Dental Quality Improvement

Generative Artificial Intelligence applications – Generative AI – are spreading quickly. In Generative AI, a computer application produces outputs based on algorithms. The algorithms are developed from analysis of training data, and the AI learns to produce responses that fit with the patterns in the original data.

AI is not perfect. If the data set on which it is ‘trained’ has biases, these will be replicated in the AI. Out of date or inaccurate information can result in hallucinations, where the AI produces wrong results and offers them with complete certainty. Keeping humans in the loop is essential.

The IHI Lucian Leape Institute, named to honour the work of a patient safety pioneer, recently considered the roles that AI might play in patient safety. An expert group identified three roles that AIs can plausibly have in patient safety in the near future:

• Clinical documentation support.
• Clinical decision support.
• Chatbots that provide patient support.

All of these have potential dental applications related to quality and safety.

1. Documentation support. Writing clinical notes takes time. The Permanente Medical Group have developed an AI application that can record notes on a consultation in real time using a smartphone. It then generates a summary, uploads it to a cloud location, and creates a plain language summary for the patient, including any warnings the clinician included in the discussion. This is far from theoretical: Permanente have reported data from over 300,000 clinical consultations using the technology. The IHI expert group noted that a similar system could also suggest actions from the appointment based on what had been said. The advantages of having a verbatim note of a consultation available without having to take time to type, and a record of what information has been given to patients, has real possibilities in dental treatment.
2. Decision Support. This could include identifying guidelines that appear relevant to the case, and making them available to the dentist during the appointment, without the clinician having to search for them. It could also include identifying patients more at risk of adverse events from a particular intervention by drawing on information in their electronic clinical record, such as age and pre-existing medical conditions. In the longer term, AIs may be able to generate potential diagnoses for the clinician to consider, incorporating information from imaging systems in the practice.
3. Patient Support. Every dental practice spends time, sometimes a lot of time, responding to patient queries. An online chatbot, or a telephone chatbot, that could obtain patient information, respond to straightforward queries, and escalate concerns to a clinician could streamline responses to texts, e-mails and calls and make more time available for patient care. An accurate AI system could also lead to patients that need an urgent dental review being identified, and therefore seen, more quickly. This feels further away, but in one trial medical system, patients reported the AI to be more empathetic than doctors.

There are many challenges in AI, not least the need to check their output. This is more difficult than it sounds. It is easy to see what you expect to see, and to miss an error. If time gained is recycled only into increased activity, there is also a risk of further stress on already busy dental teams.

AI also has wider societal level impacts. There are ethical and legal issues about the datasets on which the AIs are trained. This may be less of an issue in dentistry and other healthcare applications, where there is likely to have been agreement on the use of any dataset. Other problems include the remarkable amount of processing power, and of energy, needed by AI applications. Nvidia became the most valuable company in the world for a reason. The knock on impact on worldwide energy consumption is obvious.

In dentistry, recording and summarising advice given by the clinician feels more achievable than AI assisted clinical decision making. It is worth remembering, however, that the now ubiquitous Chat GPT was only launched at the end of November 2022. The speed of spread of AI applications, and their impact, has been astonishing and the scale of impact on jobs is expected to be enormous. AI is very likely coming to your dental team, and soon. Harnessing its benefits to increase service quality and patient safety, while avoiding its downsides, will become an important consideration for dental teams.

Photo by Google DeepMind