Study Design, Participants And Setting
A cross-sectional observational design was adopted. Participants were recruited via responses to messages posted on social media groups (e.g. Facebook sports club groups), forums (e.g. London School of Economics Alumni) and personal contacts. Participants were eligible if ≥ 50 years, fully or semi-retired, owned a smart-phone, and able to attend an enrolment meeting in London or Loughborough, East Midlands (UK). Seventy-three adults met the inclusion criteria and consented to take part. Recruitment spanned February to December 2019. During their enrolment meeting, eligible participants provided informed consent, were familiarised with study procedures and equipment, assigned an initialised Actigraph activity monitor, shown how to download and login to the EMA app and asked to complete a survey.
Measurements and procedures
Ecological Momentary Assessment
Details of activities, contexts and feelings were captured using EMA (Shiffman, Stone and Hufford, 2008). Several studies have demonstrated the feasibility, acceptability, reliability and validity of smartphone-based EMA to measure behaviours and feelings in a variety of older populations (Maher, Rebar and Dunton, 2018; Paolillo et al., 2018; Liu and Lou, 2018). Ethica software (www.ethicadata.com) was selected because it had both Android and IOS versions, making it compatible with most mobile phones. The EMA protocol was piloted in six smartphone users aged 50 + and modifications made.
At the initial meeting participants completed a survey through the Ethica app which asked about age, gender, ethnicity, religion, self-assessed health, highest educational level achieved, whether living alone, whether working part-time or not, number of adults in household, and overall life satisfaction and worthwhileness of life, all of which have been previously identified as confounding factors in analyses of subjective wellbeing (Dolan, Kudrna and Stone, 2017), or associated with choice of leisure-time activities (Galenkamp et al., 2016). Principal sources for question wordings included the Office for National Statistics harmonised wordings for the 2011 Census and the Taking Part Survey questions (Ipsos MORI, 2018)
Over the seven days of monitoring, participants received six prompts at random within 150-minute windows between 06:30 and 21:30 to complete 6 questions about their main activity in the last hour and their feelings of happiness and sense of purpose during that activity. They were instructed to answer prompts immediately but only when safe and convenient. Participants were asked: What was your main activity in the last hour? (21 activities, grouped into six higher-level categories (physical, mental, social, recreational, travel and resting), similar to those used in previous studies of older peoples’ activities (Chang, Wray and Lin, 2014); Galenkamp et al., 2016; Yamashita, Bardo and Liu, 2018)). They were then asked to rate their happiness and sense of purpose on a visual analogue scale by moving a slider from the default setting of 5 to a number between 0 (not at all) and 10 (totally/wholly) (Office for National Statistics, 2015; Cabrita et.al., 2017). Participants were then asked: What was your posture while carrying out this activity? (Standing/Sitting/Lying Down/Moving about). Were you indoors or outdoors while doing it? (Indoors at home/ Indoors another venue/In a vehicle/ Outdoors/Mix of the above). Who were you doing it with? (Alone/With a pet/With people I don’t know/With people I met through this activity/With friends/family/With a service provider (e.g. doctor; shop assistant).
Accelerometry
Participants were asked to wear an Actigraph accelerometer, either GT3x or wGT3X- BT (Actigraph, 2013) over the right hip using an elasticated waist band during waking hours for seven consecutive days. These devices have been identified as having acceptable validity and reliability in older adults (Copeland and Esliger, 2009). Data were collected at a sampling rate of 100 Hz and downloaded in epochs of 60 seconds for analysis using Actilife software. Raw Actigraph data files were reprocessed to derive outcome variables, using custom data reduction software (KineSoft, V.3.3.67, Loughborough, UK). Non-wear time was defined as ≥ 60 minutes of consecutive zero counts, allowing for 2 minutes of non-zero interruptions (Tay, Chan and Diener, 2014). Participants’ accelerometer data was considered valid if they had ≥ 5 days with ≥ 10 hours of valid accelerometer wear (Pruitt et al., 2008). Vertical axis intensity cut-points derived for use in older participants were adopted (Copeland and Esliger, 2009).
Statistical Methods
The unit of analysis is a single response to an EMA prompt. Happiness and purpose outcome measures were regressed on activities undertaken, controlling for confounding influences from the social and environmental context (e.g. where they took place), accelerometer assessed physical activity/time spent sedentary, and participant characteristics. Given the use of multiple regression methods, a key consideration governing sample size was degrees of freedom. Using a two-tailed test and 95% confidence level, assuming 20 independent variables, a sample size of 1,302 responses was sufficient to detect a very small effect size of 0.01 (Faul, F., Erdfelder, E., Lang, A.-G. & Buchner, 2007). To account for the hierarchically-structured the data, the errors of the regressions were clustered on individuals. Given the number of independent variables that could potentially be included in the regressions, multicollinearity was likely. To preserve degrees of freedom, a pragmatic approach was adopted. Accordingly, backwards elimination was used to derive parsimonious estimates, and further independent variables were removed where variance inflation factors exceeded 10, which eliminated most of the confounding variables except for accelerometer assessed sedentary time. All analyses were conducted using Stata version 14.2 (StataCorp, 2018) and statistical significance was set at p < 0.05.