Abstract
There has been a lot of interest in digital mental health interventions but adherence to online programmes has been less than optimal. Chatbots that mimic brief conversations may be a more engaging and acceptable mode of delivery. We developed a chatbot, called 21-Day Stress Detox, to deliver stress management techniques for young adults. The purpose of the study was to explore the feasibility, acceptability, and potential efficacy of this low-intensity digital mental health intervention in a non-clinical population of young adults. The content was derived from cognitive behavioural therapy (CBT) and included evidence-informed elements such as mindfulness and gratitude journaling. It was delivered over 21 daily sessions using the Facebook Messenger platform. Each session was intended to last about 5-7 min and included text, animated GIFs, relaxation tracks and reflective exercises. We conducted an open single-arm trial collecting app usage through passive data collection as well as self-rated satisfaction and qualitative (open-ended) feedback. Efficacy was assessed via outcome measures of well-being (World Health Organisation (Five) Well-being Index; WHO-5; and Personal Well-being Measure; ONS4); stress (Perceived Stress Scale-10 item version; PSS-10); and anxiety (Generalized Anxiety Disorder 7-item scale; GAD-7). One hundred and ten of the 124 participants who completed baseline commenced the chatbot and 64 returned the post-intervention assessment. Eighty-one percent were female and 51% were first year students. Forty-five percent were NZ European and 41% were Asian. Mean engagement was 11 days out 21 days (SD = 7.8). Most (81%) found the chatbot easy to use. Sixty-three percent rated their satisfaction as 7 out of 10 or higher. Qualitative feedback revealed that convenience and relatable content were the most valued features. There was a statistically significant improvement on the WHO-5 of 7.38 (SD = 15.07; p < 0.001) and a mean reduction on the PSS-10 of 1.77 (SD = 4.69; p = 0.004) equating to effect sizes of 0.49 and 0.38, respectively. Those who were clinically anxious at baseline (n = 25) experienced a greater reduction of GAD-7 symptoms than those (n = 39) who started the study without clinical anxiety (-1.56, SD = 3.31 vs. 0.67, SD = 3.30; p = 0.011). Using a chatbot to deliver universal psychological support appears to be feasible, acceptable, have good levels of engagement, and lead to significant improvements in well-being and stress. Future iterations of the chatbot should involve a more personalised content.