The risks

Accuracy

Natural language processing can turn your recorded speech into text or predict what you want to say based on commonly used phrases, but its suggestions are not always reliable.

How many times have you looked back at a text to find that your message app has changed what you intended to say? Automatically adding AI-generated text to a record without checking could allow incorrect information to creep in, which could in turn compromise patient care.

Integrity

Documentation tools will typically suggest the same default wording that you can then edit or use non-specific templates. However, the discovery of identical entries in multiple records would lead to questions about whether they can possibly properly reflect the individual consultations.

That could undermine their integrity in the event of a complaint or claim, and even have the potential to give rise to allegations of dishonesty.

Confidentiality

The power of AI lies in its capacity to rapidly learn and improve, but the process of learning from masses of sensitive patient data raises privacy and data protection issues unless the data is anonymised (some AI tools do this).

Dental professionals have an ethical and legal duty to keep patients' information secure, including ensuring that third party providers have appropriate measures in place.

Patient consent

Public perceptions of AI range from enthusiasm to outright suspicion so adopting AI record-keeping tools without being upfront with patients could easily backfire. For example, not obtaining a patient's informed consent for their consultation to be recorded (which is required by some AI software) could prompt a complaint.

Risk management advice

We are not technical experts in the use or implementation of AI - but from a practical and dento-legal standpoint, there are several things you may want to consider before implementing AI record-keeping in your dental practice.

Due diligence

A thorough risk assessment will help you identify and address the specific risks. The newly formed Artificial intelligence (AI) and digital regulations service for health and social care (a collaboration between NICE, the CQC and others) has put together guidance for adopters including a data compliance checklist.

There are also general resources on AI and data protection on the Information Commissioner's Office (ICO) website, including a risk toolkit for organisations. As a data controller, you should also ask developers about their compliance with data protection regulations.

Practice policies

Have a practice AI policy ready, including the name and purpose of AI record-keeping tools, their use, who is responsible for checking the record, data protection and raising concerns. Make sure you also update other policies and documents where necessary to reflect the introduction of AI (for example, your data protection and retention policies).

Training

While AI record-keeping tools can be highly intuitive, it's important to make sure everyone using them has the appropriate training - like how to review and edit notes before adding them to the patient record, and how to correct errors (including the date, time and who made the correction).

Monitoring and audits

Once implemented, there should be a process for monitoring any AI tool to check it is working (and saving time) as intended. For example, you may want to carry out regular audits of AI generated records to make sure they meet GDC standards, particularly that they are individual to that patient and accurately reflect the nature of the appointment.

Informing patients

Your privacy notice should already include information about the types of personal data you hold, reasons for processing it, the lawful basis for doing so (and special category condition in the case of patient health data), when it might be shared and patients' legal rights.

You'll need to think about how the adoption of AI record-keeping should be reflected in your privacy notice. Much will depend on whether it involves sharing identifiable patient data, and we recommend you seek specific advice - the ICO may be a good place to start.

You should also think about how to tell patients that you are adopting AI record-keeping and what this means for them. Explaining what you are doing and why at the outset might prompt objections, but saying nothing is likely to damage trust and lead to more complaints.

Indemnity

An AI system is not a legal entity and cannot be sued or convicted, which means any claim, regulatory action or prosecution would either fall against the developer, the user or both. You should check you have the appropriate indemnity in place to cover potential claims or other complaints arising from AI systems you adopt.