Abstract
Introduction
Artificial intelligence (AI) has the potential to revolutionise medical practice to optimise patient care, and is already doing so in areas such as radiology (Hosny et al., 2018) and pathology (Cui and Zhang, 2021). General practice also has the potential to benefit from AI (Lin et al., 2019).
Recently, the freely accessible, online, AI chatbot, ChatGPT, has caught the public eye (OpenAI, 2022). This chatbot is trained on an extremely large body of text from the internet and aims to provide a human-like text conversation with users. It is constantly learning from the internet and interactions people have with it, so the answers given are usually unique.
In this article, we present an interview-style interaction with ChatGPT on the use of AI in general practice that we conducted on 22 March 2023. The questions posed to ChatGPT were agreed between all authors. The authors collectively decided to select questions and responses for inclusion in this manuscript from those posed to the system. The responses presented below are unedited.
The interview
What are the greatest challenges currently faced in general practice that AI can help to address?
One of the biggest challenges faced in general practice is the increasing workload and time constraints. AI can help by automating certain tasks such as appointment scheduling, triage and clinical decision-making, freeing up time for GPs to focus on more complex cases. AI can also help with patient stratification, identifying high-risk patients who require more attention and enabling early intervention. Another challenge is the increasing demand for personalised care, which can be met through AI-powered predictive analytics and patient monitoring systems. Lastly, AI can assist with data analysis, identifying patterns and trends to improve healthcare delivery and resource allocation.
If AI is used for tasks such as appointment scheduling and triage, does that reduce the need for employees, such as receptionists and administrative staff?
AI can certainly streamline administrative tasks such as appointment scheduling and triage, but it does not necessarily mean a reduction in staff. Instead, it can free up administrative staff to focus on more value-added tasks, such as patient communication and engagement, which require human interaction and empathy. Additionally, AI can augment the work of administrative staff by improving efficiency and accuracy, reducing the risk of errors and delays, and enabling them to handle more complex tasks. Ultimately, the use of AI can help practices to optimise their workforce and improve the quality of care they provide.
If AI can help GPs with clinical decision-making and routine tasks, will it help reduce doctor burnout?
AI has the potential to help reduce burnout among GPs by automating routine tasks, enabling more efficient and accurate clinical decision-making, and freeing up time for patient care and engagement. By reducing the burden of administrative tasks, AI can help GPs to focus on the aspects of their job that they find most rewarding and fulfilling, such as building relationships with patients, improving patient outcomes, and engaging in ongoing professional development. However, it is important to note that AI is not a panacea for burnout and that addressing this complex issue requires a multifaceted approach that includes addressing systemic issues in healthcare delivery, improving work–life balance, and fostering a culture of well-being and resilience among healthcare professionals.
What are the ethical issues faced when introducing AI into general practice?
The introduction of AI into general practice raises a number of ethical issues, including:
Privacy and security: AI systems require access to patient data, which raises concerns about data privacy and security Bias and fairness: AI systems can be biased, which may result in differential outcomes for certain patient groups Accountability and transparency: The use of AI in decision-making can make it difficult to attribute responsibility for clinical outcomes, leading to issues of accountability and transparency Autonomy and consent: Patients may not fully understand the implications of AI-assisted care, which raises questions around informed consent and patient autonomy Professional responsibility: The use of AI in clinical decision-making raises questions about the role of healthcare professionals, their responsibilities, and their accountability for decisions made with the assistance of AI Equity: The introduction of AI into general practice may widen existing inequities in healthcare delivery if certain patient groups have limited access to AI-assisted care
It is important to address these ethical issues to ensure that AI is developed and implemented in a way that maximises its potential benefits while minimising its risks and ensuring that patient rights and interests are protected.
If AI makes mistakes in clinical decision-making, who should be held responsible for the adverse outcome?
The issue of responsibility for adverse outcomes resulting from AI-assisted clinical decision-making is complex and still evolving. Responsibility may be shared among multiple parties, including the healthcare professional, the developer of the AI system, and the organisation that deployed the AI system. The specific circumstances of the adverse outcome, the degree of human oversight involved, and the level of transparency and accountability in the development and deployment of the AI system will all be relevant factors in determining responsibility.
In general, it is important to establish clear protocols for AI-assisted clinical decision-making, including guidelines for when and how to intervene in the decision-making process, as well as mechanisms for monitoring and auditing the system's performance. Ultimately, the responsibility for ensuring patient safety and well-being rests with the healthcare professional, who should exercise clinical judgment and discretion in the use of AI-assisted care and take responsibility for any adverse outcomes that result from its use.
Can AI help with communication of complex medical conditions to patients?
Yes, AI can help with the communication of complex medical conditions to patients. AI-powered tools can help to simplify complex medical information and present it in a way that is easy for patients to understand. For example, AI-powered chatbots can interact with patients in natural language and provide personalised information and guidance based on their specific medical condition.
Additionally, AI-powered visualisation tools can help to present complex medical information in a more intuitive and engaging way. For example, interactive 3D models can be used to help patients visualise the anatomy of their condition and understand the underlying mechanisms of their disease.
Overall, AI has the potential to transform the way that medical information is communicated to patients, making it more accessible, understandable, and engaging. This can help to improve patient education, increase patient engagement in their care, and ultimately improve patient outcomes.
In what way can AI assist medical education in general practice, in both the undergraduate and postgraduate setting?
AI can assist medical education in general practice by providing access to personalised and adaptive learning experiences that can help to improve knowledge retention, clinical decision-making skills, and patient outcomes. In the undergraduate setting, AI-powered educational tools can provide students with interactive and engaging learning experiences that can help them to develop foundational knowledge and skills. For example, AI-powered simulators can provide students with virtual patient encounters that simulate real-world scenarios, enabling them to practice and refine their clinical skills in a safe and controlled environment. In the postgraduate setting, AI can assist with continuing professional development by providing personalised feedback and recommendations based on the individual's learning goals and performance. Additionally, AI can assist with clinical decision-making by providing access to the latest research and clinical guidelines, enabling healthcare professionals to stay up-to-date with the latest developments in their field.
How can AI be used in general practice research?
AI can be used in general practice research in several ways, including:
Data analysis: AI-powered tools can help to analyse large datasets to identify patterns and correlations that may not be immediately apparent to human researchers. This can help to identify new research questions and areas of investigation Predictive modelling: AI-powered predictive models can be used to identify patients who are at risk of certain health outcomes, such as hospitalisation or readmission. This can help researchers to identify risk factors and develop interventions to prevent adverse outcomes Natural Language Processing (NLP): NLP is a type of AI that can be used to analyse and extract meaning from unstructured data, such as clinical notes and medical records. This can help researchers to identify key themes and trends in patient data and develop new hypotheses for investigation Precision medicine: AI-powered tools can help to identify patient subgroups with specific clinical characteristics, enabling researchers to develop more targeted and personalised treatment approaches.
Overall, AI can help to accelerate the pace of research in general practice by enabling more efficient and accurate data analysis, enabling more personalised and targeted interventions, and improving our understanding of the underlying mechanisms of disease.
In what areas should AI not be used in general practice?
Although AI has the potential to transform the field of general practice, there are certain areas where it should not be used. These include:
Diagnosing complex medical conditions: Although AI can be helpful in assisting with clinical decision-making, it should not be used to replace a trained medical professional in diagnosing complex medical conditions. A human doctor is better equipped to understand the nuances of a patient's medical history, physical exam findings, and other contextual factors that may be relevant to making an accurate diagnosis Making ethical decisions: AI should not be used to make ethical decisions, such as end-of-life care or other sensitive issues. These decisions require a human touch and should be made by a medical professional who can take into account the patient's values and preferences Prescribing medications: AI should not be used to prescribe medications without the input of a trained medical professional. Although AI can help to identify potential drug interactions or adverse reactions, it should not be relied upon to make treatment decisions without the input of a qualified doctor Replacing human interaction: Although AI can be used to enhance patient care, it should not be used to replace the human interaction between a doctor and patient. The doctor–patient relationship is a crucial component of general practice, and while AI can assist with routine tasks, it should not replace the human touch that is so important in building trust and rapport with patients
Overall, AI should be used as a tool to assist medical professionals in their work, rather than as a replacement for human decision-making and interaction.
Discussion
ChatGPT has highlighted above diverse ways in which AI could benefit general practice. It presents a balanced and realistic overview of how AI can support administrative, information- and data-driven tasks while not replacing person-centred roles. This is supported by in-depth ethnographic studies in general practices (Willis et al., 2020).
Likewise, ChatGPT approaches complex challenges, such as doctor burnout, in a balanced and informed way. Despite useful suggestions, it recognises the wider drivers behind such complex issues, calling for systemic changes and acknowledging that AI is no ‘panacea’ for these problems. It is of note that in the context of general practice at least, ChatGPT tends to offer overarching ways in which it could provide assistance. What remains unclear, however, is whether these benefits can ultimately be realised in the real-world workplace of primary care. That is because the bulk of activity in general practice lies in human interactions that require negotiation, flexibility and compromise (Cooper et al., 2015). There is a need for a solid evidence base before integrating AI in clinical care (Bowman, 2022). That is especially important in general practice, where digital solutions to routine tasks (e.g. documenting consultations) may increase GP workload and stress (Cooper et al., 2023).
Likewise, in general practice, clinical uncertainty is common, and striving for a unifying diagnosis often not required. Managing this is a key skill for general practice (Cooper M, Sornalingam S, Heath J, et al., 2022). In response to questions regarding clinical uncertainty in this interview, the chatbot tended to strive to find ways to reach a diagnosis, rather than embracing the uncertainty inherent in many consultations. This shows how AI may risk medicalising the well patient and increasing over-investigation and over-diagnosis. It thus calls into question the clinical value of AI in the context of low-risk populations and primary care more generally. That is because providing a clinical label is not always beneficial to patients (Heath, 2013). Conversely, it could be beneficial in some scenarios, for example, a diagnosis rarely encountered (or just overlooked) by the GP.
Overall, this interaction with ChatGPT provides a valuable insight into the future of AI in general practice. AI could play an important role in helping patients navigate services, undertake self-care and help them understand complex medical conditions, ultimately empowering patient autonomy. In this way, ChatGPT (and other AI services) is likely to emerge as the new ‘Dr Google’, i.e. the primary resource for patients to research their own symptoms and clinical management (Coomper et al., 2022). This could be of value in closing health inequities, such as improving access and care for non-English speakers. Further understanding is needed on how to introduce AI in a constructive, non-biased and ethical fashion into general practice. This should also consider how AI can enhance the work of GPs to improve patient care while preserving the doctor–patient relationship that is founded upon trust, continuity and empathy.
