Sage Journals: Discover world-class research

Abstract

Background

Chatbots have been extensively utilized in chronic disease management to collect real-time health data, deliver personalized educational content, and guide self-management. Nevertheless, critical research gaps persist regarding their differential implementation across specific contexts and quantified comparative effectiveness.

Objective

To synthesize existing research on the application of chatbots in chronic disease management, providing evidence-based insights to inform future clinical practice.

Methods

Following Arksey and O’Malley's framework, we systematically searched eight databases from their inception until October 20, 2024. Relevant data were extracted from eligible studies, with a focus on disease areas, application platforms, interaction methods, technical architectures, implementation elements, and evaluation indicators. The findings were then synthesized and analyzed to identify key trends and gaps in the literature.

Results

A total of 19 studies were included in this review, comprising 10 randomized controlled trials (RCTs) and 9 quasi-experimental studies. The investigated chronic conditions encompassed cancer, diabetes, hypertension, and other prevalent chronic diseases. Chatbot deployment platforms primarily included mobile applications, web-based platforms, and instant messaging software. The underlying technical architectures consisted of artificial intelligence-driven systems, rule-based systems, and hybrid models. The implementation strategies were categorized into night key dimensions. The predominant interaction modality was hybrid, with communication content focusing on self-management education, emotional support, and related domains. Outcome measures evaluated health-related indicators and user adherence indicators.

Conclusions

Chatbots hold considerable clinical application value in chronic disease management. However, current research has some limitations. Future research should further optimize interaction design, refine system functionalities, and fortify privacy protection measures to better facilitate the integration of chatbots into chronic disease management.

Keywords

Chatbot chronic diseases whole health conversation agent telehealth

Introduction

Noncommunicable diseases (NCDs), encompassing cardiovascular diseases, malignancies, chronic respiratory diseases, diabetes, hypertension, and related conditions, represent a critical global public health challenge.¹ Characterized by multifactorial etiology, prolonged disease duration, refractory nature to complete cure, and high rates of disability and mortality, NCDs exhibit persistently increasing prevalence and mortality rates. This trend imposes substantial disease burden and socioeconomic costs worldwide.^1,2 According to the World Health Organization (WHO) reports, NCDs account for approximately 41 million annual deaths globally, representing 74% of total mortality worldwide.³ This challenge is particularly acute in China, where epidemiological data reveal a substantial and growing chronic disease burden. Notably, over 85% of all deaths in China are attributable to NCDs, with these conditions contributing to more than 70% of the nation's total disease burden,^3,4 which significantly exceeds global averages. Health management serves as a pivotal strategy for chronic disease prevention and control, demonstrating efficacy in reducing disease burden and extending healthy life expectancy. However, persistent challenges in chronic disease management include inadequate patient self-management capacity, poor long-term adherence to follow-up protocols, and the need for personalized interventions.⁵

To address the challenges in chronic disease management, WHO has strongly advocated for digital health interventions as a strategic approach to enhance healthcare capacity, expand service coverage, and improve intervention efficiency, ultimately advancing universal health coverage goals.⁶ In recent years, the digital health landscape has undergone a paradigm shift, evolving from traditional domains such as eHealth and mHealth to cutting-edge interdisciplinary frontiers integrating advanced computational sciences, particularly big data analytics, genomics, and artificial intelligence (AI).⁷ The rapid advancement of AI, particularly machine learning (ML), deep learning (DL), and natural language processing (NLP) technologies,⁸ is transforming healthcare interventions in unprecedented ways. AI applications have demonstrated significant potential across critical clinical domains, including disease diagnosis, risk prediction, prognostic evaluation, and patient follow-up management. Against this technological backdrop, AI-powered chatbots have emerged as innovative solutions. Chatbots are computer systems designed to simulate human conversation through multimodal interfaces, including text, voice, and visual interactions. These systems are broadly categorized into two distinct architectures: artificial intelligence-based chatbots and rule-based conversational agents.⁹

The clinical applications of chatbots have expanded beyond basic health inquiries to encompass diverse domains, including disease screening, chronic condition management, psychological support, and clinical training. Given the characteristics of chronic diseases, including relatively stable disease status, insidious symptoms, protracted disease course, propensity for recurrence, and difficulty in achieving a complete cure, patients require long-term self-management and medical supervision. However, traditional follow-up models frequently encounter challenges such as limited resources, temporal and spatial constraints, and inadequate dynamic monitoring. A growing evidence suggests that chatbots can play a pivotal role in chronic disease management by assisting patients with disease monitoring, facilitating healthy behavior change, delivering disease-specific knowledge and skills education, enhancing adherence, and providing psychosocial support, thereby improving clinical outcomes, enhancing quality of life, and reducing healthcare costs.^10,11 Furthermore, chatbot systems demonstrate significant utility in automating the collection of patient-reported outcomes (PROs),¹² thereby providing clinicians with comprehensive, longitudinal health data that enhances clinical decision making during both consultations and follow-up care. An online survey found that doctors generally believe that healthcare chatbots have a positive effect on patients’ health self-management behaviors, which can improve patients’ disease symptoms, mental health, and health behaviors.¹³ Therefore, chatbots provide an efficient, personalized, and cost-effective digital health intervention for chronic disease management, demonstrating significant potential to delay disease progression and reduce all-cause mortality rates in chronic disease populations.

In recent years, the design of medical chatbot systems has largely shifted from traditional rule-based frameworks to sophisticated generative models powered by large language models (LLMs).¹⁴ Conventional systems rely on static decision trees and keyword matching, which, while demonstrating a certain capacity to mimic medical interactions through scripted dialogs, inherently constrain their flexibility, contextual understanding, and ability to handle diverse patient inputs. In contrast, LLMs such as GPT and Deepseek employ transformer-based architectures to comprehend and generate natural language in a dynamic and context-sensitive manner.¹⁵ This transition enables more fluid, context-aware, and linguistically complex interactions, allowing such systems to adapt to varied user needs and expand their functional scope.¹⁶ For instance, Zhang Sainan et al.¹⁷ developed a chatbot named Chat Ella using Generative Pre-trained Transformer 2 (GPT-2) to assist in diagnosing 24 chronic diseases. Their results indicated a diagnostic accuracy of 97.50%, and the system received positive feedback from 68.7% of participants. Similarly, Kim et al. (2024) developed an AI-based chatbot to support self-management in gastric cancer patients, reporting an accuracy of 85.2% and a precision of 87.6%. Despite the considerable potential of LLMs in building more intelligent and personalized chatbots for chronic disease management, their integration requires careful consideration of multiple challenges, including accuracy, fairness, and ethical and privacy concerns, to ensure safe and effective clinical application.^15,16

Nevertheless, current research exhibits substantial heterogeneity in chatbot design elements, including interaction modalities, functional capabilities, and evaluation indicators, which compromises the comparability of study outcomes and hinders the development of standardized practice guidelines.^11,18 Consequently, this study employs a scoping review methodology to systematically examine the existing literature on chatbot applications in chronic disease management. Through comprehensive analysis of relevant research, we aim to identify current research gaps and methodological limitations. This study aims to provide a foundational basis for formulating standardized clinical intervention strategies and developing clinically demand-driven chatbots.

Objectives

This scoping review aims to synthesize existing research on the application of chatbots in chronic disease management, with particular focus on technical architectures, implementation characteristics, and evaluation indicators. By identifying current research limitations and knowledge gaps, the study provides an evidence-based foundation for optimizing clinical decisions and intervention strategies in digital health.

Methods

Study design

This scoping review was conducted following the five-stage methodological framework established by Arksey and O'Malley¹⁹: (1) identifying the research question, (2) determining relevant research, (3) selecting research, (4) charting the extracted data, and (5) collating, summarizing, and reporting the results. This scoping review adhered to the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) statement,²⁰ as detailed in Supplementary File 1. This review is not registered as PROSPERO does not accept scoping reviews.

Identifying the research question

Following preliminary literature screening and analysis of current chatbot applications in chronic disease management, we formulated four key research questions: (1) What platforms and techniques support chatbot implementation in chronic disease management? (2) What implementation characterize these interventions? (3) How are chatbot-based interventions evaluated?

Determining relevant research

This scoping review conducted a systematic search of PubMed, the Cochrane Library, Web of Science, Embase, CINAHL, the China National Knowledge Infrastructure (CNKI), Wanfang Database, and the Chinese Biomedical Literature Database (CBM), covering all available records from inception through October 20, 2024. Search terms including keywords and subject headings were related to chatbot technologies (e.g. “conversational agent,” “chatbot,” “smartbot,” “dialog system,” and “dialog agent”) and chronic diseases (e.g. “chronic kidney disease,” “chronic disease,” “cardiovascular diseases,” “stroke,” “neoplasms,” “hypertension,” “diabetes mellitus,” and “dementia”). The search strategy of all databases used is presented in Supplementary File 2.

Selecting research

All records retrieved from database searches were imported into EndNote Reference Manager for deduplication. Two independent reviewers (JH and SL) conducted the screening process following a standardized protocol: initial title and abstract screening followed by full-text assessment of potentially eligible studies. Any discrepancies were resolved through consensus discussion. Inclusion criteria comprised: (1) Studies involving adult patients with chronic conditions; (2) Interventions utilizing chatbots with automated response algorithms for chronic disease management (excluding human-operated systems); (3) Study designs including randomized controlled trials, quasi-experimental studies, observational studies, or systematic reviews; and (4) Clearly defined interventions with measurable outcomes. Exclusion criteria included: (1) Unavailable full-text articles; (2) Duplicate publications; and (3) Non-English or Non-Chinese language publications.

Charting the extracted data

Information extraction was performed on the included literature, with extracted content encompassing author(s), year of publication, country, study type, type of chronic disease, sample size, application objectives, application platforms, interaction methods, implementation elements, intervention duration, and evaluation indicators. All data were entered into an Excel spreadsheet for aggregation and analysis.

Data synthesis and analysis

Through a comprehensive synthesis of the included studies, we identified and descriptively analyzed the core characteristics of chatbot interventions in chronic disease management. Given the substantial heterogeneity in interventions, outcome measures, and reporting formats among the included studies, a meta-analysis was not deemed appropriate or feasible. The findings are therefore synthesized narratively and presented in a summary table of key outcomes.

Results

Study selection

A total of 1079 studies were retrieved. After eliminating duplicates, 673 remained, of which 567 were excluded following a review of titles and abstracts. A full-text review was conducted on the remaining 106 studies, and ultimately 19 studies were included (see Figure 1).

Figure 1.

Flowchart of literature screening.

General characteristics of the included studies

Among the included studies, RCTs represented the majority (n = 10)^21–30 followed by quasi-experimental studies (n = 9).^31–39 These studies encompassed a broad area of chronic diseases, with cancer being the most frequently investigated condition (n = 6),^{28,29,33,35,36,38} followed by diabetes (n = 5).^{24,27,30,37,39} Other disease areas included mental and neurological disorders (n = 3),^22,23,25 hypertension (n = 2),^26,34 kidney diseases (n = 2),^21,32 and orthopedic conditions (n = 1).³¹ Intervention durations ranged from 1 week to 2 years, except for one study that did not report this information.³⁶ One study focusing on task completion time reported an average duration of 18 min.³⁴ The chatbot-delivered interventions served distinct clinical purposes across studies. Self-management was the primary objective (n = 11),^{21,22,26,29–31}^{,33,35,37–39} while other applications included supervision (n = 3),^23,25,34 education (n = 3),^28,32,36 and emotional support (n = 2).^24,27 Most studies account for patients’ varied disease characteristics and demographic profiles, while also requiring device support. Specifically, some research incorporates diverse racial backgrounds (n = 3),^34,37,38 smartphone usage capabilities (n = 3),^23,24,28,31 and language diversity support (n = 2)^32,39 to alleviate the digital divide. The two reviewers maintained agreement throughout all phases of the systematic review process. The general characteristics of the included studies are summarized in Table 1.

Table 1.

General characteristics of the included studies (N = 19).

Number	First author, year	Country	Disease type	Study type	Sample size (T/C, cases)	Duration	Study objective	Considerations for inclusion in the study population	Privacy	Clinical integration	Outcome type	Specific indicators and statistical results
1	Liu J et al., 2023	China	Patients with arteriovenous fistula for hemodialysis	RCT	60/60	4 months	To explore the impact of WeChat chatbot intervention on the mental state, treatment adherence, and complications of patients with arteriovenous fistula for hemodialysis.	Age, gender, BMI, pulse pressure, disease duration	Accessed through WeChat official account, patients self-register and log in	Operates as a standalone tool	Mental health	Self-Rating Anxiety Scale (SAS): The SAS score of the intervention group was lower than that of the control group (P < .01)
											Mental health	Self-Rating Depression Scale (SDS): The SDS score of the intervention group was lower than that of the control group (P < .01)
											Treatment adherence	Self-designed adherence questionnaire: The scores of dietary adherence, fluid intake adherence, medication adherence, and dialysis plan adherence in the intervention group were all higher than those in the control group (P < .05)
											Health outcome	Complication rate: 3.33% in the intervention group, lower than 18.33% in the control group (P < .01)
2	Yue RX et al., 2023	China	Patients after osteoporotic fracture surgery	RCT	150/150	12 months	To explore the application effect of AI chatbot in the post-discharge transitional care of patients after osteoporotic fracture surgery.	Gender, age, bone mineral density, education level, presence of diabetes mellitus, use of hormones, equipment utilization	Accessed through WeChat official account, patients self-register and log in	Healthcare professionals can send and receive information through the robot	Self-management ability	Osteoporosis health belief level: The scores of the intervention group at 3 months, 6 months, and 12 months were all higher than those of the control group (P = .001)
											Self-management ability	Health self-management ability: The intervention group was significantly higher than the control group at 3, 6, and 12 months (P < .05)
											Health outcome	Pain score: There was no statistically significant difference in scores between the two groups
												Joint function: There was no statistically significant difference in scores between the two groups
												Incidence of secondary fractures: The incidence of secondary fractures in the intervention group was 4.67%, and that in the control group was 11.33%. The control group was significantly higher than the intervention group (P = .033)
3	Chen NJ et al., 2023	Taiwan Province, China	Chronic kidney disease (CKD)	RCT	55/-	3 months	To develop and evaluate the feasibility of a chat-based instant messaging support health education intervention program for CKD patients.	Age, gender, marital status, education level, CKD stage, language, number of comorbidities, BMI	Not mentioned	Operates as a standalone tool	Self-management ability	CKD-Specific Communicative Literacy and Knowledge Assessment: Paired t-tests revealed significant differences before and after intervention in communicative literacy (t = 3.99; P < .001) and CKD-specific disease knowledge (t = 7.54; P < .001)
3	Chen NJ et al., 2023	Taiwan Province, China	Chronic kidney disease (CKD)	RCT	55/-	3 months			Not mentioned	Operates as a standalone tool	System usability	System Usability Scale (SUS): SUS scores were calculated at 3 time points, resulting in mean scores of 71.19 (SD 15.31) at time 1, 76.42 (SD 16.29) at time 2, and 80.23 (SD 13.39) at time 3, respectively.
4	Ogawa M et al., 2022	Japan	Parkinson's disease (PD)	RCT	10/10	5 months	To evaluate the feasibility and effectiveness of using an artificial intelligence-based chatbot to improve smiling and speech in PD patients.	Age, gender, native language, cognitive status, speech issues	Not mentioned	Operates as a standalone tool	Health outcome	Movement Disorder Society–Unified Parkinson's Disease Rating Scale (MDS-UPDRS): No significant interaction effects
												Mini-Mental State Examination (MMSE): No significant interaction effects
												Japanese version of Montreal Cognitive Assessment (MoCA-J): No significant interaction effects
											Mental health	Smile index: Chatbot conversations had a significant interaction effect on the mean and standard deviation of the smile index during smile sections (both P = .02), the maximum duration of the initial rise of the smile index (P = .04), and the frequency of filler words (P = .04).
											Mental health	Beck Depression Inventory–Second Edition (BDI-II): No significant interaction effects
											Quality of life	Parkinson's Disease Questionnaire-39 (PDQ-39): No significant interaction effects
5	Yasukawa S et al., 2024	Japan	Subthreshold depression	Quasi-experimental study	70/73	8 weeks	To examine the effect of an automated chatbot on improving the adherence rate of internet-based cognitive behavioral therapy (iCBT).	Age, gender, place of residence, smartphone proficiency	Developed by Sony Group, considering privacy and data security	Operates as a standalone tool	Treatment adherence	Completion rate of iCBT program: The intervention group (34.8%) had a higher completion rate than the control group (19.2%) (P < .05)
											Mental health	Patient Health Questionnaire-9 (PHQ-9): The change in PHQ-9 at week 8 was −2.21 in the iCBT + chatbot group and −2.30 in the iCBT group; both groups showed significant improvements.
												Generalized Anxiety Disorder-7 (GAD-7): The mean score improved from 5.93 (3.46) at baseline to 5.33 (3.51) at week 8.
												The Satisfaction with Life Scale (SWLS): The mean score improved from 21.5 (6.00) at baseline to 21.9 (6.27) at week 8.
												World Health Organization-Five (WHO-5): The mean score improved from 11.9 (4.05) at baseline to 12.2 (4.10) at week 8.
											Self-management ability	Cognitive and behavioural skills: Knowledge and skills in psychoeducation (PE), behavioural activation (BA), self-monitoring (SM), cognitive restructuring (CR), and assertiveness training (AT) were improved, while the problem-solving module was not.
											Social function	Attendance scale, Work and Social Adjustment Scale (WSAS): The mean score changed from 15.9 (7.48) at baseline to 15.6 (9.05) at week 8, showing no progress.
												Utrecht Work Engagement Scale (UWES): The control group (iCBT group) improved significantly more than the intervention group (iCBT + chatbot group) in UWES (P < .05).
												Presenteeism: The mean score changed from 5.60 (2.14) at baseline to 6.23 (1.74) at week 8.
6	Kim AR et al., 2024	South Korea	Gastric cancer patients after radical gastrectomy	RCT	56/-	1 week	To develop a knowledge-based question-answering chatbot to assist self-management and evaluate the user experience and performance of the chatbot.	Gender, age, disease stage, medical application experience	Not mentioned	Operates as a standalone tool	System usability	Accuracy rate was 85.2%, precision rate 87.6%, recall rate 96.8%, and F-score 92.0%, indicating good performance.
6	Kim AR et al., 2024	South Korea	Gastric cancer patients after radical gastrectomy	RCT	56/-	1 week		Gender, age, disease stage, medical application experience	Not mentioned	Operates as a standalone tool	User satisfaction	Self-designed questionnaire: Users were satisfied with the overall experience of the chatbot, with an average score of 4.28.
7	Griffin AC et al., 2023	United States	Hypertension	RCT	10/-	Time to complete the task	To design, develop, and evaluate the usability of a hypertension self-management chatbot prototype “Medicagent.”	Age, race, educational level, drug use status	Uses Google Cloud Firestore database, protecting privacy and data security	Operates as a standalone tool	System usability	System Usability Scale (SUS): The average SUS score was 78.8, indicating good usability.
											User engagement	Almost all tasks (98%) were successfully responded to. A total of 252 button clicks and 128 utterances were made during the test sessions. Only 8.6% of utterances were not successfully mapped to intents.
											User satisfaction	Interviews showed that users gave positive evaluations of Medicagent's visual presentation and ease of use.
8	Gomaa S et al., 2023	United States	Gastrointestinal cancer patients receiving intravenous chemotherapy	RCT	34/-	2 months	To develop and evaluate the feasibility, effectiveness, and acceptability of a system for real-time monitoring and active management of patients’ side effects.	Age, gender, marital status, income, cancer stage, duration of chemotherapy	Data de-identified, compliance with privacy regulations, information kept confidential	Self-reported information is integrated with electronic health records and clinical information systems	User engagement	Evaluate the usage of the platform: 20 patients (59%) participated in seeking more information through at least one SMS response. Interactions ranged from 1 message to 33 messages.
											Self-management ability	Patient Activation Measure (PAM): Patients in the intervention group showed a significant increase in activation measure scores, with an average increase of 3.70 (95% confidence interval: −6.919 to −0.499), P = .02
											Self-management ability	Cancer Treatment Survey (CaTS): No statistically significant changes before and after intervention
											Quality of life	Functional Assessment of Cancer Therapy (FACT) scale: No statistically significant changes before and after intervention
											Health outcome	Memorial Symptom Assessment Scale (MSAS): No statistically significant changes before and after intervention
											Social function	Multidimensional Scale of Perceived Social Support (MSPSS): No statistically significant changes before and after intervention
											User satisfaction	Self-designed satisfaction questionnaire: Participants were highly satisfied with the intervention
											User satisfaction	Through interviews to understand the experience and feedback of intervention measures: Participants found the intervention user-friendly and received valuable information.
9	Soley N et al., 2023	United States	Pancreatic cancer	RCT	60/-	Not mentioned	To conduct a feasibility assessment of the chatbot Gia that provides genetic education (GEd).	Age, gender, family history of pancreatic carcinoma	Uses the Invitae platform, compliant with “Health Insurance Portability and Accountability Act”	Operates as a standalone tool	User engagement	Evaluate participation through the time of each interaction of participants: Findings indicate that patients who reported getting a genetic test (GT) spent more time (11 min) interacting with Gia compared to those who were not sure if they wanted to have a GT (9 min). All participants spent approximately the same amount of time on the post-visit survey.
											System usability	Self-designed questionnaire: Most patients reported that genetic education through the chatbot helped them make informed decisions about genetic testing, and the robot received the highest usability score (average score 4.50)
											User satisfaction	Most patients were satisfied with the usability of the chatbot and the content of genetic education.
10	Nassar CM et al., 2023	United States	Type 2 diabetes	Quasi-experimental study	150/-	16 months	To evaluate the acceptability, satisfaction, and engagement of a chatbot for diabetes education and support.	Age, gender, ethnicity	SaaS-based Clinical decision support system web application, data stored and accessed within a secure electronic medical record	Deployed within the regional healthcare system, shares actionable information with clinicians	User engagement	77.3% clicked and agreed to the terms and conditions and completed the first chat. Among participants who completed the first chat, approximately 48% completed more than half of the subsequent chats and were considered actively engaged.
											Self-management ability	Self-care confidence scale: Almost 90% of active users agreed that the chat made them more confident in managing diabetes.
											User satisfaction	Send self-designed questionnaire after chat: Overall satisfaction was high, and most participants were satisfied with the content, length, and frequency of the chat.
											Health outcome	Blood glucose results (A1C): Participants’ A1C decreased by an average of 1.04%, while the control group's A1C increased by an average of 0.09% (P = .008).
11	Ma D et al., 2021	United States	Head and neck cancer patients	Quasi-experimental study	95/-	12 months	To evaluate the feasibility and acceptability of using a fully automated chatbot during patients’ radiation therapy.	Age, gender, ethnicity, tumor location, treatment intent	Not mentioned	Operates as a standalone tool	User engagement	Patient engagement with the chat was highest in the first month (67%) but gradually decreased to 23% at the end of the study (P = .004)
											Health outcome	Complications and adverse events: 58% reported at least one serious adverse event. The consistency between patient-reported patient-reported outcomes (PRO) and clinician-reported outcomes (CRO) was moderate, ranging from 0.10 to 0.43 (Cohen κ statistic).
											User satisfaction	89% of respondents gave positive evaluations of the chatbot's user-friendliness; however, 49% of patients thought there were too many inconsistencies in the chat.
12	MacNeill AL et al., 2024	Canada	Patients with arthritis or diabetes	RCT	34/34	4 weeks	To determine whether a mental health chatbot can improve the mental health of patients with chronic diseases.	Age, disease type, smartphone usage ability	No personal identification information required	Operates as a standalone tool	Mental health	Patient Health Questionnaire-9 (PHQ-9): Patients in the intervention group reported a reduction in depression severity during the study (P < .001).
												Generalized Anxiety Disorder Scale-7 (GAD-7): Patients in the intervention group reported a reduction in anxiety levels (P < .001)
												Perceived Stress Scale-10 (PSS-10): Neither group reported changes in stress.
											User satisfaction	Users liked the app's functions and features, overall design, and user experience. However, dissatisfaction was mostly focused on the chatbot's conversational ability.
13	Ulrich S et al., 2024	Switzerland	Headache patients	Quasi-experimental study	110/88	20–60 days	To develop a smartphone-based intervention for headache patients and evaluate its effectiveness, engagement, and acceptability in promoting a healthy lifestyle and enhancing mental health among headache patients.	Age, verbal ability, headache frequency	Communication is encrypted	Operates as a standalone tool	Mental health	Patient Health Questionnaire for Anxiety and Depression (PHQ-ADS): There was a significant treatment effect on PHQ-ADS in the intervention group after intervention, with a moderate effect size improvement from pre-intervention to post-intervention (Cohen d = 0.62, 95%CI −0.84 to −10.39).
											Mental health	Depression PHQ-9 (t = −3.56; P < .001, 95%CI −2.79 to −0.80), anxiety GAD-7 (t = −2.73; P = .007, 95%CI −2.50 to −0.40), psychosomatic symptoms PHQ-15 (t = −4.48; P < .001, 95%CI −3.35 to −1.31), and Perceived Stress Scale-10 (t = −3.07; P = .003, 95%CI −4.25 to −0.94) all showed significant treatment effects.
											Self-management ability	Headache Management Self-Efficacy Scale (HMSE-G-SF) (t = 4.08; P < .001, 95%CI 2.10–5.99), intention to change behavior (t = 4.82; P < .001, 95%CI 0.45–1.07), and Pain Coping Inventory (t = 4.96; P < .001, 95%CI 3.38–7.79) all showed significant treatment effects.
											Social function	Attendance and absence (t = −3.00; P = .003, 95%CI −7.96 to −1.68) showed a significant treatment effect.
											User engagement	86.3% of participants completed all modules, with an average of 6.9 days required to complete each module.
											User satisfaction	Self-designed satisfaction questionnaire: Participants had high acceptability and satisfaction with the BalanceUP app.
14	Echeazarra L et al., 2021	Spain	Hypertension patients	Quasi-experimental study	55/57	2 years	To improve the feasibility of self-blood pressure management in hypertension patients by developing and testing the TensioBot chatbot, and to evaluate its effects and potential benefits in application.	Gender, age, employment, education, diagnosis, comorbidities	Phone number not required, protecting privacy	Operates as a standalone tool	Self-management ability	Questionnaire on knowledge and skills of blood pressure measurement: There was a significant difference in knowledge acquisition between the two groups, with a mean score of 24.126 in the intervention group and 17.591 in the control group (P < .05)
											Self-management ability	Difference in blood pressure values: Both systolic and diastolic values in the intervention group showed small differences but were not significant.
											User engagement	More than 55 patients successfully participated in the experiment. There was no significant difference in blood pressure check adherence between the intervention group and the control group.
											User satisfaction	Interviews showed that 92.5% of patients thought TensioBot was easy to use, and 72.5% thought it was very useful. 85% of patients continued to use TensioBot after the experiment.
15	Bruijnes M et al., 2023	Netherlands	Diabetes	Quasi-experimental study	79/77	3 weeks	To determine the feasibility and preliminary effects of an automated conversational agent providing personalized psychoeducation to diabetes patients to help them cope with chronic disease-related social distress.	Gender, age, diabetes mellitus pain level	Data stored on secure servers, processed anonymously	Operates as a standalone tool	Mental health	Diabetes Distress Scale (DDS): Compared with the control group (M = 0.002, SD = 0.743), those using the conversational agent had a greater reduction in diabetes distress (M = −.305, SD = 0.865), and the difference between the two groups was statistically significant (t = 2.377, P = .019)
											Mental health	Feeling of Being Heard (FBH): There was no statistically significant difference in the feeling of being heard between participants in the agent group (M = −0.305, SD = 0.865) and the control group (M = 0.002, SD = 0.743) (t = 0.60, P = .55).
											User satisfaction	Client Satisfaction Questionnaire-8 (CSQ-8): Participants expressed satisfaction.
											System usability	System Usability Scale (SUS): The average system usability score was 81.6 (standard deviation 12.0), indicating good usability.
											User satisfaction	Interviews showed that most people gave positive comments (n = 46), while negative comments (n = 26) mainly included comments on limited conversation content (n = 17) and technical issues or limitations (n = 9).
16	Baumgartner K et al., 2024	Germany	Male patients with prostate cancer	Quasi-experimental study	59/53	4 weeks	To evaluate the impact of the chatbot “Prostate Cancer Conversational Agent” (PROSCA).	Age, equipment usage, Gleason score	Uses conversational AI platform, protecting privacy and data security	Operates as a standalone tool	Self-management ability	Self-designed questionnaire to survey information needs: Patients’ information needs decreased over time; compared with the control group, patients using the chatbot had a more significant reduction in information needs (P = .035)
											User engagement	The overall participation rate during the study was 71%. 57% of participants completed most of the scheduled conversations, with a median of 7 completed conversations (range 0–16).
											System usability	Mobile Health APP Usability Assessment Questionnaire: 73.2% of participants fully or partially agreed that they obtained a lot of information about the disease and treatment; 83.7% of participants stated that they did not need any help when using the chatbot.
17	Tawfik E et al., 2023	Egypt	Female breast cancer patients receiving chemotherapy	Quasi-experimental study	50/100	10 months	To evaluate the usability and effectiveness of ChemoFreeBot.	Age, educational level, marital status, occupation, income, disease stage	Built on Microsoft Azure platform, compliant with international security standards	Operates as a standalone tool	Self-management ability	Memorial Symptom Scale (MSAS): Women in the ChemoFreeBot group experienced significantly fewer physical and psychological symptom frequency, severity, and distress (P < .001)
											Self-management ability	Modified Self-Care Behavior Diary (SCBD): Women in the ChemoFreeBot group reported the highest effectiveness in self-care behaviors (P < .001)
											System usability	Chatbot Usability Questionnaire (CUQ): Most women (94%) reported that the chatbot was easy to use, and its answers were useful, appropriate, and informative.
18	Mash R et al., 2022	South Africa	Type 2 diabetes	Quasi-experimental study	81/58	6 months	To evaluate the implementation of a chatbot diabetes education program during the epidemic to improve patient education, self-management, and blood glucose control levels.	Age, gender, language	Contact information protected by law, not shared	Operates as a standalone tool	User engagement	A total of 8158 people logged into the chatbot, and only 4716 (57.83%) responded to the terms and conditions and started using it.
											Self-management ability	Self-designed self-management questionnaire: More than 90% of patients thought the messages they received were useful.
											User satisfaction	Interviews showed that 87.6% of patients were more confident in self-care. All respondents agreed that the chatbot service should continue.
19	Gong E et al., 2020	Australia	Type 2 diabetes	RCT	93/94	12 months	To evaluate the usability and effectiveness of the My Diabetes Coach program, which aims to support diabetes self-management in the home environment.	Age, gender, educational level, employment status, etc.	Not mentioned	Operates as a standalone tool	User engagement	92 participants in the intervention group completed 1942 conversations with Laura, averaging 243 min per person. Participants uploaded blood glucose level data an average of 181.8 times. However, usage decreased over time. The proportion of participants who chatted with Laura at least once a month dropped from 87% in the first month to 15% in the 12th month, and the proportion of participants who recorded blood glucose in the app dropped from 78% to 23%.
											Health outcome	Changes in glycosylated hemoglobin (HbA1c): Compared with baseline, the mean estimated HbA1c decreased in both groups at 12 months (intervention group: 0.33%, control group: 0.20%), but the difference in HbA1c changes between the two groups (−0.04%, 95%CI −0.45 to 0.36; P = .83) was not significant.
											Health outcome	Health-related quality of life (HRQoL): At 12 months, compared with the control group, the HRQoL utility score of the intervention group improved (between-group difference 0.04, 95%CI 0.00 to 0.07; P = .04)
											Mental health	Anxiety and depression symptoms: A significant between-group difference was observed in the mean change in HADS anxiety score at 6 months (0.89, 95%CI 1.74 to 0.04; P = .04), but not at 12 months.

Platforms and technical architectures of chatbots

The included studies deployed chatbots across three primary platforms. Instant messaging applications (e.g. WeChat, WhatsApp, LINE, Telegram) were the most ubiquitous (n = 9) due to their widespread adoption and low operational complexity.^{21,23,26,29,31,32,35,38,39} Mobile applications enabled more sophisticated and integrated self-management functionalities (n = 4).^22,24,25,30 Web-based platforms served well for information education and retrieval (n = 6).^{27,28,33,34,36,37}

Regarding technical architectures, artificial intelligence (AI) systems were employed in 12 studies to develop chatbots capable of personalized and conversational interactions,^21–24^{,26–28,30–33,37} while rule-based systems were used in two studies for managing structured tasks.^36,39 Furthermore, two studies adopted hybrid architectures to balance flexibility with control, with one integrating rule-based grammatical matching with ML,³⁴ and the other combining rule-based scripted dialog content with artificial intelligence technologies designed to enhance question comprehension.³⁸ The technical architectures were not explicitly specified in three studies.^25,29,35 See Table 2 for details.

Table 2.

Comparison of platforms and technical architectures.

Category		Number of included studies (references)	Advantages	Disadvantages
Platform	Instant messaging application	9 ^{21,23,26,29,31,32,35,38,39}	High penetration rate, no installation required, easy to use	Limited functionality, privacy relies on third-party platforms
	Mobile application	4 ^22,24,25,30	Powerful functionality, smooth user experience	High development cost, requireing user download
	Web-based platform	6 ^{27,28,33,34,36,37}	Cross-platform compatibility, no installation required	Low user stickiness, weak interactivity
Technical architecture	AI-driven	12^18–21^{,23–25,27–30,34}	Flexible interaction, high adaptability	Requiring large amounts of data
	Rule-based	2 ^36,39	Accurate responses, low cost, easy to control	Cannot handle unscripted queries
	hybrid architecture	2 ^34,38	Balancing accuracy and flexibility	Complex design
	Not specified	3 ^25,29,35	—	—

Interaction methods of chatbots

Among the 19 included studies, hybrid interaction modes dominated chatbot designs. With advancing technologies and evolving patient needs, chatbots relying on a single interaction mode have shown limited effectiveness in clinical management. Conversational chatbots (n = 4)^22,23,25,27 provide personalized emotional support and open-ended Q&A through bidirectional, open dialog with patients, which is conducive to improving users’ mental health and facilitating behavioral change. Task-oriented chatbots (n = 2)^36,39 guide patients through self-management tasks via structured steps, significantly improving treatment adherence. These studies utilized pre-visit surveys for symptom monitoring, requiring patients to submit health data regularly and receive strategy feedback, which led to marked improvements in self-management compliance. Informational chatbots (n = 5)^{28,29,32,33,38} depend on the accuracy and accessibility of knowledge bases. The studies built specialized knowledge bases as information sources, allowing patients to query symptoms and self-management issues via inputting questions or keywords, and receiving standardized information in return. The remaining eight studies featured hybrid chatbots,^{21,24,26,30,31,34,35,37} integrating the strengths of different interaction modes. Three studies combined informational and task-oriented features,^34,35,37 four studies merged conversational and informational functions,^21,24,26,31 and one study integrated all three modes.⁴⁰ See Table 3 for details.

Table 3.

Interaction methods and core functions of chatbots.

Interaction methods	Number of included studies (references)	Core functions	Appropriate intervention contents
Conversational	4 ^22,23,25,27	Emotional support, open-ended Q&A, motivational interviewing	Improving mental health, promoting behavior change
Task-oriented	2 ^36,39	Structured task guidance, surveys, reminders	Enhancing treatment adherence, enabling regular data collection
Informational	5 ^{28,29,32,33,38}	Disease-specific Q&A, information push, keyword-based queries	Providing patient education, addressing specific information needs
Hybrid	8 ^{21,24,26,30,31,34,35,37}	Combines the above functions	Providing comprehensive self-management support

Implementation elements of chatbots

Effective chatbot implementation strategies should be multifaceted, encompassing: user needs-oriented personalized design, multidisciplinary team collaboration, development of evidence-based knowledge bases, friendly personas and multimodal interaction design, embodiment of shared decision-making, privacy and data security, implementation of incentive mechanisms, simplified operation and interface design, and integration with clinical workflows. As detailed in Table 1 and Table 4, the most emphasized implementation elements were a user needs-oriented personalized design (n = 14),^21–23^{,25–32,34,35,37} privacy and data security (n = 14)^21,23–30^,32–37,39 and friendly personas with multimodal interaction design (n = 14).^{21,23,25–34}^,37,39 Furthermore, multidisciplinary team collaboration^{27,29,33,34,37} and evidence-based knowledge bases^24,27–29^{,31–33,35,36} were identified as foundational to maintaining scientific rigor and content accuracy in chatbot development. Notably, eight studies^{25,26,28,29,33–35}^,39 point out that simplifying the operation interface (e.g. by adding navigation bars or supporting font scaling) can significantly lower the usage barriers for elderly patients or digitally disadvantaged groups. Six studies^{22,23,25,27,37,38} mentioned that chatbots should support patient participation in health decision-making. Five studies^{24,27,32,36,39} suggested that providing incentives such as gift cards, monetary rewards, and awards can improve patient engagement to a certain extent. By integrating chatbots with clinical information systems (e.g. electronic medical record, clinical decision support system), three studies^31,35,37 demonstrated that this approach supported real-time clinician review of electronic patient-reported information and enhanced proactive follow-up care. Other details are provided in Table 1.

Table 4.

Implementation elements of chatbots.

Implementation Elements	Specific measures	Number of included studies (references)
User needs-oriented personalized design	It addresses the needs of chronic disease patients by providing information, symptom management, and psychological support through personalized messaging and emotional support modules, while continuously optimizing its functions from user feedback.	14 ^21–23^{,25–32,34,35,37}
Friendly personas and multimodal interaction design	A friendly visual identity and encouraging, non-judgmental language are adopted, while multimodal input methods such as voice, text, and button selections are supported to enhance user trust and engagement.	14 ^{21,23,25–34}^,37,39
Privacy and data security	Employ encrypted communication, anonymous data storage, and strict access controls to protect patient privacy and data security, while obtaining informed consent by signing data usage agreements with patients.	14 ^21,23–30^,32–37,39
Multidisciplinary team collaboration	Form a multidisciplinary team of experts from fields including clinical medicine, nursing, informatics, behavioral psychology, and pharmacology to ensure the scientific validity and practical applicability of the content.	5 ^{24,27,32,36,39}
Development of evidence-based knowledge bases	The knowledge base is built on evidence-based guidelines, expert-validated content, and patient needs, and is regularly updated to maintain information accuracy.	9 ^24,27–29^{,31–33,35,36}
Simplified operation and interface design	Optimize the user interface by incorporating features such as navigation bars, usage guides, font scaling support, and multiplatform compatibility.	8 ^{25,26,28,29,33–35}^,39
Embodiment of shared decision-making	The system supports patient participation in health decision-making and helps develop personalized management plans through functions such as semi-open-ended questions, personalized goal setting, and strategy selection.	6 ^{22,23,25,27,37,38}
Implementation of incentive mechanisms	Provide incentives such as gift cards, monetary rewards, and awards to enhance patient engagement to a certain extent.	5 ^{24,27,32,36,39}
Integration with clinical workflows	Integrating patient-reported data into clinical systems and facilitating actionable information exchange within the healthcare team.	3 ^31,35,37

Evaluation indicators for chatbots

Given the heterogeneity of the included studies, a meta-analysis was not performed. Instead, a narrative synthesis was conducted, with outcomes quantified in Table 1. Chatbot interventions were evaluated across four domains: health outcomes, user engagement, self-management ability, system usability and user satisfaction. Fifteen studies^21–27^{,29,30,35,37,38} indicated that chatbots could improve health-related outcomes, including health-related quality of life^22,30,35 mental health,^21,23–25^,27,30 disease-specific indicators,^{21,22,25,26,29–31}^,35,37,38 and social function indicators.^23,25,35 Ten studies^{26,28,30,33–39} assessed user engagement through analysis of usage data, including frequency of use, task completion rates, and interaction duration. However, a decline in usage over time was observed in two studies.^30,38 Eleven studies^{21,23,25,26,28–30}^,32,35,37,39 addressed the impact of chatbots on self-management indicators, such as self-management capabilities,^29,31,37 treatment adherence,^21,23,26,39 self-efficacy,^25,35,39 knowledge or skill acquisition,^23,26,30,32 information needs,²⁸ and treatment readiness.³⁵ Seven studies^27–29^,32,34,36,38 evaluated system usability and feasibility using tools including the System Usability Scale (SUS),^27,32,34 the Chatbot Usability Questionnaire (CUQ),²⁹ self-developed usability questionnaires,^28,36,38 and system performance metrics (e.g. accuracy, specificity, FI score).³³ Overall, usability ratings were high. User satisfaction was investigated in seven studies^{24,25,33,35–37} via self-developed questionnaires and the Client Satisfaction Questionnaire-8 (CSQ-8).²⁷ Additionly, six studies employed interviews to gain deeper insights.^{24,26,27,34,35,39}

Discussion

This scoping review synthesized evidence from 19 studies to map the current landscape of chatbot applications in chronic disease management. The analysis revealed diverse application platforms, technical architectures, and implementation strategies, while also identifying significant trends and gaps in the critical, cross-cutting themes of equity, privacy, and clinical integration. The clinical value of this study lies in providing a theoretical foundation and practical guidance for optimizing chatbot technology, formulating standardized clinical implementation pathways, and promoting the sustainable development of chatbots in chronic disease management.

This review found that instant messaging applications were the most widely used platforms among the included studies, followed by web-based platforms and mobile applications. Instant messaging applications benefit from their extensive user base and high penetration rate.^{21,23,26,29,31,32,38,39} Since patients can utilize chatbot features without learning new operations, these platforms enhance participation rates while significantly lowering the usage barrier for older adults. Mobile applications offer comprehensive functionalities, supporting complex monitoring and management features, but incur higher development and maintenance costs. Web-based platforms facilitate centralized information display and retrieval, yet their limited adaptability to mobile devices poses operational inconveniences for patients. These differences suggest that platform selection should be based on a comprehensive consideration of specific clinical contexts, target population characteristics, and intervention complexity.

In terms of technical architecture, this study confirms that artificial intelligence (AI) systems represent the predominant framework,^21–24^{,26–28,30–33,37} demonstrating advantages in natural language interaction and personalized feedback.⁴¹ With ongoing advances in natural language generation and emotion recognition technologies, chatbots are progressively achieving more natural and empathetic interactions.^42,43 Nevertheless, AI systems still carry potential risks, such as output uncertainty and lack of transparency in decision-making logic.¹⁵ In contrast, rule-based systems offer greater controllability and safety in handling structured tasks such as medication reminders and risk assessments. To balance usability and safety, future systems should adopt layered architectures and modular designs tailored to the characteristics and needs of different chronic disease patient groups, so as to better support their self-management behaviors.

Current research has accumulated significant experience in user-centered design, interdisciplinary collaboration, and evidence-based knowledge base construction. However, to achieve the scalability and sustainable application of chatbots in chronic disease management, it is still necessary to develop a multidimensional integration strategy while balancing technical feasibility and clinical practicality. First, at the user level, an equity-oriented design philosophy should be advocated. Users of different ages, digital literacy levels, racial and ethnic backgrounds, socioeconomic statuses, and language preferences should be included in the evaluation and testing phases during the early development stages to lower the barrier to technology use and avoid exacerbating health inequities due to the digital divide. Some studies have begun to address these aspects, but these efforts are often isolated and not systematically integrated. In the short term, measures such as incentive mechanisms and interface simplification can effectively enhance user engagement. In the short term, measures such as incentive mechanisms and interface simplification can effectively enhance user engagement.^{24,27,32,36,39} However, long-term engagement still depends on patients’ intrinsic motivation, the functional design of the chatbot, and its clinical value. Second, at the system functionality level, multidisciplinary teams and evidence-based knowledge bases are the foundation for ensuring information accuracy.^44,45 However, predefined scripts often lack flexibility when dealing with complex symptoms. Therefore, technologies such as NLP need to be introduced to enhance semantic understanding and interactive capabilities.⁴³ The current hybrid interaction model, which integrates informational, task-based, and conversational elements, has become the most widely adopted approach. It covers the workflow of screening, education, implementation, and feedback, thereby improving adherence and clinical utility.⁴⁶ Future research should prioritize the systematic integration of equity principles by ensuring that chatbots are rigorously tested and optimized with diverse user groups during the design phase to mitigate the digital divide. It is also essential to establish interdisciplinary teams throughout the development process, fostering the creation of hybrid interaction modes tailored to user needs. Such an approach will enhance the reliability of knowledge bases and improve the overall technical adaptability of chatbot systems.

The application of chatbots may raise ethical problems, primarily related to patient data security and privacy. As chatbots collect and store substantial amounts of personal and medical information (e.g. voice recordings and geolocation data), the highly sensitive nature of healthcare data poses significant risks. Unauthorized disclosure of such information could lead to stigmatization or discrimination against participants, thereby amplifying their concerns regarding data privacy and security.^14–16 Consequently, it is essential to strengthen privacy safeguards in digital health through measures such as data encryption and anonymization to establish and maintain patient trust. Regarding clinical integration, current progress remains limited. True integration should enable data sharing between patients and providers, supporting real-time monitoring and proactive interventions. Future research should strengthen privacy protection technologies, including measures such as communication encryption and data anonymization, to ensure the security of chatbots; secondly, a clinical integration framework should be established. Since chatbots collect extensive management data from cancer patients, including medication, symptoms, psychology, and information, the use of chatbots can be integrated into hospital electronic medical record systems to enable data sharing. Additionally, an ethical protection mechanism should be established, with transparent informed consent processes developed to ensure fairness and transparency.

Chatbot-mediated interventions demonstrate potential in enhancing chronic disease management, yet their effectiveness is closely tied to alignment with patient-specific needs and clinical contexts. Evidence confirms benefits across several domains, particularly in mental health, symptom control, social functioning, and self-management ability. Improvements were observed in psychological indicators and selected clinical outcomes such as complication and fracture rates. However, benefits in areas including cognitive function, pain management, and psychosocial support remained limited or inconsistent. Furthermore, some benefits, such as reductions in anxiety, were not sustained in the long term, as evidenced by non-significant differences at the 12-month follow-up in one study.³⁰ In the management of breast cancer chemotherapy side effects, Tawfik et al.reported that the ChemoFreeBot intervention led to significant reductions in the frequency, severity, and distress of physical and psychological symptoms, with self-care behavior effectiveness scores significantly higher than in the nurse-led education group.²⁹ However, it is crucial to note that this finding is not universally generalizable. Notably, studies by Gong et al.³⁰and Ma et al.³⁸ observed declining usage frequency over time, potentially attributed to misalignment between interaction design and patients’ health literacy or usage preferences, declining long-term needs, and insufficient intrinsic motivation. While usability and information utility receive high patient ratings.^28,29 These findings highlight a need for more adaptive and person-centered interaction strategies to sustain participation. User satisfaction was generally high, often attributed to perceived information value, interface usability, and clear visual design.^24,34 Nevertheless, limitations in conversational naturalness and technical reliability were consistently identified as barriers to a positive user experience.^24,27 System usability assessments reveal critical gaps. Only one study employed the chatbot-specific CUQ scale,²⁹ while others relied on SUS or custom questionnaires, overlooking core conversational competencies like dialog fluency, affective interaction naturalness, and clinical semantic precision.^17,47 Current research suffers from fragmented metrics, lack of safety assessments, absence of standardized tools, and insufficient long-term verification, hindering effective comparison of results across studies.⁴⁶ Thus, future research should establish standardized, multidimensional evaluation systems, verify chatbots’ long-term efficacy in chronic disease management via high-quality studies, use mixed methods to reveal feasibility-influencing mechanisms, clarify ethical guidelines, and conduct safety assessments, ultimately enhancing the comparability of intervention effects and the clinical translational value of chatbots in chronic disease management.

Implications for future research

Future research should concentrate on four priorities: First, establishing interdisciplinary collaborative teams that integrate expertise from clinical medicine, informatics, and psychology throughout the chatbot design, development, and evaluation process to enable data-driven precision management. Second, advancing AI-driven optimization by leveraging LLMs for complex query processing, developing personalized multimodal chatbots integrated with electronic health records, and implementing senior-friendly designs (e.g. simplified user interface, gamification) with cross-device compatibility. Third, conducting RCTs to establish core outcome measures and validate long-term effectiveness through multicenter studies. Finally, strengthening regulatory and privacy protection frameworks by implementing rigorous privacy-preserving protocols and robust data security mechanisms to build patient trust.

Strengths and limitations

The innovation of this review lies in being the first to systematically extract and construct an night-dimensional implementation strategy framework for the use of chatbots in chronic disease management, along with an analysis of their characteristics, thereby providing a critical theoretical foundation for establishing standardized clinical intervention pathways in the future. Another advantage is reflected in the methodological rigor. By integrating evidence from multiple countries through a scoping review approach and addressing the current challenges of high heterogeneity among studies, it offers a practice-oriented direction for developing standardized solutions driven by clinical needs.

A potential limitation of the current study is the small sample sizes of the included studies and the limited number of studies on the same diseases, coupled with significant variations (e.g. in disease types and outcome measures), which precluded a meta-analysis of effects across studies. Secondly, the single-center study designs limit the generalizability of the findings. Future research should involve multicenter, large-sample randomized controlled trials, and meta-analyses are necessary. Finally, Most prior studies were conducted as short-term pilot investigations; future research should adopt longitudinal designs, and qualitative methods could be employed to explore potential barriers and facilitators during implementation.

Conclusion

Chatbots hold considerable clinical value in chronic disease management, yielding positive effects on health outcomes, self-management, and other domains. However, current research has some limitations. Moving forward, it will be necessary to further optimize interaction design, enhance system functionality, and strengthen privacy protection to better facilitate the application of chatbots in chronic disease management.

Supplemental Material

sj-docx-1-dhj-10.1177_20552076251411287 - Supplemental material for Application of chatbots in chronic disease management: A scoping review

Supplemental material, sj-docx-1-dhj-10.1177_20552076251411287 for Application of chatbots in chronic disease management: A scoping review by Jiayi Hou, Shineng Lin, Peimeng Teng, Yuyuan Han, Yijia Luo and Guijuan He in DIGITAL HEALTH

Footnotes

Acknowledgments

We would like to acknowledge all of the participants and all of the researchers in this study.

ORCID iDs

Jiayi Hou

Shineng Lin

Peimeng Teng

Yijia Luo

Guijuan He

Ethical considerations

Not applicable.

Consent to participate

Not applicable.

Contributorship

J. Hou and S. Lin: conceptualization, methodology, investigation, data curation, formal analysis, writing—original draft, writing—review & editing. P. Teng, Y. Han, and Y. Luo: investigation, data curation, writing—review & editing. G. He: conceptualization, methodology, resources, supervision, project administration, funding acquisition, writing—review & editing.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by Zhejiang Provincial Natural Science Foundation of China under Grant No.LTGY23H250001.

Declaration of conflicting interests

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

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

World Health Organization. Noncommunicable diseases. 2024. https://www.who.int/health-topics/noncommunicable-diseases#tab=tab_1 .

Hacker

Briss

Richardson

, et al. COVID-19 and chronic disease: the impact now and in the future. Prev Chronic Dis 2021; 18: E62.

World Health Organization. World health statistics 2024: monitoring health for the SDGs, sustainable development goals. https://www.who.int/publications/i/item/9789240094703.

State Council of the People's Republic of China. Opinions of the State Council on Implementing the Healthy China Initiative. https://www.gov.cn/zhengce/content/2019-07/15/content_5409492.htm.

Guan

Meng

. Health Providers’ perceptions and experiences of using mHealth for chronic noncommunicable diseases: qualitative systematic review and meta-synthesis. J Med Internet Res 2023; 25: e45437.

van de Vijver

Tensen

Asiki

, et al. Digital health for all: how digital health could reduce inequality and increase universal health coverage. Digit Health 2023; 9: 20552076231185434.

Thacharodi

Singh

Meenatchi

, et al. Revolutionizing healthcare and medicine: the impact of modern technologies for a healthier future-A comprehensive review. Health Care Sci 2024; 3: 329–349.

Liang

Pei

Staibano

, et al. Clinical applications of large language models in medicine and surgery: a scoping review. J Int Med Res 2025; 53: 3000605251347556.

Xue

Zhang

Zhao

, et al. Evaluation of the current state of chatbots for digital health: scoping review. J Med Internet Res 2023; 25: e47217.

10.

Wang

Yan

Zhu

, et al. Application progress of conversational agents in chronic disease management. J Nurs Admin 2023; 23: 365–369.

11.

Schachner

Keller

Wangenheim

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

12.

Wen

Sun

Yang

, et al. Research progress on the application of AI chatbots in cancer nursing. J Nurs 2025; 32: 23–26.

13.

Palanica

Flaschner

Thommandram

, et al. Physicians’ perceptions of chatbots in health care: cross-sectional web-based survey. J Med Internet Res 2019; 21: e12887.

14.

Chow

JCL

. Ethical considerations in human-centered AI: advancing oncology chatbots through large language models. JMIR Bioinform Biotechnol 2024; 5: e64406.

15.

Chow

JCL

. Large language models in medical chatbots: opportunities, challenges, and the need to address AI risks. Information 2025; 16: 549.

16.

Singhal

Gottweis

, et al. Toward expert-level medical question answering with large language models. Nat Med 2025; 31: 943–950.

17.

Zhang

Song

. A chatbot based question and answer system for the auxiliary diagnosis of chronic diseases based on large language model. Sci Rep 2024; 14: 17118.

18.

Altom

Awad Taha

Mahmoud Hussein

AAA

, et al. Artificial intelligence-based chatbots in chronic disease management: a systematic review of applications and challenges. Cureus 2025; 17: e81001.

19.

Arksey

O'malley

. Scoping studies: towards a methodological framework. Int J Soc Res Methodol 2005; 8: 19–32.

20.

Tricco

Lillie

Zarin

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

21.

Liu

Zhao

Zhang

. Influence of WeChat chatbot intervention on psychological status, treatment compliance, and complications in patients with hemodialysis arteriovenous fistula. Acta Academiae Medicinae Bengbu 2023; 48: 1121–1125.

22.

Ogawa

Oyama

Morito

, et al. Can AI make people happy? The effect of AI-based chatbot on smile and speech in Parkinson's disease. Parkinsonism Rel Disord 2022; 99: 43–46.

23.

Yasukawa

Tanaka

Yamane

, et al. A chatbot to improve adherence to internet-based cognitive-behavioural therapy among workers with subthreshold depression: a randomised controlled trial. BMJ Mental Health 2024; 27: e300881.

24.

MacNeill

Doucet

Luke

. Effectiveness of a mental health chatbot for people with chronic diseases: randomized controlled trial. JMIR Formative Research 2024; 8: e50025.

25.

Ulrich

Gantenbein

Zuber

, et al. Development and evaluation of a smartphone-based chatbot coach to facilitate a balanced lifestyle in individuals with headaches (Balanceup app): randomized controlled trial. J Med Internet Res 2024; 26: e50132.

26.

Echeazarra

Pereira

Saracho

. Tensiobot: a chatbot assistant for self-managed in-house blood pressure checking. J Med Syst 2021; 45: 54.

27.

Bruijnes

Kesteloo

Brinkman

. Reducing social diabetes distress with a conversational agent support system: a three-week technology feasibility evaluation. Front Digit Health 2023; 5: 1149374.

28.

Baumgartner

Byczkowski

Schmid

, et al. Effectiveness of the medical chatbot PROSCA to inform patients about prostate cancer: results of a randomized controlled trial. Europ Urol Open Sci 2024; 69: 80–88.

29.

Tawfik

Ghallab

Moustafa

. A nurse versus a chatbot ‒ the effect of an empowerment program on chemotherapy-related side effects and the self-care behaviors of women living with breast cancer: a randomized controlled trial. BMC Nurs 2023; 22: 102.

30.

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.

31.

Yue

, et al. Application of AI chatbots in the extended nursing care for patients after osteoporotic fracture surgery. J Nurs Sci 2023; 38: 13–17.

32.

Chen

Huang

Fan

, et al. User evaluation of a chat-based instant messaging support health education program for patients with chronic kidney disease: preliminary findings of a formative study. JMIR Form Res 2023; 7: e45484.

33.

Kim

Park

. A question answering chatbot for gastric cancer patients after curative gastrectomy: development and evaluation of user experience and performance. Comput Inform Nurs 2024; 42: 829–839.

34.

Griffin

Khairat

Bailey

, et al. A chatbot for hypertension self-management support: user-centered design, development, and usability testing. JAMIA Open 2023; 6: ooad073.

35.

Gomaa

Posey

Bashir

, et al. Feasibility of a text messaging-integrated and chatbot-interfaced self-management program for symptom control in patients with gastrointestinal cancer undergoing chemotherapy: pilot mixed methods study. JMIR Form Res 2023; 7: e46128.

36.

Soley

Klein

Taylor

, et al. Feasibility of the genetic information assistant chatbot to provide genetic education and study genetic test adoption among pancreatic cancer patients at Johns Hopkins Hospital. AMIA Jt Summits Transl Sci Proc 2023; 2023: 497–504.

37.

Nassar

Dunlea

Montero

, et al. Feasibility and preliminary behavioral and clinical efficacy of a diabetes education chatbot pilot among adults with type 2 diabetes. J Diabetes Sci Technol 2023; 19: 54–62.

38.

Orner

Ghaly

, et al. Automated health chats for symptom management of head and neck cancer patients undergoing radiation therapy. Oral Oncol 2021; 122: 105551.

39.

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.

40.

Kim

. The effects of artificial intelligence chatbots on women's health: a systematic review and meta-analysis. Healthcare (Basel) 2024; 12: 534.

41.

Ullah

Ali

. Role of artificial intelligence in healthcare settings: a systematic review. J Med Artif Intellig 2025; 08: 24.

42.

Wang

Zeng

. Tracing intricate cues in dialogue: joint graph structure and sentiment dynamics for multimodal emotion recognition. IEEE Trans Pattern Anal Mach Intell 2025; 47: 8786–8803.

43.

Sorin

Brin

Barash

, et al. Large language models and empathy: systematic review. J Med Internet Res 2024; 26: e52597.

44.

Wang

Lian

, et al. Development and application of chatbots in the medical field: from traditional methods to large language models. Med J Peking Union Med Coll 2025; 16: 1–11.

45.

Liang

Stockinger

de Farias

, et al. Querying knowledge graphs in natural language. J Big Data 2021; 8: 3.

46.

Laranjo

Dunn

Tong

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

47.

Holmes

Moorhead

Bond

, et al.

Usability testing of a healthcare chatbot: can we use conventional methods to assess conversational user interfaces?

Proceedings of the 31st European Conference on Cognitive Ergonomics 2019; 47: 207–214.

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

0.04 MB