Assessing activity levels of individuals in a large, self-selecting dataset Nik Lomax and Michelle Morris University of Leeds ECQTG | York | 8 September 2017 

Context: sedentary populations • Obesity is a global health concern, adult obesity levels increasing almost universally. • Net-cost of healthcare and welfare for overweight and obese people in the UK is £2.47 billion per annum. • The use of activity tracking devices, as an intervention to decrease sedentary behaviour is well researched. • Many previous studies use relatively small samples. 

Bounts: lifestyle app data 

Bounts: lifestyle app data • App built by Active Inspiration Technologies • >500k unique users • 13 months activity data • Continuously updated • Daily user activity including step count • GPS trace data 

Gender differences • What does the literature tell us? • Women are more likely to use a fitness tracker than men • In our sample? • True – in the Bounts data 75% users are female. 

Steps by gender • What does the literature tell us? • On average, men take slightly more steps than women each day. • In our sample? • True – but fairy large difference. Women = ~6k. Men = ~7.5k. 

Enough steps? • 10,000 steps per day often used as a target (despite criticism, e.g. Tudor- Locke and Bassett 2004)1. • Large variation in Bounts data but many not achieving ‘enough’ steps. 1Tudor-Locke, C. and Bassett, D.R., 2004. How many steps/day are enough?. Sports medicine, 34(1), pp.1-8. 

Steps by age • What does the literature tell us? • Number of steps taken declines as age increases. • In our sample? • False. We see a surprising positive correlation between age and number of steps. 

Seasonality and marketing • What does the literature tell us? • Levels of physical activity peak in summer; energy expenditure decreases in winter. • In our sample? • Some evidence of seasonal difference and a ‘holiday period’ effect. Clocks go back Clocks go forward Holiday lull? 

Seasonality and marketing • What does the literature tell us? • Levels of physical activity peak in summer; energy expenditure decreases in winter. • In our sample? • Some evidence of seasonal difference and a ‘holiday period’ effect. 

Seasonality and marketing • But patterns better explained by changes in marketing and terms/ conditions of app use? TV campaign Ts&Cs/ user experience change 

Geographical variation • Many of the areas averaging over 7,000 steps per day are peripheral or coastal areas. • The most active areas can be seen in Outer London, e.g. Sutton and Enfield. Partnership with Liverpool Council 

Key findings and message • Activity patterns and levels in these data are different from the general population (i.e. findings reported in previous studies). • Bounts participants are self-selecting and likely motivated to be more active. • Encouraging people to use a rewards scheme/ track their activity levels can help increase activity levels and decrease risk. 

Further work • Investigation of other activity types (gyms, tennis clubs, etc.) in relation to user demographics and seasonal activity. • Monitor effect of partnerships and marketing (e.g. links with Liverpool City Council). • Investigation of activity spaces and obesogenic environments using GPS trace data. 

Assessing activity levels of individuals in a large, self-selecting dataset Nik Lomax and Michelle Morris University of Leeds [email protected] ECQTG | York | 8 September 2017