Designing for adherence: modelling use intention in digital mental health tools

• Main Author: Tan, Benny Toh Hsiang
• Other Authors: Vun Chan Hua, Nicholas
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Digital alliance
• Online Access: https://hdl.handle.net/10356/175564
• Institution: Nanyang Technological University

While digital interventions and tools hold the potential to address many of the issues found in traditional approaches to mental health interventions such as accessibility and scalability, the problem of adherence continues to undermine the efficacy of this approach. Of the many approaches to address the issue of adherence, design-based approaches stand out as the most promising. To guide the design of mental health applications, a thorough understanding of the reasons people make use of technology is required. A review of commonly used technology acceptance and use models indicates that numerous issues such as inconsistent predictive power, applicability, and relevance of included constructs, and unnecessary complexity persist. To better understand a user’s intention to make use of technology, we propose the inclusion of digital alliance as a mediator of intention to use, bringing the concept of working alliance from mental health research to the field of technology acceptance and use research. Building on this, we propose a novel technology use model consisting of task-technology fit, technology self-efficacy, perceived usefulness, digital alliance, attitude, and intention to use, arranged in four layers corresponding to external stimuli, cognitive response, affective response and behavioural response, to explain and predict how the design of an application impacts a user’s intention to use. To validate our proposed model, a three-pronged study was carried out. In the first study, structural equation modelling was carried out on data from a survey conducted. Results from the structural equation modelling confirm the theoretical and statistical validity of our proposed model as well as the reliability and validity of our instrument, shedding light on the relationships between task-technology fit, technology self-efficacy, perceived usefulness, digital alliance, and attitude, as well as how these factors influence the intention of university students to make use of digital mental health technologies. Our results also show that our proposed model demonstrated higher variance explained, higher predictive power, and better fit compared to other existing models. To gain a more in-depth understanding of how the constructs proposed in our model apply in the real world, as well as to develop the system requirements for a test application, study two took the form of a qualitative study on the barriers and facilitators of mental health resource utilization among university students. Thematic analysis was applied to responses collected using an a priori framework based on our proposed model. The results of this study indicate the importance of effort expectancy, professionalism, relevance, and role when designing mental health tools for university students. Informed by the findings of study two, a mental health literacy intervention application, MyPocketPal, was developed, and a pilot randomised control trial was conducted to understand the impact, feasibility and acceptability of our model-guided design. Findings from our study indicated that intervention group participants demonstrated statistically significant increases in adherence, engagement, effectiveness measures, as well as acceptability compared to the control group. Findings from this pilot study help establish causal validity, highlighting the potential for using our proposed model to understand and address the issues of adherence to digital mental health tools for university students. Overall, the results from study one and study three converge, providing empirical support for the mechanisms proposed in our model, indicating that the theoretical relationships proposed in our model are not only statistically supported but also have a causal basis.