boosting daily activity levels in older people. We have shown that, although the time spent in daily activity is slightly lower in winter than in summer, the intensity of activity is much lower in winter months. Indoor leisure facilities, better transport links and working with older people to modify their daily routine or increase their leisure activities in winter are just some of the ways that could potentially have an impact and deserve closer study.

What is already known on this subject

  • Physical activity is critically important for disease prevention and the maintenance of independence in older adults.

  • Previous studies on physical activity levels and weather have focused on predominantly younger populations.

  • Previous studies have used self-report measures of physical activity, which are notoriously inaccurate.

What this study adds

  • Day length and weather conditions had a profound effect on objectively measured physical activity levels in older people.

  • This has important methodological implications for the timing and interpretation of all future interventions and strategies aimed at promoting physical activity participation.

Acknowledgments

With thanks to all the patients involved in the study, and the staff of the Royal Victoria Day Hospital, Dundee, UK. DS performed the original trial, assisted with data analysis and co-wrote the paper. MW helped design the trial, collected weather data, performed the analyses and co-wrote the paper. AS and MM co-designed the trial, and critically revised the manuscript. MM is the guarantor of the article.

REFERENCES

Footnotes

  • Competing interests: None.

  • Funding: This study was funded by the Chief Scientist Office, Scottish Government. The funding source had no role in the study design, collection or interpretation of data, writing the report or in the decision to submit the paper.

  • Ethics approval: Approval was obtained from the Tayside Committee on Medical Research Ethics, Dundee, UK.