Exploring within-subject associations with mixed effects models: How does weather affect physical activity?

Physical activity is one of the most important factors for human health, but around 28 % of the world's population does not meet the World Health Organisation's recommendations. Physical activity can be increased in everyday life, for example through exercise or walking, with weather being a facilitator or a barrier. Due to climate change, extreme weather events are becoming more frequent, highlighting the relevance of further research into the associations between weather and physical activity. The future increase in heat waves may become a problem, particularly for vulnerable groups. Current research shows that higher air temperatures, more daylight hours, no precipitation and low wind speeds are associated with more physical activity. The present study aims to investigate these associations at the within-subject level (i.e., differences within individuals over time) by further exploring individuals’ fitness level as a moderating factor.

The Days in Motion study examined 676 participants (aged 18-80) from 338 heterosexual couples living in Berlin (Germany) at a baseline week and a 12-month follow-up week between 2013 and 2015. This secondary analysis used data from a sub-sample of 228 participants who did not receive an intervention, provided longitudinal data, and did not report to be ill, injured or on vacation during the assessment period. Minutes of moderate-to-vigorous physical activity were measured using accelerometers. Hourly physical activity data were linked to meteorological variables (air temperature, relative humidity, precipitation, wind speed, sunshine duration and cloud cover) from a central weather station in Berlin. Physical fitness level was assessed objectively by the 2-km indoor walking test. Mixed effects models were used for the analysis, with the following data structure: hours nested within days, weeks, and persons. Between-subject (i.e., differences between individuals) and within-subject associations can be analysed in mixed effects models.

Significant between-subject associations occurred only for wind speed, with higher mean wind speed reducing physical activity. Positive significant within-subject associations with physical activity were found for air temperature and sunshine duration, whereas a significant negative association was found for precipitation and no association for cloud cover. However, the effects were moderated by the fitness index. In the future more data is needed during heat waves to be able to adequately analyse when physical activity might start to decrease again, or when physical activity in heat might have a detrimental effect on health.