The tapestry of health care is in a constant state of evolution. Among the myriad innovations that have emerged in recent decades, artificial intelligence (AI) holds the most transformative potential. AI could provide clinical decision support tools, make predictive analyses, and expedite administrative tasks.[1] Family medicine, anchored in personalized and holistic care, is at an exciting crossroad as this digital force spreads across specialties. Introducing AI into family medicine poses the question: How does technology match up with a deeply human-centred practice? By understanding the interplay between AI and the family physician, we can enhance the practice of family medicine and ensure that its foundational principles remain firm in the face of technological advancements.

What makes up the continuum of patient care in family medicine?

In family medicine, the continuum of patient care can be used to outline a patient’s comprehensive journey, from the initial point of contact to long-term health management [Box]. It starts with patient engagement, where practitioners establish trust and open channels of communication. This progresses to clinical encounter preparation, where thorough reviews of patient histories equip physicians for informed consultations; patient examination, which delves into history taking and physical assessment; and diagnosis, where practitioners distill insights from examinations to identify health conditions. Postdiagnosis, the focus shifts to treatment planning, ensuring tailored interventions are charted. Finally, long-term health management emphasizes continuous care, ensuring patients adhere to treatment regimens and receive guidance for sustained well-being. 

How can AI influence the continuum of patient care in family medicine? 

Patient engagement

An AI-enhanced approach allows clinicians to tailor care to each patient’s health narrative with better precision and empathy. While traditional methods of dissecting massive data sets often fall short, AI offers a new lens to detect patterns and derive insights. One of the most tangible impacts of this capability is the precise identification of at-risk patient characteristics. For example, AI has been shown to accurately predict individual suicide risk.[2] AI-driven algorithms can also create personalized health recommendations by combining genetic predispositions and lifestyle data.[3] The Rothman Index, a scoring system using electronic health records to predict overall patient health, is one example of AI-driven precision medicine.[4] Personalized advisories foster adherence and position the patient at the centre of their health journey, promising more informed decision making and proactive health management. In particular, patients with chronic diseases can benefit from nudge-inspired AI-driven behavior intervention. For example, a macronutrient detection algorithm can analyze images of food taken by a patient and communicate nutritional information to the patient.[5]

Clinical encounter preparation

An enduring patient–physician relationship provides a mosaic of longitudinal data that is ripe for AI’s analytical abilities. AI excels in efficiently processing vast amounts of data, saving clinicians countless hours spent reviewing results, discharge summaries, and imaging reports. A prognostic study in a gastroenterology department revealed that clinicians who used AI systems that specialized in organizing patient records saved 18% of the time to address clinical questions without compromising accuracy.[6] More than just collating information, this is about deep-diving into the data, spotlighting recurring health patterns, unearthing latent risks, and predicting future complications. For example, AI-based methods can be used to optimize medication alerts for possible drug interactions, allergies, or adherence issues using information from medical records.[7] AI-enabled previsit planning tools can automate routine tasks such as appointment reminders, questionnaire integration, and tracking of age-appropriate screenings, further freeing up a clinician’s time[8] and potentially alleviating administrative burnout.

Patient examination

Patient examination combines history taking and physical examination, representing a deeply personal aspect of health care. Patients share intimate details, while physicians strive to interpret an intricate blend of health experiences. AI may provide guidance and assistance in this area. Ambient AI is one tool that can seamlessly capture physician–patient conversations while operating in the background, reducing the likelihood of details being missed.[9] Natural language processing AI, such as Dragon Ambient eXperience (DAX) Express, is already being incorporated into clinicians’ workflows to reduce documentation burden.[10] Beyond observing, AI may also predict potential areas of concern, guiding the physician toward pertinent questions based on the patient’s narrative. 

The application of AI to physical examinations addresses the variability and subjectivity inherent in diagnostic techniques, bringing standardization and consistency. For example, in pulmonary conditions, where traditional auscultation varies based on clinician experience, AI-powered auscultation devices can reduce variability, offering more consistent results than those based on the human ear alone.[11] In dermatological examinations, AI systems perform nuanced real-time image classifications.[12] These advanced algorithms compare patient data against extensive databases, highlighting patterns that human observation may miss.[13] Integrating AI with imaging and point-of-care ultrasound has the potential to enhance the sensitivity and specificity of physical exams.[14] This streamlines the diagnostic process and brings physical examination techniques closer to the objective precision sought in family medicine.  

Diagnosis and treatment planning

AI especially shows its value in medical cases where symptoms resemble a jigsaw puzzle, with pieces that are scattered, nonspecific, or indicative of multiple overlapping conditions. The use of AI-assisted image analysis has already shown promising results in improving early detection of invasive and small-sized breast cancer.[15] Using natural language processing, AI can also analyze unstructured clinical notes and lab reports to highlight patterns that may aid in diagnosis.[16] Given the relevant clinical features, AI may also generate a list of must-not-miss differentials to support clinicians’ investigative process.[17] As AI technology advances and becomes more integrated into health care systems, it could lead to a re-evaluation of existing guidelines, enabling a shift toward more personalized and age-specific screening strategies. This has profound implications for primary care, where early detection and efficient allocation of screening resources are paramount. 

AI also shows significant potential in treatment planning. Harnessing the power of predictive analytics, AI platforms can simulate various treatment trajectories tailored to a patient’s unique medical and genetic blueprint.[18] Physicians can then visualize the likely therapeutic impacts and be forewarned of potential side effects, drug interactions, and the longer-term prognosis.[19] Such a broad view ensures that treatment plans are effective and aligned with individual patient needs and preferences and has the potential to lead to enhanced resource allocation. 

Continual care and long-term management

AI’s potential in this space is visible in the emergence of wearable health devices,[20] which monitor numerous health parameters, from heart rhythms to glucose levels. Their strength lies in the ability to detect and flag deviations or anomalies in real time. Prompt alerts, when acted upon, can lead to timely medical interventions, circumvent complications, and forestall potential health crises. With AI-enabled wearables as a catalyst, health care shifts from reactive to proactive. AI can improve patient adherence to treatment plans, addressing forgetfulness through applications that provide reminders, tailored dosage advice, and feedback on medication efficacy based on real-time patient feedback.[21-23] These systems can also provide personalized lifestyle recommendations, such as activity prompts or dietary advice, based on sedentary patterns and glucose levels from data-driven insights.[18-24] AI-powered chatbots can also provide 24/7 support for patients, offering health recommendations, assistance with mental health concerns, and symptom tracking.[25,26]

What are the obstacles to integrating AI into family medicine?

There are four major challenges to integrating AI into diagnosis and treatment: erosion of human touch, clinician and patient overreliance on technology, threats to privacy, and spread of bias in AI algorithms.

Erosion of human touch

As AI-driven tools mediate patient interactions with physicians, is the human touch that defines family medicine being diluted? Automated advisories, while personalized, may lack the empathy and understanding of a human practitioner, potentially undermining the core values of family medicine and eroding patient trust. As patients grow used to AI-driven assistance, the traditional, human-centric medical consultation risks being supplanted, leading to a sense of isolation and a feeling of being managed by an impersonal algorithm, which can weaken the therapeutic relationship. In the worst cases, patients may be reduced to mere data points, undermining the trust and the sense of personal touch that lies at the heart of family medicine.

Overdependence on technology

Intertwining AI and medicine comes with the risk of physicians becoming overly reliant on AI and potentially overlooking their clinical expertise.27 While the quantity of data that AI can process is impressive, it can paint an incomplete picture while leading clinicians to undervalue their own judgment. Previously, overdependence on technologies like electronic health records has led to automation-induced errors.[28] Additionally, AI’s emphasis on precision may lead to overdiagnosis and unnecessary tests, increasing health care costs and patient anxiety.[29] Medicine is an art as well as a science—a delicate blend of intuition, accumulated experience, patient narratives, and context-driven judgment. This exposes a crucial medicolegal question: If a physician misses critical information due to an AI error, who is medicolegally at fault? While AI enhances diagnostics and treatment, its integration must be carefully managed, ensuring it supports rather than replaces human expertise.

Threats to privacy

Integrating AI into health care causes unease about patient privacy as the public grapples with concerns over the security, transparency, and responsible use of their medical information. AI algorithms typically train on databases that potentially contain confidential health information. Some algorithms may operate as “black boxes” and arrive at decisions through an unexplainable process.[30] This may challenge privacy regulations such as the Health Insurance Portability and Accountability Act and the Personal Information Protection and Electronic Documents Act, which require transparency in data processing to protect a patient’s right to know how their health information is used.[31] Another privacy concern is the adequate removal of personally identifiable information from data sets. Through complex data patterns, AI algorithms may inadvertently re-identify individuals through quasi-identifiers, linkage attacks, or techniques that nullify data perturbation.[32] Addressing these challenges requires a comprehensive approach, including robust data security measures, ethical AI development practices, transparent algorithms, and ongoing education for health care professionals and the public about the privacy implications of AI in health care. 

AI biases

The maxim “garbage in, garbage out” applies to AI’s role in the continuum of patient care. Incomplete records, biased data sets, and minor inaccuracies can severely compromise AI’s outputs, leading to potentially erroneous recommendations. The dangers of relying on AI algorithms are well documented in dermatology: underrepresentation of diverse skin types in training data threatens the external validity of such tools.33 AI deep learning models have been shown to readily predict patient race from radiographs, even when clinical experts failed to do the same.[34] Therefore, human oversight of AI models may be of limited use in identifying race-specific errors generated by AI algorithms.[34] Without careful consideration, the use of AI risks exacerbating existing biases and worsening outcomes for marginalized groups. In the age of information inundation, discerning data that are of clinical relevance from noise and ensuring that essential insights are not buried can be an overwhelming task. 

Conclusions

The impact of AI stretches across the continuum of patient care, and, as with any significant technological leap, the journey is dotted with challenges. The essence of family medicine, with its emphasis on human touch, empathy, and shared decision making, risks being overshadowed in a health care paradigm that is overly dependent on AI. As we venture into this AI-augmented era, continuous introspection is imperative. The promise of AI in family medicine is about not just efficient care, but care that is more attuned, responsive, and, ultimately, human.

Competing interests

None declared.
   

BOX. Artificial intelligence–enabled continuum of patient care in family medicine.

Patient engagement:

• Prediction of at-risk individuals.
• Personalized health care.
• Patient empowerment.

Clinical encounter preparation:

• Patient profile construction.
• Predicting health complications.
• Automating routine tasks.

Patient examination:

• Ambient artificial intelligence and real-time exam interpretation.
• Eliminating subjectivity.
• Increased sensitivity and specificity.

Diagnosis and treatment planning:

• Personalized medical plan.
• Flagging contraindications.
• Enhanced resource allocation.

Continual care and long-term management:

• Continuous monitoring.
• Fortifying medical adherence.
• Prompting preventive medicine.

This article has been peer reviewed.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Ms Hui is a second-year medical student in the Faculty of Medicine at the University of British Columbia. Dr Raff is a family medicine resident in the Faculty of Medicine at UBC. Dr Hu is an internal medicine resident in the Faculty of Medicine at UBC. Dr Yau is a family medicine resident in the Faculty of Medicine at UBC. Mr Singla is an MD/PhD student in the School of Biomedical Engineering and the Faculty of Medicine at UBC.