Ecological Momentary Assessment to Identify Within-Person Influences on Health-Related Behaviours: A Workshop to Inform Cancer Prevention Research
By Emma Norris, on 22 November 2018
By Olga Perski & Claire Stevens – UCL, UK
Ecological Momentary Assessment (EMA), or ‘experience sampling methodology’, uses repeated real-time measures of cognitions, emotions and behaviour over a period of time in naturalistic settings. By virtue of generating temporally rich and deeply contextualised datasets, this method has the potential to increase the validity of research findings and reduce biases associated with retrospective data collection. EMA is increasingly used to study processes at the within-person level. In order to understand and change behaviour, our theories and interventions need to work for individuals, and not just for the ‘average person’ in larger groups. However, EMA comes with a unique set of challenges including study design, recruitment and retention of participants, and data management and analysis. The aim of this workshop, which was held at University College London (UCL) on October 12th 2018 and kindly sponsored by the UCL Grand Challenges Doctoral Students’ Small Grant Scheme (https://www.ucl.ac.uk/grand-challenges/), was to build understanding of EMA as a novel data collection method and to bring together researchers from across UCL and beyond to form cross-disciplinary collaborations. This blog post provides a summary of the key topics discussed during the workshop.
The workshop started with a captivating presentation from Dr. Jaclyn Maher (https://kin.uncg.edu/about-us/our-faculty/jaclyn-maher/), who is Assistant Professor at the University of North Carolina at Greensboro (Presentation slides: Jaclyn Maher UCL EMA Workshop). Jaclyn gave a very clear overview of what EMA entails, highlighting that it can improve the ecological validity of findings, reduce recall bias and allow for the analysis of psychological and behavioural processes over time. She also pointed out that traditional methods tend to aggregate and summarise participants’ experiences over a given period of time (e.g. average daily step count over 2 weeks), but that an individual participant’s mean may not accurately represent their behaviour on any given day. The core of Jaclyn’s presentation was on the 3 primary ways in which EMA can help advance our understanding of health behaviours (Dunton, 2017), using different examples from her own research to illustrate these. The first is the concept of sequentiality, which allows researchers to ask questions related to the temporal sequence of antecedents to, and consequences of, health behaviours. A key question that Jaclyn and colleagues addressed in one of their research studies was whether there are bi-directional relationships between momentary affective and physical feeling states and physical activity in older adults. They found that on occasions where older adults engaged in more stepping and standing behaviours than was typical for them in the 15 minutes before the EMA prompt, they tended to report greater feelings of energy at the prompt. However, the reverse was also true: on occasions when older adults reported greater feelings of energy than was typical for them, they tended to engage in more stepping and standing behaviours in the 15 minutes following the EMA prompt. Interestingly, these patterns were not observed for positive affect. The second concept is that of synchronicity, which allows researchers to study the extent to which explanatory factors co-occur in time and space with health behaviours. For example, does being alone or with others influence a person’s affective experience during physical activity and sedentary behaviour? The third concept is that of instability, which allows researchers to study the extent to which fluctuations in explanatory factors influence health behaviours. For example, Jackie and colleagues assessed whether within-person variability in affective and physical feeling states influenced the probability of meeting physical activity guidelines (Maher et al., 2018).
We then listened to a very engaging presentation from Dr. Dan Powell (https://www.abdn.ac.uk/iahs/research/health-psychology/profiles/daniel.powell), who is Lecturer at the University of Aberdeen (Presentation slides: Dan Powell UCL EMA Workshop 12.10.18). The overarching theme of Dan’s presentation was on threats to data quality in EMA studies, particularly with regards to issues of task attention and the temporal sensitivity of biological measures, and how to mitigate these. The first example was an EMA study which assessed whether within-person fluctuations in inhibitory control (a facet of executive functioning) are associated with unhealthy snacking behaviours. To measure inhibitory control, Dan and colleagues developed a Go/No-Go test, which was deployed via a wrist-worn smart device. Participants were prompted hourly between 7am and 10pm for 7 consecutive days to complete the real-time Go/No-Go test and a short measure of snacking consumption. Results from this study indicated that within-person fluctuations in inhibitory control (i.e. 100 ms slower reaction times than usual) were associated with increased snacking consumption in the subsequent hour (Powell et al., 2017). However, this study also highlighted the importance of anticipating issues of non-compliance prior to conducting your EMA study, as such issues are otherwise likely to crop up during the analysis stage. When looking at the data from this study, Dan noticed that some participants had multiple data entries per Go/No-Go trial, indicating that they might have initiated the task when prompted, but instead of paying attention to it, they mindlessly tapped on the smart device until the task was completed. This response pattern was labelled ‘multi-tappers’, and a procedure was subsequently developed for how to best manage data from these trials within the main analysis. Dan emphasised the importance of carefully piloting your task and exploring what your data may look like given different response patterns so that you can set pre-specified rules for where a test result is insufficiently or inappropriately completed.
Dan also talked us through an example from his PhD research, which assessed whether fatigue severity in individuals diagnosed with multiple sclerosis is associated with an attenuated cortisol awakening response. This study involved 4 consecutive days of EMA prompts, which assessed momentary fatigue severity ratings and salivary cortisol at 0, 30 and 45 minutes post-awakening. As timing is crucial in salivary cortisol studies, Dan ensured that participants collected samples as promptly as possible when asked by using the EMA device as an alarm clock, which presented a time-limited code to participants. This code could then be transferred to the plastic tube used to store each saliva sample to ensure that these were collected within the desired time window (Powell et al., 2015). However, Dan pointed out that a remaining challenge in this type of EMA study is the possibility of spontaneous awakenings.
These two thought-provoking presentations were followed by a moderated panel discussion, which covered areas of how to determine adequate sample sizes and sampling frequencies in EMA studies, how to evaluate study quality (e.g. risk of bias) in EMA studies and the issue of ‘measurement reactivity’, which refers to the phenomenon that frequent measurement of psychological and behavioural processes might in fact alter those processes (Wilding et al., 2016). This was followed by a networking reception, in which attendees discussed current EMA projects and ideas for future research and cross-disciplinary collaborations.
Some broader questions arising from these discussions were:
· In what situations can EMA provide insight that other methods cannot? Is it more suited to some research questions that others?
· What are the key barriers to implementing EMA methods in health research?
This workshop was organised by Claire Stevens (Department of Behavioural Science and Health) and Dr. Olga Perski (Department of Clinical, Educational and Health Psychology).
Claire Stevens’ PhD work is focused on exploring whether population based cancer screening programmes provide a teachable moment for health behaviour change. This includes determining whether behaviour change occurs spontaneously following participation in cancer screening, and whether cancer screening is a suitable opportunity to provide advice, and deliver interventions to improve a range of lifestyle related cancer risk factors: @claire_stevens_ email@example.com
Dr. Olga Perski’s research is cross-disciplinary in scope, drawing on theories and methodologies from behavioural science, computer science and human-computer interaction to develop and evaluate digital behaviour change interventions for smoking cessation and alcohol reduction. Specifically, her doctoral work was focused on the conceptualisation, measurement and promotion of ‘engagement’ with digital behaviour change interventions: @OlgaPerski firstname.lastname@example.org