Conversational Agents for Chronic Disease Self-Management: A Systematic Review (Preprint)

Abstract

BACKGROUND Conversational agents have the potential to deliver patient-centered interventions that focus on goal setting, feedback, and education. However, little is known how these agents may be utilized to assist patients with chronic disease self-management. OBJECTIVE The objective of this study was to systematically review the literature to assess how conversational agents have been used to facilitate chronic disease self-management and to identify gaps in the evidence base. METHODS The systematic review utilized Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines across five databases. We included full-text journal articles or conference proceedings that contained primary research findings for text-based conversational agents focused on self-management for chronic diseases in adults. Two reviewers independently extracted information from the included studies, analyzed the data, grouped the studies into themes, and rated the quality of the studies. RESULTS 1,606 studies were identified, and 12 studies met inclusion criteria. There were five randomized controlled trials, five quasi-experimental studies, and two non-experimental studies. Studies were heterogeneous in design and many lacked methodological rigor. Participants mostly reported positive attitudes towards conversational agents, and there were improvements on the Patient Health Questionnaire (P<.05), Generalized Anxiety Disorder Scale (P=.004), Perceived Stress Scale (P=.048), Flourishing Scale (P=.032), and Overall Anxiety Severity and Impairment Scale (P<.05) between the conversational agent intervention and control groups in three randomized controlled trials. Major themes related to user perceptions were personalization, engagement, and social support. The majority of conversational agents were based upon theoretical grounding, but few studies reported using established design principles and no studies used heuristic evaluation. CONCLUSIONS Our findings suggest that conversational agents are acceptable and have the potential to support self-management particularly for mental health conditions. As this is an emerging area of research, efficacy of long-term use of conversational agents for chronic disease self-management is unknown but shows promise. Future studies should assess the characteristics of agents that may be most useful for self-management based on the patient’s clinical and sociodemographic characteristics, motivation, and preferences.

A Systematic Review on Healthcare Artificial Intelligent Conversational Agents for Chronic Conditions

“…In the last decade, CAs’ popularity has increased, particularly those that use unconstrained natural language [ 3 , 4 , 5 ]. For example, consumers can talk to CAs on their smartphones for daily tasks, such as managing their calendars and retrieving information [ 6 , 7 ].…”

“…Hence, CAs can play an useful role in health care, improving consultations by assisting clinicians and patients, supporting consumers with behavior change, and assisting older people in their living environments [ 13 , 14 , 15 ]. They can also help in completing specific tasks such as self-monitoring and overcoming obstacles for self-management, which is important in chronic disease management and in the fight against pandemics [ 6 , 16 ].…”

“…Some research studies have shown advantages from the use of AI-enabled CAs in different healthcare settings, such as enabling behavior change, coaching to support a healthy lifestyle, helping breast cancer patients, and self-anamnesis for therapy patients [ 7 , 22 , 23 ]. Prior systematic literature reviews explored a variety of CAs in general health care [ 1 , 6 , 24 ] and aspects of the personalization of health care chatbots using AI [ 25 ]. However, there is little evidence on the use of AI-based CAs in chronic disease health care.…”

Tailoring coaching conversations with virtual health coaches

“…Many studies with conversational agents, especially when they have a physical embodiment or are computationally heavy, still take place in the controlled settings of a lab (Griffin et al, 2020;. Such evaluations in controlled conditions provide valuable information and should be performed during development, for example, to investigate relations between personalisation of conversational agents and user experience (Kocaballi et al, 2019).…”

“…They also reported that agents often had to ask questions more than once to be able to process the response. In addition to interpretation challenges, free text input also raises concerns with respect to privacy and safety (Griffin et al, 2020;Tudor Car et al, 2020). In contrast, it is argued that speech input might make applications more accessible and intuitive (Dingler et al, 2021), for example, for those who have difficulty reading or typing (Tudor Car et al, 2020).…”