Conversational Agents in Health Care: Scoping Review and Conceptual Analysis

Lorainne Tudor Car, Dhakshenya Ardhithy Dhinagaran, Bhone Myint Kyaw, Tobias Kowatsch, Shafiq Joty, Yin Leng Theng, Rifat AtunJournal of Medical Internet Research 2020 · DOI

Abstract

The literature on conversational agents in health care is largely descriptive and aimed at treatment and monitoring and health service support. It mostly reports on text-based, artificial intelligence-driven, and smartphone app-delivered conversational agents. There is an urgent need for a robust evaluation of diverse health care conversational agents' formats, focusing on their acceptability, safety, and effectiveness.

Introduction

Conversational agents, commonly known as chatbots, are increasingly positioned as scalable solutions for improving health care accessibility, personalization, and efficiency. However, the rapid pace of health-tech development has created a "wild west" environment where many tools are deployed without a clear understanding of their clinical efficacy or safety. This scoping review maps the current landscape of health care chatbots, identifying a critical shift from simple rule-based systems to sophisticated, AI-driven tools, while highlighting a significant "evaluation gap" in the existing literature.

Methodology

The authors conducted a comprehensive scoping review following the Arksey and O’Malley framework. A broad literature search was performed across five major databases: MEDLINE (Ovid), EMBASE (Ovid), PubMed, Scopus, and Cochrane Central, supplemented by gray literature sources.

The search strategy utilized 63 terms covering synonyms for conversational AI and chatbots. Inclusion was restricted to primary research studies that implemented and evaluated a health-specific conversational interface, while Embodied Conversational Agents (ECAs) featuring visual avatars were excluded to focus strictly on the conversational modality.

PRISMA flow diagram of the study selection process

As shown in the selection process, the review ultimately synthesized data from 47 study reports, comprising 45 peer-reviewed articles and 2 ongoing clinical trials, representing the state-of-the-art in health chatbot research as of early 2019.

The Evolving Landscape of Health Chatbots

The field has undergone a significant evolution, moving away from rigid, rule-based systems toward "smart" agents powered by Machine Learning (ML) and Natural Language Processing (NLP). This transition allows for more fluid interactions but introduces the "black box" concern, where the reasoning behind an agent's clinical advice becomes opaque, posing potential safety risks.

Timeline of conversational agent development from 1966 to 2019

A key design trend identified is the use of specific "personality" traits—such as being "coach-like" or "human-like"—to drive user engagement, particularly in mental health contexts. Despite the rise of voice assistants like Alexa and Siri, the majority of health-specific agents remain text-heavy. Of the studies reviewed, 19 utilized free-text input and 30 relied on text-only output, suggesting that speech-based interaction remains an underutilized modality for populations with motor or visual impairments.

Key Findings and Results

The review categorized the primary applications of these agents into three dominant areas:

  1. Treatment and Monitoring: Managing chronic diseases and medication adherence.
  2. Health Care Service Support: Triaging symptoms and booking appointments.
  3. Patient Education: Providing accessible health information.

While the technology is advancing, the quality of evidence remains lopsided. Most literature consists of descriptive development case studies; only 11 of the 47 identified studies were Randomized Controlled Trials (RCTs). Furthermore, a significant geographic bias exists, with 40 of the 47 studies originating in high-income countries, leaving a research gap in low- and middle-income regions where these tools could significantly improve care access.

Discussion and Future Directions

To move the field forward, the authors identify several critical hurdles. First, there is a glaring lack of safety data regarding adverse effects or inappropriate AI-generated advice. Second, the authors advocate for "hybrid systems" where AI supplements rather than replaces human clinicians, escalating complex cases to professionals when necessary.

Finally, as chatbots increasingly take on roles akin to health care professionals, there is an urgent need for economic evaluations and more stringent regulatory frameworks to ensure these tools are not only innovative but also cost-effective and safe for patient use.

Summary updated

Abstract

The literature on conversational agents in health care is largely descriptive and aimed at treatment and monitoring and health service support. It mostly reports on text-based, artificial intelligence-driven, and smartphone app-delivered conversational agents. There is an urgent need for a robust evaluation of diverse health care conversational agents' formats, focusing on their acceptability, safety, and effectiveness.
BibTeX
@article{car2020conversational,
  author = {Lorainne Tudor Car and Dhakshenya Ardhithy Dhinagaran and Bhone Myint Kyaw and Tobias Kowatsch and Shafiq Joty and Yin Leng Theng and Rifat Atun},
  title = {{Conversational Agents in Health Care: Scoping Review and Conceptual Analysis}},
  journal = {Journal of Medical Internet Research},
  year = {2020},
  doi = {10.2196/17158},
}
Figure