Dr Alexis Whitton (UNSW)

Abstract

In the face of escalating rates of depression among young people, effective and scalable treatments are urgently required. Although a range of different interventions have been found to be effective, the pivotal question is Which treatments yield the greatest benefit, and for whom?

This talk explores a new approach to answering this question, describing the potential of artificial intelligence (AI)-driven adaptive trials to deliver more efficient and personalised treatments. Compared to traditional randomised controlled trials, AI-driven adaptive trials require fewer participants, reach a conclusion earlier, and can identify interactions between intervention effects and individual characteristics, making them a powerful trial design for personalised treatment research.

The ’Vibe Up’ trial is the first application of AI-driven adaptive trial methodology in digital psychological therapy research. Over 12 sequential ‘mini-trials’, >1200 university students with elevated symptoms of depression, anxiety and stress were allocated to receive one of three digitally-delivered psychological or behavioural therapies – mindfulness, physical activity, sleep hygiene – or an ecological momentary assessment control. AI-driven response adaptive randomisation was used to optimise allocation of participants to each trial arm. Over the 12 sequential mini-trials, an underlying mathematical model learned which intervention was most effective for individuals with different symptom profiles. Results indicated that treatment effects differed significantly between subgroups of individuals, supporting a personalised treatment approach. Importantly, treatment effects estimated by the AI model were found to differ from the clinical predictions made by an independent sample of mental health clinicians.

This talk will distil key insights gained from the Vibe Up study, and spotlight the potential of AI-driven adaptive trials in personalising scalable interventions for common mental health conditions.