Barriers to the implementation of artificial intelligence in primary care: a scoping review, 2020–2025

Introduction

Background

Artificial intelligence (AI) has emerged as a transformative technology in healthcare, offering promising avenues to enhance clinical decision-making, improve patient outcomes, and address systemic inefficiencies (1,2). AI is broadly defined as a technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity, and autonomy (3,4). This includes various methods such as expert systems, machine learning (ML), deep learning (DL), and natural language processing (NLP) each contributing uniquely to healthcare applications (5). Increasing interest has been observed in applying large language models (LLMs) as innovative tools within primary care settings, owing to their unique capabilities and diverse applications (6,7).

Primary care is characterized by first-contact access, longitudinal continuity, comprehensive care, and coordination across settings. The American Academy of Family Physicians (AAFP) defines primary care as “the provision of integrated, accessible health care services by physicians and their health care teams who are accountable for addressing a large majority of personal health care needs, developing a sustained partnership with patients, and practicing in the context of family and community” (8). Primary care’s high-volume first-contact care, longitudinal records, and coordination demands make it well-suited for AI-enabled support: diverse undifferentiated presentations benefit from triage/decision support, longitudinal data enable risk prediction and earlier signal detection, and coordination tasks create opportunities for automation to reduce documentation burden and close care gaps. The potential for AI to support primary care has been well-documented. Applications such as ambient documentation systems, risk prediction models, and diagnostic decision support have demonstrated potential to reduce clinician burden and enhance service delivery (9). A systematic review reported improved documentation quality and clinician satisfaction with the use of AI scribes (10). However, translating these innovations into routine primary care practice remains challenging due to numerous implementation barriers.

Rationale and knowledge gap

Barriers to AI implementation in healthcare span technical, organizational, and human domains. He et al. identified seven key challenges of AI implementations: Data sharing, Transparency, Patient safety, Data standardization and integration, Financial issues, Training needs, and Policy and the regulatory environment (11). In primary care specifically, general practitioners have expressed concerns about diagnostic accountability, algorithmic bias, and generalizability of AI tools across diverse patient populations (12). Allied health professionals have cited concerns including lack of AI knowledge and fear of job loss (13). These findings underscore the importance of addressing both system-level and human-centered challenges to AI adoption.

Prior reviews have assessed AI use in primary care. Abbasgholizadeh Rahimi et al. synthesized 90 studies evaluating AI applications in community-based settings, noting that ML, NLP, and expert systems were the most common modalities, primarily focused on diagnosis or surveillance (14). However, their review emphasized applications rather than barriers.

Emerging tools like generative AI and LLMs are now being piloted in clinical contexts, such as automated triage and note generation, raising both opportunities and new risks related to accuracy, bias, and responsible use (15). Given the diversity of study designs, stakeholder perspectives, and conceptual frameworks involved, a scoping review was deemed appropriate to comprehensively map and categorize the range of barriers to AI implementation in primary care.

Many scoping reviews have explored barriers to AI implementation in healthcare (16-18). In our review, we aimed to build on this work by not only mapping these barriers using an established theoretical framework—the Unified Theory of Acceptance and Use of Technology (UTAUT) (19), but also by examining how they manifest across different phases of AI implementation. By categorizing the literature into pre-implementation, implementation, and post-implementation stages, we found that the types and frequencies of reported barriers varied significantly depending on the phase. This phase-specific analysis offers deeper insight into the nature of implementation challenges and provides an in-depth understanding of where targeted interventions or policy adjustments may be most effective.

Objectives

This scoping review aims to systematically examine the barriers to AI implementation in primary care settings by trying to answer (I) What types of AI methods have been evaluated? (II) What categories of barriers are most reported? (III) Which stakeholder groups are represented? (IV) How have contextual factors influenced adoption? and (V) What types of primary care settings were studied? During our analysis, we also identified and analyzed the phase of implementation described in each study, offering an additional perspective on how specific barriers may evolve across the AI lifecycle. This draws attention to not just what barriers are encountered, but also when they are most likely to arise. We present this article in accordance with the PRISMA-ScR reporting checklist (available at https://jphe.amegroups.com/article/view/10.21037/jphe-2025-1-37/rc) (20).

Methods

This scoping review was conducted in accordance with the Joanna Briggs Institute (JBI) methodology for scoping reviews (21). This scoping review protocol was registered on the Open Science Framework (OSF Registration string: JFXPC).

Eligibility criteria

The eligibility criteria for this review were structured using the Population-Concept-Context (PCC) framework recommended by the JBI.

• Population: The review included studies involving stakeholders in primary care, such as healthcare providers (e.g., primary care physicians, nurse practitioners, physician assistants, nurses), health information technology professionals (digital health decision-makers), and patients or caregivers. Studies focusing solely on specialty care, including those focused on radiologists and ophthalmologists, and studies dealing with hospitalized patients were excluded unless they referenced integration with primary care.
• Concept: The central concept was the identification of barriers, challenges, or limitations to the adoption, implementation, or sustained use of AI tools in primary care. Eligible studies were required to report on at least one implementation-related barrier, including technical, ethical, legal, infrastructural, organizational, or human factors. Studies that focused exclusively on AI algorithm development or model performance without addressing implementation barriers were excluded. The included studies were then assessed to note if the barriers were implied as a hypothetical concept if empirically evaluated.
• Context: Studies were included if conducted in primary care settings, defined according to the AAFP as the provision of integrated, accessible health care services that address a large majority of personal health care needs, develop sustained partnerships with patients, and are practiced in the context of family and community (8). Both in-person and virtual care settings were eligible. Studies situated exclusively in emergency or specialty-only settings were excluded. Those studies that dealt with the broader healthcare context were not excluded if primary care settings were part of their evaluation or scope.
• Types of sources: Eligible sources included peer-reviewed journal articles reporting primary research (qualitative, quantitative, and mixed-methods studies), systematic and scoping reviews, and grey literature (e.g., conference abstracts). Only studies published in English from January 1, 2020 onward were included to capture the contemporary use and evaluation of AI tools in primary care, particularly in the context of post COVID-19 digital transformation. Study or trial protocols with incomplete or partial results were excluded. This period was chosen because of the rapid increase in the perceived importance of AI tools at a time when alternative care delivery models were needed.

Search strategy

The search strategy followed the three-step process recommended by the JBI for scoping reviews. First, an initial limited search of PubMed was conducted to identify relevant keywords and Medical Subject Headings (MeSH) related to artificial intelligence, primary care, and barriers to implementation. K.S.A. and S.B. utilized the MeSH on Demand tool to identify additional controlled vocabulary terms within the PCC framework. Insights from this initial search informed the development of a comprehensive search strategy.

Second, this strategy was adapted and executed across multiple electronic databases, including PubMed, IEEE Xplore and Google Scholar, covering publications from January 1, 2020, to the date of the search. A publication date filter was applied to ensure the relevance of findings to future research. With increasing interest in alternative and digitally enabled primary care delivery models following 2020, this restriction was considered warranted to reflect the evolving implementation landscape of AI in primary care. All searches were conducted between May 17, 2025, and June 11, 2025. Only studies published in English were included. No geographic limitations were applied.

Third, citation chasing was performed. The reference lists of full-text articles that met inclusion criteria were screened using Litmaps (22,23) to identify additional relevant sources (backward citation tracking), and forward citation tracking was performed to identify newer studies that cited included articles. This step was completed on July 13, 2025.

The complete search strings used for each database are provided in supplementary tables in Appendix 1.

Study selection

All identified citations were imported into Zotero for deduplication and reference management. Following deduplication, title and abstract screening was conducted using ASReview, an open-source ML-based tool designed to streamline evidence selection through the principle of active learning. In our workflow, all relevance decisions (relevant vs. irrelevant) were made by the first author, K.S.A.; ASReview used these author-provided labels to iteratively train a supervised model and continuously re-rank the remaining records from most to least likely relevant. Specifically, the tool utilizes a combination of Term Frequency-Inverse Document Frequency (TFIDF) as the feature extractor, and a support vector machine (SVM) classifier to distinguish relevant from irrelevant studies. We applied a pre-specified heuristic stopping rule (h50), whereby screening was stopped after 50 consecutively presented records were labeled irrelevant. The sequential recall metrics of relevant vs. irrelevant records can be found in Figures S1,S2. The tool continually refines predictions as the reviewer labels relevant and irrelevant studies. This approach has demonstrated efficiency in several recent evidence synthesis projects and is increasingly being adopted in health research for systematic and scoping reviews (24-26).

Following the ASReview-assisted screening, studies deemed potentially relevant were exported to Rayyan and Zotero for full-text retrieval and review. The full texts were assessed independently by K.S.A., S.B., H.G., and L.K. using pre-established inclusion criteria. Any disagreements were resolved through discussion. Reasons for exclusion at the full-text stage were documented, and the overall study selection process is illustrated using a PRISMA-ScR flow diagram.

Data extraction

Data were extracted from all included sources of evidence using a structured data charting form developed by K.S.A. and S.B. in accordance with the review objectives and inclusion criteria. Using a deductive approach, codes were developed through discussion between the authors; these were iteratively changed after piloting on various study types. The codes for barriers were chosen to reflect a broad range of terms in alignment with the domains of the UTAUT framework. UTAUT is a theory of technology acceptance that explains intention to use and use behavior through four core determinants: performance expectancy, effort expectancy, social influence, and facilitating conditions. In this review, we used UTAUT as a sensitizing framework to organize barrier codes during charting; however, because core UTAUT constructs do not explicitly capture prominent AI-specific barriers such as trust/explainability and related ethical-legal considerations, we applied a modified framework that additionally included the domains “Attitude Toward Using” and “Ethical and Legal Concerns” (27). Following this process, a code form was created and distributed among the authors for comments and revisions. The form was pilot tested on a subset of 10 studies to ensure clarity, consistency, and comprehensiveness and was iteratively refined based on feedback. The following variables were charted from each source of evidence:

• Bibliographic information: Author(s), year of publication, and country/region of origin or country/region of study focus.
• Study characteristics: the type of evidence source, such as qualitative study, quantitative observational or interventional study, review/evidence synthesis (systematic reviews, scoping reviews, and policy statements), commentary or opinion piece (letters to the editor).
• Primary care context: Description of the care setting (e.g., in-person clinic, community health center, telehealth, home-based, remote/rural settings).
• AI application: Type of AI method discussed [e.g., ML, expert systems, NLP, or retrieval-augmented generation (RAG)], and the intended use or function of the AI tool (e.g., diagnosis, triage, clinical decision support, administrative optimization).
• Study participant type: The study participants were coded to reflect a broad range of stakeholder groups, including primary care physicians, nurses, patients, caregivers, researchers, digital health decision-makers, and other stakeholders with an interest in AI. These categories were used to identify and extract information on which stakeholder groups were represented and engaged in each study.
• Implementation barriers: All reported barriers, challenges, or limitations were recorded from the sources and coded according to the predesigned code form. Barriers were coded as empirical or hypothetical, with empirical barriers referring to those identified through data collection within the studies, and hypothetical barriers referring to those discussed conceptually or proposed without empirical evaluation, such as in editorials, opinion pieces, or speculative commentary.
• Phase of implementation: Studies were accordingly classified into one of three categories: pre-implementation, which included studies reporting on anticipated or perceived barriers prior to deployment or use of a tool; implementation, comprising studies that examined real-time challenges encountered during the initial rollout of an AI tool; and post-implementation, representing studies that analyzed barriers in the context of AI tools already in clinical use. Studies that retrospectively analyzed the implementation of a tool were also coded in this category. This categorization was applied only to studies that provided empirical evidence of either qualitative, quantitative, or mixed methods on implementation barriers. Studies that discussed potential challenges without empirical evaluation (e.g., opinion pieces or theoretical commentaries) were excluded from this phase-based analysis. This operational framework was incorporated into the data extraction process and used to enhance the thematic synthesis of implementation challenges. Based on this, a total of 88 studies among the included studies were evaluated for phase of implementation.

AI-assisted data extraction using AIDE

To streamline and standardize the data extraction process, we employed a novel tool called AIDE (AI-Assisted Data Extraction)—an open-source, graphical user interface (GUI) package built in R that integrates LLM-based extraction with mandatory human oversight.

AIDE was developed by Schroeder et al. (28) to operationalize a human-in-the-loop (HIL) framework for systematic review data extraction, ensuring that LLM-assisted extraction is accountable and accurate. To the best of our knowledge, our scoping review represents the first documented instance of this tool being applied in a scoping review context.

AIDE architecture and functionality

AIDE allows users to connect to multiple freely available LLMs (e.g., Google Gemini, Mistral, or even local models such as Ollama) through API keys. The interface is designed to be accessible without advanced programming knowledge.

The program reads a structured Excel-based ‘code form’, in which each column header represents a specific extraction prompt. Both the code form and the corresponding article PDFs are uploaded to AIDE. When the user initiates analysis, the R application sends a structured request to the chosen LLM’s API—transmitting the PDF content, the global instruction, and the prompts from the code form in a single request. The interface displays the PDF document on one side and the model’s generated responses on the other, enabling direct, point-by-point comparison. Each model response can be traced to its textual source within the PDF via the “Source” function, which highlights the evidence used to generate the answer.

AIDE’s design enforces human verification before data is finalized. The “Record” button must be manually clicked by the user to accept an LLM-generated response; only then is the entry written back to the locally stored code form in R. The LLM does not autonomously finalize any extracted fields. Users may edit or correct any field before recording, and no data are saved automatically, allowing us to verify and correct the responses to each prompt. This was particularly relevant for the variable ‘phase of implementation’, where classification was performed manually by the authors based on full-text assessment; LLM-generated suggestions were reviewed but were largely inconsistent with our operational framework. This structure ensures that all extracted variables undergo explicit human approval—embodying a robust HIL approach that preserves human judgment while leveraging the efficiency of machine extraction. The full set of prompts used to assist the extraction process is provided in Appendix 1.

The initial data extraction form included both closed-ended fields and open-text entries, including an “Other” option for categories that did not fit predefined codes. Responses in these open-text fields were reviewed inductively and mapped to expand the coding scheme. Following data extraction, all entries were compiled and cleaned using Python 3.13.5, with the support of relevant packages for data wrangling, validation, and transformation. The analysis focused on mapping the frequency and distribution of reported barriers, stakeholder groups, AI methods, and implementation contexts across studies. LLMs (Gemini 2.5 Pro and Claude Sonnet 4) were used to support code generation and troubleshoot technical issues during the data processing and cleaning stages. Descriptive statistics were calculated, and visualizations—including heatmaps, pie charts, and bar graphs—were created using Python to summarize patterns in the extracted data. A PRISMA 2020 compliant flow diagram was compiled at the end using the web-based Shiny app and its associated open-source R package (29).

Results

A total of 15,362 records were identified through database searches, including 3,064 from PubMed, 5,930 from Google Scholar, and 6,368 from IEEE Xplore. Of these, 4,064 records were exported for screening (database-specific export details are provided in Appendix 1). After removing 547 duplicates, 3,517 records remained for title and abstract screening using ASReview. Following screening, 166 full-text reports were sought for retrieval. Of these, three reports could not be retrieved. The remaining 163 full-text articles were assessed for eligibility. After full-text review, 29 reports were excluded for reasons such as tool not based on ML or prediction model (n=10), studies unrelated to primary care settings (n=7), non-English language (n=3), and other reasons. In addition, seven records were identified through citation mapping; all seven were retrieved and met eligibility criteria. The full selection process is detailed in Figure 1.

Figure 1 PRISMA flow diagram.

Summary of review findings

A descriptive synthesis of the key study variables is presented in Table 1. A total of 141 studies (See Table 2) were included in this scoping review, reflecting a diverse range of evidence types, AI methodologies, stakeholder perspectives, and primary care contexts post-2020.

Types of evidence

Qualitative studies were the most prevalent type of evidence source, constituting 36.8% of the included literature. This was followed by mixed-methods studies, which accounted for 20.5% of the total. Review/evidence synthesis articles made up 17% of the studies, while quantitative (observational) studies represented 15.6%. Opinion pieces comprised a smaller proportion at 6.3%, with quantitative (interventional) studies being the least common, at 3.5%.

AI methods used

A total of 151 instances of AI-based tools were noted. The most mentioned AI methods were Clinical Decision Support Systems (43 times, 28.4%), ML (11 times, 7.2%), DL (6 times, 3.9%), and NLP (5 times, 3.3%). A considerable number of studies in our review did not evaluate a particular AI-based tool or failed to specify the details of the tools used. These instances were coded as ‘Model not specified’, accounting for 81 or 53.6% of studies.

Countries of origin

The publication data reveal a significant concentration from high-income countries, with the United Kingdom and the United States representing the largest shares at 19.1% each, followed by Canada (7.8%), Australia (5.6%), and Germany (4.9%). In contrast, publications from low- and middle-income countries (LMICs), as defined by the World Bank (https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups), collectively account for approximately 11.8% of the total. Among them, the most frequent contributors include China (2.1%), India (2.1%), South Africa (2.1%), and Pakistan (1.4%).

Objectives of AI implementation

The analysis of objectives for AI implementation reveals a total of 178 mentions. The most frequently observed objectives were Diagnosis (63 times, 35.3%), Quality improvement (45 times, 25.2%), Administrative automation (27 times, 15.1%), Risk prediction (24 times, 13.4%), and History and physical examination (10 times, 5.6%). Approximately 4.4% of studies did not specify a particular intended objective of implementation.

Stakeholder representation

Our scoping review recorded a total of 396 mentions of several types of participants. The most frequently observed participant types are: Primary Care Physician (119 times, 30%), Patients (63 times, 15.9%), Researchers (55 times, 13.8%), Nurses (53 times, 13.3%), and stakeholders with AI Interest (49 times, 12.3%).

Primary care settings

In-person clinics were the most represented (34.3%), followed by community health centers (13.2%), rural/remote care (9.0%), and hybrid or telemedicine settings (6.6% combined). No setting was specified in 54 studies (32.5%), most of which were of the Review/Evidence synthesis type.

Key barriers identified

Trust or Explainability Issues were the most frequently cited barrier, appearing in 94 studies (see Table 3). This was followed by Workflow Disruption, mentioned in 72 studies, and Data Quality Issues, present in 63 studies. Lack of Integration was identified in 55 studies, while Regulatory or Legal Factors were reported in 51 studies and excess Training Needs were reported in 58 studies. Less commonly cited barriers included Low Usability (42 studies), Equity and Bias (30), and Patient Acceptability (29). While Cost was explicitly identified in a smaller number of studies, the broader domain, Facilitating Conditions, was among the most frequently identified barrier domains in the reviewed studies. Temporal analysis demonstrated changing barrier emphasis over time. Trust/explainability concerns increased substantially after 2022, while workflow disruption, data quality, and integration barriers also rose steadily over the study period (Figure 2).

Figure 2 Temporal trend of key barriers across the years. AI, artificial intelligence.

The correlation matrix of coded barriers reveals several relationships that suggest thematic clustering of challenges encountered during AI implementation in primary care (see Figure 3). The strongest positive correlation was observed between Lack of Integration and Low Usability (r=0.34), indicating that studies identifying poor system integration often also reported user interface or workflow compatibility issues. Trust or Explainability Issues were moderately correlated with both Regulatory/Legal Factors (r=0.26) and Equity & Bias (r=0.23), suggesting that concerns about transparency often co-occurred with ethical and legal apprehensions. In contrast, Workflow Disruption exhibited a modest negative correlation with both Regulatory/Legal Factors (r=−0.37) and Trust or Explainability Issues (r=−0.26), indicating that studies focused on technical workflow barriers rarely emphasized broader normative concerns.

Figure 3 Correlation matrix.

Domain mapping

Each barrier code was mapped to one or more conceptual domains, as outlined in Table 4. To structure the analysis of implementation barriers, we developed a domain-based coding framework informed by the UTAUT framework. While UTAUT served as a conceptual foundation for its constructs of performance expectancy, effort expectancy, and facilitating conditions, we modified the framework to better reflect emerging themes in the AI in healthcare literature. The original UTAUT domain of Social Influence was restructured and distributed across these two new domains: Attitude Toward Using, capturing general user receptiveness or resistance to AI tools through exploring concepts of trust, acceptability and peer influence, and ‘Ethical and Legal Concerns’, encompassing issues such as privacy, bias, and regulatory challenges (see Table 4). Barriers identified across studies were deductively mapped to these five domains based on explicit descriptions or contextually relevant cues in the texts.

Domain-specific findings

Based on our modified domain framework, five key categories of barriers were identified and analysed across the included studies. Attitude Toward Using emerged as the most frequently mapped domain, appearing in 73.76% of studies (see Table 4). This reflects the crucial importance of user acceptance—particularly among clinicians and patients—as a key theme in the literature. Barriers mapped to the Performance Expectancy domain were noted in 69.50% of studies, suggesting that concerns around the clinical utility, reliability, and added value of AI systems in clinical settings are important considerations to most stakeholders. In 60.99% of studies, Effort Expectancy, a domain that addresses concerns regarding ease of use and workflow integration, was noted. Facilitating Conditions, reported in 50.35% of studies, highlighted the critical role of infrastructure, technical support, and organizational readiness in enabling successful implementation. Lastly, Ethical and Legal Concerns were identified in 44.68% of studies, reflecting a growing awareness of challenges related to data privacy, equity, and regulatory compliance—issues that are particularly salient in early-phase planning and system design. Together, these findings emphasize that effective AI integration in primary care depends not only on technical performance but also on alignment with user needs, organizational capacity, and ethical standards.

Phases of implementation

The pre-implementation phase included studies that explored physician perspectives or potential barriers prior to the deployment of an AI tool. Among the empirically evaluated studies, 58.0% (n=51) were conducted in the pre-implementation phase, 30.7% (n=27) in the implementation phase, and 11.4% (n=10) in the post-implementation phase.

Understanding patterns across phases

In the pre-implementation phase, Attitude towards Using appears to be most prominent, suggesting that issues of Patient Acceptability and Trust or Explainability Issues are the key concerns in planning the implementation of a tool (see Figure 4). In contrast, the proportion of studies that report barriers in the same domain dropped to 51.85% during implementation phase studies. The most frequently reported barriers in the post-implementation studies were workflow disruption (9 mentions), trust or explainability (7 mentions) and training needs (5 mentions). Effort Expectancy—reflecting usability and workflow fit—peaked during implementation (92.59%) and remained high post-implementation (90.0%), suggesting that practical challenges emerge more prominently once tools are in use. Performance Expectancy remained a consistent concern across phases, while Facilitating Conditions were most emphasized pre-implementation (58.82%).

Figure 4 Domain distribution across phases of artificial intelligence implementation.

Discussion

Key findings and explanations

A notable finding from this scoping review was the predominance of qualitative studies (37%) and mixed-methods research (21%), with a smaller proportion of quantitative observational studies and only 4% employing interventional designs. This distribution suggests a significant emphasis on stakeholder perspectives, perceived barriers, and conceptual frameworks. This observation is also supported by a rapid increase in the proportion of studies in the pre-implementation phase across the years (170). This pattern in the type of research may stem from a shifting concern towards ethical and regulatory aspects of AI implementation. Of note, one of the key barriers described in the studies included in our review was ‘Trust or Explainability issues’. This finding is consistent with other scoping reviews and systematic reviews that identified Ethical concerns or Trust as one of the key barriers to implementation of AI in the broader healthcare context (44,46,171).

A key limitation observed in the literature was the lack of specificity in reporting AI methods, with 54% of studies failing to describe the underlying algorithm or provide specific model details. Without clear delineation of the AI approach—whether rule-based, ML, or DL—stakeholders are left with limited insight into the appropriateness, scalability, and limitations of the systems under study. Concerns have been raised by previous assessments of AI reporting standards, such as the CONSORT-AI standard, regarding the precise reporting of AI methods utilized in clinical trials. They recommend that specific terminology relating to the type of model should be detailed in the abstract (172). Among the studies that did specify an AI method, the most common were clinical decision support systems (CDSS) (28%), followed by ML and DL algorithms. This suggests that while traditional or rule-based AI systems have seen some integration into primary care contexts, newer and more dynamic technologies have yet to gain significant traction in either research or implementation. The limited presence of LLMs and chatbots is particularly notable given their rapid proliferation in parallel domains such as radiology and ophthalmology (173,174). The most frequently reported objective for AI implementation in primary care was Diagnostic purposes (35%), followed by Quality Improvement and Administrative automation, 25% and 15%, respectively. These trends reflect a growing interest in leveraging AI to enhance clinical accuracy, care coordination, and workflow efficiency—all of which align with the broader goal of achieving the quintuple aim of healthcare: improving population health, enhancing patient experience, reducing costs, and supporting clinician well-being and health equity through improved accessibility (175). To enhance the practical applicability of these findings, we synthesized strategies reported in the included studies to address common implementation barriers (Table 5). Across studies, trust/explainability concerns were most frequently addressed through improving model transparency, local validation prior to deployment, staged autonomy with clinician oversight, and routine feedback on performance. Workflow disruption was commonly mitigated through workflow- and EHR-integrated design, co-design with end-users, minimizing additional documentation burden, and change-management support. Training needs were addressed through structured AI literacy initiatives, incorporation into curricula, and ongoing point-of-care training.

Underexplored care settings

In-person clinics were the most represented care setting, while a substantial proportion of studies (33%) were not mapped to a particular setting at all. One explanation for this is that 17% of the studies in our review were themselves scoping, literature, or systematic reviews, which often aggregated findings from diverse or unspecified primary care settings, thereby diluting context-specific detail. Alternative delivery models, including telehealth, and virtual or hybrid modalities, were markedly underrepresented (collectively <7%).

Comparison with similar research

Emerging literature continues to document the expansion and normalization of telehealth in primary care. Bhatia et al. (176) and Shachar et al. (177) discuss the rapid integration of virtual modalities in response to shifting patient expectations and logistical constraints. Future research should aim to target underrepresented settings and stakeholder groups, particularly as healthcare delivery continues to evolve beyond the walls of traditional clinics. Without such inclusion, AI development risks being anchored to outdated assumptions about where and how primary care is delivered.

Implications and actions needed

Our analysis of co-occurring barriers reveals important insights into how challenges are reported in primary care AI literature. While most barriers were discussed in relative isolation, a few moderate correlations might suggest the presence of thematic clusters. The strongest positive correlation was observed between Lack of Integration and Low Usability (r=0.34), indicating that studies citing poor system integration often simultaneously reported challenges related to user interface design or workflow compatibility.

Although most barrier relationships were weak or near-zero, moderate-level associations warrant comparative analysis which could reveal deeper interdependencies among technical, ethical, and organizational barriers. Such work is beyond the scope of this review but could inform more targeted interventions that address multiple implementation challenges simultaneously rather than in isolation.

Ethical & Legal Concerns are prominent in pre-implementation but do not feature in the post-implementation phase. Usability (effort expectancy) becomes a central barrier once AI tools are actively used. Performance expectancy becomes more important in the implementation phase. Effort expectancy remains the most important barrier domain in the implementation and post-implementation phases. The predominance of ethical concerns in the planning stage may reflect a growing trend among physicians and end-users to ask one of the most vital questions regarding AI: “Should AI be used for this purpose?” These trends collectively highlight the value of a phase-specific approach to barrier mitigation and suggest that implementation strategies must evolve in tandem with the AI maturity curve—from pre-deployment planning to post-deployment optimization. Future research should explore these dynamics using longitudinal, mixed methods designs that track how different barrier domains emerge, recede, or transform over time. Moreover, there is a need to move beyond cataloguing barriers toward designing, testing, and scaling interventions that target co-occurring or temporally clustered challenges in real-world primary care environments.

Strengths & limitations

A key strength of our review is the systematic identification of barriers and their subsequent comprehensive mapping to specific domains of a framework. By categorizing barriers within a recognized framework, our findings offer practical utility for stakeholders aiming to design more effective implementation strategies. By focusing on papers published from 2020 to mid-2025, this review captures recent developments, emerging challenges, and contemporary perspectives on AI implementation in primary care. However, as a scoping review, our methodology is designed to map the existing literature and identify the breadth and nature of evidence, rather than to conduct a critical appraisal or risk of bias assessment of individual studies. Therefore, we cannot comment on the methodological quality or strength of evidence of the included studies. The findings should be interpreted as a synthesis of reported barriers in the literature, not as definitive evidence of their prevalence or impact based on rigorous evaluation. Our review identified fewer studies originating from LMICs; while this may be a trend among the publications in this field, future research should prioritize evidence from these countries. A pragmatic deductive framework for coding barriers may have limited our review’s ability to capture subtle nuances of various forms of representation of potential challenges faced. In addition, our review included studies published in English language only, which limits its ability to capture publications reporting differing opinions and themes in other languages.

Conclusions

The use of AI has increased significantly in recent years. Many aspects of healthcare have been affected or influenced in several ways. Like most technological advances, adopting AI in healthcare has faced notable challenges. A common theme in studies evaluating AI is that trust in these systems is especially important for successful adoption. Along with issues like data quality and concerns about equity or bias, these factors are more significant for AI systems than for other technological developments in the past. Our review found that most publications focused on the pre-implementation phase, which evaluates physicians’ and stakeholders’ perspectives. Given the phase-specific approach to assessment, this work can help future researchers and stakeholders identify potential barriers they might face, depending on the stage of implementation they plan to examine. We believe that this approach can contribute to a future where AI is implemented in primary care in an equitable, ethical, and evidence-based manner.

Acknowledgments

Artificial Intelligence Tools including ChatGPT (OpenAI, GPT-4o & o3), Claude (Anthropic, Sonnet 3.7), Gemini (Google, 2.5 pro), ASReview (2.0), and AIDE (AI-Assisted Data Extraction) were used in the following stages:

Conceptualization: ChatGPT was used to generate ideas for visualizing the extracted data (e.g., correlation matrices, domain frequencies), and to support the interpretation of implementation phase trends across studies.

Data Analysis: Gemini and Claude were used to develop Python code for data preprocessing and creating visualizations. Claude was used to troubleshoot Python-related errors and adjust formatting for figures and visualizations.

Following deduplication, title and abstract screening was performed using ASReview, a machine learning-based active learning tool. Data extraction was operationalized through AIDE (AI-Assisted Data Extraction)—a novel human-in-the-loop tool integrating large language model (LLM) capabilities with mandatory human verification.

Writing—Review & Editing: ChatGPT was used to revise sections of the manuscript, including the Discussion, for clarity, academic tone, and flow.

No identifiable data or unpublished results from individual studies were shared with any AI tools. All data used with these tools was publicly accessible. Only web-based tools (chat.openai.com and Claude via Poe.com) were used, and no institutional or secure datasets were uploaded.

Footnote

Reporting Checklist: The authors have completed the PRISMA-ScR reporting checklist. Available at https://jphe.amegroups.com/article/view/10.21037/jphe-2025-1-37/rc

Peer Review File: Available at https://jphe.amegroups.com/article/view/10.21037/jphe-2025-1-37/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jphe.amegroups.com/article/view/10.21037/jphe-2025-1-37/coif). V.O.K. serves as an unpaid editorial board member of Journal of Public Health and Emergency from April 2025 to March 2027. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Gilfoyle M, Bosworth KT, Adesanya TMA, et al. Exploring Artificial Intelligence and the future of primary care. Ann Fam Med 2024;22:174-5.
2. Jiang F, Jiang Y, Zhi H, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol 2017;2:230-43. [Crossref] [PubMed]
3. IBM. What is artificial intelligence (AI)? IBM; 2024. Available online: https://www.ibm.com/topics/artificial-intelligence
4. Dahlke J. A.I. go by many names: towards a sociotechnical definition of artificial intelligence. arXiv preprint; arXiv: 2410.13452.
5. Saibene A, Assale M, Giltri M. Expert systems: Definitions, advantages and issues in medical field applications. Expert Syst Appl 2021;177:114900.
6. Gencer G, Gencer K. Large Language Models in Healthcare: A Bibliometric Analysis and Examination of Research Trends. J Multidiscip Healthc 2025;18:223-38. [Crossref] [PubMed]
7. Andrew A. Potential applications and implications of large language models in primary care. Fam Med Community Health 2024;12:e002602. [Crossref] [PubMed]
8. American Academy of Family Physicians. Primary Care [Internet]. Leawood (KS): American Academy of Family Physicians; [cited 2025 Jul 1]. Available online: https://www.aafp.org/about/policies/all/primary-care.html
9. Keng C, DiGiorgio A, Ehrenfeld JM, et al. Unburdening Patients and Clinicians Through Automation and Artificial Intelligence: Informatics Strategies for Reducing Administrative Burden. J Med Syst 2025;49:128. [Crossref] [PubMed]
10. Hassan H, Zipursky AR, Rabbani N, et al. Clinical Implementation of Artificial Intelligence Scribes in Health Care: A Systematic Review. Appl Clin Inform 2025;16:1121-35. [Crossref] [PubMed]
11. He J, Baxter SL, Xu J, et al. The practical implementation of artificial intelligence technologies in medicine. Nat Med 2019;25:30-6. [Crossref] [PubMed]
12. Blease C, Kaptchuk TJ, Bernstein MH, et al. Artificial Intelligence and the Future of Primary Care: Exploratory Qualitative Study of UK General Practitioners' Views. J Med Internet Res 2019;21:e12802. [Crossref] [PubMed]
13. Abdulazeem HM, Meckawy R, Schwarz S, et al. Knowledge, attitude, and practice of primary care physicians toward clinical AI-assisted digital health technologies: Systematic review and meta-analysis. Int J Med Inform 2025;201:105945. [Crossref] [PubMed]
14. Abbasgholizadeh Rahimi S, Légaré F, Sharma G, et al. Application of Artificial Intelligence in Community-Based Primary Health Care: Systematic Scoping Review and Critical Appraisal. J Med Internet Res 2021;23:e29839. [Crossref] [PubMed]
15. Reddy S. Generative AI in healthcare: an implementation science informed translational path on application, integration and governance. Implement Sci 2024;19:27. [Crossref] [PubMed]
16. Hassan M, Kushniruk A, Borycki E. Barriers to and Facilitators of Artificial Intelligence Adoption in Health Care: Scoping Review. JMIR Hum Factors 2024;11:e48633. [Crossref] [PubMed]
17. Khattak M, Bowness JS, Yonis R, et al. Navigating the barriers and facilitators to implementation of AI in healthcare: a scoping review. Bone Joint J 2025;107-B:666-72. [Crossref] [PubMed]
18. Chomutare T, Tejedor M, Svenning TO, et al. Artificial Intelligence Implementation in Healthcare: A Theory-Based Scoping Review of Barriers and Facilitators. Int J Environ Res Public Health 2022;19:16359. [Crossref] [PubMed]
19. Marikyan M, Papagiannidis P. Unified theory of acceptance and use of technology. TheoryHub book, 2021.
20. Tricco AC, Lillie E, Zarin W, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med 2018;169:467-73. [Crossref] [PubMed]
21. Peters MDJ, Marnie C, Tricco AC, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evid Synth 2020;18:2119-26. [Crossref] [PubMed]
22. Gong J, Maitra D. WIP: Leveraging AI for Literature Reviews: A Guide for New Researchers. 2025 ASEE Annual Conference & Exposition.
23. Jones J, Armstrong R, Kowatch A, et al. 514 Navigating the Research Landscape: A Study of AI Search Tools with Literature Mapping Capabilities. Ann Emerg Med 2025;86:S219.
24. van de Schoot R, de Bruin J, Schram R, et al. An open source machine learning framework for efficient and transparent systematic reviews. Nat Mach Intell 2021;3:125-33.
25. Oude Wolcherink MJ, Pouwels XGLV, van Dijk SHB, et al. Can artificial intelligence separate the wheat from the chaff in systematic reviews of health economic articles? Expert Rev Pharmacoecon Outcomes Res 2023;23:1049-56. [Crossref] [PubMed]
26. Oami T, Okada Y, Sakuraya M, et al. Efficiency and Workload Reduction of Semi-automated Citation Screening Software for Creating Clinical Practice Guidelines: A Prospective Observational Study. J Epidemiol 2024;34:380-6. [Crossref] [PubMed]
27. Bile Hassan I, Murad MAA, El-Shekeil I, et al. Extending the UTAUT2 Model with a Privacy Calculus Model to Enhance the Adoption of a Health Information Application in Malaysia. Informatics 2022;9:31.
28. Schroeder NL, Jaldi CD, Zhang S. Large language models with human-in-the-loop validation for systematic review data extraction. arXiv preprint; arXiv: 2501.11840.
29. Haddaway NR, Page MJ, Pritchard CC, et al. PRISMA2020: An R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and Open Synthesis. Campbell Syst Rev 2022;18:e1230. [Crossref] [PubMed]
30. Ahluwalia VS, Schapira MM, Weissman GE, et al. Primary Care Provider Preferences Regarding Artificial Intelligence in Point-of-Care Cancer Screening. MDM Policy Pract 2025;10:23814683251329007. [Crossref] [PubMed]
31. Al-Bikkalli Z, Douven R, Wallenburg I, et al. The Evolving Landscape of Primary Healthcare: Exploring AI Readiness of the Dutch General Practitioners. Stud Health Technol Inform 2025;323:31-5. [Crossref] [PubMed]
32. Al-Qudimat AR, Alqudimat MR, Singh K, et al. Perception and Knowledge of Hospital Workers Toward Using Artificial Intelligence: A Descriptive Study. Health Sci Rep 2025;8:e70623. [Crossref] [PubMed]
33. Alanzi T, Alotaibi R, Alajmi R, et al. Barriers and Facilitators of Artificial Intelligence in Family Medicine: An Empirical Study With Physicians in Saudi Arabia. Cureus 2023;15:e49419. [Crossref] [PubMed]
34. Alasiri SF, Douiri A, Altukistani S, et al. The Role of Clinical Decision Support Systems in Preventing Stroke in Primary Care: A Systematic Review. Perspect Health Inf Manag 2023;20:1d.
35. Allen MR, Webb S, Mandvi A, et al. Navigating the doctor-patient-AI relationship - a mixed-methods study of physician attitudes toward artificial intelligence in primary care. BMC Prim Care 2024;25:42. [Crossref] [PubMed]
36. Alsudairy N, Alahdal A, Alrashidi M, et al. Artificial Intelligence in Primary Care Decision-Making: Survey of Healthcare Professionals in Saudi Arabia. Cureus 2025;17:e81960. [Crossref] [PubMed]
37. Armitage R. Using AI to improve skin cancer detection in primary care: the vision and barriers. Br J Gen Pract 2025;75:168-9. [Crossref] [PubMed]
38. Armitage RC. How do GPs Want Large Language Models to be Applied in Primary Care, and What Are Their Concerns? A Cross-Sectional Survey. J Eval Clin Pract 2025;31:e70129.
39. Ash JS, Chase D, Baron S, et al. Clinical Decision Support for Worker Health: A Five-Site Qualitative Needs Assessment in Primary Care Settings. Appl Clin Inform 2020;11:635-43. [Crossref] [PubMed]
40. Beccia C, Hunter B, Manski-Nankervis JA, et al. Exploring the User Acceptability and Feasibility of a Clinical Decision Support Tool Designed to Facilitate Timely Diagnosis of New-Onset Type 1 Diabetes in Children: Qualitative Interview Study Among General Practitioners. JMIR Form Res 2024;8:e60411. [Crossref] [PubMed]
41. Bentley KH, Zuromski KL, Fortgang RG, et al. Implementing Machine Learning Models for Suicide Risk Prediction in Clinical Practice: Focus Group Study With Hospital Providers. JMIR Form Res 2022;6:e30946. [Crossref] [PubMed]
42. Brancaccio G, Balato A, Malvehy J, et al. Artificial Intelligence in Skin Cancer Diagnosis: A Reality Check. J Invest Dermatol 2024;144:492-9. [Crossref] [PubMed]
43. Canny A, Donaghy E, Murray V, et al. Patient views on asthma diagnosis and how a clinical decision support system could help: A qualitative study. Health Expect 2023;26:307-17. [Crossref] [PubMed]
44. Chew HSJ, Achananuparp P. Perceptions and Needs of Artificial Intelligence in Health Care to Increase Adoption: Scoping Review. J Med Internet Res 2022;24:e32939. [Crossref] [PubMed]
45. d’EliaAGabbayMFrithLArtificial intelligence and health equity in primary care: A qualitative study with key stakeholders.medRxiv 2023. doi: .
46. d'Elia A, Gabbay M, Rodgers S, et al. Artificial intelligence and health inequities in primary care: a systematic scoping review and framework. Fam Med Community Health 2022;10:e001670. [Crossref] [PubMed]
47. Daines L, Canny A, Donaghy E, et al. Use and acceptability of an asthma diagnosis clinical decision support system for primary care clinicians: an observational mixed methods study. NPJ Prim Care Respir Med 2024;34:40. [Crossref] [PubMed]
48. Daines L, Donaghy E, Canny A, et al. Clinician views on how clinical decision support systems can help diagnose asthma in primary care: a qualitative study. J Asthma 2024;61:377-85. [Crossref] [PubMed]
49. Darcel K, Upshaw T, Craig-Neil A, et al. Implementing artificial intelligence in Canadian primary care: Barriers and strategies identified through a national deliberative dialogue. PLoS One 2023;18:e0281733. [Crossref] [PubMed]
50. Davis VH, Qiang JR, Adekoya MacCarthy I, et al. Perspectives on Using Artificial Intelligence to Derive Social Determinants of Health Data From Medical Records in Canada: Large Multijurisdictional Qualitative Study. J Med Internet Res 2025;27:e52244. [Crossref] [PubMed]
51. De Raeve P, Davidson PM, Shaffer FA, et al. Leveraging the trust of nurses to advance a digital agenda in Europe: a critical review of health policy literature. Open Res Eur 2021;1:26. [Crossref] [PubMed]
52. Dey AK. Navigating the Artificial Intelligence Revolution: The Future of General Practice in India. J Assoc Physicians India 2025;73:84-5. [Crossref] [PubMed]
53. Elgin CY, Elgin C. Ethical implications of AI-driven clinical decision support systems on healthcare resource allocation: a qualitative study of healthcare professionals' perspectives. BMC Med Ethics 2024;25:148. [Crossref] [PubMed]
54. Escalé-Besa A, Vidal-Alaball J, Miró Catalina Q, et al. The Use of Artificial Intelligence for Skin Disease Diagnosis in Primary Care Settings: A Systematic Review. Healthcare (Basel) 2024;12:1192. [Crossref] [PubMed]
55. Evans R, Bryant L, Russell G, et al. Clinical Trust in Data-Driven Decision Support Tools: Qualitative Interview Study. Stud Health Technol Inform 2025;327:410-1. [Crossref] [PubMed]
56. Ford E, Edelman N, Somers L, et al. Barriers and facilitators to the adoption of electronic clinical decision support systems: a qualitative interview study with UK general practitioners. BMC Med Inform Decis Mak 2021;21:193. [Crossref] [PubMed]
57. Ganapathi S, Duggal S. Exploring the experiences and views of doctors working with Artificial Intelligence in English healthcare; a qualitative study. PLoS One 2023;18:e0282415. [Crossref] [PubMed]
58. Garies S, Liang S, Weyman K, et al. Artificial intelligence in primary care practice: Qualitative study to understand perspectives on using AI to derive patient social data. Can Fam Physician 2024;70:e102-9. [Crossref] [PubMed]
59. Garies S, Liang S, Weyman K, et al. Developing an AI Tool to Derive Social Determinants of Health for Primary Care Patients: Qualitative Findings From a Codesign Workshop. Ann Fam Med 2024;22:317-24. [Crossref] [PubMed]
60. Gezer M, Hunter B, Hocking JS, et al. Informing the design of a digital intervention to support sexually transmissible infection care in general practice: a qualitative study exploring the views of clinicians. Sex Health 2023;20:431-40. [Crossref] [PubMed]
61. Gold R, Middendorf M, Heintzman J, et al. Challenges involved in establishing a web-based clinical decision support tool in community health centers. Healthc (Amst) 2020;8:100488. [Crossref] [PubMed]
62. Goldberg DG, Soylu TG, Hoffman CF, et al. Clinicians’ perspectives on the adoption and implementation of EMR-integrated clinical decision support tools in primary care. Digit Health 2025;11:20552076251334043. [Crossref] [PubMed]
63. Greenhalgh S, Guo L, Yeowell G. 'In the Midst of Every Crisis, Lies Great Opportunity': Perceptions of the Future Use of Artificial Intelligence in the UK NHS Primary Care. Musculoskeletal Care 2025;23:e70092. [Crossref] [PubMed]
64. Hanna K, Chartash D, Liaw W, et al. Family Medicine Must Prepare for Artificial Intelligence. J Am Board Fam Med 2024;37:520-4. [Crossref] [PubMed]
65. Held LA, Wewetzer L, Steinhäuser J. Determinants of the implementation of an artificial intelligence-supported device for the screening of diabetic retinopathy in primary care - a qualitative study. Health Informatics J 2022;28:14604582221112816. [Crossref] [PubMed]
66. Helenason J, Ekström C, Falk M, et al. Exploring the feasibility of an artificial intelligence based clinical decision support system for cutaneous melanoma detection in primary care - a mixed method study. Scand J Prim Health Care 2024;42:51-60. [Crossref] [PubMed]
67. Hendrix N, Hauber B, Lee CI, et al. Artificial intelligence in breast cancer screening: primary care provider preferences. J Am Med Inform Assoc 2021;28:1117-24. [Crossref] [PubMed]
68. Herter WE, Khuc J, Cinà G, et al. Impact of a Machine Learning-Based Decision Support System for Urinary Tract Infections: Prospective Observational Study in 36 Primary Care Practices. JMIR Med Inform 2022;10:e27795. [Crossref] [PubMed]
69. Hong G, Smith M, Lin S. The AI Will See You Now: Feasibility and Acceptability of a Conversational AI Medical Interviewing System. JMIR Form Res 2022;6:e37028. [Crossref] [PubMed]
70. Jensen C, McKerrow NH, Wills G. Acceptability and uptake of an electronic decision-making tool to support the implementation of IMCI in primary healthcare facilities in KwaZulu-Natal, South Africa. Paediatr Int Child Health 2020;40:215-26. [Crossref] [PubMed]
71. Jimenez G, Tyagi S, Osman T, et al. Improving the Primary Care Consultation for Diabetes and Depression Through Digital Medical Interview Assistant Systems: Narrative Review. J Med Internet Res 2020;22:e18109. [Crossref] [PubMed]
72. Jones OT, Matin RN, van der Schaar M, et al. Artificial intelligence and machine learning algorithms for early detection of skin cancer in community and primary care settings: a systematic review. Lancet Digit Health 2022;4:e466-76. [Crossref] [PubMed]
73. Jones OT, Calanzani N, Saji S, et al. Artificial Intelligence Techniques That May Be Applied to Primary Care Data to Facilitate Earlier Diagnosis of Cancer: Systematic Review. J Med Internet Res 2021;23:e23483. [Crossref] [PubMed]
74. Jones OT, Calanzani N, Scott SE, et al. User and Developer Views on Using AI Technologies to Facilitate the Early Detection of Skin Cancers in Primary Care Settings: Qualitative Semistructured Interview Study. JMIR Cancer 2025;11:e60653. [Crossref] [PubMed]
75. Jørgensen NL, Merrild CH, Jensen MB, et al. The Perceptions of Potential Prerequisites for Artificial Intelligence in Danish General Practice: Vignette-Based Interview Study Among General Practitioners. JMIR Med Inform 2025;13:e63895. [Crossref] [PubMed]
76. Kaymakçı V, Kasap İ, Sevindi M, et al. The role and future of artificial intelligence in primary care. Jour Turk Fam Phy 2024;15:26-37.
77. Katsakiori PF, Kagadis GC, Mulita F, et al. Implementing Artificial Intelligence in Family Medicine: Challenges and Limitations. Cureus 2024;16:e75518. [Crossref] [PubMed]
78. Kocaballi AB, Ijaz K, Laranjo L, et al. Envisioning an artificial intelligence documentation assistant for future primary care consultations: A co-design study with general practitioners. J Am Med Inform Assoc 2020;27:1695-704. [Crossref] [PubMed]
79. Koch R, Steffen MT, Wetzel AJ, et al. Exploring Laypersons' Experiences With a Mobile Symptom Checker App as an Interface Between eHealth Literacy, Health Literacy, and Health-Related Behavior: Qualitative Interview Study. JMIR Form Res 2025;9:e60647. [Crossref] [PubMed]
80. Kowalewska E. Physicians and AI in healthcare: insights from a mixed-methods study in Poland on adoption and challenges. Front Digit Health 2025;7:1556921. [Crossref] [PubMed]
81. Krogh M, Hentze M, Jensen MSA, et al. Valuable insights into general practice staff's experiences and perspectives on AI-assisted diabetic retinopathy screening-An interview study. Front Med (Lausanne) 2025;12:1565532. [Crossref] [PubMed]
82. Kueper J, Lizotte D, Brown J, et al. Identifying priorities for artificial intelligence and primary care in ontario: A multi-stakeholder engagement event. Ann Fam Med 2022;20:2943. [Crossref] [PubMed]
83. Kueper JK, Terry A, Bahniwal R, et al. Connecting artificial intelligence and primary care challenges: findings from a multi stakeholder collaborative consultation. BMJ Health Care Inform 2022;29:e100493. [Crossref] [PubMed]
84. Kueper JK, Terry AL, Zwarenstein M, et al. Artificial Intelligence and Primary Care Research: A Scoping Review. Ann Fam Med 2020;18:250-8. [Crossref] [PubMed]
85. Kueper J, Rayner J, Bhatti S, et al. Data-Driven Decision Support Tool Co-Development with a Primary Health Care Practice Based Learning Network. F1000Res 2024;13:336. [Crossref] [PubMed]
86. Laka M, Milazzo A, Merlin T. Factors That Impact the Adoption of Clinical Decision Support Systems (CDSS) for Antibiotic Management. Int J Environ Res Public Health 2021;18:1901. [Crossref] [PubMed]
87. Liaw W, Hischier B, James CA, et al. Bridging the Gap: Transforming Primary Care Through the Artificial Intelligence and Machine Learning for Primary Care (AIM-PC) Curriculum. Ann Fam Med 2024;22:570-1. [Crossref] [PubMed]
88. Liaw W, Kueper JK, Lin S, et al. Competencies for the Use of Artificial Intelligence in Primary Care. Ann Fam Med 2022;20:559-63. [Crossref] [PubMed]
89. Lin S. A Clinician's Guide to Artificial Intelligence (AI): Why and How Primary Care Should Lead the Health Care AI Revolution. J Am Board Fam Med 2022;35:175-84. [Crossref] [PubMed]
90. Lowe C, Atherton L, Lloyd P, et al. Improving Safety, Efficiency, Cost, and Satisfaction Across a Musculoskeletal Pathway Using the Digital Assessment Routing Tool for Triage: Quality Improvement Study. J Med Internet Res 2025;27:e67269. [Crossref] [PubMed]
91. Mahlknecht A, Engl A, Piccoliori G, et al. Supporting primary care through symptom checking artificial intelligence: a study of patient and physician attitudes in Italian general practice. BMC Prim Care 2023;24:174. [Crossref] [PubMed]
92. Manski-Nankervis JA, Biezen R, Thursky K, et al. Developing a Clinical Decision Support Tool for Appropriate Antibiotic Prescribing in Australian General Practice: A Simulation Study. Med Decis Making 2020;40:428-37. [Crossref] [PubMed]
93. Marcolino MS, Oliveira JAQ, Cimini CCR, et al. Development and Implementation of a Decision Support System to Improve Control of Hypertension and Diabetes in a Resource-Constrained Area in Brazil: Mixed Methods Study. J Med Internet Res 2021;23:e18872. [Crossref] [PubMed]
94. Medina-Lara A, Grigore B, Lewis R, et al. Cancer diagnostic tools to aid decision-making in primary care: mixed-methods systematic reviews and cost-effectiveness analysis. Health Technol Assess 2020;24:1-332. [Crossref] [PubMed]
95. Mehta N, Gupta S, Kularathne Y. The Role and Impact of Artificial Intelligence in Addressing Sexually Transmitted Infections, Nonvenereal Genital Diseases, Sexual Health, and Wellness. Indian Dermatol Online J 2023;14:793-8. [Crossref] [PubMed]
96. Mikkelsen JG, Sørensen NL, Merrild CH, et al. Patient perspectives on data sharing regarding implementing and using artificial intelligence in general practice - a qualitative study. BMC Health Serv Res 2023;23:335. [Crossref] [PubMed]
97. Minian N, Noormohamed A, Lingam M, et al. Integrating a brief alcohol intervention with tobacco addiction treatment in primary care: qualitative study of health care practitioner perceptions. Addict Sci Clin Pract 2021;16:17. [Crossref] [PubMed]
98. Mondal H, De R, Mondal S, et al. A large language model in solving primary healthcare issues: A potential implication for remote healthcare and medical education. J Educ Health Promot 2024;13:362. [Crossref] [PubMed]
99. Morrison K. Artificial intelligence and the NHS: a qualitative exploration of the factors influencing adoption. Future Healthc J 2021;8:e648-54. [Crossref] [PubMed]
100. Morton K, Santillo M, Van Velthoven MH, et al. Promoting the implementation of clinical decision support systems in primary care: A qualitative exploration of implementing a Fractional exhaled Nitric Oxide (FeNO)-guided decision support system in asthma consultations. PLoS One 2025;20:e0317613. [Crossref] [PubMed]
101. Moschogianis S, Darley S, Coulson T, et al. Seven Opportunities for Artificial Intelligence in Primary Care Electronic Visits: Qualitative Study of Staff and Patient Views. Ann Fam Med 2025;23:214-22. [Crossref] [PubMed]
102. Mwogosi A. AI-driven optimisation of EHR systems implementation in Tanzania’s primary health care. Transforming Government: People Process and Policy 2025;19:288-315.
103. Ngoc Nguyen O, Amin D, Bennett J, et al. GP or ChatGPT? Ability of large language models (LLMs) to support general practitioners when prescribing antibiotics. J Antimicrob Chemother 2025;80:1324-30.
104. Nguyen K, Wilson DL, Diiulio J, et al. Design and development of a machine-learning-driven opioid overdose risk prediction tool integrated in electronic health records in primary care settings. Bioelectron Med 2024;10:24. [Crossref] [PubMed]
105. Nolan B, Daybranch ER, Barton K, et al. Patient and Provider Experience with Artificial Intelligence Screening Technology for Diabetic Retinopathy in a Rural Primary Care Setting. J Maine Med Cent 2023;5:2. [Crossref] [PubMed]
106. Peek N, Capurro D, Rozova V, et al. Bridging the Gap: Challenges and Strategies for the Implementation of Artificial Intelligence-based Clinical Decision Support Systems in Clinical Practice. Yearb Med Inform 2024;33:103-14. [Crossref] [PubMed]
107. Peiffer-Smadja N, Descousse S, Courrèges E, et al. Implementation of a Clinical Decision Support System for Antimicrobial Prescribing in Sub-Saharan Africa: Multisectoral Qualitative Study. J Med Internet Res 2024;26:e45122. [Crossref] [PubMed]
108. Peiffer-Smadja N, Poda A, Ouedraogo AS, et al. Paving the Way for the Implementation of a Decision Support System for Antibiotic Prescribing in Primary Care in West Africa: Preimplementation and Co-Design Workshop With Physicians. J Med Internet Res 2020;22:e17940. [Crossref] [PubMed]
109. Petersen GB, Joensen LE, Kristensen JK, et al. How to Improve Attendance for Diabetic Retinopathy Screening: Ideas and Perspectives From People With Type 2 Diabetes and Health-care Professionals. Can J Diabetes 2025;49:121-7. [Crossref] [PubMed]
110. Petersson L, Larsson I, Nygren JM, et al. Challenges to implementing artificial intelligence in healthcare: a qualitative interview study with healthcare leaders in Sweden. BMC Health Serv Res 2022;22:850. [Crossref] [PubMed]
111. Pratt R, Saman DM, Allen C, et al. Assessing the implementation of a clinical decision support tool in primary care for diabetes prevention: a qualitative interview study using the Consolidated Framework for Implementation Science. BMC Med Inform Decis Mak 2022;22:15. [Crossref] [PubMed]
112. Preiser C, Radionova N, Ög E, et al. The Doctors, Their Patients, and the Symptom Checker App: Qualitative Interview Study With General Practitioners in Germany. JMIR Hum Factors 2024;11:e57360. [Crossref] [PubMed]
113. Qin H, Tong Y. Opportunities and Challenges for Large Language Models in Primary Health Care. J Prim Care Community Health 2025;16:21501319241312571. [Crossref] [PubMed]
114. Razai MS, Al-Bedaery R, Bowen L, et al. Implementation challenges of artificial intelligence (AI) in primary care: Perspectives of general practitioners in London UK. PLoS One 2024;19:e0314196. [Crossref] [PubMed]
115. Schütze D, Holtz S, Neff MC, et al. Requirements analysis for an AI-based clinical decision support system for general practitioners: a user-centered design process. BMC Med Inform Decis Mak 2023;23:144. [Crossref] [PubMed]
116. Scipion CEA, Manchester MA, Federman A, et al. Barriers to and facilitators of clinician acceptance and use of artificial intelligence in healthcare settings: a scoping review. BMJ Open 2025;15:e092624. [Crossref] [PubMed]
117. Seol HY, Shrestha P, Muth JF, et al. Artificial intelligence-assisted clinical decision support for childhood asthma management: A randomized clinical trial. PLoS One 2021;16:e0255261. [Crossref] [PubMed]
118. Shah SJ, Crowell T, Jeong Y, et al. Physician Perspectives on Ambient AI Scribes. JAMA Netw Open 2025;8:e251904. [Crossref] [PubMed]
119. Sharma V, McDermott J, Keen J, et al. Pharmacogenetics Clinical Decision Support Systems for Primary Care in England: Co-Design Study. J Med Internet Res 2024;26:e49230. [Crossref] [PubMed]
120. Shour AR, Jones GL, Anguzu R, et al. Development of an evidence-based model for predicting patient, provider, and appointment factors that influence no-shows in a rural healthcare system. BMC Health Serv Res 2023;23:989. [Crossref] [PubMed]
121. SidesTKbaierDFarrellTExploring the Potential of Artificial Intelligence in Primary Care: Insights from Stakeholders’ Perspectives.Preprints 2023, 2023110995.
122. Siira E, Johansson H, Nygren J. Mapping and Summarizing the Research on AI Systems for Automating Medical History Taking and Triage: Scoping Review. J Med Internet Res 2025;27:e53741. [Crossref] [PubMed]
123. Siira E, Tyskbo D, Nygren J. Healthcare leaders' experiences of implementing artificial intelligence for medical history-taking and triage in Swedish primary care: an interview study. BMC Prim Care 2024;25:268. [Crossref] [PubMed]
124. Singareddy S, Sn VP, Jaramillo AP, et al. Artificial Intelligence and Its Role in the Management of Chronic Medical Conditions: A Systematic Review. Cureus 2023;15:e46066. [Crossref] [PubMed]
125. Smak Gregoor AM, Sangers TE, Eekhof JA, et al. Artificial intelligence in mobile health for skin cancer diagnostics at home (AIM HIGH): a pilot feasibility study. EClinicalMedicine 2023;60:102019. [Crossref] [PubMed]
126. Steerling E, Svedberg P, Nilsen P, et al. Influences on trust in the use of AI-based triage-an interview study with primary healthcare professionals and patients in Sweden. Front Digit Health 2025;7:1565080. [Crossref] [PubMed]
127. Šteinmiller J, Ross P. General practitioners' user experience of the nationwide digital decision support system in primary care. Digit Health 2024;10:20552076241271816. [Crossref] [PubMed]
128. Stroud AM, Curtis SH, Weir IB, et al. Physician Perspectives on the Potential Benefits and Risks of Applying Artificial Intelligence in Psychiatric Medicine: Qualitative Study. JMIR Ment Health 2025;12:e64414. [Crossref] [PubMed]
129. Sunjaya AP, Martin A, Jenkins C. A Design Thinking Approach to Developing a Clinical Decision Support System for Breathlessness in Primary Care. Stud Health Technol Inform 2022;290:839-43. [Crossref] [PubMed]
130. Tegenaw GS, Amenu D, Ketema G, et al. Evaluating a clinical decision support point of care instrument in low resource setting. BMC Med Inform Decis Mak 2023;23:51. [Crossref] [PubMed]
131. Terry AL, Kueper JK, Beleno R, et al. Is primary health care ready for artificial intelligence? What do primary health care stakeholders say? BMC Med Inform Decis Mak 2022;22:237.
132. Upshaw TL, Craig-Neil A, Macklin J, et al. Priorities for Artificial Intelligence Applications in Primary Care: A Canadian Deliberative Dialogue with Patients, Providers, and Health System Leaders. J Am Board Fam Med 2023;36:210-20. [Crossref] [PubMed]
133. Witkowski K, Dougherty RB, Neely SR. Public perceptions of artificial intelligence in healthcare: ethical concerns and opportunities for patient-centered care. BMC Med Ethics 2024;25:74. [Crossref] [PubMed]
134. Wojda T, Hoffman C, Kindler K, et al. Ethics of Artificial Intelligence: Implications for Primary Care and Family Medicine Residency Programs [Internet]. Artificial Intelligence. IntechOpen; 2024. Available online: https://doi.org/10.5772/intechopen.114907
135. Young RA, Martin CM, Sturmberg JP, et al. What Complexity Science Predicts About the Potential of Artificial Intelligence/Machine Learning to Improve Primary Care. J Am Board Fam Med 2024;37:332-45. [Crossref] [PubMed]
136. Zhou J, Zhang J, Wan R, et al. Integrating AI into clinical education: evaluating general practice trainees' proficiency in distinguishing AI-generated hallucinations and impacting factors. BMC Med Educ 2025;25:406. [Crossref] [PubMed]
137. Alanazi AS, Salah Alanazi A, Benlaria H. Enhancing Healthcare for People with Disabilities Through Artificial Intelligence: Evidence from Saudi Arabia. Healthcare (Basel) 2025;13:1616. [Crossref] [PubMed]
138. AlSerkal YM, Ibrahim NM, Alsereidi AS, et al. Real-Time Analytics and AI for Managing No-Show Appointments in Primary Health Care in the United Arab Emirates: Before-and-After Study. JMIR Form Res 2025;9:e64936. [Crossref] [PubMed]
139. Barry B, Zhu X, Behnken E, et al. Provider Perspectives on Artificial Intelligence-Guided Screening for Low Ejection Fraction in Primary Care: Qualitative Study. JMIR AI 2022;1:e41940. [Crossref] [PubMed]
140. Beals D, Simon L, Rogers F, et al. Revolutionizing Diabetic Retinopathy Screening: Integrating AI-Based Retinal Imaging in Primary Care. J CME 2025;14:2437294. [Crossref]
141. Bowman MA, Seehusen DA, Britz J, et al. Research to Improve Clinical Care in Family Medicine: Big Data, Telehealth, Artificial Intelligence, and More. J Am Board Fam Med 2024;37:161-4. [Crossref] [PubMed]
142. Canaway R, Dai L, Hallinan C, et al. The feasibility of integrating an alcohol screening clinical decision support tool into primary care clinical software: a review and Australian key stakeholder study. BMC Prim Care 2024;25:408. [Crossref] [PubMed]
143. Dehnavieh R, Inayatullah S, Yousefi F, Nadali M. Artificial Intelligence (AI) and the future of Iran's Primary Health Care (PHC) system. BMC Prim Care 2025;26:75. [Crossref] [PubMed]
144. Escalé-Besa A, Yélamos O, Vidal-Alaball J, et al. Exploring the potential of artificial intelligence in improving skin lesion diagnosis in primary care. Sci Rep 2023;13:4293. [Crossref] [PubMed]
145. Fedele DA, Ray JM, Mallela JL, et al. Development of a Clinical Decision Support Tool to Implement Asthma Management Guidelines in Pediatric Primary Care: Qualitative Study. JMIR Form Res 2025;9:e65794. [Crossref] [PubMed]
146. Hamari L, Parisod H, Siltanen H, et al. Clinical decision support in promoting evidence-based nursing in primary healthcare: a cross-sectional study in Finland. JBI Evid Implement 2023;21:294-300. [Crossref] [PubMed]
147. Horwood C, Luthuli S, Mapumulo S, et al. Challenges of using e-health technologies to support clinical care in rural Africa: a longitudinal mixed methods study exploring primary health care nurses’ experiences of using an electronic clinical decision support system (CDSS) in South Africa. BMC Health Serv Res 2023;23:30. [Crossref] [PubMed]
148. Jensen C, McKerrow NH. The feasibility and ongoing use of electronic decision support to strengthen the implementation of IMCI in KwaZulu-Natal, South Africa. BMC Pediatr 2022;22:80. [Crossref] [PubMed]
149. Khan H, Bokhari SFH. Integrating Artificial Intelligence (AI) Chatbots for Depression Management: A New Frontier in Primary Care. Cureus 2024;16:e66857. [Crossref] [PubMed]
150. Lagerin A, Törnkvist L, Fastbom J, et al. District nurses' experiences of using a clinical decision support system and an assessment tool at elderly care units in primary health care: a qualitative study. Prim Health Care Res Dev 2021;22:e45. [Crossref] [PubMed]
151. Li J, Guan Z, Wang J, et al. Integrated image-based deep learning and language models for primary diabetes care. Nat Med 2024;30:2886-96. [Crossref] [PubMed]
152. Li Q, Drinkwater JJ, Woods K, et al. Implementation of A New, Mobile Diabetic Retinopathy Screening Model Incorporating Artificial Intelligence in Remote Western Australia. Aust J Rural Health 2025;33:e70031. [Crossref] [PubMed]
153. Menchaca JT. For AI in Primary Care, Start With the Problem. Ann Fam Med 2025;23:5-6. [Crossref] [PubMed]
154. Meunier PY, Raynaud C, Guimaraes E, et al. Barriers and Facilitators to the Use of Clinical Decision Support Systems in Primary Care: A Mixed-Methods Systematic Review. Ann Fam Med 2023;21:57-69. [Crossref] [PubMed]
155. Nash DM, Thorpe C, Brown JB, et al. Perceptions of Artificial Intelligence Use in Primary Care: A Qualitative Study with Providers and Staff of Ontario Community Health Centres. J Am Board Fam Med 2023;36:221-8. [Crossref] [PubMed]
156. Neff MC, Schütze D, Holtz S, et al. Development and expert inspections of the user interface for a primary care decision support system. Int J Med Inform 2024;192:105651. [Crossref] [PubMed]
157. Rubin G, Walter FM, Emery J, et al. Electronic clinical decision support tool for assessing stomach symptoms in primary care (ECASS): a feasibility study. BMJ Open 2021;11:e041795. [Crossref] [PubMed]
158. Schmitz T, Beynon F, Musard C, et al. Effectiveness of an electronic clinical decision support system in improving the management of childhood illness in primary care in rural Nigeria: an observational study. BMJ Open 2022;12:e055315. [Crossref] [PubMed]
159. Thirunavukarasu AJ, Hassan R, Mahmood S, et al. Trialling a Large Language Model (ChatGPT) in General Practice With the Applied Knowledge Test: Observational Study Demonstrating Opportunities and Limitations in Primary Care. JMIR Med Educ 2023;9:e46599. [Crossref] [PubMed]
160. Thomsen CHN, Kronborg T, Hangaard S, et al. Developing an AI-Based clinical decision support system for basal insulin titration in type 2 diabetes in primary Care: A Mixed-Methods evaluation using heuristic Analysis, user Feedback, and eye tracking. Int J Med Inform 2025;195:105783. [Crossref] [PubMed]
161. Ursin F, Timmermann C, Orzechowski M, et al. Diagnosing Diabetic Retinopathy With Artificial Intelligence: What Information Should Be Included to Ensure Ethical Informed Consent?. Front Med (Lausanne) 2021;8:695217. [Crossref] [PubMed]
162. Vidal-Alaball J, Panadés Zafra R, Escalé-Besa A, et al. The artificial intelligence revolution in primary care: Challenges, dilemmas and opportunities. Aten Primaria 2024;56:102820. [Crossref] [PubMed]
163. Waheed MA, Liu L. Perceptions of Family Physicians About Applying AI in Primary Health Care: Case Study From a Premier Health Care Organization. JMIR AI 2024;3:e40781. [Crossref] [PubMed]
164. Wang L, Yin Y, Glampson B, et al. Transformer-based deep learning model for the diagnosis of suspected lung cancer in primary care based on electronic health record data. EBioMedicine 2024;110:105442. [Crossref] [PubMed]
165. Wewetzer L, Held LA, Goetz K, et al. Determinants of the implementation of artificial intelligence-based screening for diabetic retinopathy-a cross-sectional study with general practitioners in Germany. Digit Health 2023;9:20552076231176644. [Crossref] [PubMed]
166. Elliott TE, Asche SE, O'Connor PJ, et al. Clinical Decision Support with or without Shared Decision Making to Improve Preventive Cancer Care: A Cluster-Randomized Trial. Med Decis Making 2022;42:808-21. [Crossref] [PubMed]
167. Huguet N, Ezekiel-Herrera D, Gunn R, et al. Uptake of a Cervical Cancer Clinical Decision Support Tool: A Mixed-Methods Study. Appl Clin Inform 2023;14:594-9. [Crossref] [PubMed]
168. Wiedermann CJ, Mahlknecht A, Piccoliori G, et al. Redesigning Primary Care: The Emergence of Artificial-Intelligence-Driven Symptom Diagnostic Tools. J Pers Med 2023;13:1379. [Crossref] [PubMed]
169. Willis M, Duckworth P, Coulter A, et al. Qualitative and quantitative approach to assess of the potential for automating administrative tasks in general practice. BMJ Open 2020;10:e032412. [Crossref] [PubMed]
170. Khan SD, Hoodbhoy Z, Raja MHR, et al. Frameworks for procurement, integration, monitoring, and evaluation of artificial intelligence tools in clinical settings: A systematic review. PLOS Digit Health 2024;3:e0000514. [Crossref] [PubMed]
171. Giaxi P, Vivilaki V, Sarella A, et al. Artificial Intelligence in Midwifery: A Scoping Review of Current Applications, Future Prospects, and Midwives' Perspectives. Healthcare (Basel) 2025;13:942. [Crossref] [PubMed]
172. Liu X, Cruz Rivera S, Moher D, et al. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Lancet Digit Health 2020;2:e537-48. [Crossref] [PubMed]
173. Bhayana R. Chatbots and Large Language Models in Radiology: A Practical Primer for Clinical and Research Applications. Radiology 2024;310:e232756. [Crossref] [PubMed]
174. Antaki F, Touma S, Milad D, et al. Evaluating the Performance of ChatGPT in Ophthalmology: An Analysis of Its Successes and Shortcomings. Ophthalmol Sci 2023;3:100324. [Crossref] [PubMed]
175. Itchhaporia D. The Evolution of the Quintuple Aim: Health Equity, Health Outcomes, and the Economy. J Am Coll Cardiol 2021;78:2262-4. [Crossref] [PubMed]
176. Bhatia RS, Chu C, Pang A, et al. Virtual care use before and during the COVID-19 pandemic: a repeated cross-sectional study. CMAJ Open 2021;9:E107-14. [Crossref] [PubMed]
177. Shachar C, Engel J, Elwyn G. Implications for Telehealth in a Postpandemic Future: Regulatory and Privacy Issues. JAMA 2020;323:2375-6. [Crossref] [PubMed]