Just-In-Time Adaptive Interventions (JITAIs) are not self-help
By Artur Direito, on 15 June 2018
By Dr. Felix Naughton & Dr. Wendy Hardeman, University of East Anglia
Defining tech that can reach parts humans cannot
Over the last few decades, digital behaviour change interventions have evolved substantially. Most people carry around mobile devices that can run dynamic systems that predict your behaviour and deliver behavioural support in response in real time. For example, if someone is approaching a high-risk situation for engaging in a ‘negative’ health behaviour like smoking such a system can forearm them with some strategies for dealing with that situation e.g. using self-talk: ‘I can handle this!’. Or when someone has an opportunity to engage in a ‘positive’ behaviour like walking instead of taking public transport to their work place they might receive a prompt to do so. Such systems are often referred to as Just-In-Time Adaptive Interventions or JITAIs. They are adaptive, usually based on data collected from sensors in the smartphone, for example accelerometers, or other devices or sources. They also deliver support Just-In-Time or at least close in time to events, situations or moments when an individual is considered to be in need of or most receptive to support.
There has yet to be a clear definition to help identify when an intervention is or is not a JITAI. For example, some descriptions of JITAIs include systems that only deliver Just-In-Time support when requested or triggered by the user, or systems which regularly prompt users during the day to enquire whether they might need support and then deliver support in response (Nahum-Shani et al, 2016; Ann Behav Med). In our view, these do not reflect a pure JITAI, which we consider should be automated and adaptive, and not reliant on the user to trigger support.
Does it matter? We think so. Relying on the user either at the beginning of an intervention or in real time to identify high-risk moments, vulnerabilities or opportunities to guide support delivery is essentially an advanced form of self-help. This differs from a system that predicts when support is likely to be needed and beneficial, using incoming data about the user. These approaches would typically be classified differently when using a ‘mode of delivery of behaviour change intervention’ taxonomy (Carey et al, 2016), with JITAIs categorised as sensor driven. These approaches would also likely differ in their complexity, cost, burden, precision etc. Importantly, users rarely request on-demand Just-In-Time support (e.g. texting ‘HELP’ for craving management advice from a text message system (Naughton, 2017; Nic Tob Res). Users are even less likely to request support when there is an opportunity to engage in a positive health behaviour. Prompt-triggered Just-In-Time support may be used more, however, although this is a less common feature of interventions and less likely to be timely. Another key reason for having a clear definition is to facilitate evidence synthesis about the emerging literature on JITAIs. For this reason we developed a definition of a behavioural JITAI, which contains three key features:
- The intervention aims to provide behavioural support that directly corresponds to a need in real time when the user is at risk of engaging in a negative health behaviour or has an opportunity to engage in a positive behaviour in line with their health goal(s).
- The content or timing of behavioural support is adapted or tailored according to input (data) collected by the system since the support was initiated (e.g., location sensing identifying the user as having an opportunity to be physically active, or identifying the user as in a bar when the behavioural goal is to avoid binge drinking).
- The behavioural support is triggered by the system and not directly by the users themselves (e.g. an unsolicited mobile app notification providing strategies for being active in specific situations or contexts, as opposed to users opening an app or requesting text message support to access strategies to be more physically active).
JITAIs must meet all three criteria. ‘Just-in-time’ interventions meet criteria 1 and 2 but not criterion 3; or they meet criteria 1 and 3 but not criterion 2. Tailored support interventions meet criteria 2 and 3, but not criterion 1.
- Are there additional JITAI features that should be included in the definition?
- How can we best reach consensus about what constitutes a JITAI? We encourage further debate about the classification of different types of digital behaviour change support.
Felix (@FelixNaughton) is a Health Psychologist and a Senior Lecturer in Health Psychology within the School of Health Sciences, University of East Anglia. He has a key interest in the development and evaluation of mobile phone interventions to promote and support health behaviour change (mHealth), particularly those promoting smoking cessation.
Wendy is a Senior Lecturer in Health Psychology, Health Promotion Research Group, School of Health Sciences, University of East Anglia. Her research programme focuses on scalable interventions (e.g., very brief interventions for physical activity) and more intensive interventions targeting at-risk groups, with behaviour change techniques and process evaluation as methodological interests.