Tailoring digital health interventions: Different strategies, different effects
By Emma Norris, on 24 September 2019
By Maria Altendorf – University of Amsterdam, Netherlands
In September 2019, the annual conference of the European Health Psychology Society (EHPS) took place in beautiful Dubrovnik at the Adriatic coast in Croatia. During this conference, the first symposium from the EHPS Special Interest Group Digital Health & Computer-Tailoring (@EHPSDigiHealth) entitled “Tailoring digital health interventions: Different strategies, different effects” was organized and chaired by Eline Smit. The symposium was dedicated to introducing the audience to a wide variety of novel approaches to tailoring in digital health, as well as their different effects on health behavior and health behavior change. The tailoring strategies covered by the presentations in the symposium were: message frame tailoring, mode tailoring, customization, as well as data-driven tailoring. Moreover, a systematic review of tailored digital interventions for weight loss, especially focusing on the differential effects of different tailoring types was presented and discussed.
Image courtesy of @EHPSDigiHealth
From message framing towards message frame tailoring in online computer-tailored smoking cessation communication
The first presenter was Maria Altendorf. She showcased an experimental research on the effects of autonomy-supportively framed smoking cessation messages in an online computer-tailored health communication intervention, compared with messages that used a controlling frame. Earlier research showed that people may be more motivated when they perceive support for their autonomy. Therefore, autonomy-supportive message frames seem to be a promising manner to enhance people’s motivation and eventually change their behaviour, such as to move them towards quitting smoking. In this study, autonomy-supportive messages comprised the use of suggestive language and words as “could” and “would”, as well as offering choice. Controlling messages in contrast made use of commands and words as “must” and no provision of choice. Result showed that participants’ motivation to quit smoking was not significantly enhanced through the autonomy-supportively framed messages. However, on average, participants had a positive opinion about the smoking cessation messages and also perceived high levels of autonomy-support – regardless of the condition they were randomized into. Furthermore, participants reported a need for autonomy, which means that the people participating in our study generally preferred to make their own choices regarding health decisions.
The findings result in many more questions, such as:
- Is it the internet environment that leads to these relatively high levels of perceived autonomy-support?
- Do people with a higher need for autonomy generally perceive higher levels of autonomy-support?
These are some questions that should – and will – be investigated further. More information (in Dutch) about the research project this study was part of, can be found here.
Text, images, video? Tailoring the modality of presentation in online health information for older patients.
Next, Minh Hao Nguyen pointed towards the potential benefit of providing patients with preparatory online information, which can improve patients’ processing of medical information. She presented a randomized controlled trial among 232 younger and older Dutch patients with cancer which aimed to test the effectiveness of a tailored website (vs. non-tailored website) on pre- and post-hospital visit patient outcomes. Different from traditional tailoring research focusing mainly on content, this educational website tailored the information to patients preferences for the mode of presentation (i.e., by self-selecting text, images and/or videos). The findings showed that the mode-tailored website increased website satisfaction and decreased anxiety in younger patients (<65 years). These patterns were not found among older patients (≥65 years). Both the tailored and non-tailored websites were surprisingly well used by patients (on average 34 minutes). Additional analyses showed that higher website involvement and greater website satisfaction explained the level of knowledge patients acquired from the website, before going to the hospital. Higher knowledge, in turn, predicted patients’ recall of information from the consultations with their healthcare provider. To conclude, offering online health information may benefit patients’ information processing. Moreover, offering this information in a tailored manner can be optimal for some patients. Thus, mode tailoring is a promising strategy to optimize the effectiveness of online health information for cancer patients.
Based on these results, Minh Hao recommended future research to further disentangle the active ingredients of tailored websites for different patient populations:
- How, when, and for whom exactly does mode tailoring have an added value?
Does combining mode tailoring with other strategies, such as content tailoring, lead to greater effects that exceed the sum of their individual effects?
Customizable digital environments: can customization in mobile apps support physical activity?
Research showed that health behaviour change can be promoted by allowing users to customize their mobile health apps. Therefore, Nadine Bol investigated in an experiment the question of why and for whom the customization of mobile health apps is most effective. It appeared that participants with a higher need for autonomy increased their intention to engage in physical activity after engaging with a customizable health app, whereas participants with a lower need for autonomy did not. This finding suggests that differences in the need for autonomy should be considered to optimise the impact of mobile health apps.
Based on these results, one could raise the question of how to deal with participants with a lower need for autonomy:
- Is customisation even an option for people with a lower need for autonomy or should they be guided by clear-cut expert advice, as suggested elsewhere?
A systematic review of tailored eHealth interventions for weight loss: a focus on tailoring methodology
Then, Kathleen Ryan provided insights into the current evidence for tailored eHealth weight-loss interventions. She presented results of a systematic review about how tailoring was implemented and whether these tailored approaches were effective in producing weight loss compared with generic or inactive control. Tailoring was carried out in a number of ways, such as based on anthropometric data, health-related behaviours (e.g. dietary intake, physical activity), and participant location. Also, the data for tailoring was acquired from a range of manners, for instance from online questionnaire administration, dynamic gathering of data from web-based diaries, or mobile applications. Overall, tailored interventions were more effective in supporting weight loss than generic or waitlist control groups but effect sizes were very small to moderate, with evidence for fluctuations in effect sizes and differences of effect between tailoring and non-tailoring interventions, and between tailoring types, over time.
Based on these findings, Kathleen developed a model of tailoring depth, which categorized the many different approaches to tailoring used within weight loss interventions and provided a suite of examples of tailoring methods. This model highlighted the concept of ‘tailoring depth’ or the degree to which an intervention is personalized, contrary to viewing interventions as either ‘tailored’ or not, which is unhelpful for determining the effectiveness of different tailoring strategies. Moreover, one could ask:
- What are the mechanisms at play among the fundamental differences behind different tailoring approaches?
Quality assessment of artificial intelligence to tailor a digital health intervention for smoking cessation.
The last presentation was given by Santiago Hors-Fraile, who could not attend the symposium in-person. Via a virtual presentation – matching the digital moto of the event perfectly – he highlighted the potential of health recommender systems based on Artificial Intelligence (AI). In his talk, he showed how health recommendations can be provided that learn over time based on AI, instead of using traditional static decision rules as it is common practice in most computer-tailored interventions nowadays. In his study, he validated this innovative data-driven tailoring approach to assess the quality of an AI algorithm which tailored motivational messages for people willing to quit smoking. Preliminary results show that machine learning-based health recommender systems can be used to provide relevant messages to support smoking cessation patients. Moreover, the used algorithm learnt from patients’ preferences to recommend and tailor health messages and participants stated high levels of satisfaction with the system.
In the general discussion following Santiago’s talk, same questions were posed:
- Will humans (i.e. nurses or coaches) eventually be replaced through computer-tailoring?
- And should these web, mobile app and data-driven approaches be considered the “holy grail” for the future of behaviour change? – Should they? Or do we need empathy and a human being to also motivate us, or even support us in certain situations?
- What would happen to those for whom computerized tailoring is not effective? How could we reach those people?
Maria Altendorf is a PhD-student at the department of Communication Science, University of Amsterdam, the Netherlands. In her PhD research she investigates the effects of message frame tailoring in online health communication for individuals who intend to change their unhealthy lifestyle. Moreover, Maria is interested in the assessment and the implementation of health behaviour change interventions on a broader scale. She is supervised by Prof. Julia van Weert, Dr. Ciska Hoving, and Dr. Eline Smit and currently finalizing her dissertation.
Her research is supported by the Dutch Cancer Society (grant number: KWF 2015-7913).
You can find Maria here:
Linked in: Maria Altendorf (https://www.linkedin.com/in/maria-altendorf-377b2495/)
Research gate: https://www.researchgate.net/profile/Maria_Altendorf