This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Symptom checkers (SCs) for laypersons’ self-assessment and preliminary self-diagnosis are widely used by the public. Little is known about the impact of these tools on health care professionals (HCPs) in primary care and their work. This is relevant to understanding how technological changes might affect the working world and how this is linked to work-related psychosocial demands and resources for HCPs.
This scoping review aimed to systematically explore the existing publications on the impacts of SCs on HCPs in primary care and to identify knowledge gaps.
We used the Arksey and O’Malley framework. We based our search string on the participant, concept, and context scheme and searched PubMed (MEDLINE) and CINAHL in January and June 2021. We performed a reference search in August 2021 and a manual search in November 2021. We included publications of peer-reviewed journals that focused on artificial intelligence- or algorithm-based self-diagnosing apps and tools for laypersons and had primary care or nonclinical settings as a relevant context. The characteristics of these studies were described numerically. We used thematic analysis to identify core themes. We followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist to report the study.
Of the 2729 publications identified through initial and follow-up database searches, 43 full texts were screened for eligibility, of which 9 were included. Further 8 publications were included through manual search. Two publications were excluded after receiving feedback in the peer-review process. Fifteen publications were included in the final sample, which comprised 5 (33%) commentaries or nonresearch publications, 3 (20%) literature reviews, and 7 (47%) research publications. The earliest publications stemmed from 2015. We identified 5 themes. The theme
The scoping review approach was suitable for this new field of research. The heterogeneity of technologies and wordings was challenging. We identified research gaps in the literature regarding the impact of artificial intelligence– or algorithm-based self-diagnosing apps or tools on the work of HCPs in primary care. Further empirical studies on HCPs’ lived experiences are needed, as the current literature depicts expectations rather than empirical findings.
Medical laypersons’ symptom self-assessment and deducing the necessity and urgency of contacting a physician is a core moment in medical care. Although self-diagnosis and self-triage in acute medical conditions or self-allocation to health care services beyond acute situations are not new, the internet has increased this phenomenon owing to the rapid and accessible health information that is available to consumers on the web. Web-based searches are a means for medical laypersons in general and for patients to inform themselves in this process. Representative studies in Germany have shown that 74% of laypersons use internet searches to inform themselves about health-related issues [
A novelty in this process is the plethora of artificial intelligence (AI)- or algorithm-based tools such as symptom checkers (SCs;
“Chatbots, also known as conversational agents, interactive agents, virtual agents, virtual humans, or virtual assistants, are artificial intelligence programs designed to simulate human conversation via text or speech...In the context of health care, chatbots or
“[Symptom checkers] generally provide people with possible alternative diagnoses based on their reported symptoms and/or suggest a course of action (eg, self-care, make a general practitioner (GP) appointment or go to an emergency department [ED])” [
From the perspective of occupational health, we are particularly interested in understanding how technological changes might affect the working world and how this is linked to work-related psychosocial demands and resources [
As SCs are a rather new technology, we expected limited knowledge and heterogeneous sources with regard to our research interests. Therefore, we chose to conduct a scoping review as a suitable method to gain an overview of existing knowledge [
This scoping review aimed to systematically explore the existing knowledge on the impacts of SC on HCPs in general practice and to identify gaps in knowledge. This review was guided by the following research question:
A scoping review approach allowed us to identify, retrieve, and summarize publications relevant to a particular topic and to identify the key concepts underpinning a research area and the main sources and types of evidence available [
Our scoping review included the following key phases [
We followed the population, concept, and context scheme [
We included publications in peer-reviewed journals and did not restrict by study design to capture the wide array of relevant publications. The time frame from 2000 to 2021 was chosen to minimize the inclusion of irrelevant publications of older tools for self-diagnosis. The publications included were published in English, German, French, Turkish, or Russian and were obtainable in full text. We excluded studies that did not fit the topic and setting or focused on technologies that were designed for a professional target group such as tools for clinical decision support (
Publication date from 2000 to 2021
Languages: English, German, French, Turkish, and Russian
All study types, views, editorials, etc
Symptom checkers for general conditions
General practice and primary care
Health care professionals and physicians in general
Used by lay public or patient-facing
Artificial intelligence– or algorithm-based
For self-diagnosis or self-triage
For example, blogs, thesis, and dissertations
Symptom checkers focused on specific medical conditions
Hospital
Used by physicians, nurses, etc
Diagnostics with sensors, wearables, etc
Clinical decision support systems, etc
As we expected limited results with regard to our specific research interests, we conceptualized a broad search strategy. We considered 2 medical databases, PubMed (MEDLINE) and CINAHL, as comprehensive databases for our research question and searched them systematically. We worked with a librarian at the University of Tübingen to construct a search term to ensure comprehensive results. We developed the search term using keywords and MeSH (Medical Subject Headings) terms from thematically relevant empirical studies [
The initial database search was conducted in January 2021, and the follow-up search was conducted in June 2021 to identify the publications published after the initial search. Subsequently, we reviewed the reference lists of the included publications and references of known relevant reviews on SCs [
We imported the database search results into the reference management system Citavi [
The whole process (screening and selection) was conducted and constantly discussed by independent reviewers (NR and EÖ) to assess interrater reliability [
We developed the data charting form following the Arksey and O’Malley framework [
We extracted key information from the included publications pertaining to publication information (first author, year, type of publication, and country of origin); study information, if applicable (aim, setting, population data, sample size, methods, self-diagnosis tool types, and limitations); main findings; and implications related to our research question. All findings concerning the impact of SCs on GPs or HCPs were included. The main outcomes of interest were (impacts on) work content, organization or service use, job satisfaction of HCPs, perceived stress, HCP-patient or lay interaction or relationship and communication, and professional identity. Data extraction was performed simultaneously by 2 reviewers (NR and CP) and checked reciprocally for accuracy and consistency.
We used thematic analysis [
A numerical analysis of the selection process and the included publications is presented. Of the 2729 publications found in the database search, 385 (14.11%) duplicates and an additional 211 (7.73%) publications with unsuitable publication types were removed. Of the remaining 78.16% (2133/2729) of publications, 2090 (97.98%) articles were removed after screening for inclusion, and 43 (2.02%) articles were screened for eligibility, of which 9 (21%) publications were considered eligible. A total of 69 publications were included through manual search and screened for eligibility, which resulted in further 8 (12%) included publications. The number of included publications for data extraction was 17. On the basis of the detailed feedback during the peer-review process in this journal, we excluded 2 more publications. The final number of included publications was 15 (
PRISMA (Preferred Reporting Item for Systematic Reviews and Meta-Analyses) flowchart of the study selection process.
Most publications (5/15, 33%) were published in 2019 [
We identified 5 themes in the heterogeneity of the included publications.
The charting table (
Symptom checkers (SCs) and health care professionals (HCPs): overview of main themes in the publications included.
Theme, topic, and subtopic | Research | Literature review | Nonresearch | |||||
|
||||||||
|
|
|||||||
|
|
Lower accuracy of SCs in diagnostics and triage | [ |
[ |
N/Aa | |||
Physicians consider various sources of information | N/A | N/A | [ |
|||||
SCs provide a first sense of urgency | [ |
N/A | N/A | |||||
|
||||||||
|
SCs lack human factor | [ |
N/A | [ |
||||
Human factor as a strength of physicians | [ |
[ |
N/A | |||||
|
||||||||
|
|
|||||||
|
|
SCs might provide laypersons with knowledge | N/A | N/A | [ |
|||
SCs might have layperson activating potential | N/A | N/A | [ |
|||||
|
||||||||
|
SCs' quality and liability still unclear | [ |
N/A | N/A | ||||
SCs might increase anxieties in laypersons | N/A | N/A | [ |
|||||
|
||||||||
|
|
|||||||
|
|
Potential of changing roles | [ |
N/A | [ |
|||
|
||||||||
|
Improvement of relationship | N/A | N/A | [ |
||||
HCPs remain the primary source of trustworthy information | [ |
N/A | N/A | |||||
|
||||||||
|
|
|||||||
|
|
SCs could support repetitive and less-complicated tasks | [ |
N/A | [ |
|||
SCs could enable physicians to concentrate on other tasks | [ |
N/A | [ |
|||||
|
||||||||
|
SCs might cause unnecessary contacts | [ |
[ |
[ |
||||
SCs might cause additional tasks during consultation | [ |
N/A | [ |
|||||
|
||||||||
|
|
|||||||
|
|
SCs transform the roles of HCPs | [ |
N/A | N/A | |||
SCs complement HCPs | [ |
[ |
[ |
|||||
SCs in competition with HCPs | [ |
[ |
N/A | |||||
|
||||||||
|
Improved supply for patients | [ |
N/A | [ |
||||
Potential overuse of service through patients | [ |
N/A | N/A |
aN/A: not applicable.
Charting table of included publications.
Publication details | Main themes | ||||||||
First author, year, reference | Setting in health care | Type of publication and study typea | SCb or AIc-based technologyd | Finding prediagnosis: comparison SCs and HCPse | Patient-technology-relationship | (Impacts on) physician-patient relationship | Impact on HCPs’ tasks | (Future) role of SCs in health care | |
Blease et al [ |
Primary care | Research: qualitative approach | AI technology | ✓ | N/Af | N/A | ✓ | ✓ | |
Chambers et al [ |
Setting not specified | Literature review | SC and triage services | ✓ | N/A | N/A | N/A | N/A | |
Denecke et al [ |
Setting not specified | Nonresearch: commentary | Chatbots or intelligent conversational agents | N/A | ✓ | ✓ | ✓ | N/A | |
Gottliebsen and Petersson [ |
Primary care | Literature review | Digital tools with triage service | ✓ | N/A | N/A | N/A | N/A | |
Jutel et al [ |
Setting not specified | Research: qualitative approach | Medical diagnosis smartphone apps | ✓ | N/A | ✓ | N/A | N/A | |
Kuhn et al [ |
Setting not specified | Nonresearch: commentary | Chatbots | N/A | ✓ | N/A | ✓ | ✓ | |
Kujala et al [ |
Primary care | Research: mixed methods approach | Web-based SC | N/A | N/A | ✓ | ✓ | ✓ | |
Lupton et al [ |
Setting not specified | Research: qualitative approach | Self-diagnosis smartphone apps or SC | ✓ | N/A | ✓ | N/A | N/A | |
Merz et al [ |
Setting not specified | Nonresearch: commentary | (Self-)diagnosis apps | ✓ | N/A | N/A | N/A | N/A | |
Miles et al [ |
Primary care | Research: quantitative approach | Chatbots (for diagnosis and triage) | N/A | ✓ | ✓ | N/A | N/A | |
Müschenich et al [ |
Setting not specified | Nonresearch: commentary | Chatbots | ✓ | ✓ | ✓ | N/A | N/A | |
Palanica et al [ |
Primary care | Research: quantitative approach | Health care chatbots | ✓ | N/A | N/A | ✓ | ✓ | |
Semigran et al [ |
Setting not specified | Research: quantitative approach | SC | ✓ | N/A | N/A | N/A | ✓ | |
Summerton and Cansdale [ |
Primary care | Nonresearch: commentary | AI-based diagnosis | ✓ | N/A | N/A | ✓ | ✓ | |
Wattanapisit et al [ |
Primary care | Literature review | Mobile health apps for self-diagnosis | ✓ | N/A | N/A | N/A | ✓ |
aIf applicable.
bSC: symptom checker.
cAI: artificial intelligence.
dTools or technology concerning research questions and eligibility criteria.
eHCP: health care professional.
fN/A: not applicable.
One theme that repeatedly occurred in the publications can be grouped under the
Some authors attested that SCs have a
According to some authors, SCs
Some authors addressed the impacts of SCs on layperson users as (potential) patients. We consider this relevant to our research question, as it might indirectly transform HCPs’ work content.
Authors saw the potential of SCs to
According to some authors, SCs could be potentially harmful for laypersons, as the
Some authors addressed the impacts of SCs on the physician-patient relationship. We consider this to be relevant because the physician-patient relationship is also part of the HCPs’ work content (whereas the relationship between practice team members is part of social relationships).
Some authors addressed the
Some authors indicated that SCs might have an impact but that the roles of HCPs would remain unchallenged. Some authors voiced that SC might have
The impacts of SCs on HCPs’ tasks is another theme in the literature relevant to our research question, as it is linked to work content. We included those parts that might have an impact on HCPs’ work.
In some publications, SCs were seen as reducing HCPs’ workload through providing or
Some authors assumed that SCs might increase HCPs’ workload.
Some authors made prognoses for the future of SCs and their impacts on health care. We consider those visions, which indicate an impact on HCPs’ work, as relevant to our topic.
Some publications addressed the future role that SCs could play in health care with regard to HCPs and the health care system. SCs might
Some authors addressed the potential impacts of SCs on the health care system. Some assumed the potential for
Looking beyond the respective themes, we identified a spectrum of 4 main discourses in which SCs were discussed as inferior, harmful, beneficial, or superior to humans (
In the discourse of
Symptom checkers (SCs): technology in relation to humans.
Theme, topic, and subtopic | SCs as inferior | SCs as harmful | SCs as beneficial | SCs as superior | ||||||
|
||||||||||
|
|
|||||||||
|
|
Lower accuracy of SCs in diagnostics and triage | ✓ | N/Aa | N/A | N/A | ||||
Physicians consider various sources of information | ✓ | N/A | N/A | N/A | ||||||
SCs provide a first sense of urgency | N/A | N/A | ✓ | N/A | ||||||
|
||||||||||
|
SCs lack human factor | N/A | N/A | N/A | ✓ | |||||
Human factor as a strength of physicians | ✓ | N/A | N/A | N/A | ||||||
|
||||||||||
|
|
|||||||||
|
|
SCs might provide laypersons with knowledge | N/A | N/A | ✓ | N/A | ||||
SCs might have layperson activating potential | N/A | N/A | ✓ | N/A | ||||||
|
||||||||||
|
SCs' quality and liability still unclear | ✓ | N/A | N/A | N/A | |||||
SCs might increase anxieties in laypersons | N/A | ✓ | N/A | N/A | ||||||
|
||||||||||
|
|
|||||||||
|
|
Potential of changing roles | N/A | N/A | N/A | ✓ | ||||
|
||||||||||
|
Improvement of relationship | N/A | N/A | ✓ | N/A | |||||
HCPs remain the primary source of trustworthy information | ✓ | N/A | N/A | N/A | ||||||
|
||||||||||
|
|
|||||||||
|
|
SCs could support repetitive and less-complicated tasks | N/A | N/A | ✓ | N/A | ||||
SCs could enable physicians to concentrate on other tasks | N/A | N/A | ✓ | N/A | ||||||
|
||||||||||
|
SCs might cause unnecessary contacts | N/A | ✓ | N/A | N/A | |||||
SCs might cause additional tasks during consultation | N/A | ✓ | N/A | N/A | ||||||
|
||||||||||
|
|
|||||||||
|
|
SCs transform the roles of HCPs | N/A | N/A | N/A | ✓ | ||||
SCs complement HCPs | N/A | N/A | ✓ | N/A | ||||||
SCs in competition with HCPs | N/A | ✓ | N/A | N/A | ||||||
|
||||||||||
|
Improved supply for patients | N/A | N/A | ✓ | N/A | |||||
Potential overuse of service through patients | N/A | ✓ | N/A | N/A |
aN/A: not applicable.
bHCP: health care professional.
SCs are designed for laypersons but might have direct and indirect impacts on HCPs, as patients may use SCs instead of, before, or in addition to the consultation with an HCP. We conducted a scoping review to identify existing knowledge about the impacts of SCs on HCPs in primary care, especially with regard to work content and organization, HCP-patient or lay interaction or relationship, professionals’ identity, job satisfaction, and perceived stress of HCPs. We derived these perspectives from established models of work-related stress and mental health in the workplace [
We included 15 publications and identified 5 main themes concerning the impacts of SCs on physicians in general practice and other settings. The literature in our scoping review consists of research publications, literature reviews, and nonresearch publications such as commentaries. We found the main themes across all 3 types of literature, with clear differences in the levels of the subtopics. Most subtopics in the themes
Interestingly, we could see that SCs entering the field of (pre)diagnosis was not enlarged from the perspective of “professionals’ identity,” “job satisfaction,” and “perceived work-related stress” that were further perspectives on our research question. Subtopics such as
The existing literature turned out to be heterogeneous and inconsistent with regard to the concepts and wording that are used (
The strengths and limitations of our scoping review are related to the methods, search process, findings, and quality assessment.
We used the scoping review methodology using the Arksey and O’Malley framework [
We faced some challenges during the search process. A lack of common keywords and inconsistent database indexing for the topics “SCs/AI in self-diagnosis” made a reproducible database search difficult. We conceive that despite using comprehensive search terms, some potentially relevant publications could not be found. These challenges lead to a rather complex search string for database searches and an iterative search process involving an additional literature search [
In addition, sometimes GPs were only one group of physicians among several included in a study or the background of the physicians was not specified. In both cases, when in doubt, we included a publication that fits other relevant dimensions of our research question. No publication could be found that explicitly addressed the work of practice assistants.
To ensure the quality of the thematic analysis [
The themes we identified resonated with the wider context of SCs and AI in health care.
We identified the first theme as relevant to our research question, as it revolves around the question of who can conduct a core work content of physicians better: trained physicians or SCs. In the wider literature on SCs, this has repeatedly been addressed as a technical question of diagnostic accuracy. With regard to diagnosis performance, the first empirical studies have shown that the quality varied substantially between SCs [
Patient work is a core work content of HCPs and the impacts of SCs on laypersons and (potential) patients might thus also affect HCPs’ work content and workload [
The impact of SCs on the physician-patient relationship is prominently addressed in the publications included in our scoping review and centers on the question of whether the roles of physicians and patients might change through SCs. According to a study on the impacts of patients’ use of social media on their interaction with HCPs, discussing health information collected on the web with HCPs led to perceived “tacit opposition” [
The publications in our scoping review show that SCs are expected to transform the roles of GPs. This is similar to the expected impacts of AI on the roles of HCPs [
With regard to future roles in health care systems, publications on COVID-19–related SCs show that unnecessary in-person visits may be eliminated [
On the basis of the literature, SCs can partially outsource and improve preliminary diagnoses and enhance diagnostic decision-making. Furthermore, they could become tools for “appropriate triage advice” by the patients themselves [
SCs are currently not integrated into daily routines; therefore, GPs cannot control patient risks from using SCs for self-diagnosis and evaluation of a suggested course of action [
Although the topic of bias reduction on HCPs’ side has been addressed, there is a research gap concerning exacerbating inequalities [
This is the first scoping review that centers on the perspective of occupational health in the context of the potential impacts of SCs on the work of physicians in primary care. A current ethical reflection on SC chatbots in clinical settings indicates potential impacts on the work of HCPs such as impacts on HCPs’ practical wisdom, workload, and motivation at work [
PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist.
Search strings for PubMed and CINAHL.
artificial intelligence
general practitioner
health care professional
Medical Subject Headings
Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews
symptom checker
The authors thank Dr Diana Mader, the librarian at the University of Tübingen, for assisting in developing and testing the database search strategy and search string. The authors would like to thank Dr Regina Müller and Dr Malte Klemmt for their valuable feedback on the research process. The authors would also like to thank the anonymous reviewers for their helpful comments.
We acknowledge the support from the Open Access Publishing Fund of the University of Tübingen.
This study used own funds from the Institute of Occupational Medicine, Social Medicine, and Health Services Research. The work of the Institute of Occupational and Social Medicine and Health Services Research Tübingen is supported by an unrestricted grant of the employers’ association of the metal and electric industry Baden-Württemberg (Südwestmetall e.V). The funding bodies had no role in the design of this study or during its execution, analyses, interpretation of the data, or in the decision to submit the results.
The data sets generated and analyzed during this study are available from the corresponding author upon reasonable request.
NR contributed to the planning (scoping review conception); scoping review conduction (search strategy development, literature search and inclusion; data extraction, thematic analysis, data synthesis); and reporting (writing the original draft, feedback to the revised manuscript).
EÖ contributed to the planning (scoping review conception); scoping review conduction (search strategies development, literature search, literature inclusion, and thematic analysis); and reporting (commenting on the original draft and providing feedback to the revised manuscript).
AJW contributed to conducting the scoping review (feedback to search strategy and data synthesis) and reporting (writing original draft and providing feedback to the revised manuscript).
MAR contributed to planning (scoping review conception and protocol development); scoping review conduction (quality assessment of search strategy, thematic analysis, and data synthesis); and reporting (feedback to the original draft and the revised manuscript).
CP contributed to the planning (project coordination, scoping review conception, and protocol development); scoping review conduction (quality assessment of search process, search strategy development, literature inclusion, data extraction, thematic analysis, and data synthesis); and reporting (writing the original draft and revised manuscript).
All authors read and improved the final manuscript.
None declared.