Introduction
Artificial Intelligence (AI) is rapidly integrating into and transforming the delivery of health care, including physiotherapy. For example, AI may be used to streamline administrative processes, support the identification of patterns in patient data and enhance clinical decision-making.
The College recognizes that AI has the potential to improve both provider and patient experiences. At the same time, there are important ethical and legal challenges that must be managed to ensure patient trust is upheld and professional responsibilities are met.
This guidance document does not set out new professional requirements. Rather, it identifies key obligations through a core set of principles to guide the integration of AI into practice in a manner that serves patients’ best interests.
Definition
Artificial Intelligence generally refers to computer systems that can perform tasks commonly associated with human intelligence including learning, reasoning, decision-making, pattern recognition and problem-solving.
Existing Professional Requirements
Like any new technology, the introduction and adoption of AI in physiotherapy does not change physiotherapists’ fundamental obligations set out in:
- The College’s Code of Ethical Conduct;
- Existing Standards of Practice, including but not limited to Assessment, Diagnosis, Treatment, Communication, Evidence-Informed Practice, and Record Keeping;
- Relevant legislation including but not limited to the Physiotherapy Act, 1991 (in particular the Professional Misconduct regulation) and the Personal Health Information Protection Act, 2004.
Physiotherapists must continue to fulfill these obligations as they use new technologies such as AI.
Guiding Principles
Artificial Intelligence is a valuable tool that can support administrative efficiencies and clinical decision-making, but it does not replace the professional judgment, expertise and compassion of physiotherapists.
To support physiotherapists in fulfilling their obligations, the College has developed the following memory aid to help physiotherapists recall their core obligations as they adopt AI in practice. These principles do not replace but rather support identifying obligations that are set out elsewhere.
Patient trust is central to the provision of high-quality physiotherapy care. To serve patients’ best interests with the adoption of AI in physiotherapy practice, physiotherapists can remember:
TRUST
T – Transparency & Consent
R – Responsibility
U – Understanding Bias & Limitations
S – Security
T – Training
| Principle | Commitment | Actions |
|---|---|---|
| Transparency & Consent | Openly discuss and obtain informed consent from patients for the use of AI in their care. | > Be transparent about how you are using AI > Clearly explain the risks and benefits to the use of AI and how their personal health information will be kept secure > Obtain consent prior to using AI to aid in the delivery of clinical care, including for record keeping > Be ready to use alternate options if a patient is uncomfortable with the use of AI as part of their treatment |
| Responsibility | Physiotherapists are responsible and accountable for the accuracy of any inputs and checking and verifying outputs of AI when used in care. | > Closely monitor the use of AI in practice > Critically assess AI generated output (e.g., notes in a chart, recommendations, etc.) > Exercise sound clinical and professional judgment when assessing or responding to the outputs of AI (e.g., comparing to best practices, questioning suspicious results, etc.) > Approach the integration of AI in a similar manner to supervising delegates (e.g., students or PTAs) where you are ultimately accountable for their involvement in care |
| Understanding Bias & Limitation | Assess and adjust your use of AI when applying to different populations. | > Recognize the bias inherent in AI training data sets > Adjust the outputs of AI to reflect your patient’s unique characteristics or circumstances > Be aware of the potential for improper influence of AI on your professional judgment and patient autonomy |
| Security | Comply with legal requirements to protect your patients’ personal health information. | > Ensure your use of AI complies with the Personal Health Information Protection Act, 2004 > Work closely with AI vendors, and/or employers to understand security protocols and compliance measures of tools used in practice > Ensure patient personal health information is not used for other purposes without informed consent |
| Training | Engage in ongoing professional development to maintain or enhance your competence with AI. | > Understand basic AI principles and limitations > Stay updated on AI advancements > Develop proficiency in creating prompts or directions to AI > Apply AI responsibly to enhance patient care rather than replace the role you play |
For more information on the benefits and risks of AI along with examples of how AI can be integrated into practice, please review the AI Backgrounder below.
Backgrounder for Artificial Intelligence in Physiotherapy
Introduction
Artificial intelligence (AI) is rapidly transforming healthcare, including physiotherapy services. AI encompasses a wide range of applications and tools, making it challenging to define precisely. For the College’s purposes, AI is broadly defined to refer to all forms of artificial intelligence, both traditional and generative, where computer systems are able to simulate or otherwise undertake tasks that are typically associated with humans.
For example, AI systems are able to exhibit behaviours that look like learning, reasoning, decision-making, pattern recognition, and problem-solving.
Practice Potential
AI has the potential to significantly support and transform the way healthcare is delivered, including physiotherapy. Importantly, it has the potential to benefit both providers and patients to support increased efficiency, accuracy, decision-making, and even access.
Some common examples of the way AI is being integrated into practice include:
- Patient Record Development: Automating and enhancing the creation and management of patient records.
- Diagnostic Support: Improving diagnostic accuracy through analysis of clinical data.
- Treatment Planning: Assisting in developing personalized treatment strategies tailored to each patient.
- Patient Monitoring: Tracking progress and alerting physiotherapists to important changes.
- Administrative Tasks: Streamlining scheduling, follow-ups, and operational workflows.
- Research Analysis: Reviewing clinical research datasets to support evidence-based decisions.
AI can be integrated into physiotherapy practice in both visible and invisible ways. Visible uses include AI-supported record keeping and treatment recommendations based on patient information. Invisible uses involve AI working behind the scenes to, for example, optimize patient appointments, staff scheduling, and streamline billing processes, rather than directly supporting the delivery of patient care.
Risks Associated with AI
While AI in healthcare offers many benefits, it also comes with risks that need careful management. Key risks for physiotherapists include:
- Privacy and Cybersecurity
AI systems process patient health information (PHI). There are risks of data being accessed by unauthorized individuals or handled improperly or used for purposes other than the provision of healthcare, which can compromise patient confidentiality. The PHI stored in these systems may be vulnerable to cybersecurity threats or other breaches.
- Patient Trust
Patients see healthcare providers with an understanding that they are accessing the expertise and judgment of a qualified and trained professional. With the emergence of AI patients may not understand and/or trust the use of this emerging technology and its potential role in supporting their care. Additionally, patients may feel deceived if they come to learn that AI was used without their consent and perceive its use as substituting for the healthcare provider they are seeing.
- Bias
All AI systems have some built-in biases because of the data they learn from. A key risk is that physiotherapists may not be aware of these biases and how they can influence the advice given by AI. If the data used to train the AI contains biases related to race, gender, or social status, the AI might make unfair or inaccurate recommendations. - Decision Making and Clinical Judgment
While AI may be able to analyze vast amounts of data or complex datasets, the outputs generated and the process by which those outputs are generated may not be immediately apparent or understandable to humans which creates challenges for validating the accuracy of the information. Additionally, some research shows that AI may not always be correct in its decision-making and may provide false or inaccurate information (often referred to as hallucination in AI research). This creates further uncertainty as it relates to the output created by the AI. - Overreliance and Training
Overdependence on AI without applying sound clinical judgment can pose safety risks to patients as AI may replace the judgment of healthcare professionals rather than enhance their approach to patient care. Additionally, the responses generated by AI are dependent on the nature and quality of the prompts provided by the human user. A poor prompt may lead to poor results, necessitating some skill on the part of the user.
Regulatory Considerations and Implementation
As outlined in the Guidance, the use of AI does not alter the fundamental obligations physiotherapists have. The TRUST principles articulated are intended to remind physiotherapists of the key risks that need to be mitigated and the core actions that need to be taken in order to safely integrate AI into practice.
To put these principles into practice, clinic owners or physiotherapists may wish to develop policies or procedures within their practice.
For example, by providing training or education for clinic staff, developing internal policies regarding patient consent, and engaging in quality assurance processes both individually and collectively to share learnings and reflect on the accuracy of the outputs being generated by the AI tools being used.
Patient Engagement
As the availability of online healthcare resources increased, healthcare practitioners needed to adjust to the reality that patients may arrive at appointments with ideas about their condition based on what they’ve sourced themselves online. The advent of AI and the availability of these tools to patients at home creates new challenges and opportunities for healthcare practitioners. While AI can empower patients and help them navigate healthcare issues they are experiencing, providers will have to be prepared to respectfully navigate and work with patients who have engaged with AI prior to appointments. This may involve providing some education regarding the importance of a hands-on assessment by a trained professional while listening carefully to the patient’s sense of their condition.
At the same time, patient feedback indicates a clear desire and expectation that patients themselves get to decide when and how AI is used in their healthcare. As noted in the Guidance document, transparency about the way AI is being integrated into care and empowering patients to make decisions about AI’s involvement is a cornerstone element of maintaining trust in the provider-patient relationship. Different patients may have different expectations and may assess the specific use differently (e.g., might be less interested in treatment recommendations, but satisfied with AI chat bots for booking appointments).
Conclusion
AI has the potential to transform physiotherapy by reducing administrative burden, supporting physiotherapists in exercising their clinical judgment, and even improving access to care. However, its use comes with significant responsibilities and physiotherapists must ensure they are meeting the ethical, professional, and legal responsibilities that are set out for them. In doing so, physiotherapists will continue to provide high-quality patient care while adopting this transformative new tool.
References
- CPO. (2024, April). The A.I. will see you now: Thinking about the A.I. implications on practice [Blog post].
- Office of the Privacy Commissioner of Canada. (2023, December 7). Principles for responsible, trustworthy, and privacy-protective generative AI technologies.
- College of Physicians and Surgeons of British Columbia. (2024, October). Ethical principles for artificial intelligence in medicine.
- Canadian Medical Protective Association. (2023, May). The emergence of AI in health care.
- College of Physicians and Surgeons of Alberta. (2023, September). A.I. in generated patient record content.
- COTO. (2024, October). What should occupational therapists consider if using artificial intelligence (AI) in practice?
- College of Physicians and Surgeons of Ontario. (2024, June). Advice to the profession: AI scribes in clinical practice.
AI Checklist
| Transparency | Have you discussed the AI tool with your patients? | [ ] |
| Have you obtained informed consent from patients regarding AI use? | [ ] | |
| Responsibility | Have you clearly defined how you plan to integrate AI into your physiotherapy practice? | [ ] |
| Have you consulted with your hospital or facility administration regarding AI implementation? | [ ] | |
| Is the purpose and objective of the AI technology in physiotherapy well understood? | [ ] | |
| Do you believe the AI tool is appropriate for your physiotherapy services? | [ ] | |
| Understanding Bias | Have you reviewed the vendor’s information on the AI’s intended use, performance, and limitations? | [ ] |
| Is the training data used to develop the AI representative of your patient population? | [ ] | |
| Have you verified that the training data reflects your patient demographic and condition? | [ ] | |
| Security and Privacy | Have you reviewed the vendor’s privacy policy and contractual terms? | [ ] |
| Are there adequate privacy safeguards, and is the AI tool compliant with relevant legislation? | [ ] | |
| Have you verified data handling practices, including data retention and storage location? | [ ] | |
| Training and Education | Have you and your team received proper training on how to effectively use the AI tool? | [ ] |
| Is there an ongoing education plan to stay updated on AI advancements and best practices? | [ ] |