Persuasive chatbot-based interventions for depression: A list of recommendations for improving reporting standards

Kerstin Denecke^1*Octavio Rivera Romero^2,3Rolf Wynn^4,5Elia Gabarron^6,7

• ¹Bern University of Applied Sciences, Bern, Switzerland
• ²Institute of Computer Engineering, Sevilla University, Sevilla, Spain
• ³Department of Electronic Technology, Superior Polytechnical School, University of Seville, Seville, Spain
• ⁴UiT The Arctic University of Norway, Tromsø, Troms, Norway
• ⁵Østfold University College, Halden, Østfold, Norway
• ⁶Norwegian Centre for E-health Research, Tromsø, Troms, Norway
• ⁷Department of Education, ICT and Learning, Faculty of Teacher Education and Language, Østfold University College, Halden, Østfold, Norway

Background: Depression is the leading cause of disability worldwide. Digital interventions based on chatbots could be an alternative or complementary approach to the treatment of depression. However, the absence of technical information in papers on depression-related chatbots often obstructs study reproducibility and hampers evaluating intervention efficacy. Objective: This study aims to identify specific characteristics of chatbots for depression and formulate recommendations for improving reporting standards.Methods: In an initial step, a list of items that must be reported was defined based on a previous review on digital interventions for depression, the Behaviour Change Wheel framework, and a taxonomy for defining archetypes of chatbots. To capture the existing knowledge on the development of chatbots for depression, a literature review was conducted in a second step. From the identified studies, we tried to extract information related to the items from our initial list and described in this way the chatbots and their evaluation. As a third step, the findings of the literature review were analyzed, leading to an agreement on a list of recommendations for reporting chatbot-based interventions for depression.Results: The items of the recommendation list for reporting fall into four dimensions: General information; Chatbot-based depression intervention functions; Technical data; and Study. Through a literature review, a total of 23 studies on chatbots for depression were identified. We found that a lot of information as requested by our initial reporting list was missing, specifically regarding the involvement of natural language processing, data privacy handling, data exchange with third-party providers, and hosting. Additionally, technical evaluation details were often unreported in many papers.Conclusion: Studies on chatbots for depression can improve reporting by specifically adding more technical details and chatbot evaluation. Such reporting of technical details is important even in papers on clinical trials that utilize chatbots in order to allow reproducibility and advance this field. Future work could obtain expert consensus on the recommended reporting items for chatbot-based interventions for depression.