JMIR Publications

ABSTRACT

Globally, the need for mental health care greatly surpasses the availability of services, especially in low- and middle-income countries (LMICs). Adolescents living with HIV (ALWH) are especially vulnerable because their developmental life stage predisposes them to higher rates of mental health morbidity compared to other age groups, which, left untreated, can also lead to suboptimal HIV outcomes. However, due to a lack of mental health services, most adolescent mental health morbidity—especially in LMIC—goes unaddressed. Among the most prevalent mental health conditions experienced by ALWH is depression, which often initially presents with mild to moderate symptoms that, if addressed early on, can prevent escalation to more severe presentations of depression. Such early interventions for depression include depression education, learning self-help skills, and understanding when depressive symptoms require specialized treatment by a mental health professional. Chatbots, which are increasingly used in both commercial and healthcare settings, present an opportunity to deliver depression education and self-help skills to ALWH in a highly scalable and cost-effective manner. As part of a larger study to explore the acceptability and feasibility of a mental health chatbot for ALWH, this tutorial explains the iterative, Human-Centered Design (HCD) and programming of a chatbot named “EVA,” which was co-developed with ALWH in Peru. Using the chatbot programming platform Smarbot360 and guided by the principles of HCD, we began by collecting preliminary data on the perceived acceptability of a mental health chatbot among ALWH. Next, we convened a Youth Advisory Board (YAB), comprised of six ALWH, who met monthly for five months to guide, test, critique, and finally approve a final version of the chatbot EVA for future testing. During the iterative development phase, ALWH identified key topics of interest and the preferred methods for presenting information in the chatbot. For example, in addition to depression information and self-help skills, ALWH requested that EVA include information on stigma, bullying, and eating disorders. Furthermore, emphasis was placed on using empathetic and friendly language that is both motivating and encouraging. Despite some challenges with meeting all the YAB’s desired chatbot characteristics, their participation allowed us to build a mental health chatbot that was highly responsive to their stated needs and preferences. We demonstrate that tailored chatbots like EVA can be easily and rapidly developed by teams with no specific expertise in chatbot programming and, importantly, may eventually be used to address the crucial mental health service gap experienced by ALWH.