This paper is in the following e-collection/theme issue:

Chatbots and Conversational Agents (458) Digital Health Reviews (1174) Web-based and Mobile Health Interventions (3092) mHealth for Wellness, Behavior Change and Prevention (2728) Behavior Change (652)

Published on in Vol 24, No 10 (2022): October

This is a member publication of Imperial College London (Jisc)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/39243, first published .
Conversational Agents in Health Care: Scoping Review of Their Behavior Change Techniques and Underpinning Theory

Conversational Agents in Health Care: Scoping Review of Their Behavior Change Techniques and Underpinning Theory

Conversational Agents in Health Care: Scoping Review of Their Behavior Change Techniques and Underpinning Theory

Authors of this article:

Laura Martinengo1 Author Orcid Image ;   Ahmad Ishqi Jabir1 Author Orcid Image ;   Westin Wei Tin Goh1 Author Orcid Image ;   Nicholas Yong Wai Lo1 Author Orcid Image ;   Moon-Ho Ringo Ho2 Author Orcid Image ;   Tobias Kowatsch3, 4, 5, 6 Author Orcid Image ;   Rifat Atun7, 8, 9, 10 Author Orcid Image ;   Susan Michie11 Author Orcid Image ;   Lorainne Tudor Car1, 12 Author Orcid Image
Article Authors Cited by (10) Tweetations (9) Metrics

    Review

    1Lee Kong Chian School of Medicine, Nanyang Technological University Singapore, Singapore, Singapore

    2School of Social Sciences, Nanyang Technological University Singapore, Singapore, Singapore

    3Institute for Implementation Science in Health Care, University of Zurich, Zurich, Switzerland

    4School of Medicine, University of St.Gallen, St. Gallen, Switzerland

    5Centre for Digital Health Interventions, Department of Management, Technology, and Economics, ETH Zurich, Zurich, Switzerland

    6Future Health Technologies Programme, Campus for Research Excellence and Technological Enterprise, Singapore-ETH Centre, Singapore, Singapore

    7Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Harvard University, Cambridge, MA, United States

    8Department of Global Health and Social Medicine, Harvard Medical School, Harvard University, Cambridge, MA, United States

    9Health Systems Innovation Lab, Harvard T.H. Chan School of Public Health, Harvard University, Cambridge, MA, United States

    10Department of Health Policy and Management, Harvard T.H. Chan School of Public Health, Harvard University, Cambridge, MA, United States

    11UCL Centre for Behaviour Change, University College London, London, United Kingdom

    12Department of Primary Care and Public Health, School of Public Health, Imperial College London, London, United Kingdom

    Corresponding Author:

    Lorainne Tudor Car, MD, PhD

    Lee Kong Chian School of Medicine

    Nanyang Technological University Singapore

    11 Mandalay Road

    Clinical Sciences Building Level 18

    Singapore, 308232

    Singapore

    Phone: 65 69047142

    Email: lorainne.tudor.car@ntu.edu.sg


    Background: Conversational agents (CAs) are increasingly used in health care to deliver behavior change interventions. Their evaluation often includes categorizing the behavior change techniques (BCTs) using a classification system of which the BCT Taxonomy v1 (BCTTv1) is one of the most common. Previous studies have presented descriptive summaries of behavior change interventions delivered by CAs, but no in-depth study reporting the use of BCTs in these interventions has been published to date.

    Objective: This review aims to describe behavior change interventions delivered by CAs and to identify the BCTs and theories guiding their design.

    Methods: We searched PubMed, Embase, Cochrane’s Central Register of Controlled Trials, and the first 10 pages of Google and Google Scholar in April 2021. We included primary, experimental studies evaluating a behavior change intervention delivered by a CA. BCTs coding followed the BCTTv1. Two independent reviewers selected the studies and extracted the data. Descriptive analysis and frequent itemset mining to identify BCT clusters were performed.

    Results: We included 47 studies reporting on mental health (n=19, 40%), chronic disorders (n=14, 30%), and lifestyle change (n=14, 30%) interventions. There were 20/47 embodied CAs (43%) and 27/47 CAs (57%) represented a female character. Most CAs were rule based (34/47, 72%). Experimental interventions included 63 BCTs, (mean 9 BCTs; range 2-21 BCTs), while comparisons included 32 BCTs (mean 2 BCTs; range 2-17 BCTs). Most interventions included BCTs 4.1 “Instruction on how to perform a behavior” (34/47, 72%), 3.3 “Social support” (emotional; 27/47, 57%), and 1.2 “Problem solving” (24/47, 51%). A total of 12/47 studies (26%) were informed by a behavior change theory, mainly the Transtheoretical Model and the Social Cognitive Theory. Studies using the same behavior change theory included different BCTs.

    Conclusions: There is a need for the more explicit use of behavior change theories and improved reporting of BCTs in CA interventions to enhance the analysis of intervention effectiveness and improve the reproducibility of research.

    J Med Internet Res 2022;24(10):e39243

    doi:10.2196/39243

    Keywords

    behavior change; behavior change techniques; conversational agent; chatbot; mHealth 


    Conversational agents (CAs), or chatbots, are computer programs that simulate conversations with humans [1]. Although the first CAs were developed in the mid-1960s, it was not until the early 2000s that their availability and popularity markedly increased [2]. CAs can be used to automate a variety of tasks, such as the provision of news or weather forecasts and the facilitation of web-based shopping [3]. CAs may be deployed as stand-alone apps or websites, integrated into multifunctional apps, or included in messaging apps such as Telegram, Facebook Messenger, and Slack [2]. They may use text or voice-assisted interfaces or may include an embodied agent using virtual characters to simulate both verbal and nonverbal aspects of human communication [4]. CAs can be further classified as simple rule-based agents or smart, artificial intelligence (AI)–based agents using natural language processing or machine learning to generate the responses [2].

    Following the trends in other industries, health care has seen increasing adoption of CAs in recent years [1]. Health care CAs are versatile tools able to cater to several health needs, such as providing timely information [5], supporting mental health disorder management [6,7], assisting with triage in clinical settings [8,9], supporting chronic disease self-management, or delivering lifestyle change interventions, such as physical activity [10] and dietary changes, that increasingly incorporate elements of behavior change in the intervention design. In general, health care CAs appear to be effective in improving individuals’ outcomes [11,12] and are acceptable to users, who often describe them as friendly and trustworthy.

    Increasingly, health care CAs are used to deliver behavior change interventions, defined as complex interventions, comprising an interplay of 1 or several heterogeneous behavior change techniques (BCTs) [13]. BCTs are “observable and replicable components designed to change behavior” [13]. BCTs are considered the smallest active ingredient in an intervention, and can be used alone or in combination with other BCTs [13]. Adequate categorization of the BCTs included in an intervention allows for more efficient coding, leading to easier replication when designing similar interventions [13]. Several methods to classify BCTs have been developed, of which the Behavior Change Technique Taxonomy version 1 (BCTTv1) [14] is the most established and commonly used.

    Several reviews have synthesized the evidence about behavior change interventions delivered by digital health tools and CAs, such as a systematic review reporting on the use of BCTs in effective digital diabetes prevention interventions [15], a mapping review offering a description of the current uses of CAs for behavior change [16], and a scoping review describing the use of embodied CAs to support healthy lifestyle [17]. These reviews presented descriptive data, without an in-depth analysis of the type of BCTs used in the interventions, the use of behavior change theories to guide the interventions, the frequency with which each BCT was used, and potential associations between BCTs and intervention effectiveness. Therefore, this scoping review aims to analyze the use of BCTs in behavior change interventions delivered by CAs; specifically, it describes the health behaviors and disorders targeted by the intervention, describes the types of CAs used to deliver the behavior change interventions, identifies the theories or frameworks guiding the design of the behavior change interventions, identifies the most common type of BCTs used in CA-delivered interventions in health care, compares the BCTs employed in different types of CAs and for different health disorders, and compares the BCTs employed in the experimental and comparison interventions of studies evaluating CA-delivered behavior change interventions.


    Overview

    The scoping review was performed according to the Joanna Briggs Institute guidelines [18] and reported in alignment with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) reporting guidelines (Multimedia Appendix 1) [19]. The protocol was registered in Open Science Framework Registries [20] in April 2021 and was published in a peer-reviewed journal in July 2021 [21].

    Search Strategy

    The search strategy was designed using a comprehensive list of words and phrases that define CAs (Multimedia Appendix 2). We searched PubMed, Embase (Ovid), and CENTRAL (Cochrane Central Register of Controlled Trials), from their inception, and the first 10 pages of Google and Google Scholar [22,23] on April 26, 2021.

    Eligibility Criteria

    This scoping review included primary, experimental studies in English evaluating the use of CAs to deliver health care interventions focusing on behavior change. Eligible study designs were randomized controlled trials (RCTs), quasi-RCTs, cluster-randomized trials, controlled before-and-after studies, uncontrolled before-and-after studies, interrupted time series, and pilot and feasibility studies. We excluded nonexperimental study designs, such as observational studies, qualitative studies, opinion pieces, editorials, conference abstracts, and secondary studies.

    We included studies on text-based, voice-based, and embodied CAs, defined as conversational interfaces featuring a human-like avatar able to mimic the verbal and nonverbal components of a face-to-face conversation [24]. The eligible studies reported any health care intervention focused on behavior change to improve or promote a healthy lifestyle, or to support the management of physical or mental health conditions. Lastly, behavior change was an essential aspect of the eligible studies, with or without reference to an associated behavior change theory, in line with previous research in this area [25]. The BCTs were coded according to the BCTTv1 [14]. The taxonomy consists of 93 BCTs grouped into 16 distinct categories, aimed at providing a cross-domain template to facilitate research and intervention replication.

    Screening, Data Extraction, and Analysis

    Screening

    Screening for eligibility was performed in 2 stages. First, 2 researchers (NYWL and WWTG) worked independently to screen the titles and abstracts of all retrieved studies using Covidence [26]. Studies were excluded if their focus or study design did not align with our predefined eligibility criteria. Studies included in the first round of screening were uploaded to EndNote X9 (Clarivate), and the full-text papers were retrieved and screened for eligibility by 3 researchers working independently (AIJ, NYWL, and WWTG). Discrepancies in any screening stage were resolved through discussions between the reviewers, or by engaging a fourth reviewer (LM). The search and screening processes were documented in a study selection flowchart [27].

    Data Extraction

    The data were extracted using a Microsoft Excel (Microsoft Corporation) form developed by the research team, based on a data extraction form used in a previous scoping review [2], and a section on behavior change was added. The form was piloted in 3 studies and amended according to team members’ feedback before being used for data extraction. Reviewers worked in pairs (AIJ worked with LM and NYWL worked with WWTG) to extract data from 10 papers (20%) and individually for the remaining 42 papers (80%). Data extracted by all reviewers were subsequently reexamined by 2 researchers (LM and AIJ). Reviewers met regularly during this process to ensure a common understanding of the data extraction process and the concordance of the extracted data. The data extracted by each pair of reviewers were compared, and any disagreements were resolved through consensus or consultation with a third reviewer, acting as an arbiter.

    The data extraction form contained the following items: first author, year of publication, title of the article, study design, target disorder, description of the behavior change intervention, CA name, delivery channel, dialog technique, input and output modalities, end goal of the intervention, use of behavior change theories or frameworks, and BCTs mapped according to the BCTTv1 [14].

    Data Analysis

    Data were analyzed using descriptive statistics and frequent itemset mining (FIM) to explore possible BCT clustering [28]. Data were presented in a diagrammatic or tabular form accompanied by a narrative summary.

    Frequent Itemset Mining

    The FIM analysis was performed by implementing the Apriori algorithm using the arules package version 1.7-1 [29] in R version 4.1.2 (R Foundation for Statistical Computing) [30]. FIM aims to find patterns or associations in a group of items (itemset) by sorting the items that frequently appear together in the data set. The analysis starts by calculating support (how frequently an item appears in the data set) and confidence (number of times individual items “x” and “y” appear together in the data set) thresholds and discarding any itemset with support or confidence values below the predetermined minimum threshold.

    For this analysis, we assessed the 10 most frequently appearing patterns, for the overall data set and for each clinical domain. For the overall data set, the minimum threshold for algorithm support and confidence was set at 0.10 and 0.90, respectively, or itemset appearing in at least 10% of the data set (≥4 studies) and appearing together at least 90% of the time. For each clinical domain, the minimum thresholds were 0.20 for support and 0.90 for confidence to account for the fewer number of studies in each sub data set [31].


    Overview of Search Strategy

    The search strategy retrieved 2579 papers after removing duplicates, of which 349 were eligible for full-text screening. Among these, 52 papers were finally included in this review. We reported 47 studies, as 4 studies were reported in 2 papers each and 1 study included a corrigendum. Figure 1 presents the study selection process.

    ‎
    Figure 1. Study selection flowchart. BCT: behavior change technique; CA: conversational agent.
    View this figure

    Characteristics of the Included Studies

    Multimedia Appendix 3 presents a summary of the studies included in this review [6,11,32-79]. Over half of the studies (26/47, 55%) were published from 2019 onward [11,32,34,37,40,42-46,48-55,58-61,65,66,71,72,76-78], including 6 published in the first quarter of 2021 [42,46,49,54,55,60]. All papers except 1 [32] were published in high-income countries, and 24/47 studies (51%) were published in the United States [6,32,34,36,39,43,45,47, 48,51,52,54,56-58,61-64,67,69-75].

    Most studies included a control group except 5/47 (11%) single-group pretest posttest trials [43,46,55,58,65,66], 3/47 (6%) feasibility studies [59-61], and 1/47 (2%) pilot study [48]. A total of 26/47 studies (55%) were RCTs [6,11,33,35-37,39-41,44,45,49,50,53,54,62-64,68-75,77,78]. In 36/47 studies (77%), the primary outcomes were associated with improvement of the target disorder [6,33,36, 38-45,47-59,62-64,67,68,70-75,77-80], 5/47 studies (11%) reported technical-related primary outcomes (eg, technical performance, system crashes) [11,60,65,66,69,76], and 6/47 studies (13%) reported primarily user experience outcomes (eg, engagement with the CA, user satisfaction) [32,34,35,37,46,61]. Most interventions aimed to support treatment or monitoring (22/47, 47%) [6,33,35-44,46,48-50,53-55,59,60,80] or to promote healthy lifestyle change (18/47, 38%) [11,32,34, 45,61-66,68-76,78,79]. Table 1 presents a summary of the included studies.

    Table 1. Characteristics of included studies (N=47).
    Study characteristicsStudies, n (%)
    Year of publication

    		Before 201921 (45)

    		2019 or after26 (55)
    Country

    		United States24 (51)

    		United Kingdom6 (13)

    		Japan3 (6)

    		Korea3 (6)

    		Switzerland3 (6)

    		Australia2 (4)

    		France1 (2)

    		Germany1 (2)

    		India1 (2)

    		Netherlands1 (2)

    		Spain1 (2)

    		Sweden1 (2)
    Study design

    		Randomized controlled trial26 (55)

    		Pilot study9 (19)

    		Single-group pretest posttest trial5 (11)

    		Feasibility study5 (11)

    		Microrandomized controlled trials1 (2)

    		Nonrandomized comparison study1 (2)
    Study outcomes

    		Clinical23 (49)

    		Clinical; user experience12 (26)

    		User experience; clinical6 (13)

    		Technical; clinical3 (6)

    		Technical; clinical; user experience2 (4)

    		Clinical; technical1 (2)
    Clinical focus of the interventions

    		Lifestyle behavior change17 (36)

    		Treatment and monitoring16 (34)

    		Treatment and monitoring + education4 (9)

    		Education4 (9)

    		Education + lifestyle behavior change3 (6)

    		Treatment and monitoring + lifestyle behavior change2 (4)

    		Education + treatment and monitoring1 (2)

    		Lifestyle behavior change + education1 (2)
    Clinical domains

    		Mental health19 (40)

    		Chronic disorders14 (30)

    		Lifestyle modification14 (30)

    Clinical Domains

    Mental Health Interventions

    Most CAs focused on mental health (19/47, 40%) [6,32-47,79,80], either supporting mental well-being (5/19, 26%) for healthy individuals [46,47,79,80] or patients recovering from cancer [33]; enabling self-improvement interventions such as problem solving [34] or communications skills [35]; or assisting participants in the management of a mental health disorder (14/19, 74%) [6,36-46], including depression (with or without anxiety; 3/19, 16%) [6,36,37], emotional distress (2/19, 11%) [38,39], bipolar disorder [40], panic disorder [41], fear of heights [42], adult attention deficit disorder [43], substance use disorder [44], gambling [45], and social exclusion [46].

    All except 2 interventions [44,47] included a control group, and 10/19 studies (53%) were RCTs [6,33-37,39,41,45,46]. A total of 6 studies included an active comparison with another digital intervention [34,38,39,46], a paper-based version of the CA intervention [40], or mood monitoring [33]. Besides, 6 studies provided information about the target disorder [6,35,37,41,43,48], and 10 experimental interventions (10/17, 59%) were reported as more effective than the comparisons [6,33-37,39,41,45,46].

    Chronic Disorder Management Interventions

    A total of 14/47 studies (30%) offered interventions focusing on a chronic disease other than mental illness [49-63]. Most studies (4/14, 29%) targeted a metabolic disorder including obesity (n=1) [63], prediabetes (n=1) [62], or type 2 diabetes (n=2) [51,56]. Three studies evaluated a pain management intervention for osteoarthritis (n=2) [57,58] or for general management of chronic pain (n=1) [54]. Other studies focused on asthma [61], atrial fibrillation [52,53], HIV [49], hypertension [50], insomnia [60], irritable bowel syndrome [55], and prostate cancer [59]. The interventions aimed to support treatment and monitoring tasks (8/14, 57%) or provide education (4/14, 29%).

    Half of the included studies were feasibility or pilot studies, and 5/14 studies (36%) were RCTs [49,50,53,54,62]. Comparison interventions included a nurse-led instruction mirroring the CA intervention [50], physical activity monitoring using a pedometer [63], provision of information [57,58], treatment as usual [51-53], and waitlist controls [54,55]. Furthermore, 6/14 studies (43%) were single-group interventions without a comparison group [48,55,58-61]. Only 2 studies described the experimental interventions as more effective than the comparisons (2/8, 25%) [51,52,54].

    Lifestyle Change Interventions

    A total of 14/47 studies (30%) included interventions to support lifestyle modification [11,64-79], particularly increasing physical activity (10/14, 71%), either as the sole intervention (n=6) [64,69,74-77,79] or in combination with another approach such as diet improvement (n=2) [65-67], or diet improvement plus stress relief (n=1) [70]. Four studies (4/14, 29%) targeted an aspect of women’s health including preconception care (n=3) [71-73,78] and breastfeeding support (n=1) [68]. One study offered a smoking cessation intervention [11]. In 12 studies, the interventions aimed to facilitate lifestyle change (12/14, 86%) [11,63-76,78], while 2 studies offered education [67,77].

    Among this, 1/14 (7%) study was a single-group pretest-posttest trial [65,66], while most studies (11/14, 79%) were RCTs [11,63,64,68-75,77,78]. In 7/13 studies (54%) comparison interventions consisted of face-to-face versions of the intervention [74-76], abridged interventions that excluded the CA [11,64,65,70], or a similar version of the intervention with differing reward systems [77,79]. Other comparisons included information-only interventions (3/13, 23%), treatment as usual (1/13, 8%), or waitlists (2/13, 15%). Most experimental interventions were reported to be more effective than the comparisons (9/13, 69%).

    Characteristics of CAs

    Table 2 summarizes the characteristics of the included CAs.

    A total of 39 CAs were included. Six CAs were reported in 2 or more manuscripts. Four CAs (Carmen [74-76], Tanya [52,53,68], Tess [37,62], and Todaki [41,43]) were reported in 2 papers each, and 2 CAs (Gabby [70-73] and MYLO [34,38,39]) were reported in 3 manuscripts. Three CAs were adapted for different target disorders. Embodied CA Tanya was used as an educational tool for patients with atrial fibrillation [52,53] and to offer breastfeeding support [68], CA Tess was used for mental health [37] and diabetes care [62], and Todaki was used to deliver CBT for panic disorder [41] and to manage adults with attention deficit disorder [43]. Finally, MYLO was used in student and older adult [38] populations by 2 distinct research groups.

    The majority of CAs featured 1 or more anthropomorphic characteristics, such as the assignation of gender, name, or a human-like display. Most CAs (41/47, 87%) responded to a name, 27/47 CAs (57%) were presented as female agents, and 20/47 (43%) were embodied CAs. Most CAs used rule-based algorithms to design the flow of conversations, either by themselves (35/47, 75%) or complemented with AI (2/47, 4%). CAs were more often available through a smartphone app (14/47, 30%) or web page (13/47, 28%). In all but 3 CAs (44/47, 94%), the primary method for users’ inputs was text; 7/47 of these CAs (15%) also accepted verbal or visual inputs, whereas 3/47 CAs (6%) received only verbal inputs. Almost 80% of all CAs (36/47, 77%) displayed a “coach-like” personality, characterized by an encouraging, motivating, and nurturing conversational style.

    Table 2. Characteristics of CAsa (N=47).
    CA characteristicsValues, n (%)
    Type of CA

    		Embodied CAs20 (43)

    		No visual representation12 (26)

    		Human-like cartoon avatar10 (21)

    		Nonhuman cartoon avatar5 (11)
    Gender

    		Female27 (57)

    		No gender assigned (no avatar/no human avatar)16 (34)

    		Male2 (4)

    		Defined by the user2 (4)
    CA “level of intelligence”

    		Rule-based CAs34 (72)

    		Artificial intelligence CAs9 (19)

    		Rule-based + artificial intelligence CAs4 (9)
    Dialog modality

    		Predetermined text28 (60)

    		Free text8 (17)

    		Predetermined and free text7 (15)

    		Not specified4 (9)
    Delivery channel

    		Smartphone app14 (30)

    		Web based13 (28)

    		Desktop7 (15)

    		Messaging apps6 (13)

    		Two or more delivery channels6 (13)

    		Tablet computer1 (2)
    Users’ input modalities

    		Text37 (79)

    		Text + others (voice, images, video)7 (15)

    		Voice (± video)3 (6)
    CA output modalities

    		Text + others (voice, images, video)29 (62)

    		Text15 (32)

    		Voice (± images, video)3 (6)
    CA personality

    		Coach like36 (77)

    		Health care professional like9 (19)

    		Not specified2 (4)

    aCA: conversational agent.

    Type of CA and Clinical Domains

    Embodied CAs were used to deliver almost two-thirds (9/14, 64%) of the interventions promoting lifestyle modification [64,65,68-76], 43% (6/14) of the chronic disease management interventions [49,51-53,59,60,63] and only 26% (5/19) of the mental health interventions.

    By contrast, most mental health CAs did not include an avatar (8/19, 42%) [34,35,38-40,45,47,81], or they were represented by a nonhuman avatar (5/19, 26%) [6,33,41,43,44]. Human-like avatars were present in 1/19 (5%) mental health intervention [37], 6/14 (43%) chronic disease management interventions [54,55,57,58,61,62], and 3/14 (21%) lifestyle change interventions [66,67,77,78].

    Behavior Change Theories and Techniques

    Behavior Change Theories

    A total of 12/47 (26%) studies incorporated a behavior change theory to guide the CA intervention design, including 4/14 (29%) studies targeting a chronic disorder [51,54,59,61], 7/14 (50%) studies [65,71-76,78,79] evaluating a lifestyle change intervention, and 1/19 study (5%) [37] on mental health. The Transtheoretical Model was the most used behavior change theory, either alone [37,71-73,78] or together with the Social Cognitive Theory [51,65,74-76]. In addition, 4/19 (21%) mental health studies and 2/14 (14%) studies targeting a chronic disorder based their interventions on theories derived from the behavior [34,38,39], communication [57,58], learning [59], or psychological domains [33] (Table 3).

    The use of theories aimed to guide the design of the intervention or to monitor participants’ stages of change as they progressed through the intervention, as exemplified by 3 studies [71-73,78] using the Transtheoretical Model and 1 study using the Health Action Process Approach [54]. It was not clear how the use of theories influenced the intervention design or the choice of BCTs. For example, 4 studies using the Transtheoretical Model included a wide variety of BCTs, ranging from 3 [78] to 10 [72,73]. Similarly, 4 studies [51,65,74-76] using the Transtheoretical Model and the Social Cognitive Theory incorporated between 6 [51] and 19 [75,76] BCTs.

    Table 3. Behavior change theories informing the CAa-based interventions (N=47).
    Theories guiding CA interventionsStudies, n (%)
    No theory29 (62)
    Behavior change theories11 (23)

    		Transtheoretical Model4 (9)

    		Transtheoretical Model + Social Cognitive Theory4 (9)

    		Theory of Planned Behavior + Self-Determination Theory + Technology1 (2)

    		Acceptance theories

    		Health Action Process Approach1 (2)

    		Habit Formation Model1 (2)
    Behavior change theories + other theories1 (2)

    		Unified Theory of Acceptance and Use of Technology + Cognitive Theory1 (2)

    		Multimedia Learning
    Other theories6 (13)

    		Perceptual Control Theory3 (6)

    		Communication Accommodation Theory2 (4)

    		Stress and Coping Theory + Broaden and Build Theory of Positive Emotion1 (2)

    aCA: conversational agent.

    Incorporated BCTs

    The experimental interventions incorporated 63 BCTs from 15 categories, whereas the comparison interventions included 32 BCTs from 10 categories. However, only 24 BCTs were incorporated into experimental interventions in 5 or more studies, whereas 12 BCTs were reported in only 1 study each. The most incorporated BCT across interventions was 4.1 “Instruction on how to perform a behavior” (34/47, 72%), followed by 3.3 “Social support (emotional)” (27/47, 57%) and 1.2 “Problem solving” (24/47, 51%), whereas only 1 study included a BCT from category 14 (14.4 “Reward approximation”) in the experimental intervention, and none included BCTs from category 16 “Covert learning.” Figure 2 shows the frequency of presentation of all 63 BCTs in experimental and comparison interventions.

    The average number of BCTs included in the experimental interventions was 9 (range 2-21 BCTs). By contrast, comparison interventions (n=38) included an average of 2 BCTs (range 0-17 BCTs).

    ‎
    Figure 2. Number of studies using each BCT in the experimental and comparison interventions. BCT: behavior change technique; Int: intervention.
    View this figure
    Use of BCTs According to the Clinical Domain

    The number of BCTs in experimental interventions was consistent across all clinical domains. Mental health interventions included an average of 8 BCTs (range 3-16 BCTs), chronic disorder management interventions included an average of 9 BCTs (range 2-18 BCTs), and lifestyle change interventions included an average of 10 BCTs (range 3-21 BCTs). The number of BCTs included in comparison interventions varied from an average of 2 BCTs in chronic disorder management (range 1-3 BCTs) and mental health interventions (range 1-2 BCTs) to a mean of 6 BCTs (range 1-17 BCTs) in lifestyle change interventions.

    Mental health interventions incorporated 41 BCTs in experimental interventions. The most common BCTs were 3.3 “Social support (emotional)” (12/19, 63%), 11.2 “Reduce negative emotions” (11/19, 58%), 4.1 “Instruction on how to perform a behavior” (9/19, 47%), and BCTs 1.1 “Goal setting (behavior),” 1.2 “Problem solving,” 2.2 “Feedback on behavior,” 7.1 “Prompts/cues,” 8.1 “Behavioral practice/rehearsal,” and 8.3 “Habit formation” that were included in 7/19 (37%) studies each.

    Lifestyle change interventions included 46 BCTs. The most common BCT was 1.2 “Problem solving” (11/14, 79%), followed by 4.1 “Instruction on how to perform a behavior” (10/14, 71%) and BCTs 1.1 “Goal setting (behavior),” 1.4 “Action planning,” and 2.3 “Self-monitoring of behavior,” included in 9/14 (64%) studies each.

    Chronic disorder management interventions included a total of 41 BCTs. Almost all studies included BCT 4.1 “Instruction on how to perform a behavior” (13/14, 93%), followed by 7.1 “Prompts/cues” (8/14, 57%), 3.3 “Social support (emotional)” (7/14, 50%), and BCTs 1.2 “Problem solving,” 8.1 “Behavioral practice/rehearsal,” and 8.3 “Habit formation,” all included in 6/14 studies (43%).

    Figure 3 presents a summary of the most commonly used BCTs according to the clinical domain. Multimedia Appendix 4 presents a table summarizing the use of each BCT according to the clinical domain.

    ‎
    Figure 3. Commonly used BCTs according to the clinical domain. BCT: behavior change technique.
    View this figure
    BCT Clustering According to the Clinical Domain Using FIM

    The overall data set (n=47) generated 206 rules with an average support of 0.12, suggesting that the rules applied to at least 12% of the data set or about 6 studies. In general, 26% of the studies included BCTs 4.1 “Instruction on how to perform a behavior” and 8.1 “Behavioral practice/rehearsal,” whereas 23% of the studies included BCTs 4.1 “Instruction on how to perform a behavior,” 7.1 “Prompts/cues,” and 8.3 “Habit formation.”

    The mental health domain (n=19) generated 45 rules with an average support of 0.22. About one-quarter of studies (26%) included 1 of 3 rules: the first itemset included BCTs 1.5 “Review behavior goal(s),” 2.2 “Feedback on behavior,” and 3.3 “Social support”; followed by the itemset comprising BCTs 3.3 “Social support” and 12.6 “Body changes”; and the itemset containing BCTs 3.3 “Social support,” 4.1 “Instruction on how to perform a behavior,” and 11.2 “Reduce negative emotions.” Conversely, the lifestyle change domain (n=14) generated 1322 rules with an average support of 0.24. About 64% of the studies included BCTs 1.2 “Problem solving” and 2.3 “Self-monitoring of behavior,” whereas 57% of the studies also included BCT 1.1 “Goal-setting (behavior).” Finally, the chronic disorder management domain (n=14) generated 230 rules with an average support of 0.23. Most studies (93%) included BCT 4.1 “Instruction on how to perform a behavior,” whereas 57% also included BCT 7.1 “Prompts/cues.”

    Multimedia Appendix 5 presents a table describing the top 10 itemsets for all included papers and each clinical domain.

    Use of BCTs According to the CA Type

    Interventions delivered by any type of CA included an average of 9 BCTs. However, the number of BCTs in experimental interventions varied by type of CA: embodied CAs included 2-19 BCTs, CAs represented by an avatar included 3-14 BCTs, and CAs with nonspecified or nonvisual representation incorporated 4-21 BCTs.

    Embodied CAs included a total of 49 BCTs in the interventions. The most common BCTs were 3.3 “Social support (emotional) (14/20, 70%), and BCTs 1.2 “Problem solving,” 2.3 “Self-monitoring of behavior,” and 4.1 “Instruction on how to perform a behavior,” which were found in 13/20 (65%) studies each. By contrast, CAs represented by an avatar included a total of 38 BCTs in the interventions. The most common BCTs were 4.1 “Instruction on how to perform a behavior” (13/15, 87%), and BCTs 3.3 “Social support (emotional)” and 7.1 ”Prompts/cues” included in 10/15 (67%) studies each. Finally, CAs with nonspecified or nonvisual representation incorporated a total of 47 BCTs. Four BCTs (1.2 “Problem solving,” 4.1 “Instruction on how to perform a behavior,” 7.1 ”Prompts/cues,” and 8.3 “Habit formation”) were included in 6/12 (50%) studies, and BCT 11.2 “Reduce negative emotions” was included in 5/12 (42%) studies. Multimedia Appendix 6 provides further information about the use of BCTs according to the type of CA.


    Principal Findings

    This scoping review included 47 studies reporting behavior change interventions delivered by CAs, targeting chronic disorders, lifestyle change, and mental health. The interventions included a total of 63 BCTs, but only 24 were consistently found in 5 or more interventions. The BCTs represented aspects of health education (BCT 4.1), self-management (BCTs 1.1, 1.2, and 2.3), and social support (BCT 3.3). Several behavior change theories informed the intervention design in 12/47 (26%) studies of the included studies. However, studies informed by the same theory employed different sets of BCTs. Our findings align with previous systematic reviews reporting that similar BCTs were frequently incorporated into effective lifestyle change interventions [82], or into digitally delivered interventions [15].

    We did not find a relationship between the use of theories, the type of theory used, and the number and type of BCTs included in the interventions. Furthermore, a small number of studies [11,61] guided the intervention design, using modified BCT taxonomies that addressed smoking cessation [11] and diet modification [61]. These data suggest that the choice of BCTs may be primarily determined by the target behavior rather than the use of a behavior change theory. The impact of using a behavior change theory is nevertheless unclear. A 2010 systematic review [83] reported that the use of a behavior change theory was associated with increased effectiveness of the interventions, although just over 20% of studies included a theory. Conversely, a systematic review by Van Rhoon et al [15] reported the use of theories in 16/21 (76%) studies but did not assess intervention effectiveness. In addition, a recent overview of systematic reviews [84] reported the use of theories in the intervention design of 19%-52% of the included studies, although there was no clear association with the intervention effectiveness.

    The categorization of studies in 3 distinct clinical domains suggested different prioritizations in mental health, lifestyle change, and chronic disorders, although the delivery of health education, evidenced by the frequent occurrence of BCTs 4.1 “Instruction on how to perform a behavior,” 8.1 “Behavioral practice/rehearsal,” and 8.3 “Habit formation,” was consistent across all clinical domains.

    Mental health interventions frequently included BCTs 3.3 “Social support (emotional)” and 11.2 “Reduce negative emotions.” Specifically, BCT 3.3 may be associated with the use of psychotherapeutic techniques such as cognitive behavioral therapy or motivational interviewing, while the inclusion of BCT 11.2 suggests the use of relaxation techniques and mindfulness to support stress management and emotional regulation. Therefore, behavior change in mental health settings appeared to be closely interlinked with the therapeutic strategies. Concurrently, the inclusion of other BCTs, such as instructions to perform a behavior (BCT 4.1), goal setting (BCT 1.1) and reviews (BCT 1.5), problem solving (BCT 1.2), and feedback (BCT 2.2), may be aligned with general principles of patient participation in decision making [85], as well as highlight the importance of health education [86,87], particularly relevant in self-initiated digital interventions.

    Lifestyle change interventions frequently included problem-solving (BCT 1.2) techniques to help users better understand their barriers to behavior change, and goal setting (BCT 1.1) and self-monitoring (BCT 2.3) to work toward the target behavior. These BCTs were often included together and this may suggest a synergistic relationship. At the same time, the importance of ensuring adequate health literacy to improve population outcomes was emphasized by the frequent inclusion of BCT 4.1 “Instruction on how to perform a behavior.”

    Chronic disorder management interventions favored not only the inclusion of instructional BCTs, such as guidance to perform a target behavior (BCT 4.1) but also reminders (BCT 7.1 “Prompts/cues”) to facilitate the acquisition of new routines (BCT 8.3 “Behavioral practice/rehearsal”). Self-management of chronic illnesses is essential to ensure improved patient outcomes and adequate quality of life but requires that individuals engage in a steep learning curve as they adapt to living with a long-term condition and develop new habits.

    In general, the relationship between the number and type of BCTs and the effectiveness of the interventions was inconsistent and appeared to be determined by the clinical domain. Effective lifestyle change interventions tended to include a higher number of BCTs, a finding that was not replicated in the other clinical domains. At the same time, lifestyle change interventions were comparatively more effective than those in other clinical domains, particularly chronic disorders. Effective interventions in the lifestyle change and mental health domains frequently included BCTs related to goal setting and planning, timely provision of feedback, health education, and rewards on completed tasks. Previous studies reported varied results. A 2017 systematic review of 48 studies [82] evaluating the management of overweight and obesity in adults found small pooled effect sizes for short- and long-term diet and physical activity interventions. Effective interventions included a larger number of BCTs, particularly BCTs encouraging goal setting and self-monitoring of behavior. Similarly, a systematic review on the BCTs and technical features of digital interventions for the prevention of type 2 diabetes [15] found that effective interventions included a larger number of BCTs or BCTs related to social support, goal setting, and feedback.

    There was an unexpected relationship between the CA types and the clinical domain, manifested by a predominance of embodied CAs in lifestyle change interventions, and the use of nonhuman or nonavatar CAs in mental health interventions. The reasons for these findings are unclear and beyond the scope of this review; however, further research may help clarify the role of avatars, or virtual humans, if any, in delivering behavior change interventions. Other reviews have reported the use of embodied CAs to support mental health interventions, particularly autism [20,24], but methodological differences limit the comparisons with our findings. Provoost et al’s scoping review [4] used a broader definition of embodied CA, while a systematic review by Laranjo et al [87] included only AI-based CAs.

    Strengths and Limitations

    This scoping review has several strengths. First, we used a comprehensive literature search of peer-reviewed and gray literature that prioritized the sensitivity of the search terms to capture a broad range of publications reporting the use of CAs in health care. However, relevant studies may have been omitted. Second, we included studies reporting on a wide variety of physical and mental health conditions, and categorized the studies into 3 distinct clinical domains, revealing differences in the type of BCTs selected in each domain.

    There are also some limitations. First, many studies did not provide exact BCT codes when describing the interventions, therefore categorization of BCTs was inferred from the paper’s description by the research team, based on thorough analysis, rigorous team discussion, and reviews to establish consensus. Second, given the descriptive nature of scoping reviews, we were unable to explore in more depth the relationship between the choice of BCTs and the effectiveness of the intervention, or the type of CA used to deliver the intervention.

    Future Research and Practice Recommendations

    This review has highlighted several areas that warrant further research. First, reporting guidelines to ensure accurate reporting of the BCTs included in behavior change interventions according to standardized taxonomies, such as the BCTTv1 [14], should be implemented. Such guidelines would facilitate reproducibility of research, assessment of active intervention components, and evidence synthesis. Second, further research is needed to increase our understanding of the impact of behavior change theories in the design of interventions, the choice of BCTs, and the effectiveness of the intervention. Third, the impact of CAs to deliver behavior change interventions should be further explored, particularly the influence of a conversational interface on engagement, adherence, and effectiveness of the intervention when compared with less interactive digital technologies. Furthermore, comparisons between rule-based CAs and those incorporating machine learning or natural language processing should be further investigated. Fourth, the possible role of the type of CA in delivering behavior change interventions, as suggested in our findings, should be further explored. Fifth, the relationship between the ideal combination of BCTs required to design effective interventions may be evaluated using data mining techniques such as FIM or multiple correspondent analysis. Lastly, the relationship between behavior change interventions and mental health requires further evaluation.

    The use of CAs to deliver behavior change interventions appears promising, particularly to support lifestyle change, although better reporting of BCTs included in the interventions is warranted to facilitate analysis of active components, design more effective interventions, and ensure reproducibility of research. The role of CA types in delivering behavior change interventions should be further explored.

    Acknowledgments

    This research is supported by the Singapore Ministry of Education under Singapore Ministry of Education Academic Research Fund Tier 1 (RG36/20). The research was conducted as part of the Future Health Technologies program, which was established collaboratively between ETH Zurich and the National Research Foundation, Singapore. This research is supported by the National Research Foundation, Prime Minister’s Office, Singapore, under its Campus for Research Excellence and Technological Enterprise program.

    Authors' Contributions

    LTC conceptualized the study and provided supervision at all steps of research. LTC and LM designed the study. LM, AIJ, WWTG, and NYWL extracted data and conducted the analysis. LM and AIJ wrote the manuscript. MHRH, TK, RA, and SM provided critical review of the manuscript. All authors approved the final version of the manuscript and take accountability for all aspects of the work.

    Conflicts of Interest

    TK is affiliated with the Centre for Digital Health Interventions, a joint initiative of the Department of Management, Technology, and Economics at ETH Zurich and the Institute of Technology Management at the University of St.Gallen, which is funded in part by CSS, a Swiss health insurer. TK is also a cofounder of Pathmate Technologies, a university spin-off company that creates and delivers digital clinical pathways. However, neither CSS nor Pathmate Technologies was involved in this research. The other authors declare that they have no competing interests.

    ‎
    Multimedia Appendix 1

    PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist.

    DOCX File , 108 KB
    ‎
    Multimedia Appendix 2

    PubMed search strategy.

    DOCX File , 17 KB
    ‎
    Multimedia Appendix 3

    Characteristics of included studies.

    DOCX File , 42 KB
    ‎
    Multimedia Appendix 4

    Use of BCTs according to the clinical domain. BCT: behavior change technique.

    DOCX File , 25 KB
    ‎
    Multimedia Appendix 5

    Frequent Itemset Mining (FIM).

    DOCX File , 20 KB
    ‎
    Multimedia Appendix 6

    Use of BCTs according to the CA type. BCT: behavior change technique; CA: conversational agent.

    DOCX File , 426 KB
    1. Conversational agents or chatbots. Oxford Learner's Dictionaries.   URL: https://www.oxfordlearnersdictionaries.com/definition/english/chatbot [accessed 2021-03-23]
    2. Tudor Car L, Dhinagaran DA, Kyaw BM, Kowatsch T, Joty S, Theng Y, et al. Conversational Agents in Health Care: Scoping Review and Conceptual Analysis. J Med Internet Res 2020 Aug 07;22(8):e17158 [FREE Full text] [CrossRef] [Medline]
    3. Diederich S, Brendel A, Kolbe L. Towards a Taxonomy of Platforms for Conversational Agent Design. In: Proceedings of Internationale Tagung Wirtschaftsinformatik. 2019 Presented at: Internationale Tagung Wirtschaftsinformatik; February 24-27, 2019; Siegen, Germany   URL: https:/​/www.​researchgate.net/​publication/​329739579_Towards_a_Taxonomy_of_ Platforms_for_Conversational_Agent_Design/​
    4. Provoost S, Lau HM, Ruwaard J, Riper H. Embodied Conversational Agents in Clinical Psychology: A Scoping Review. J Med Internet Res 2017 May 09;19(5):e151 [FREE Full text] [CrossRef] [Medline]
    5. Amith M, Zhu A, Cunningham R, Lin R, Savas L, Shay L, et al. Early Usability Assessment of a Conversational Agent for HPV Vaccination. Stud Health Technol Inform 2019;257:17-23 [FREE Full text] [Medline]
    6. Fitzpatrick KK, Darcy A, Vierhile M. Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial. JMIR Ment Health 2017 Jun 06;4(2):e19 [FREE Full text] [CrossRef] [Medline]
    7. Inkster B, Sarda S, Subramanian V. An Empathy-Driven, Conversational Artificial Intelligence Agent (Wysa) for Digital Mental Well-Being: Real-World Data Evaluation Mixed-Methods Study. JMIR Mhealth Uhealth 2018 Nov 23;6(11):e12106 [FREE Full text] [CrossRef] [Medline]
    8. Gilbert S, Mehl A, Baluch A, Cawley C, Challiner J, Fraser H, et al. How accurate are digital symptom assessment apps for suggesting conditions and urgency advice? A clinical vignettes comparison to GPs. BMJ Open 2020 Dec 16;10(12):e040269 [FREE Full text] [CrossRef] [Medline]
    9. Miller S, Gilbert S, Virani V, Wicks P. Patients' Utilization and Perception of an Artificial Intelligence-Based Symptom Assessment and Advice Technology in a British Primary Care Waiting Room: Exploratory Pilot Study. JMIR Hum Factors 2020 Jul 10;7(3):e19713 [FREE Full text] [CrossRef] [Medline]
    10. Eisele A, Schagg D, Krämer LV, Bengel J, Göhner W. Behaviour change techniques applied in interventions to enhance physical activity adherence in patients with chronic musculoskeletal conditions: A systematic review and meta-analysis. Patient Educ Couns 2019 Jan;102(1):25-36. [CrossRef] [Medline]
    11. Perski O, Crane D, Beard E, Brown J. Does the addition of a supportive chatbot promote user engagement with a smoking cessation app? An experimental study. Digit Health 2019;5:2055207619880676 [FREE Full text] [CrossRef] [Medline]
    12. Milne-Ives M, de Cock C, Lim E, Shehadeh MH, de Pennington N, Mole G, et al. The Effectiveness of Artificial Intelligence Conversational Agents in Health Care: Systematic Review. J Med Internet Res 2020 Oct 22;22(10):e20346 [FREE Full text] [CrossRef] [Medline]
    13. Michie S, Wood CE, Johnston M, Abraham C, Francis JJ, Hardeman W. Behaviour change techniques: the development and evaluation of a taxonomic method for reporting and describing behaviour change interventions (a suite of five studies involving consensus methods, randomised controlled trials and analysis of qualitative data). Health Technol Assess 2015 Nov;19(99):1-188 [FREE Full text] [CrossRef] [Medline]
    14. Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W, et al. The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Ann Behav Med 2013 Aug;46(1):81-95. [CrossRef] [Medline]
    15. Van Rhoon L, Byrne M, Morrissey E, Murphy J, McSharry J. A systematic review of the behaviour change techniques and digital features in technology-driven type 2 diabetes prevention interventions. Digit Health 2020 Dec;6:2055207620914427 [FREE Full text] [CrossRef] [Medline]
    16. Pereira J, Díaz. Using Health Chatbots for Behavior Change: A Mapping Study. J Med Syst 2019 Apr 04;43(5):135. [CrossRef] [Medline]
    17. Kramer LL, Ter Stal S, Mulder BC, de Vet E, van Velsen L. Developing Embodied Conversational Agents for Coaching People in a Healthy Lifestyle: Scoping Review. J Med Internet Res 2020 Feb 06;22(2):e14058 [FREE Full text] [CrossRef] [Medline]
    18. Peters MDJ, Godfrey CM, Khalil H, McInerney P, Parker D, Soares CB. Guidance for conducting systematic scoping reviews. Int J Evid Based Healthc 2015 Sep;13(3):141-146. [CrossRef] [Medline]
    19. Tricco AC, Lillie E, Zarin W, O'Brien KK, Colquhoun H, Levac D, Tunçalp, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med 2018 Oct 02;169(7):467-473. [CrossRef] [Medline]
    20. The choice of behavioural change techniques in conversational agents in healthcare: a scoping review protocol. OSF Registry. 2021 Apr 28.   URL: https://osf.io/487jd [accessed 2022-09-21]
    21. Martinengo L, Lo NYW, Goh WIWT, Tudor Car L. Choice of Behavioral Change Techniques in Health Care Conversational Agents: Protocol for a Scoping Review. JMIR Res Protoc 2021 Jul 21;10(7):e30166 [FREE Full text] [CrossRef] [Medline]
    22. Gusenbauer M, Haddaway NR. Which academic search systems are suitable for systematic reviews or meta-analyses? Evaluating retrieval qualities of Google Scholar, PubMed, and 26 other resources. Res Synth Methods 2020 Mar;11(2):181-217 [FREE Full text] [CrossRef] [Medline]
    23. Haddaway NR, Collins AM, Coughlin D, Kirk S. The Role of Google Scholar in Evidence Reviews and Its Applicability to Grey Literature Searching. PLoS One 2015;10(9):e0138237 [FREE Full text] [CrossRef] [Medline]
    24. Kwasnicka D, Dombrowski SU, White M, Sniehotta F. Theoretical explanations for maintenance of behaviour change: a systematic review of behaviour theories. Health Psychol Rev 2016 Sep;10(3):277-296 [FREE Full text] [CrossRef] [Medline]
    25. Covidence.   URL: https://www.covidence.org/about-us/ [accessed 2021-03-28]
    26. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009 Jul 21;6(7):e1000097 [FREE Full text] [CrossRef] [Medline]
    27. Luna JM, Fournier‐Viger P, Ventura S. Frequent itemset mining: A 25 years review. WIREs Data Mining Knowl Discov 2019 Jul 16;9(6):1329. [CrossRef]
    28. Hahsler M, Grün B, Hornik K. arules - A Computational Environment for Mining Association Rules and Frequent Item Sets. J. Stat. Soft 2005;14(15):1-25. [CrossRef]
    29. R Foundation for Statistical Computing. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2021.
    30. Luna JM, Ondra M, Fardoun HM, Ventura S. Optimization of quality measures in association rule mining: an empirical study. IJCIS 2018;12(1):59. [CrossRef]
    31. World Bank Country and Lending Groups. World Bank.   URL: https:/​/datahelpdesk.​worldbank.org/​knowledgebase/​articles/​906519-world-bank-country-and-lending-groups [accessed 2021-11-01]
    32. Greer S, Ramo D, Chang Y, Fu M, Moskowitz J, Haritatos J. Use of the Chatbot "Vivibot" to Deliver Positive Psychology Skills and Promote Well-Being Among Young People After Cancer Treatment: Randomized Controlled Feasibility Trial. JMIR Mhealth Uhealth 2019 Oct 31;7(10):e15018 [FREE Full text] [CrossRef] [Medline]
    33. Bird T, Mansell W, Wright J, Gaffney H, Tai S. Manage Your Life Online: A Web-Based Randomized Controlled Trial Evaluating the Effectiveness of a Problem-Solving Intervention in a Student Sample. Behav Cogn Psychother 2018 Sep;46(5):570-582. [CrossRef] [Medline]
    34. Ali R, Hoque E, Duberstein P, Schubert L, Razavi SZ, Kane B, et al. Aging and Engaging: A Pilot Randomized Controlled Trial of an Online Conversational Skills Coach for Older Adults. Am J Geriatr Psychiatry 2021 Aug;29(8):804-815. [CrossRef] [Medline]
    35. Burton C, Szentagotai TA, McKinstry B, Matheson C, Matu S, Moldovan R, et al. Pilot randomised controlled trial of Help4Mood, an embodied virtual agent-based system to support treatment of depression. J Telemed Telecare 2016 Sep;22(6):348-355. [CrossRef] [Medline]
    36. Fulmer R, Joerin A, Gentile B, Lakerink L, Rauws M. Using Psychological Artificial Intelligence (Tess) to Relieve Symptoms of Depression and Anxiety: Randomized Controlled Trial. JMIR Ment Health 2018 Dec 13;5(4):e64 [FREE Full text] [CrossRef] [Medline]
    37. Bennion MR, Hardy GE, Moore RK, Kellett S, Millings A. Usability, Acceptability, and Effectiveness of Web-Based Conversational Agents to Facilitate Problem Solving in Older Adults: Controlled Study. J Med Internet Res 2020 Mar 12:e16794 [FREE Full text] [CrossRef] [Medline]
    38. Gaffney H, Mansell W, Edwards R, Wright J. Manage Your Life Online (MYLO): a pilot trial of a conversational computer-based intervention for problem solving in a student sample. Behav Cogn Psychother 2014 Nov;42(6):731-746. [CrossRef] [Medline]
    39. Depp CA, Ceglowski J, Wang VC, Yaghouti F, Mausbach BT, Thompson WK, et al. Augmenting psychoeducation with a mobile intervention for bipolar disorder: a randomized controlled trial. J Affect Disord 2015 Mar 15;174:23-30. [CrossRef] [Medline]
    40. Oh J, Jang S, Kim H, Kim J. Efficacy of mobile app-based interactive cognitive behavioral therapy using a chatbot for panic disorder. Int J Med Inform 2020 Aug;140:104171. [CrossRef] [Medline]
    41. Freeman D, Haselton P, Freeman J, Spanlang B, Kishore S, Albery E, et al. Automated psychological therapy using immersive virtual reality for treatment of fear of heights: a single-blind, parallel-group, randomised controlled trial. Lancet Psychiatry 2018 Dec;5(8):625-632. [CrossRef] [Medline]
    42. Jang S, Kim J, Kim S, Hong J, Kim S, Kim E. Mobile app-based chatbot to deliver cognitive behavioral therapy and psychoeducation for adults with attention deficit: A development and feasibility/usability study. Int J Med Inform 2021 Jun;150:104440. [CrossRef] [Medline]
    43. Prochaska JJ, Vogel EA, Chieng A, Kendra M, Baiocchi M, Pajarito S, et al. A Therapeutic Relational Agent for Reducing Problematic Substance Use (Woebot): Development and Usability Study. J Med Internet Res 2021 Mar 23;23(3):e24850. [CrossRef]
    44. So R, Furukawa TA, Matsushita S, Baba T, Matsuzaki T, Furuno S, et al. Unguided Chatbot-Delivered Cognitive Behavioural Intervention for Problem Gamblers Through Messaging App: A Randomised Controlled Trial. J Gambl Stud 2020 Dec;36(4):1391-1407. [CrossRef] [Medline]
    45. de Gennaro M, Krumhuber EG, Lucas G. Effectiveness of an Empathic Chatbot in Combating Adverse Effects of Social Exclusion on Mood. Front Psychol 2019;10:3061 [FREE Full text] [CrossRef] [Medline]
    46. Bendig E, Erb B, Meißner D, Bauereiß N, Baumeister H. Feasibility of a Software agent providing a brief Intervention for Self-help to Uplift psychological wellbeing ("SISU"). A single-group pretest-posttest trial investigating the potential of SISU to act as therapeutic agent. Internet Interv 2021 Apr;24:100377 [FREE Full text] [CrossRef] [Medline]
    47. Hudlicka E. Virtual training and coaching of health behavior: example from mindfulness meditation training. Patient Educ Couns 2013 Aug;92(2):160-166 [FREE Full text] [CrossRef] [Medline]
    48. Dworkin MS, Lee S, Chakraborty A, Monahan C, Hightow-Weidman L, Garofalo R, et al. Acceptability, Feasibility, and Preliminary Efficacy of a Theory-Based Relational Embodied Conversational Agent Mobile Phone Intervention to Promote HIV Medication Adherence in Young HIV-Positive African American MSM. AIDS Educ Prev 2019 Feb;31(1):17-37. [CrossRef] [Medline]
    49. Echeazarra L, Pereira J, Saracho R. TensioBot: a Chatbot Assistant for Self-Managed in-House Blood Pressure Checking. J Med Syst 2021 Mar 15;45(4):54. [CrossRef] [Medline]
    50. Gong E, Baptista S, Russell A, Scuffham P, Riddell M, Speight J, et al. My Diabetes Coach, a Mobile App-Based Interactive Conversational Agent to Support Type 2 Diabetes Self-Management: Randomized Effectiveness-Implementation Trial. J Med Internet Res 2020 Nov 05;22(11):e20322 [FREE Full text] [CrossRef] [Medline]
    51. Guhl E, Althouse AD, Pusateri AM, Kimani E, Paasche-Orlow MK, Bickmore TW, et al. The Atrial Fibrillation Health Literacy Information Technology Trial: Pilot Trial of a Mobile Health App for Atrial Fibrillation. JMIR Cardio 2020 Sep 04;4(1):e17162 [FREE Full text] [CrossRef] [Medline]
    52. Magnani JW, Schlusser CL, Kimani E, Rollman BL, Paasche-Orlow MK, Bickmore TW. The Atrial Fibrillation Health Literacy Information Technology System: Pilot Assessment. JMIR Cardio 2017;1(2):e7 [FREE Full text] [CrossRef] [Medline]
    53. Hauser-Ulrich S, Künzli H, Meier-Peterhans D, Kowatsch T. A Smartphone-Based Health Care Chatbot to Promote Self-Management of Chronic Pain (SELMA): Pilot Randomized Controlled Trial. JMIR Mhealth Uhealth 2020 Apr 03;8(4):e15806 [FREE Full text] [CrossRef] [Medline]
    54. Hunt M, Miguez S, Dukas B, Onwude O, White S. Efficacy of Zemedy, a Mobile Digital Therapeutic for the Self-management of Irritable Bowel Syndrome: Crossover Randomized Controlled Trial. JMIR Mhealth Uhealth 2021 May 20;9(5):e26152 [FREE Full text] [CrossRef] [Medline]
    55. Krishnakumar A, Verma R, Chawla R, Sosale A, Saboo B, Joshi S, et al. Evaluating Glycemic Control in Patients of South Asian Origin With Type 2 Diabetes Using a Digital Therapeutic Platform: Analysis of Real-World Data. J Med Internet Res 2021 Mar 25;23(3):e17908 [FREE Full text] [CrossRef] [Medline]
    56. McDonald DD, Walsh S, Vergara C, Gifford T. Effect of a virtual pain coach on pain management discussions: a pilot study. Pain Manag Nurs 2013 Dec;14(4):200-209. [CrossRef] [Medline]
    57. McDonald DD, Walsh S, Vergara C, Gifford T, Weiner DK. The effect of a Spanish virtual pain coach for older adults: a pilot study. Pain Med 2012 Nov;13(11):1397-1406. [CrossRef] [Medline]
    58. Owens OL, Felder T, Tavakoli AS, Revels AA, Friedman DB, Hughes-Halbert C, et al. Evaluation of a Computer-Based Decision Aid for Promoting Informed Prostate Cancer Screening Decisions Among African American Men: iDecide. Am J Health Promot 2019 Feb;33(2):267-278. [CrossRef] [Medline]
    59. Philip P, Dupuy L, Morin CM, de Sevin E, Bioulac S, Taillard J, et al. Smartphone-Based Virtual Agents to Help Individuals With Sleep Concerns During COVID-19 Confinement: Feasibility Study. J Med Internet Res 2020 Dec 18;22(12):e24268 [FREE Full text] [CrossRef] [Medline]
    60. Kowatsch T, Schachner T, Harperink S, Barata F, Dittler U, Xiao G, et al. Conversational Agents as Mediating Social Actors in Chronic Disease Management Involving Health Care Professionals, Patients, and Family Members: Multisite Single-Arm Feasibility Study. J Med Internet Res 2021 Feb 17;23(2):e25060 [FREE Full text] [CrossRef] [Medline]
    61. Stephens TN, Joerin A, Rauws M, Werk LN. Feasibility of pediatric obesity and prediabetes treatment support through Tess, the AI behavioral coaching chatbot. Transl Behav Med 2019 May 16;9(3):440-447. [CrossRef] [Medline]
    62. Watson A, Bickmore T, Cange A, Kulshreshtha A, Kvedar J. An internet-based virtual coach to promote physical activity adherence in overweight adults: randomized controlled trial. J Med Internet Res 2012 Jan 26;14(1):e1 [FREE Full text] [CrossRef] [Medline]
    63. Bickmore TW, Silliman RA, Nelson K, Cheng DM, Winter M, Henault L, et al. A randomized controlled trial of an automated exercise coach for older adults. J Am Geriatr Soc 2013 Oct;61(10):1676-1683. [CrossRef] [Medline]
    64. Bickmore TW, Schulman D, Sidner C. Automated interventions for multiple health behaviors using conversational agents. Patient Educ Couns 2013 Aug;92(2):142-148 [FREE Full text] [CrossRef] [Medline]
    65. Davis CR, Murphy KJ, Curtis RG, Maher CA. A Process Evaluation Examining the Performance, Adherence, and Acceptability of a Physical Activity and Diet Artificial Intelligence Virtual Health Assistant. Int J Environ Res Public Health 2020 Dec 07;17(23):1-14 [FREE Full text] [CrossRef] [Medline]
    66. Maher CA, Davis CR, Curtis RG, Short CE, Murphy KJ. A Physical Activity and Diet Program Delivered by Artificially Intelligent Virtual Health Coach: Proof-of-Concept Study. JMIR Mhealth Uhealth 2020 Jul 10;8(7):e17558 [FREE Full text] [CrossRef] [Medline]
    67. Edwards RA, Bickmore T, Jenkins L, Foley M, Manjourides J. Use of an interactive computer agent to support breastfeeding. Matern Child Health J 2013 Dec;17(10):1961-1968. [CrossRef] [Medline]
    68. Friederichs S, Bolman C, Oenema A, Guyaux J, Lechner L. Motivational interviewing in a Web-based physical activity intervention with an avatar: randomized controlled trial. J Med Internet Res 2014 Feb 13;16(2):e48 [FREE Full text] [CrossRef] [Medline]
    69. Gardiner PM, McCue KD, Negash LM, Cheng T, White LF, Yinusa-Nyahkoon L, et al. Engaging women with an embodied conversational agent to deliver mindfulness and lifestyle recommendations: A feasibility randomized control trial. Patient Educ Couns 2017 Sep;100(9):1720-1729. [CrossRef] [Medline]
    70. Jack B, Bickmore T, Hempstead M, Yinusa-Nyahkoon L, Sadikova E, Mitchell S, et al. Reducing Preconception Risks Among African American Women with Conversational Agent Technology. J Am Board Fam Med 2015;28(4):441-451 [FREE Full text] [CrossRef] [Medline]
    71. Jack BW, Bickmore T, Yinusa-Nyahkoon L, Reichert M, Julce C, Sidduri N, et al. Improving the health of young African American women in the preconception period using health information technology: a randomised controlled trial. Lancet Digit Health 2020 Sep;2(9):e475-e485 [FREE Full text] [CrossRef] [Medline]
    72. Gardiner P, Bickmore T, Yinusa-Nyahkoon L, Reichert M, Julce C, Sidduri N, et al. Using Health Information Technology to Engage African American Women on Nutrition and Supplement Use During the Preconception Period. Front Endocrinol (Lausanne) 2020;11:571705 [FREE Full text] [CrossRef] [Medline]
    73. King AC, Bickmore TW, Campero MI, Pruitt LA, Yin JL. Employing virtual advisors in preventive care for underserved communities: results from the COMPASS study. J Health Commun 2013;18(12):1449-1464 [FREE Full text] [CrossRef] [Medline]
    74. King AC, Campero I, Sheats JL, Castro Sweet CM, Garcia D, Chazaro A, et al. Testing the comparative effects of physical activity advice by humans vs. computers in underserved populations: The COMPASS trial design, methods, and baseline characteristics. Contemp Clin Trials 2017 Oct;61:115-125 [FREE Full text] [CrossRef] [Medline]
    75. King AC, Campero MI, Sheats JL, Castro Sweet CM, Hauser ME, Garcia D, et al. Effects of Counseling by Peer Human Advisors vs Computers to Increase Walking in Underserved Populations: The COMPASS Randomized Clinical Trial. JAMA Intern Med 2020 Nov 01;180(11):1481-1490 [FREE Full text] [CrossRef] [Medline]
    76. Kramer J, Künzler F, Mishra V, Smith SN, Kotz D, Scholz U, et al. Which Components of a Smartphone Walking App Help Users to Reach Personalized Step Goals? Results From an Optimization Trial. Ann Behav Med 2020 Jun 12;54(7):518-528 [FREE Full text] [CrossRef] [Medline]
    77. Maeda E, Miyata A, Boivin J, Nomura K, Kumazawa Y, Shirasawa H, et al. Promoting fertility awareness and preconception health using a chatbot: a randomized controlled trial. Reproductive BioMedicine Online 2020 Dec;41(6):1133-1143. [CrossRef] [Medline]
    78. Piao M, Ryu H, Lee H, Kim J. Use of the Healthy Lifestyle Coaching Chatbot App to Promote Stair-Climbing Habits Among Office Workers: Exploratory Randomized Controlled Trial. JMIR Mhealth Uhealth 2020 May 19;8(5):e15085 [FREE Full text] [CrossRef] [Medline]
    79. Ly KH, Ly A, Andersson G. A fully automated conversational agent for promoting mental well-being: A pilot RCT using mixed methods. Internet Interv 2017 Dec;10:39-46 [FREE Full text] [CrossRef] [Medline]
    80. Suganuma S, Sakamoto D, Shimoyama H. An Embodied Conversational Agent for Unguided Internet-Based Cognitive Behavior Therapy in Preventative Mental Health: Feasibility and Acceptability Pilot Trial. JMIR Ment Health 2018 Jul 31;5(3):e10454 [FREE Full text] [CrossRef] [Medline]
    81. Samdal GB, Eide GE, Barth T, Williams G, Meland E. Effective behaviour change techniques for physical activity and healthy eating in overweight and obese adults; systematic review and meta-regression analyses. Int J Behav Nutr Phys Act 2017 Mar 28;14(1):42 [FREE Full text] [CrossRef] [Medline]
    82. Webb TL, Joseph J, Yardley L, Michie S. Using the internet to promote health behavior change: a systematic review and meta-analysis of the impact of theoretical basis, use of behavior change techniques, and mode of delivery on efficacy. J Med Internet Res 2010 Feb 17;12(1):e4 [FREE Full text] [CrossRef] [Medline]
    83. Dalgetty R, Miller CB, Dombrowski SU. Examining the theory-effectiveness hypothesis: A systematic review of systematic reviews. Br J Health Psychol 2019 May;24(2):334-356. [CrossRef] [Medline]
    84. Slade M. Implementing shared decision making in routine mental health care. World Psychiatry 2017 Jun;16(2):146-153 [FREE Full text] [CrossRef] [Medline]
    85. Furnham A, Swami V. Mental Health Literacy: A Review of What It Is and Why It Matters. International Perspectives in Psychology 2018 Oct;7(4):240-257. [CrossRef]
    86. Jorm AF. Mental health literacy: empowering the community to take action for better mental health. Am Psychol 2012 Apr;67(3):231-243. [CrossRef] [Medline]
    87. Laranjo L, Dunn AG, Tong HL, Kocaballi AB, Chen J, Bashir R, et al. Conversational agents in healthcare: a systematic review. J Am Med Inform Assoc 2018 Sep 01;25(9):1248-1258 [FREE Full text] [CrossRef] [Medline]

    ‎
    AI: artificial intelligence
    BCT: behavior change technique
    BCTTv1: Behavior Change Technique Taxonomy version 1
    CA: conversational agent
    CENTRAL: Cochrane Central Register of Controlled Trials
    FIM: frequent itemset mining
    PRISMA-ScR: Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews
    RCT: randomized controlled trial

    Edited by A Mavragani; submitted 04.05.22; peer-reviewed by L Van Rhoon, M Jalan; comments to author 01.07.22; revised version received 05.08.22; accepted 23.08.22; published 03.10.22

    Copyright

    ©Laura Martinengo, Ahmad Ishqi Jabir, Westin Wei Tin Goh, Nicholas Yong Wai Lo, Moon-Ho Ringo Ho, Tobias Kowatsch, Rifat Atun, Susan Michie, Lorainne Tudor Car. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 03.10.2022.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.