Raj Rattan MBE, Global Adviser at MPS, addresses some key questions around the use of AI in dental practice.
______
We know dental professionals are keen to explore and adopt Artificial Intelligence (AI) tools which may enhance patient care and help to facilitate more efficient working. However, we also know that members are increasingly aware of some risks associated with the use of AI, and want to better understand these risks.
Having reviewed the range of queries Dental Protection has received in relation to AI, I have set out some advice on the most common themes.
AI is a fast-evolving field and at Dental Protection we aim to remain at the heart of it – helping to ensure members and their patients reap the potential benefits, and providing advice and support which helps members mitigate any emergent dentolegal risks.
1. When should I consider using AI?
AI can be used to support specific tasks, such as documentation, assessment, diagnosis, measurement, monitoring, or treatment planning, when it is expected to improve decision accuracy and safety. Before implementation, the clinician should be trained in its use to clearly identify the purpose of adopting AI, its role in the workflow, and how it will enhance clinical reasoning, rather than using it by default.
2. How should I select an AI system?
Clinicians should select AI systems based on clinical fit, safety, and compliance with the needs of their specific setting and patient population. The chosen system should:
- Have a clearly defined clinical purpose, such as assessment, diagnostic support, monitoring, or treatment planning.
- Comply with all relevant regulatory, safety, and data security standards.
- Demonstrate evidence of validation, preferably including real-world performance data from populations and settings comparable to the intended clinical environment.
- Be appropriate for the target patient population, with careful consideration of potential bias and the representativeness of the training data.
- Integrate safely into existing clinical workflows, allow human oversight, and enable clinicians to review and override outputs as necessary.
Users should undertake comprehensive training to understand the system’s intended use, input requirements, benefits, and limitations, including when the system should not be used.
3. What types of outputs can AI produce?
AI systems can generate various types of outputs. These include:
- Diagnostic – analysing images or data to suggest possible diagnoses.
- Predictive – estimating risk or likelihood of a future clinical outcome (for example, periodontal deterioration, and caries progression).
- Simulative – generating virtual, 3D previews of potential treatment outcomes (e.g. digital smile design and orthodontic movement). Remember simulations provide predictive insights but do not guarantee outcomes.
- Analytical – processing and summarising data to highlight patterns and anomalies.
- Assistive – organising information, flagging items for review, or providing clinical recommendations.
Understanding the type of AI output is essential for judging how much confidence you can place in its recommendations.
4. How should AI findings be interpreted and explained?
AI outputs must be critically reviewed alongside the full clinical picture, including the patient’s dental and medical history, clinical presentation, examination findings, and symptoms. This information should be compared with AI-generated suggestions before finalising the diagnosis or treatment plan.
Clinicians should understand how the AI system reached its conclusions. This is achieved by:
- Studying the manufacturer's supporting documentation
- Training
- Studying the features of the software
When explaining this to patients, it is important to use patient-friendly language, emphasising that AI tools can analyse data, and highlight areas of concern, but that the dental professional uses this information alongside all the other clinical findings to then makes the treatment decisions.
5. Who makes the final clinical decision?
The final decision in all clinical matters should rest with the clinician, not the AI system. Whilst AI can process data, it cannot assess patient preferences and values in a way that a human can. The clinician should be accountable for accepting, questioning or overriding AI recommendations and must document the rationale for the clinical decisions made.
6. What if the AI output conflicts with clinical findings?
If an AI output conflicts with clinical findings, clinicians should review the quality of the input data, re-evaluate the patient’s history, examination, and supporting evidence, and weigh all factors, as in any clinical decision. The rationale for accepting or overriding the AI’s suggestion must be clearly documented in the clinical record.
AI systems improve through feedback from real-life experience. As part of risk management, significant discrepancies should also be reported through the system’s feedback process to monitor and improve performance. Such discrepancies help developers refine algorithms and datasets.
7. How important is data quality?
Clinicians must ensure that only high-quality, accurate clinical data is entered into the AI system prior to relying on its outputs. It is worth noting that bias can arise when the data used to train an AI system is incomplete, unrepresentative, or skewed, which may lead to outputs that are less accurate for certain patients or populations.
Poor or incomplete data inputs compromise reliability, regardless of the technology's sophistication. This reflects the GIGO principle (garbage in, garbage out). For example, entering outdated or incorrect patient information may cause the AI system to recommend an inappropriate treatment plan. Vigilance regarding data quality will reduce the risk of avoidable harm.
8. What must patients be told about AI?
General Data Protection Regulations (GDPR) requires that there is a lawful basis for processing personal data. Consent is the commonly used lawful basis in dentistry.
All practices should have policies in place regarding the use of AI for any patient interaction. Policies should include how and when patient consent should be obtained, and what information should be included within the consent process. Dental Protection advise that informed consent should be sought from patients before using AI tools that require the sharing of their personal data with a third party. The way in which consent is obtained could involve verbal consent during consultations, or a physical consent form provided prior to consultations etc. Whatever the method, it is important to document the type of consent given within a patient’s record.
The Information Commissioners Office has published guidance on AI, including a section on consent. The guidance suggests that a clinician “…must ensure that consent is freely given, specific, informed and unambiguous, and involves a clear affirmative act on the part of the individuals.”. Additionally, “…for consent to apply, individuals must have a genuine choice about whether you can use their data” and “…for consent to be valid, individuals must also be able to withdraw consent as easily as they gave it.”.
9. What risks of AI should be explained?
Clinicians should communicate limitations to patients, including the risks of false positives and false negatives, as well as potential bias that may influence diagnostic or treatment decisions.
Patients cannot provide valid consent unless they understand how the technology influences their care.
10. Who holds the duty of care when AI is used?
The legal duty of care obligates clinicians to provide treatment with reasonable skill and care, thereby protecting patients from foreseeable harm. Clinicians are likely to remain legally and ethically responsible for all diagnostic and treatment decisions, even when AI is used. This responsibility applies regardless of the type of AI system used – whether it is a ‘black box’ model with limited transparency or a ‘glass box’ model.
The principle of non-delegable duty applies; the duty of care cannot be transferred to algorithms under current legislative frameworks.
At Dental Protection, we appreciate the huge potential that AI may provide in supporting clinical practice. Members using AI as an adjunct to the delivery of care can request assistance when using AI systems provided these are not a fully autonomous systems and a human retains oversight and final decision-making authority.
Further reading and learning
Safer Practice Framework
Dental Protection has launched the AI Safer Practice Framework to help dental professionals integrate AI safely and responsibly into practice. The Framework is made up of two parts - INFORMED and RECORDS - and has been structured around these acronyms to ensure it is practical and memorable. It can be accessed at: AI Safer in Practice
The Learning Hub
Dental Protection members can access the ‘Managing the risks of AI in dentistry’ course on The Learning Hub, as a benefit of membership. The course introduces the AI Safer Practice Framework to help dental professionals integrate AI into practice safely, ethically, and effectively. The interactive course can be accessed any time: https://prism.medicalprotection.org/course/view.php?id=1834