Sage Journals: Discover world-class research

Abstract

Objective

With the increasing global burden of chronic diseases, there is the potential for conversational agents (CAs) to assist people in actively managing their conditions. This paper reviews different types of CAs used for chronic condition management, delving into their characteristics and the chosen study designs. This paper also discusses the potential of these CAs to enhance the health and well-being of people with chronic conditions.

Methods

A search was performed in February 2023 on PubMed, ACM Digital Library, Scopus, and IEEE Xplore. Studies were included if they focused on chronic disease management or prevention and if systems were evaluated on target user groups.

Results

The 42 selected studies explored diverse types of CAs across 11 health conditions. Personalization varied, with 25 CAs not adapting message content, while others incorporated user characteristics and real-time context. Only 12 studies used medical records in conjunction with CAs for conditions like diabetes, mental health, cardiovascular issues, and cancer. Despite measurement method variations, the studies predominantly emphasized improved health outcomes and positive user attitudes toward CAs.

Conclusions

The results underscore the need for CAs to adapt to evolving patient needs, customize interventions, and incorporate human support and medical records for more effective care. It also highlights the potential of CAs to play a more active role in helping individuals manage their conditions and notes the value of linguistic data generated during user interactions. The analysis acknowledges its limitations and encourages further research into the use and potential of CAs in disease-specific contexts.

Keywords

Conversational agent chatbot‌chronic disease dialogue ‌mobile health

Introduction

In recent years, there has been a significant rise in the prevalence of chronic diseases, resulting in a growing number of individuals living with at least one chronic health condition.¹ Chronic conditions have enduring and persistent effects, necessitating patients and healthcare professionals to navigate complex lifestyle and behavioral adjustments and engage in long-term management.² Moreover, these conditions reduce quality of life and life expectancy and can escalate personal healthcare expenses due to disability, frequent hospitalizations, and the need for multiple treatment procedures.

In healthcare, the continuum of care³ is the concept of an integrated care system that guides and tracks patients over time through every subsequent step of health services. The care system contains a range of healthcare and social services, from preventive care to acute treatment, rehabilitation, and long-term care. Over time, the patient goes through various parts of the system, depending on the stage of the treatment or the need for care. The continuum of care aims to provide patients with the right care at the right time and in the most appropriate setting (e.g., home, rehabilitation centers, or hospitals), promoting better health outcomes and patient satisfaction.

Alongside work performed by healthcare professionals, there are often health-related activities that patients and their informal caregivers undertake in managing their health conditions, sometimes referred to patient work.⁴ The activities include various processes, from cognitive to physical, performed individually or collaboratively with others, including family and community members, and can be classified as visible when they are acknowledged and valued by others or invisible when they are taken for granted by others and consequently undervalued.⁵

Digital technologies have the potential to empower people, giving them a sense of agency and control, allowing them to extend their skills and knowledge and giving them access to experiences and functions that people did not have before.⁶ Technologies can reduce the burden of patient work or self-management of health conditions and serve as accessible alternatives to in-person support and supervision. One such technology is conversational agents (CAs), computer systems that imitate human-like conversations using natural language user interfaces involving images, text, and voice.⁷ CAs potentially offer the advantages of scalability, reduced costs, lowered stigma, and personalized health support available at any time.^8,9 CAs can be delivered through text or speech, making them versatile for different target groups, including children and older individuals. CAs can address various healthcare needs, such as mental health management support,¹⁰ aid in chronic disease self-management,¹¹ and lifestyle change facilitation, e.g., physical activity and dietary modifications.¹²

However, existing digital technologies designed for patients with chronic conditions face challenges adapting to changing health needs and goals. Specific subgroups significantly differ in healthcare preferences and goals and require different information and recommendations.¹³ Moreover, patients evolve and change the apps they use throughout the trajectory of their condition, from diagnosis to long-term care, as their health goals change over the course of their disease.¹³ The chronic illness trajectory model¹⁴ describes how the course of illness varies for each patient and changes over time. Patients may shift between different illness phases repeatedly in unpredictable and inconsistent ways, as their conditions fluctuate.

Digital systems that are not only focused on one specific stage of the disease but also adapt to the changing needs of users can enable sustained app usage by reducing the necessity to search for more appropriate apps. Such systems would facilitate continuous tracking of healthcare data over time, supporting individuals in effectively managing their health and reducing the disruption caused by switching between different apps.

Despite the extensive research on the application of CAs in healthcare, to our knowledge, there has been no scoping review of different types of CAs in chronic disease management with no constraints on demographics. Existing reviews often restricted their focus to specific health areas, such as mental health,^15–17 smoking cessation,¹⁸ physical activity¹⁹ or body weight management^20,21; population, e.g., young (25 years and younger),¹⁷ or adults (18 years and older)²²; or agent types, e.g., embodied,^23,24 voice-based,²⁵ artificial intelligence-based^2,21,26,27 CAs or CAs with unconstrained natural language input,²⁸ e.g., free text or speech. Other reviews report solely on the evaluation outcomes, e.g., effectiveness and acceptability,^18,27,29 or on one of the CAs’ aspects, e.g., personalization³⁰ or design features.²⁴

This paper provides the results of a literature analysis aimed at addressing several research questions related to the use of CAs in the context of chronic disease management. The research questions explored in this article are:

What are the health domains, and the characteristics of the end users, targeted for CA interventions?

What are the characteristics of the CA studies?

Do CAs address continuum of care and patient work concepts for self-management? Do CAs adapt to the changing needs of users?

What are the CAs’ characteristics? How do different types of CAs map with patients’ profiles and health domains?

By addressing these research questions, this study aims to contribute valuable insights into the design, implementation, and optimization of CAs, as supportive tools in managing chronic health conditions. The findings presented in this paper seek to inform healthcare professionals and technology developers about the potential of these digital solutions to empower patients, improve health outcomes, and enhance the overall quality of life for different groups of people living with chronic diseases.

Methods

This study employs a scoping review approach,³¹ which provides a comprehensive overview of the evidence on the chosen research topic. Our approach, therefore, is to give a relative breadth in our review rather than to focus on evidence around a single clinical or systematic question.

Search strategy

The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines³² with the PRISMA-ScR checklist available as a Supplemental Material. We conducted a search of the literature using the electronic databases PubMed, ACM Digital Library, Scopus, and IEEE Xplore. These databases were chosen as they cover relevant aspects of health, technology, and interdisciplinary research and have also been used in other reviews covering similar topics. The main keywords used were “conversational agent” and “health” which were searched in titles and abstracts (Table 1).

Table 1.

Search terms.

Term [conversational agent]		Term [health]
chatbot* OR “conversational agent” OR “conversational system” OR “conversational assistan” OR “dialog system” OR “digital assistan” OR “digital coach” OR “relational agent” OR “virtual agent” OR “virtual assistan” OR “virtual coach” OR “automated messag” OR “assistance technolog” OR “smart speaker” OR “automated assistan” OR “intelligent personal assistan"	AND	health* OR mhealth* OR uhealth* OR ehealth* OR healthcare OR illness* OR disease* OR treatment OR prevent* OR lifestyle OR wellbeing OR well-being OR medic* OR patien* OR caregiv*

Study selection criteria

The review is based on literature dated between the beginning of January 2018 and the end of January 2023 published in English with full text available. We included articles on chronic disease management (e.g., treatment or monitoring) and prevention of specific chronic diseases. Articles that addressed general well-being were not included. Articles must provide some description of the theoretical basis, choice of the intervention components, or CA development process. We included both quantitative and qualitative studies, without any constraints on approach. Studies employing experimental designs that involve group comparisons must include details about the comparators used. Studies focusing only on technical aspects and design features of CAs (e.g., language models, systems, chatbot's personality) and studies using the Wizard of Oz method weren’t included.

Conversational agents

The review considers studies that involve interventions provided by a CA. This review does not include studies where CA focuses on video- and image-based diagnosis (e.g., skin cancer, gait, correct execution of exercises) and whose aim is screening before appointments, filling hospital forms, checking doctors’ availability, and giving diagnoses from the symptoms provided by the user.

Intervention

This review considers studies that evaluate intervention programs that include strategies to provide educational materials, help achieve health goals, and monitor health conditions. Interventions must not be intended for a hospital setting but can be tested in a laboratory environment. The intervention must be directly targeted toward the patients (e.g., not for emergency medical services, clinicians, or medical students).

Validity

This review considers studies that were tested on targeted user groups and have a report on the interventions’ impact on participants and/or participants’ experiences with the CA. Additionally, articles must have evaluation or validation of outcomes (e.g., acceptability, effectiveness) measured by reliable tools.

Results

The literature search process from the four databases resulted in a total of 3151 research papers. The steps of the screening methodology are presented in Figure 1, along with the number of studies that were excluded in each step. The first step consisted of removing duplicate records of common studies that were found across all databases using the reference manager Zotero, which resulted in 2406 unique articles. All remaining unique studies were then considered in the initial title- and abstract-based screening. Following the proposed exclusion criteria described above, 409 articles remained. Full-text assessment of the remaining papers was then performed, resulting in a final list of 42 articles. The most relevant exclusion factor was validity, as many studies included only preliminary methods and results or presented CAs without any evaluation results (e.g., studies at the pre-pilot stage), or CAs that were evaluated with subjects different to the intended population.

Figure 1.

PRISMA-ScR³² flow diagram of included studies.

Characteristics of included studies

The full list of included studies is provided in Tables 2–7. Article publication dates ranged from 2018 to 2023. Eighteen studies were conducted in the USA,^8,10,33–48 three each in Switzerland^49–51 and China,^52–54 two each in Australia,^55,56 France,^57,58 and the UK,^59,60 and one each in Canada,⁶¹ Japan,⁶² Singapore,⁶³ Saudi Arabia,⁶⁴ South Africa,⁶⁵ India,⁶⁶ Germany,⁶⁷ Italy,⁶⁸ Norway,⁶⁹ Spain,⁷⁰ the Netherlands,⁷¹ and across Europe.⁷² The total number of participants ranged from 6 to 4737 and one study⁶⁹ didn’t provide information on participant numbers. Participants aged between 5 and 86 years but not all articles provided information on age ranges.

Table 2.

Health domains and targeted users’ characteristics.

Health domain	Indication	Phase in chronic disease continuum	Targeted users	References
Addictions and substance abuse	Smoking cessation	During	Veterans	³⁴
	Substance use disorders	During	Alcohol users	³⁶
			Adults positive for substance misuse	³³
			Patients with methamphetamine use disorder	⁵²
Asthma	Asthma	During	Children with asthma	⁵⁰
Autoimmune disease	Celiac disease	During	Celiac patients	⁶⁴
Cancer and obesity	Cancer and obesity	After-care	Overweight or obese cancer survivors	⁴¹
Cancer	Cancer	Prevention	Patients without cancer who are eligible for cancer genetic evaluation	³⁷
	Cancer	After-care	Young people after cancer treatment	³⁹
	Breast	Prevention	Women	⁴⁸
	Breast	During	Women	^57,58
	Multiple myeloma	During	Patients with multiple myeloma	⁶⁷
Cardiovascular disease	Atrial fibrillation	During	People with atrial fibrillation	^8,40
Cardiovascular disease	Heart failure	During	Patients with heart failure	^35,61
Chronic pain and mental health	Musculoskeletal pain and depression and/or anxiety	During	Adults with musculoskeletal condition and self-reported symptoms of depression and/or anxiety	^10,43
Chronic pain	Chronic pain	During	People with chronic pain	⁴⁹
Chronic pain	Neck/shoulder pain/stiffness and low back pain	During	Workers with neck/shoulder pain/stiffness and low back pain	⁶²
Diabetes and mental health	Comorbid type 2 diabetes and depressive disorder	During	Patients with comorbid type 2 diabetes and depressive disorder	⁷⁰
Diabetes	Diabetes	Prevention	General population	⁶³
	Gestational	During	Pregnant women	⁶⁹
	Type 1	During	Adolescents with type 1 diabetes	⁷¹
	Type 1	During	Adolescents with type 1 diabetes and their parents	⁴⁷
	Type 2	During	Patients with type 2 diabetes	^55,56,65,66
	Types 1 and 2	During	Adults with diabetes	⁶⁸
Functional bowel disorder	Irritable bowel syndrome	During	People with irritable bowel syndrome	⁴²
Genetic condition	Sickle cell disease	During	Adults and young adults with sickle cell disease	⁷²
Mental health	Depression and/or anxiety	During	Students	³⁸
	Depression and/or anxiety	During	Adolescents with depression and anxiety	⁴⁵
	Complex, difficult-to-treat mental disorders	During	People with complex, difficult- to-treat mental disorders	⁴⁶
	Depressive symptoms	During	Young adults with depressive symptoms	⁵³
	Depression	During	Students with PHQ-9 score of 9 or higher^a	⁵⁴
	Mental health	During	Adults with mental health problems	⁶⁰
Obesity	Obesity	During	Adolescents with obesity	⁵¹
	Obesity	During	Children with obesity and their parents	⁵⁹
	Weight management	During	Postpartum African American/ Black women	⁴⁴

PHQ-9 score equal to or greater than nine was decided according to the average inclusion PHQ-9 score in previous depression trials.⁵⁴

Table 3.

Studies’ characteristics.

References	Participants’ characteristics					Duration	Study arms	Study approach	Measures
References	Description	N	Mean age (range)	Gender/sex	Race	Duration	Study arms	Study approach	Measures
⁸	With atrial fibrillation	151	71 (N/R)	N/R: female 49%	N/R	30 days	One	Qualitative, quantitative	AEI, satisfaction
¹⁰	With musculoskeletal condition	61	55 (18–83)	Sex: female 87%, male 12%, transgender 2%	White 90%	2 months	One	Quantitative	Feasibility, usability
³³	With >1 on the CAGE-AID^a	180	40 (18–65)	Sex: female 65%, male 35%	White 78%	8 weeks	Randomized Two: immediate and waitlist	Quantitative	Acceptability, feasibility, HO, TA
³⁴	Smokers attempting to quit smoking	6	56.3 (37–76)	Sex: female 33%, male 67%	White 50% and AA 50%	14 days	One	Qualitative, quantitative	Acceptability, feasibility
³⁵	Heart failure for 7 years on average	57	55 (>18)	Gender: female 35%, male 61%, missing 4%	Black 61%	90 days	Nonrandomized Two: smart speaker and visually animated and voice-enabled avatar	Quantitative	AEI
³⁶	Alcohol users	51	28 (21–55)	Gender: female 32%, male 62%	White 42%, Black 24%	1 week	One	Qualitative, quantitative	AEI, acceptability, feasibility
³⁷	Without cancer but with cancer family history	30	43.3 (25–60)	Sex: female 80%, male 19%	White 93%	10 minutes	One	Qualitative, quantitative	AEI
³⁸	Students	74	22.9 (>18)	Gender: female 70%, male 28%, nonconforming 1%	Asian 50%, White 43%	2–4 weeks	Randomized Three: control (information), 2 weeks and 4 weeks intervention	Quantitative	AEI, HO, satisfaction
³⁹	After cancer treatment with average 1.6 years post-completion of active cancer treatment	45	25 (19–29)	Gender: female 80%, male 20%	N/R	8 weeks	Randomized Two: immediate and waitlist	Qualitative, quantitative	Acceptability, AEI, feasibility, HO
⁴⁰	With chronic atrial fibrillation	120	72.1 (N/R)	Gender/sex: female 52%	White 92.5%	30 days	Randomized Two: control (usual care) and intervention	Qualitative, quantitative	Acceptability, AEI, HL, HRQOL, MA
⁴¹	With mean BMI^b 32.9 kg/m², 85.7% had stage 1 or 2 breast cancer	42	62.1 (N/R)	Sex: female 90%	N/R	4 weeks	Randomized Three: control (information), smart speaker and text-based intervention	Quantitative	Effectiveness
⁴²	With gastrointestinal symptoms	121	32 (18–63)	Gender: female 75%, male 25%	White 76%	8 weeks	Randomized Two: immediate and waitlist	Quantitative	AEI, HO, HRQOL
⁴³	With musculoskeletal condition	153	55 (18–86)	Sex: female 84%, male 16%	White 85%	2 months	Nonrandomized Three: control (usual care), usual care with in-person psychological counseling and with digital mental health intervention	Quantitative	HO
⁴⁴	Postpartum AA/Black women with BMI^b 32.5 ± 4.3 kg/m²	136	27.8 (18–40)	N/R: female 100%	AA/Black 100%	12 weeks	Randomized Two: control (usual care) and intervention	Quantitative	Acceptability, AEI, HO, self-efficacy, SS
⁴⁵	Had a new diagnosis of depression and anxiety in the past 3 months	18	14.7 (13–17)	Sex: female 88%, male 6%, missing 6%	White 88%	12 weeks	Randomized Two: immediate and waitlist	Qualitative, quantitative	Acceptability, AEI, feasibility, HO, safety^c, usability
⁴⁶	Enrolled in psychotherapy	73	37.3 (18–63)	Gender: female 89%, male 10%, genderqueer/androgynous 1%	N/R	4 weeks	One	Qualitative, quantitative	Feasibility, FDBT, HO, usability, usefulness
⁴⁷	With T1D, disease duration 2–13 years, and their parents	26	N/R (10–15)	Gender: female 73%, male 27%	White 85%	3 months	One	Qualitative, quantitative	AEI, feasibility, satisfaction, TA, usability
⁴⁸	Healthy females, 20% participants were identified as having high risk for breast cancer	30	22.5 (18–33)	Gender: female 100%	White 46.7%, Asian 36.7%	1 day	Randomized Three: control version of ECA, adaptive and non-adaptive ECA	Quantitative	Adoption, IMCB, HL, satisfaction
⁴⁹	With chronic pain	102	43.77 (>18)	Gender: female 80%, male 20%	N/R	8 weeks	Randomized Two: intervention and motivational messages unrelated to chronic pain	Quantitative	Acceptance, AEI, HO, IMCB, TA
⁵⁰	With an average 5.6 years since receiving the asthma diagnosis	49	12.04 (10–15)	Gender: female 33%, male 67%	N/R	30 days	One	Qualitative, quantitative	Acceptance, HL, TA, usability
⁵¹	With overweight or obesity	41	13.6 (10.9–16.9)	Sex: female 42%, male 58%	N/R	5.5 months	Randomized Two: control (usual care) and intervention	Quantitative	AEI, HO, usability
⁵²	With methamphetamine use disorder, accumulated months of methamphetamine use 99.89	49	38.85 (18–55)	N/R	N/R	30–45 minutes	One	Quantitative	HO, IMCB, SP
⁵³	Students with depressive symptoms, mean PHQ-9^d score 10.02	148	18.78 (17–21)	Gender: female 37%, male 63%	N/R	1 week	Randomized Three: an e-book, a general chatbot and mental health chatbot	Qualitative, quantitative	Acceptability, AEI, effectiveness, feasibility, HO, TA, usability
⁵⁴	Students with a PHQ-9^d score of nine or higher	83	23.08 (19–28)	Gender: female 55%, male 45%	N/R	16 weeks	Randomized Two: bibliotherapy and chatbot	Qualitative, quantitative	AEI, HO, satisfaction, TA
⁵⁵	T2D with mean HbA1c 7.3%, 56 mmol/mol	93	55.4 (N/R)	Gender/sex: female 47%	N/R	12 months	One	Qualitative, quantitative	Acceptability
⁵⁶	T2D with the baseline mean HbA1c 7.3%, 56 mmol/mol	187	57 (>18)	Sex: female 41%, male 58%	N/R	12 months	Randomized Two: control (usual care) and intervention	Quantitative	AEI, effectiveness, HO, HRQOL, usability
⁵⁷	With breast cancer in treatment or remission	142	42 (>18)	N/R: female 100%	N/R	1 day	Randomized Two: responses from chatbot and responses from physician	Quantitative	AEI, CCP
⁵⁸	With breast cancer	4737	48 (13–65+)	N/R: female 89%, male 11%	N/R	1 year	One	Qualitative, quantitative	AEI, MA, usability
⁵⁹	26 families with at least one child	34 children	8.4 (5–12)	Sex: female 50%	N/R	12 weeks	Randomized Two: control (no treatment) and intervention	Qualitative, quantitative	Acceptability, HO
⁶⁰	With self-reported mental health problems	17	33.4 (22–67)	N/R: female 65%, male 35%	N/R	2 weeks	One	Qualitative, quantitative	AEI, HO
⁶¹	Patients with heart failure	20	57.8 (21–80)	Sex: female 45%, male 55%	N/R	4 weeks	One	Qualitative, quantitative	Acceptability, feasibility
⁶²	With neck/shoulder pain/stiffness and low back pain	94	42 (20–64)	Sex: female 23%, male 77%	N/R	12 weeks	Randomized Two: control (usual care) and intervention	Quantitative	AEI, HO
⁶³	With mean BMI^b 22.3 kg/m²	75	33.7 (>21)	Gender: male 38%	Chinese 80%	4 weeks	One	Quantitative	Acceptability, AEI, EIQOL, feasibility, HL, HO
⁶⁴	Celiac patients	60	N/R	N/R	N/R	90 days	Randomized Two: control (no treatment) and intervention	Quantitative	AEI, feasibility
⁶⁵	T2D (3708/4577 provided demographic information)	4577	51 (N/R)	Gender: female 55.7%, male 44%, other 0.3%	N/R	32 days	One	Qualitative, quantitative	Acceptability, adoption, appropriateness, feasibility, coverage^e, HO
⁶⁶	T2D	102	50.8 (N/R)	Gender: female 31%, male 69%	N/R	16 weeks	One	Quantitative	AEC, HO
⁶⁷	Users (patients, caregivers, HCPs, etc.)	141	N/R	N/R	N/R	N/R	One	Quantitative	Usability, UX
⁶⁸	T1D and T2D with mean duration of 10 years	13	30.08 (18–51)	N/R: female 77%	N/R	12 sessions of 10–20 minutes each	One	Qualitative, quantitative	AEI, HO, UX
⁶⁹	610 dialogues, tool was available to all pregnant women in Norway	N/R	N/R	N/R	N/R	Open ended	One	Quantitative	AEI, UIN
⁷⁰	T2D and depressive disorder	13	63.8 (44–83)	N/R: female 69%	N/R	9 months	One	Qualitative, quantitative	Acceptance, AEI, HO, IHR, MA, usability, usefulness
⁷¹	T1D	21	13.9 (12–18)	Gender: female 52%, male 48%	N/R	6–8 weeks	One	Qualitative, quantitative	HO, usability, UX
⁷²	SCD, 64% affected by the most clinically severe SCD genotypes (Hb SS, HbSβ0)	33	37.37 (19–59)	Gender: female 67%, male 33%	N/R	1 day	One	Qualitative, quantitative	AEI, usability

Note. N/R: not reported; BMI: body mass index; HCP: healthcare professionals; SCD: patients with sickle cell disease; T1D: type 1 diabetes; T2D: type 2 diabetes; AA: African American; ECA: embodied conversational agents. AEC: association between program engagement and the change in HbA1c, FBG, and PPBG levels; AEI: adherence, engagement, and interaction; CCP: comparison of the average scores obtained by the chatbot and by the physicians for each individual; EIQOL: efficacy of the intervention with regard to the participants’ quality of life; FDBT: frequency of coping via dialectical behavior therapy skills; HL: health literacy; HO: health outcomes; HRQOL: health-related quality of life; IHR: impact on healthcare resources (number of medical appointments per month for each patient); IMCB: intention and motivation to change behavior; MA: medication adherence; SS: social support; SP: self-perception on the importance of drug abstinence and confidence in stopping the drug use; TA: therapeutic alliance; UIN: users’ informational needs; UX: user experience.

A cut-point of >1 on the CAGE-AID (cut down, annoyed, guilty, eye opener-adapted to include drugs) has a sensitivity of 70% and specificity of 85% for identifying individuals with substance use disorders.³³

BMI between 25 and 29.9 kg/m² falls within the overweight range, and BMI 30 kg/m² or higher falls within the obese range. For the Asian and South Asian populations, overweight is when BMI is between 23 and 24.9 kg/m², and obesity is when BMI is greater than 25 kg/m².⁷³

Parents were requested to provide information on any instances of their child being hospitalized or seeking medical attention for depression or anxiety-related issues in the preceding 2 weeks up to 1 month.

PHQ-9 assesses depression symptoms; scores of 5–9 predominantly represented patients with either no depression or subthreshold depression; patients with scores of 10 and higher are more likely to be diagnosed with depression.⁷⁴

Number of people accessing the service and their demographic characteristics (age, gender, and language preferences).

Table 4.

CAs’ purposes in the selected studies.

Purpose	References
Collect data	^{35,51,52,61,64}
Deliver therapy and provide counseling	^{10,33,36,38,41–43,45,49,51,52,54,60}
Encourage to think about quitting	³⁴
Encourage physical activity	^41,42
Provide feedback, summary, and suggestions	^{35,36,41,44,52,55,70,71}
Provide information and education	^{9,11,16,18–22,24,26,28–33,36,38–42,45–55}
Provide support	^{10,38,43,45,51,54–56,68}
Send motivational and self-care messages	^61,62,66,71
Send notifications and reminders	^{10,50,58,59,62,70,71}
Symptom monitoring	^40,56

Table 5.

Factors impacting CA’s role in self-management.

References	CA personalization	Medical records	Human involvement
¹⁰	No	Use data in the study	Chat with expert
³⁵	No	Use data in the study	Study nurse monitored alerts and evaluated participants’ stability
³⁶	Provided individualized feedback about user's drinking habits, including information about risk factors and consequences	N/R	N/R
³⁷	Provided more or less information based on user's preferences	Use data in the study, collect new data in the CA, and store it in the EHR	N/R
³⁸	No	Possibility to integrate with existing EHR systems	N/R
⁴⁰	The content was tailored for individual use by using the user's name and appropriate time context	Use data in the study	N/R
⁴¹	The intervention intensity and frequency depended on participant's motivation to seek coaching. MyCoach used a reinforced recommendation system to learn about participant behavior to maximize rewards. The unidirectional coach customized messages based on individual factors like schedule, sensors, preferences, and progress.	Use data in the study	N/R
⁴³	No	Use data in the study	Chat with expert
⁴⁴	Provided personalized feedback on weight changes specific to the data entered	N/R	N/R
⁴⁵	Tailored the conversation to the present situation to help the adolescent develop emotion regulation skills in the context of their everyday life, offered, and guided the user through cognitive behavioral therapy-based psychoeducation and tools, tailored to the reported need	Collect and record data	N/R
⁴⁹	Personalized text messages	N/R	N/R
⁵⁰	Asked about user's emotional state and provided personalized feedback based on their answers	N/R	Communicate with HCP, patients, and family members
⁵¹	No	N/R	Monitor, include standardized counseling and ability to chat through app, patients were able to chat with each other
⁵²	No	N/R	Psychotherapists or social workers can add new subjects and refer to patients’ assessment and treatment records
⁵⁴	No	N/R	Professionals will intervene by telephone when the participants report a psychological emergency
⁵⁵	Delivered personalized support, monitoring, and motivational coaching. Algorithms were tailored according to the clinical targets and recommendations provided by each participant's general practitioner	N/R	N/R
⁵⁶	Personalized support, monitoring, and motivational coaching	Use data in the study	N/R
⁵⁷	Personalized text messages	N/R	N/R
⁵⁸	Personalized text messages	N/R	N/R
⁶¹	Generated an automated self-care message based on the data inputted and the patient's medical history	Use and store data in the app	N/R
⁶²	Offered tailored replies depending on responses	N/R	N/R
⁶⁶	Provided educational, behavioral, and motivational messaging specific to the data entered and in the context of the patient's previous clinical, lifestyle, and behavioral data	Use and store data in the app	Voice calls and chat with expert
⁶⁷	No	N/R	No, but has a function to search for local German support groups
⁷⁰	No	N/R	HCP could monitor patients
⁷¹	Provided personalized feedback, rewarded goal-aligned behaviors, and tailored medical content and coaching messages through shared decision-making with healthcare professionals	Integrate the findings in existing patient web portals for T1D	Message exchange between patients, their peers and their caregivers
⁷²	Addressed participants’ accountability by referring to earlier data entered, tasks, or activities performed	N/R	N/R

Note. The table includes only studies that contain at least one of the discussed components. N/R: not reported; T1D: type 2 diabetes; HCP: healthcare professionals.

Table 6.

Underpinning frameworks in the selected studies.

Health domain	Underpinning framework	References
Addictions and substance abuse	US Clinical Practice Guidelines	³⁴
	Brief motivational interviewing, technology acceptance model	³⁶
	Cognitive behavioral therapy, dialectical behavior therapy, motivational interviewing, mindfulness training	³³
	Cognitive behavioral therapy, motivational enhancement therapy, mindfulness-based relapse prevention	⁵²
Asthma	Behavior change techniques, experiential learning theory, self-determination theory, theory of planned behavior	⁵⁰
Autoimmune diseases	Chronic-disease extended model	⁶⁴
Cancer	N/R	^57,58,67
	Cognitive theory of multimedia learning, adult learning theory	³⁷
	Stress and coping theory, broaden-and-build theory of positive emotions	³⁹
	Heuristic-systematic model of information processing	⁴⁸
Cancer and obesity	N/R	⁴¹
Cardiovascular diseases	N/R	^35,40,61
Cardiovascular diseases	Chronic care model	⁸
Chronic pain	N/R	⁶²
Chronic pain	Cognitive behavioral therapy	⁴⁹
Chronic pain and mental health	Cognitive behavioral therapy, dialectical behavior therapy, deep breathing techniques, motivational interviewing, mindfulness training, sleep meditations	¹⁰
Chronic pain and mental health	Cognitive behavioral therapy, motivational interviewing, mindfulness training	⁴³
Diabetes	N/R	^55,65,69
	Social cognitive theory, self-determination theory	⁴⁷
	Behavior change techniques, transtheoretical model, social cognitive theory, gamification	⁵⁶
	Distance learning-based stress management techniques, COM-B model	⁶³
	Digital persuasion model, American Association of Diabetes Educators framework	⁶⁶
	Obesity-related behavioral intervention trials framework, mindfulness-based cognitive therapy, American Association of Diabetes Educators framework	⁶⁸
	Behavior change techniques, goal setting theory, persuasive system design model, self-determination theory	⁷¹
Diabetes and mental	N/R	⁷⁰
Functional bowel disorder	Cognitive behavioral therapy	⁴²
Genetic condition	World Health Organization’ handbooks on how to implement text-based mHealth interventions	⁷²
Mental health	Acceptance and commitment therapy, cognitive behavioral therapy, emotionally focused therapy, interpersonal psychotherapy, mindfulness-based cognitive therapy, motivational interviewing, solution-focused brief therapy, self-compassion therapy, transtheoretical model	³⁸
	Cognitive behavioral therapy, dialectical behavior therapy, interpersonal psychotherapy for adolescents	⁴⁵
	Dialectical behavior therapy	⁴⁶
	Cognitive behavioral therapy	⁵³
	Bibliotherapy, cognitive behavioral therapy	⁵⁴
	Method of levels, perceptual control theory	⁶⁰
Obesity	N/R	^44,59
Obesity	Behavior change techniques	⁵¹

Note. N/R: not reported.

Table 7.

CAs’ technical characteristics in the selected studies.

Delivery channel	User input modality	Agent output modality	References
Amazon Alexa	Free speech	Speech, VA	^59,67
Amazon Alexa and mobile phone	Free speech	Speech, VA	⁴¹
Amazon Alexa and mobile phone	No	Text, UC	⁴¹
Amazon Alexa and tablet-based app	Free speech	Speech, VA	³⁵
Amazon Alexa and tablet-based app	Free speech	Text, speech, ECA	³⁵
Digital platform	Fixed, (sometimes free) text	Text	³⁷
Digital platform	Fixed, free text	Text	⁶⁹
Gamification platform	No	Text, UC	⁷¹
Messaging app	Fixed text	Text, image	⁶³
	Fixed text	Text	^70,72
	Fixed text	Text and audio	⁶⁵
	Fixed, free text	Text	^38,68
	Free text	Text	³⁹
	Free text, speech	Text	⁵⁴
	N/R	Text	⁶⁴
	N/R	Text, image, and voice	⁵³
Mobile app	Fixed text	Speech, ECA	⁸
	Fixed text	Speech, RA	⁴⁰
	Fixed text	Text	^46,51
	Fixed text, speech	Speech, ECA	^55,56
	Fixed text	Text, media	⁶²
	Fixed, (sometimes free) text	Text, media	⁴⁹
	Fixed, (sometimes free) text	Text	⁵⁰
	Fixed, free text	Text	^10,43
	Free speech	Speech, VA	⁶¹
	N/R	Text	^42,66
	N/R	Text, RA	^33,45
	No	Text, UC	⁴⁴
Tablet-based app	Fixed text	Speech, ECA	⁵²
Tablet-based app	N/R	Speech, ECA	³⁴
Web app	Fixed text	Speech, ECA	⁴⁸
	Fixed text	Speech, RA	⁴⁷
	Fixed, free text, speech	Speech, ECA	³⁶
	Free text	Text, RA	⁶⁰
Web, mobile, and messaging app	Fixed, free text	Text	⁵⁸
Web, mobile, and messaging app	Free text	Text	⁵⁷

Note. ECA: embodied conversational agents; RA: relational agent; VA: voice assistant; UC: unidirectional coach.

What are the health domains and the characteristics of the users, targeted for CA interventions?

When creating technology for a specific health domain, it is essential to consider symptoms, treatment requirements, and challenges associated with that domain that people may encounter at different stages. Moreover, understanding users’ characteristics, such as demographics, medical history, cultural background, goals, and preferences, allows the content and interaction strategies to be tailored. This enhances engagement, adherence, and overall health outcomes.^75,76 Table 2 shows information about the health domains addressed by the CAs in the selected articles, providing the context of the disease, including phase, e.g., prevention, after-care or during (between onset and end of the disease), as well as information about the target users.

Health domains

Among the various types of chronic conditions, different types of diabetes (type 1, type 2, gestational diabetes, and prediabetes)^{47,55,56,63,65,66,68,69,71} and mental health issues (depression, depressive symptoms, anxiety, bipolar disorder, and other complex, difficult-to-treat mental disorders)^{38,45,46,53,54,60} received the most attention. Other types of chronic conditions included six studies on cancer (multiple myeloma, breast cancer, cancer genetic evaluation, and after cancer treatment),^{37,39,48,57,58,67} four studies on addictions and substance abuse (smoking cessation, alcohol and methamphetamine use disorder, and substance misuse),^33,34,36,52 four studies on cardiovascular disease (heart failure and atrial fibrillation),^8,35,40,61 three studies on obesity,^44,51,59 two studies on chronic pain,^49,62 and one study each on asthma,⁵⁰ autoimmune (celiac disease),⁶⁴ genetic condition (sickle cell disease),⁷² and functional bowel disorder (irritable bowel syndrome).⁴² Additionally, there were four studies focused on comorbid diseases: two on musculoskeletal conditions and mental health,^10,43 one on type 2 diabetes and depressive disorder,⁷⁰ and another one on obesity and cancer.⁴⁰

Most of the systems were aimed at supporting people who already have chronic diseases (37 studies). There were only three studies that highlighted their focus on prevention: one study focused on patients without cancer who were eligible for cancer genetic evaluation,³⁷ one on communicating breast cancer risk and the recommended medical guidelines to healthy women,⁴⁸ and another on promoting healthy lifestyle behavior changes with the focus on diabetes and prediabetes knowledge in the general population.⁶³ Two studies focused on after-disease care: one study concentrated on overweight or obese cancer survivors⁴¹ and another on young people after cancer treatment.³⁹

Targeted users’ characteristics

In the selected studies, people with chronic conditions, or the risk of acquiring one were the most common final targeted interaction recipients. Only three studies targeted the interaction of patient–parent dyads.^47,50,59

The age groups of target users varied. There were six studies focused on children and teenagers.^{45,47,50,51,59,71} Four studies targeted young adults.^38,39,53,54 The remaining 32 studies were for different groups of adults.

There was no information about the duration of illness, medication usage, number and type of comorbidities, employment and marital status or any other socio-demographic information of target users that might be important for intervention development. Only a few studies were targeted recipients additionally specified; two studies targeted university students,^38,54 one study focused on veterans,³⁴ one specifically targeted African American/Black women,⁴⁴ and one had a focus on workers.⁶²

What are the characteristics of the CA studies?

When conducting studies, it can be meaningful to test interventions with multiple study arms to compare them, evaluate their efficacy and inform decision-making. Conducting studies with participants representative of the target group enhances applicability and generalizability that improves the validity and relevance of findings for the interventions’ intended users. Knowing the intervention duration can be important for the understanding of the impact and effectiveness of the interventions. Moreover, chosen measure types guide the evaluation of health apps, ensuring evidence-based interventions and objective outcome assessment. Thus, effectiveness, a fundamental aspect of intervention assessment, can be measured through various outcome indicators chosen by researchers, such as symptom reduction, changes in health behavior and health indicators (e.g., blood pressure, glucose levels), improvements in quality of life, and others. These measures serve as quantifiable evidence of an intervention's success in achieving its intended goals. The characteristics of the studies are summarized in Table 3. If there were multiple race groups participating in the study, we mention only those above 20%. We reported gender and sex as explicitly stated in the studies. If the totals do not add up to 100, it means this information was not provided in the studies.

Participants

The age of the research participants corresponded to the targeted users’ descriptions since this was a criteria for inclusion in the review. There were six studies focused on children and teenagers aged from 5 years to 18 years old.^{45,47,50,51,59,71} Four studies targeted young adults, 17–29 years old.^38,39,53,54 One study had a wide range age of targeted users and participants from 13 to 65+ years old.⁵⁸ The remaining 31 studies involved different groups of adults. Three studies didn’t provide any information about participants,^64,67,69 except that they corresponded to the target users, e.g., already had the health condition or risk of acquiring the disease. Moreover, one of these studies, Baertsch et al.,⁶⁷ besides patients, had healthcare professionals and caregivers among the users since the program was advertised and had open access.

Apart from age, gender, or sex was frequently mentioned in the articles as a demographic characteristic of the participants. There were 17 articles that used the term “gender,” 12 that used the term “sex,” 2 that used both terms, and 11 did not specify. Unfortunately, it is often impossible to determine exactly what researchers meant when they used a particular term. Furthermore, some of the articles provided additional information on specific health indicators, such as body mass index (BMI) or disease duration, employment and marital statuses, race, and ethnicity. Having a more detailed description provides valuable insights for understanding the diverse demographics and health profiles of the participants, allowing conducting in-depth analysis of the interventions’ effectiveness and relevance across different populations.

Duration

There were two studies with unspecified duration.^67,68 Of those with specified domains, there was one study⁶⁹ that collected data over 20 weeks (there were two periods: 8 and 12 weeks), though it is not clear for how long and how many times each user engaged with the chatbot since the data were anonymous and the identification of users was not permitted. Five studies took only 1 day to finish the planned intervention.^{37,48,52,57,72} Two studies each were for 1 week and 2 weeks, nine studies were about 1 month long, seven studies had a 2 month duration, and eight studies had a 3 month duration. Several studies lasted more than 3 months: two studies for about 4 months,^54,66 one study for 5.5 months,⁵¹ one study for 9 months,⁷⁰ and 3 studies for 1 year.^55,56,58

Study arms

Most of the studies were single-armed (22/42). There were 15 two-arm studies, 11 of which compared intervention and control groups (see Table 3), including five that compared intervention and usual care,^{40,44,51,56,62} four that compared immediate intervention and waitlist control groups,^33,39,42,45 and two compared intervention and control groups with no treatment.^59,64 Another setup encountered is a comparison of two alternative interventions: responses from chatbot versus responses from physicians,⁵⁷ interaction with a chatbot concerning pain management versus reception motivational messages unrelated to chronic pain,⁴⁹ bibliotherapy versus chatbot interaction,⁵⁴ and smart speaker versus visually animated and voice-enabled avatar interaction.³⁵ There were also five three-arm studies, that compared an e-book, a general chatbot, and mental health chatbot⁵³; control group that only received a link to a book, 2 and 4 weeks intervention groups³⁸; control version, adaptive, and non-adaptive embodied conversational agent (ECA)⁴⁸; control group that received printed written information, smart speaker, and text-based intervention⁴¹; and usual orthopedic care without any specific mental health intervention, usual orthopedic care with in-person psychological counseling and usual orthopedic care with digital mental health intervention.⁴³

Measures

A diverse range of measures was used to assess the interventions in the selected studies. One of the most common measures were health outcomes (23/42) which were measured by physical assessments like BMI, glycated hemoglobin level, blood pressure, and heart rate,^{44,51,56,59,66,70} and questionnaires, such as Generalized Anxiety Disorder-7, Patient Health Questionnaire-9, International Physical Activity Questionnaire, Pittsburgh Sleep Quality Index, and Perceived Stress Scale.^{33,38,39,42–46,49,51–54,56,59,60,62,63,65,68,70–72} All the studies demonstrated improvement, but not always statistically significant.

The same metrics are often used to evaluate engagement, adherence, and interaction with apps and CAs, e.g., number, lengths and context of messages sent, and this can create challenges in distinguishing the specific aspect being assessed. Among the selected studies, 26 evaluated chats with CAs and activity in the apps (we called this measure Adherence, Engagement, and Interaction in Table 3), which was measured by a variety of methods, including analysis of interview transcripts,⁸ chat transcripts and metrics, e.g., number and length of messages,^{35,37–40,45,49,51,53,56,58,60,62–64,69,70} app usage, e.g., number of accesses to the app and videos watched,^{42,44,45,56,70} different questionnaires, e.g., Godspeed for assessing human-like traits of CA,³⁶ the Usefulness Scale for Patient Information Material for evaluating perceived usefulness⁷² and the User Engagement Scale–Short Form questionnaire for measuring self-reported user engagement,⁶⁸ and other surveys to evaluate various aspects, e.g., the level of confidence, attitude, and perceived quality of the answers.^{39,47,54,57,58} Many metrics based on chat and app usage provided quantitative values that were evaluated relative to each other, such as the number of messages sent weekly. The results indicated that participants found the CAs to be useful, reported overall positive attitudes, and showed high engagement and adherence rates. All other 26 measurements identified were used in fewer than 13 articles.

Effectiveness

Most of the studies didn’t talk explicitly about the effectiveness. It can be assumed that effectiveness of the intervention is determined by the results of the selected measures, as noted in the previous paragraph, such as health outcomes, usability, and engagement. However, there are several articles that mentioned why they didn’t provide information about effectiveness; some of the studies weren’t designed to test the effectiveness,⁶³ other couldn’t establish it for different reasons, e.g., the small sample size and/or insufficient intervention duration.^{45,46,51,59,60}

Do CAs address continuum of care and patient work concepts for self-management? Do CAs adapt to the changing needs of users?

CAs play a multifaceted role in chronic disease management, addressing the continuum of care by providing, for example, education, remote monitoring, and seamless health records integration that can enable access to comprehensive patient data that contributes to creating personalized and tailored interventions. CAs contribute to patient work through symptom tracking and coaching, empowering patients and caregivers to self-manage their health. Moreover, human involvement, e.g., relatives, peers, and healthcare professionals, in chronic condition management can provide social support and motivation for patients, enhancing adherence to the interventions and overall well-being.^77,78 Furthermore, the inclusion of medical records in health applications can facilitate the development of personalized interventions, resulting in enhanced health outcomes, greater user engagement, and heightened satisfaction.

The purpose of the reviewed CAs are presented in Table 4 and the factors impacting CA's role in self-management are shown in Table 5 for any study that mentioned the CA's factors.

CA purposes

In the selected studies, CAs had a diverse range of purposes. The most common ones were providing information and education (33/42), delivering cognitive therapy (12), and providing mental health and emotional support (11) (see Table 4). All other purposes, such as providing feedback, collecting data, sending reminders, and motivational messages and monitoring symptoms, were mentioned fewer than six times.

CA personalization

Twenty-five CAs didn’t adopt message content to users’ characteristics at all (see Table 5). Five studies mentioned that CAs sent “personalized text messages” and “tailored replies depending on users’ responses” without further explanation.^{49,56–58,62} In one study participants can choose to receive more or less details on the topics based on their preferences.³⁷ Based on participants’ goals, habits, and data entered during a conversation with CA, four CAs provided personalized feedback,^36,44,50,71 and four CAs sent personalized educational, motivational, and coaching messages.^41,55,61,66 One CA tailored the conversation to the user's present situation and to the reported need at that moment,⁴⁵ one CA referred to earlier data entered, tasks, or activities performed and added user's name⁷² and another CA added user's name and appropriate time context to the messages, e.g., “Good Afternoon John.”⁴⁰

Medical records

We have only considered the usage of medical records in conjunction with the CAs. For example, cases when medical records were used for recruiting participants are not demonstrated in Table 5. Thirty studies didn’t mention any connection to patients’ medical records (see Table 5). Nine studies used medical records data for the studies.^{10,35,37,40,41,43,56,61,66} One study recorded information after the study to participants’ electronic health record (EHR)³⁷ and one study used participants’ medical charts to confirm any additional mental health diagnoses and record their initial treatment plan.⁴⁵ Moreover, one study aimed to integrate the findings in existing patient web portals that support the care of patients with type 1 diabetes,⁷¹ and one study noted the possibility of integrating the chatbot's administration panel into the existing EHR system.³⁸ One study collected medical history and stored it in the app when the patient was onboarded to the program⁶¹ and one study stored personal health record files with additional diabetes-related information (e.g., laboratory reports and details of treating physician) that users logged in the mobile application.⁶⁶ Among these 12 studies, there were three on diabetes,^56,66,71 three on cardiovascular conditions,^35,40,61 two on mental health,^38,45 two on comorbid chronic pain and mental health,^10,43 one on cancer,³⁷ and one on comorbid cancer and obesity.⁴¹

Human involvement

Despite the evidence that human involvement, e.g., family members, peers, and healthcare professionals, in the management of chronic conditions can offer patients social support and motivation,^77,78 in 31 studies there was no human involvement in the intervention groups (see Table 5). Four studies allowed the participants to communicate with health experts, two studies only through chat,^10,43 one through chat and face-to-face interaction⁵⁰ and one through chat and voice calls.⁶⁶ In three studies, healthcare professionals monitor participants’ engagement and results.^35,52,70 In one study, healthcare professionals could intervene by telephone (Mental Assistance Hotline) when the participants reported that they needed emergency psychological assistance⁵⁴ and in another study, the app alerted healthcare professionals in case of a lack of chat interaction during more than 2 days and four standardized counseling on-site visits, two via phone and 10 minutes of interaction through chat app was included.⁵¹ Family members were included in one study by Kowatsch et al.,⁵⁰ CA could send SMS text messages to them, and healthcare professionals had a possibility to communicate with family members through the app. In Stasinaki et al.⁵¹ patients were able to chat with each other and in Klaassen et al.⁷¹ there was a message exchange between patients, their peers, and their caregivers. In two of the studies without human involvement the CA motivated users to join online peer-support patient communities or request health coaching from a human⁷² or search for local support groups.⁶⁷

The targeted health domains for the studies with human involvement were diabetes,^66,71 comorbid chronic pain and mental health,^10,43 mental health,⁵⁴ comorbid diabetes and mental health,⁷⁰ asthma,⁵⁰ cardiovascular conditions,³⁵ addictions and substance abuse,⁵² cancer,⁶⁷ and obesity.⁵¹

What are the CAs’ characteristics? How do different types of CAs map with patients’ profiles and health domains?

Choosing or designing CAs suitable to target users requires an understanding of their characteristics, such as type of CA, input and output modalities, as these influence the user experience and engagement. The input modality refers to how users interact with the CA, such as through text, speech, or by choosing from the responses menu, while the output modality relates to the CA's responses, whether in text, speech, or visual forms. Moreover, different types of CAs have unique capabilities and advantages. For instance, relational agents focus on building and maintaining long-term, social–emotional relationships with users, fostering a sense of trust and empathy⁷⁹; ECAs have the same properties as humans in face-to-face conversation, including producing and responding to verbal and nonverbal communication⁸⁰; and task-oriented dialogue agents are designed to perform a specific function or deliver a particular service, such as providing educational tutorials. These agents are different from general-purpose CAs, which are more versatile and capable of handling a wide range of tasks and conversations.⁸¹

Theoretical frameworks are essential when developing health interventions and applications as they provide a structured and evidence-based foundation, guiding the design, implementation, and evaluation process to ensure effectiveness and alignment with established health behavior principles.

Underpinning frameworks are presented in Table 6 and CAs’ technical characteristics are in Table 7.

Theoretical frameworks

One-third of all the studies (14/42) did not mention any theoretical base on which the interventions and the application design were based (see Table 6). The rest of the studies had a lot of variety in the theories, therapies, techniques, and frameworks that were considered when developing the studies. A total of 40 different theories were mentioned, with cognitive behavioral therapy, mindfulness (includes training and all mindfulness-based therapy and techniques), motivational interviewing (includes brief motivational interviewing), dialectical behavior therapy, behavior change techniques, and self-determination theory being referenced most frequently (10, 6, 5, 4, 3, and 3 times, respectively), with the remaining 33 theories having no more than two mentions each. All the frameworks can be categorized into several domains: psychotherapy and counseling theories (acceptance and commitment therapy, cognitive behavioral therapy, dialectical behavior therapy, emotionally focused therapy, interpersonal psychotherapy, motivational enhancement therapy, solution-focused brief therapy), health behavior change theories (behavior change techniques, goal setting theory, self-determination theory, social cognitive theory, stress and coping theory, theory of planned behavior, transtheoretical model), educational and learning theories (adult learning theory, cognitive theory of multimedia learning, experiential learning theory), healthcare models and guidelines (American Association of Diabetes Educators (AADE) framework, chronic care model, chronic-disease extended model, obesity-related behavioral intervention trials framework, US Clinical Practice Guidelines, World Health Organization's handbooks on how to implement text-based mHealth interventions), mind-body techniques (mindfulness techniques, sleep meditations, deep breathing techniques), technology and digital health frameworks (digital persuasion model, persuasive system design model, technology acceptance model).

CAs’ characteristics

The majority of CAs in the reviewed studies were text-based (28 out of 42) (see Table 7). Seven CAs only allowed users to choose an answer from the options provided^{46,51,62,63,65,70,72} and three CAs had several free text questions, e.g., provide the user's name.^37,49,50 Among these CAs, four sent media, like audio lessons and pictures with exercises, to users.^49,62,63,65 In three studies, participants could type their responses freely^39,57,60 and, in Liu et al.,⁵⁴ participants could also send voice messages to the chatbot. In six CAs, users could both choose responses from a menu and send free text messages or choose from the responses menu.^{10,38,43,58,68,69} Six studies didn’t specify users’ input format.^{33,42,45,53,64,66} There were three studies with unidirectional coaches where the CAs sent messages to the participant, but the participant could not communicate with the agent.^41,44,71 Four studies based their intervention on Amazon's Alexa,^35,59,61,67 including one that compared Amazon's Alexa with a visually animated and voice-enabled avatar.³⁵ Additionally, one study compared unidirectional coach with Amazon voice assistant.⁴¹ Ten studies used an ECA to provide interventions,^{8,34–36,40,47,48,52,55,56} the users’ input modes varied: most systems allowed interaction with the ECA only with predefined answer options,^8,48,52 while others allowed voice and/or free text messages.^36,55,56

Among the selected studies, there were five with relational agents designed to build and maintain long-term emotional relationships with their users.^{33,40,45,47,60} Three of them were text-based for mental health and addictions and substance abuse^33,45,60 and two were ECA targeting diabetes and cardiovascular issues.^40,47

Discussion

Advancements in technologies have facilitated a plethora of mobile apps specifically designed to support and improve health habits that give patients more control and a sense of agency, empowering them to take charge of their health.⁸² Existing digital technologies for patients with chronic conditions face challenges in adapting to changing health needs and goals, requiring diverse information and recommendations for specific subgroups.¹³ Implementing adaptable digital systems can enable sustained app usage, supporting individuals in effectively managing their health and reducing disruption caused by switching between different apps.

Relevant studies were obtained from the PubMed, ACM Digital Library, Scopus, and IEEE Xplore databases. The methodology adopted in this review aligned well with the PRISMA-ScR guidelines and checklist³² process, see the Supplemental Material for further details. The review included 42 studies reporting interventions delivered by CAs, targeting chronic diseases. Diabetes, mental health, and cancer are the diseases most commonly targeted by CA interventions, as opposed to other chronic conditions, such as autoimmune, genetic conditions, functional bowel disorders, and asthma. The review shows that current CAs address a wide variety of chronic diseases. Although it provides valuable insights, the extensive range of conditions complicates the ability to compare them within and between conditions. To address this, future research should focus on evaluating CAs specifically tailored to particular chronic diseases.

The overall trends can be summarized as follows:

Focus on stable chronic condition management

The analysis reveals that the majority of studies have primarily targeted individuals who already have chronic diseases. There seems to be a disproportionate emphasis on stable chronic condition management rather than on other stages of the disease where people might have other needs, for example, newly diagnosed patients or patients after crises. Only one study mentioned it was designed to address current and future patients’ needs.⁵⁸ This finding highlights a potential gap in leveraging digital interventions that adapt to the changing needs of people with chronic diseases, which could significantly impact the overall burden of these conditions on healthcare systems.

Limited representation of specific demographics

The research highlights the lack of studies focusing on certain demographic groups, including children/teenagers and older adults, which aligns with previous reviews.²² While interventions for adults aged 18 and above are relatively abundant, there is a noticeable gap in providing support for younger and older populations. Understanding the specific challenges and requirements of these groups is crucial for designing age-appropriate and inclusive digital health solutions.

Incomplete participant information

For this review, we selected studies featuring participants who matched the targeted user group. Yet, there is often limited information about the participants beyond basic demographics such as age and gender or sex. Since studies sometimes did not explicitly report all values beyond the primary gender or sex and the terms “gender” and “sex” can be used interchangeably, with “gender” increasingly used to describe biological variations traditionally assigned to “sex,”⁸³ comparing these studies becomes challenging. Marital status, ethnicity, and other socio-economic features are often missing. Moreover, studies failed to provide a thorough comprehension of participants’ needs, aligning with the previous reviews.^22,28 This lack of comprehensive participant information might hinder a thorough understanding of the interventions’ applicability and effectiveness across diverse populations.

Importance of human involvement

While CAs offer promising directions, the limited provision for human interaction within these systems is identified as a potential drawback. Human support from healthcare professionals, family members, and peers with similar health problems can provide emotional encouragement, motivation, and vital insights that are essential for successful long-term management of chronic conditions.^77,78

Limited integration of medical records

The review indicates that limited use of medical records by the CAs potentially hinder smooth communication and collaboration between individuals and healthcare professionals. Moreover, medical records in health apps can enable the creation of personalized and tailored interventions that lead to improved health outcomes and increased user engagement and satisfaction.

Insufficient information on theoretical framework

Several studies didn’t provide information about the theoretical framework guiding the design and implementation of interventions. This omission raises concerns about the basis for the interventions and highlights the importance of incorporating robust theoretical foundations in future research and development.

Lack of unified evaluation measures

In accordance with prior reviews, we also found that the evaluation measures used to assess the effectiveness of the interventions and their impact on chronic conditions varied widely across studies.^15,28 This lack of standardization makes it challenging to compare and generalize findings.

Limited use of technologies and user information

There is potential for CAs to assist people in actively managing their conditions instead of passively consuming information. However, most current CAs have dialogue management systems that do not consider user preferences, goals, or history of interaction with the system. To better serve user needs, the design of these agents should evolve. The linguistic data generated by users during interaction with CAs holds the capacity to provide insights into users’ emotional and physical states. These data can help people with self-management and provide valuable information for the patients’ support team. A scoping review on psychology-oriented ECAs revealed that most agents were still in their initial phases of development and evaluation,²³ aligning with our results. A systematic review on health-related CAs with unconstrained language input capabilities found limited use of agent-based systems capable of handling complex dialogues and allowing users to lead conversation.²⁸ The review found only one study evaluated such systems in health contexts, but agents weren't designed for health-related queries. The authors indicated the requirement for large training datasets as a major drawback, potentially slowing their adoption in health applications.

Another factor to consider is that, at the time of writing, advances in the application of large language models offer significant opportunities in various fields, including healthcare. However, this has not yet been reflected in the available literature. It will be interesting to observe how this technological progress could potentially address the highlighted issues.

Strengths and limitations

The main strength of this review is its novelty. To the best of our knowledge, this is the first scoping review addressing different types of CAs in chronic disease management with no constraints on demographics. Nevertheless, it is essential to acknowledge that this analysis has its limitations. First, given that the research question is very broad, a detailed review of the use of CAs in different chronic diseases was not possible. Second, the review was exclusively focused on text and voice CAs, omitting those utilizing images and video. Third, we included articles in English with full text available only and the search in other databases, such as CINAHL and Web of Science, was not conducted. Fourth, our analysis only incorporated features, e.g., theoretical frameworks, explicitly mentioned in the articles. Therefore, certain characteristics may not be fully represented in our review, potentially limiting the generalizability of the conclusions due to potential bias.

Conclusions

This paper highlights the current state of CAs for chronic condition management and raises important considerations for future research and development in this field. The findings emphasize the health application requirements to adapt to the changing needs during the course of illness that varies for each patient and changes over time, customize interventions based on specific user subgroups, and improve the reporting of study participant characteristics to enhance the applicability of the findings. It also underscores the importance of incorporating human support and medical records integration within digital health solutions to provide more effective care for individuals living with chronic conditions. Additionally, studies should clearly articulate the theoretical frameworks guiding their interventions, and efforts should be made to standardize evaluation measures to facilitate meaningful comparisons between studies. Future research could more closely explore studies on each particular disease to gain a deeper understanding of CAs’ use and potential in disease management.

Supplemental Material

sj-docx-1-dhj-10.1177_20552076241277693 - Supplemental material for Exploring the characteristics of conversational agents in chronic disease management interventions: A scoping review

Supplemental material, sj-docx-1-dhj-10.1177_20552076241277693 for Exploring the characteristics of conversational agents in chronic disease management interventions: A scoping review by Ekaterina Uetova, Lucy Hederman, Robert Ross and Dympna O’Sullivan in DIGITAL HEALTH

Footnotes

Acknowledgments

We would like to thank Róisín Guilfoyle, research support librarian at Technological University Dublin, for her assistance and guidance in this research.

Contributorship

Ekaterina Uetova researched literature and wrote the first draft of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

This work was conducted with the financial support of the Science Foundation Ireland Centre for Research Training in Digitally Enhanced Reality (d-real) under Grant No. 18/CRT/6224 and the ADAPT SFI Research Centre for AI-Driven Digital Content Technology under Grant No. 13/RC/2106_P2. For the purpose of Open Access, the author has applied a CC BY public copyright license to any Author Accepted Manuscript version arising from this submission.

Ethical approval

Ethical approval was not required as this is a review paper.

Guarantor

Ekaterina Uetova.

Reporting guidelines

Completed Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) flowchart is included as a cited figure, and the PRISMA-ScR checklist is available as a .

ORCID iD

Ekaterina Uetova

Supplemental material

Supplemental material for this article is available online.

References

Anderson

Durstine

. Physical activity, exercise, and chronic diseases: a brief review. Sports Med Health Sci 2019; 1: 3–10.

Schachner

Keller

Wangenheim

. Artificial intelligence-based conversational agents for chronic conditions: systematic literature review. J Med Internet Res 2020; 22: e20701.

American Psychological Association. Continuum of care. APA Dictionary of Psychology. https://dictionary.apa.org/ (accessed 23 June 2023).

Valdez

Holden

Novak

, et al. Transforming consumer health informatics through a patient work framework: connecting patients to context. J Am Med Inform Assoc JAMIA 2015; 22: 2–10.

Corbin

Strauss

. Managing chronic illness at home: three lines of work. Qual Sociol 1985; 8: 224–247.

Schneider

Eiband

Ullrich

, et al. Empowerment in HCI – A survey and framework. In: Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, Montreal, QC, Canada, pp.1–14: ACM.

Hebbar

. Augmented intelligence: enhancing human capabilities. In: 2017 Third International Conference on Research in Computational Intelligence and Communication Networks (ICRCICN), 2017, pp.251–254.

Bickmore

Kimani

Trinh

, et al. Managing chronic conditions with a smartphone-based conversational virtual agent. In: Proceedings of the 18th International Conference on Intelligent Virtual Agents, New York, NY, USA, pp.119–124: Association for Computing Machinery.

Pereira

Díaz

. Using health chatbots for behavior change: a mapping study. J Med Syst 2019; 43: 135.

10.

Leo

Schuelke

Hunt

, et al. A digital mental health intervention in an orthopedic setting for patients with symptoms of depression and/or anxiety: feasibility prospective cohort study. JMIR Form Res 2022; 6: e34889.

11.

Ter Stal

Sloots

Ramlal

, et al. An embodied conversational agent in an eHealth self-management intervention for chronic obstructive pulmonary disease and chronic heart failure: exploratory study in a real-life setting. JMIR Hum Factors 2021; 8: e24110.

12.

Maher

Davis

Curtis

, et al. A physical activity and diet program delivered by artificially intelligent virtual health coach: proof-of-concept study. JMIR MHealth UHealth 2020; 8: e17558.

13.

Yin

Jung

Coiera

, et al. How patient work changes over time for people with multimorbid type 2 diabetes: qualitative study. J Med Internet Res 2021; 23: e25992.

14.

Corbin

. The Corbin and Strauss chronic illness trajectory model: an update. Sch Inq Nurs Pract 1998; 12: 33–41.

15.

Jabir

Martinengo

Lin

, et al. Evaluating conversational agents for mental health: scoping review of outcomes and outcome measurement instruments. J Med Internet Res 2023; 25: e44548.

16.

Gaffney

Mansell

Tai

. Conversational agents in the treatment of mental health problems: mixed-method systematic review. JMIR Ment Health 2019; 6: e14166.

17.

Balan

Dobrean

Poetar

. Use of automated conversational agents in improving young population mental health: a scoping review. Npj Digit Med 2024; 7: 1–9.

18.

Balaji

Wiers

, et al. Effectiveness and acceptability of conversational agents for smoking cessation: a systematic review and meta-analysis. Nicotine Tob Res 2023; 25: 1241–1250.

19.

Luo

Aguilera

Lyles

, et al. Promoting physical activity through conversational agents: mixed methods systematic review. J Med Internet Res 2021; 23: e25486.

20.

Noh

Won

, et al. Conversational agents for body weight management: systematic review. J Med Internet Res 2023; 25: e42238.

21.

Chew

HSJ

. The use of artificial intelligence-based conversational agents (chatbots) for weight loss: scoping review and practical recommendations. JMIR Med Inform 2022; 10: e32578.

22.

Griffin

Xing

Khairat

, et al. Conversational agents for chronic disease self-management: a systematic review. AMIA Annu Symp Proc AMIA Symp 2020; 2020: 504–513.

23.

Provoost

Lau

Ruwaard

, et al. Embodied conversational agents in clinical psychology: a scoping review. J Med Internet Res 2017; 19: e151.

24.

ter Stal

Kramer

Tabak

, et al. Design features of embodied conversational agents in eHealth: a literature review. Int J Hum–Comput Stud 2020; 138: 102409.

25.

Bérubé

Schachner

Keller

, et al.

Voice-Based conversational agents for the prevention and management of chronic and mental health conditions. Systematic literature review

. J Med Internet Res 2021; 23: e25933.

26.

Bin Sawad

Narayan

Alnefaie

, et al. A systematic review on healthcare artificial intelligent conversational agents for chronic conditions. Sensors 2022; 22: 2625.

27.

Milne-Ives

de Cock

Lim

, et al.

The effectiveness of artificial intelligence conversational agents in health care. Systematic Review

. J Med Internet Res 2020; 22: e20346.

28.

Laranjo

Dunn

Tong

, et al. Conversational agents in healthcare: a systematic review. J Am Med Inform Assoc JAMIA 2018; 25: 1248–1258.

29.

Balaji

Giani

, et al. Effectiveness and acceptability of conversational agents for sexual health promotion: a systematic review and meta-analysis. Sex Health 2022; 19: 391–405.

30.

Kocaballi

Berkovsky

Quiroz

, et al. The personalization of conversational agents in health care: systematic review. J Med Internet Res 2019; 21: e15360.

31.

Noble

Smith

. Reviewing the literature: choosing a review design. Evid Based Nurs 2018; 21: 39–41.

32.

Tricco

Lillie

Zarin

, et al. PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Ann Intern Med 2018; 169: 467–473.

33.

Prochaska

Vogel

Chieng

, et al. A randomized controlled trial of a therapeutic relational agent for reducing substance misuse during the COVID-19 pandemic. Drug Alcohol Depend 2021; 227: 108986.

34.

Abdullah

Gaehde

Bickmore

. A tablet based embodied conversational agent to promote smoking cessation among veterans: a feasibility study. J Epidemiol Glob Health 2018; 8: 225–230.

35.

Apergi

Bjarnadottir

Baras

, et al. Voice interface technology adoption by patients with heart failure: pilot comparison study. JMIR MHealth UHealth 2021; 9: e24646.

36.

Boustani

Lunn

Visser

, et al. Development, feasibility, acceptability, and utility of an expressive speech-enabled digital health agent to deliver online, brief motivational interviewing for alcohol misuse: descriptive study. J Med Internet Res 2021; 23: e25837.

37.

Chavez-Yenter

Kimball

Kohlmann

, et al. Patient interactions with an automated conversational agent delivering pretest genetics education: descriptive study. J Med Internet Res 2021; 23: e29447.

38.

Fulmer

Joerin

Gentile

, et al. Using psychological artificial intelligence (TESS) to relieve symptoms of depression and anxiety: randomized controlled trial. JMIR Ment Health 2018; 5: e64.

39.

Greer

Ramo

Chang

Y-J

, et al. 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; 7: e15018.

40.

Guhl

Althouse

Pusateri

, et al. The atrial fibrillation health literacy information technology trial: pilot trial of a mobile health app for atrial fibrillation. JMIR Cardio 2020; 4: e17162.

41.

Hassoon

Baig

Naiman

, et al. Randomized trial of two artificial intelligence coaching interventions to increase physical activity in cancer survivors. Npj Digit Med 2021; 4: 168.

42.

Hunt

Miguez

Dukas

, et al. Efficacy of Zemedy, a mobile digital therapeutic for the self-management of irritable bowel syndrome: crossover randomized controlled trial. JMIR MHealth UHealth 2021; 9: e26152.

43.

Leo

Schuelke

Hunt

, et al. Digital mental health intervention plus usual care compared with usual care only and usual care plus in-person psychological counseling for orthopedic patients with symptoms of depression or anxiety: cohort study. JMIR Form Res 2022; 6: e36203.

44.

Napolitano

Harrington

Patchen

, et al. Feasibility of a digital intervention to promote healthy weight management among postpartum African American/Black women. Int J Environ Res Public Health 2021; 18: 2178.

45.

Nicol

Wang

Graham

, et al. Chatbot-delivered cognitive behavioral therapy in adolescents with depression and anxiety during the COVID-19 pandemic: feasibility and acceptability study. JMIR Form Res 2022; 6: e40242.

46.

Schroeder

Wilkes

Rowan

, et al. Pocket skills: a conversational mobile web app to support dialectical behavioral therapy. In: Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, New York, NY, USA, pp.1–15: Association for Computing Machinery.

47.

Thompson

Callender

Gonynor

, et al. Using relational agents to promote family communication around type 1 diabetes self-management in the diabetes family teamwork online intervention: longitudinal pilot study. J Med Internet Res 2019; 21: e15318.

48.

Zhou

Bickmore

. A virtual counselor for breast cancer genetic counseling: adaptive pedagogy leads to greater knowledge gain. In: Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems, pp.1–17. New York, NY, USA: Association for Computing Machinery.

49.

Hauser-Ulrich

Künzli

Meier-Peterhans

, et al. A smartphone-based health care chatbot to promote self-management of chronic pain (SELMA): pilot randomized controlled trial. JMIR MHealth UHealth 2020; 8: e15806.

50.

Kowatsch

Schachner

Harperink

, 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; 23: e25060.

51.

Stasinaki

Büchter

Shih

C-HI

, et al. Effects of a novel mobile health intervention compared to a multi-component behaviour changing program on body mass index, physical capacities and stress parameters in adolescents with obesity: a randomized controlled trial. BMC Pediatr 2021; 21: 308.

52.

Chen

, et al. Virtual digital psychotherapist app-based treatment in patients with methamphetamine use disorder (Echo-APP): single-arm pilot feasibility and efficacy study. JMIR MHealth UHealth 2023; 11: e40373.

53.

Yang

Zhu

, et al. Mental health chatbot for young adults with depressive symptoms during the COVID-19 pandemic: single-blind, three-arm randomized controlled trial. J Med Internet Res 2022; 24: e40719.

54.

Liu

Peng

Song

, et al. Using AI chatbots to provide self-help depression interventions for university students: a randomized trial of effectiveness. Internet Interv 2022; 27: 100495.

55.

Baptista

Wadley

Bird

, et al. Acceptability of an embodied conversational agent for type 2 diabetes self-management education and support via a smartphone app: mixed methods study. JMIR MHealth UHealth 2020; 8: e17038.

56.

Gong

Baptista

Russell

, 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; 22: e20322.

57.

Bibault

J-E

Chaix

Guillemassé

, et al. A chatbot versus physicians to provide information for patients with breast cancer: blind, randomized controlled noninferiority trial. J Med Internet Res 2019; 21: e15787.

58.

Chaix

Bibault

J-E

Pienkowski

, et al. When chatbots meet patients: one-year prospective study of conversations between patients with breast cancer and a chatbot. JMIR Cancer 2019; 5: e12856.

59.

Carlin

Logue

Flynn

, et al. Development and feasibility of a family-based health behavior intervention using intelligent personal assistants: randomized controlled trial. JMIR Form Res 2021; 5: e17501.

60.

Gaffney

Mansell

Tai

. Agents of change: understanding the therapeutic processes associated with the helpfulness of therapy for mental health problems with relational agent MYLO. Digit Health 2020; 6: 2055207620911580.

61.

Barbaric

Munteanu

Ross

, et al. A voice app design for heart failure self-management: proof-of-concept implementation study. JMIR Form Res 2022; 6: e40021.

62.

Anan

Kajiki

Oka

, et al. Effects of an artificial intelligence-assisted health program on workers with neck/shoulder pain/stiffness and low back pain: randomized controlled trial. JMIR MHealth UHealth 2021; 9: e27535.

63.

Dhinagaran

Sathish

Soong

, et al. Conversational agent for healthy lifestyle behavior change: web-based feasibility study. JMIR Form Res 2021; 5: e27956.

64.

Alghamdi

Alnanih

. Chatbot design for a healthy life to celiac patients: a study according to a new behavior change model. Int J Adv Comput Sci Appl 12. Epub ahead of print 2021. DOI: https://doi.org/10.14569/IJACSA.2021.0121077.

65.

Mash

Schouw

Fischer

. Evaluating the implementation of the GREAT4Diabetes WhatsApp chatbot to educate people with type 2 diabetes during the COVID-19 pandemic: convergent mixed methods study. JMIR Diabetes 2022; 7: e37882.

66.

Krishnakumar

Verma

Chawla

, 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; 23: e17908.

67.

Baertsch

M-A

Decker

Probst

, et al. Convenient access to expert-reviewed health information via an Alexa voice assistant skill for patients with multiple myeloma: development study. JMIR Cancer 2022; 8: e35500.

68.

Bassi

Giuliano

Perinelli

, et al. A virtual coach (Motibot) for supporting healthy coping strategies among adults with diabetes: proof-of-concept study. JMIR Hum Factors 2022; 9: e32211.

69.

Sagstad

Morken

N-H

Lund

, et al. Quantitative user data from a chatbot developed for women with gestational diabetes Mellitus: observational study. JMIR Form Res 2022; 6: e28091.

70.

Roca

Lozano

García

, et al. Validation of a virtual assistant for improving medication adherence in patients with comorbid type 2 diabetes mellitus and depressive disorder. Int J Environ Res Public Health 2021; 18: 12056.

71.

Klaassen

Bul

KCM

Op den Akker

, et al. Design and evaluation of a pervasive coaching and gamification platform for young diabetes patients. Sensors 2018; 18: 402.

72.

Issom

D-Z

Hardy-Dessources

M-D

Romana

, et al. Toward a conversational agent to support the self-management of adults and young adults with sickle cell disease: usability and usefulness study. Front Digit Health 2021; 3: 600333.

73.

Weir

Jan

. BMI classification percentile and cut off points. In: StatPearls. Treasure Island (FL): StatPearls Publishing, http://www.ncbi.nlm.nih.gov/books/NBK541070/ (2023, accessed 20 October 2023).

74.

Kroenke

Spitzer

Williams

JBW

. The PHQ-9. J Gen Intern Med 2001; 16: 606–613.

75.

Kroeze

Werkman

Brug

. A systematic review of randomized trials on the effectiveness of computer-tailored education on physical activity and dietary behaviors. Ann Behav Med 2006; 31: 205–223.

76.

Noar

Grant Harrington

Van Stee

, et al. Tailored health communication to change lifestyle behaviors. Am J Lifestyle Med 2011; 5: 112–122.

77.

Peñarrieta

Flores-Barrios

Gutiérrez-Gómez

, et al. Self-management and family support in chronic diseases. J Nurs Educ Pract 2015; 5: 73.

78.

Fisher

Boothroyd

Elstad

, et al. Peer support of complex health behaviors in prevention and disease management with special reference to diabetes: systematic reviews. Clin Diabetes Endocrinol 2017; 3: 4.

79.

Bickmore

. Relational agents: effecting change through human-computer relationships . Thesis, Massachusetts Institute of Technology, https://dspace.mit.edu/handle/1721.1/36109 (2003, accessed 28 July 2023).

80.

Cassell

. Embodied conversational interface agents. Commun ACM 2000; 43: 70–78.

81.

Allouch

Azaria

Azoulay

. Conversational agents: goals, technologies, vision and challenges. Sensors 2021; 21: 8448.

82.

McKay

Wright

Shill

, et al. Using health and well-being apps for behavior change: a systematic search and rating of apps. JMIR MHealth UHealth 2019; 7: e11926.

83.

Torgrimson

Minson

. Sex and gender: what is the difference? J Appl Physiol 2005; 99: 785–787.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB