Survey finds GP practices embracing AI amid safety, privacy concerns

A survey of nearly 80 GP practices owned by primary health organisation ProCare shows almost all of them are using artificial intelligence, with only 1 percent saying they don't and don't intend to.

Despite the high uptake, barriers named by respondents include safety, privacy and cost.

The survey was run by ProCare, the country's largest primary health organisation (PHO), which owns 131 practices.

When members were asked if any of their staff were using AI in their work, 95 percent said yes. Only 5 percent said no, with 4 percent saying they were considering it, and a little over 1 percent saying they were not using it nor considering it.

Bindi Norwell, ProCare's chief executive, said the number of clinics using AI turned out to be higher than she thought.

"I think we're always going to find those practices and clinicians who just don't, [and] are not interested in using it," she said.

"What we're trying to do is just to make sure that we provide confidence and support and tools for practices to make sure they're using AI in a very safe way."

More than 40 respondents were using free clinical scribes such as Heidi and Nabla, just under 40 were using the paid versions, and a few fewer were using free AI programmes such as Copilot and ChatGPT.

Others were using tools for patient communications, clinical decision support and administration tasks - "All of the things that take them away from patients and can help make life easier for themselves and their patients," Norwell said.

Key concerns are privacy, security and accuracy

More than 50 of the ProCare survey's respondents named privacy and security as one of their biggest concerns, followed by the impact on clinical accuracy.

Following high-profile data hacks of [https://www.rnz.co.nz/news/health/2230243/pace-of-government-digitisation-a-real-cause-for-concern-says-manage-my-health-inquiry

Manage My Health and MediMap], data security was likely front-of-mind for many, Norwell said.

"We're dealing with patient data here at the end of the day, and ... it's very, very important that we protect patient data ... with utmost care, and I guess with privacy and security," she said.

ProCare was providing training to members to help with security, she said.

"We're also looking at the ways that we can help them with doing due diligence on new tools. So really knowing the questions to ask providers about their security levels. What are they investing in? How are they keeping up with the pace of change in AI?"

Cost was also a concern, according to more than 20 percent of respondents.

"We strongly recommend for practices to always be using paid-for versions for the extra security and protection," Norwell said.

Health NZ's guidance to its staff - who are separate to ProCare's practices - is to not use unapproved large language models (LLMs) or generative AI tools for any clinical decisions or for personalised advice to patients.

"The use of free AI tools (e.g. ChatGPT, Claude, Gemini) for clinical purposes is strictly prohibited due to data security, privacy and accountability concerns. You are also not allowed to use AI tools to draft notes and then transcribing it to handwritten or typed notes, even if you anonymise the patient information," the guidance says.

In March, Health NZ said staff had been caught using these free tools to write clinical notes, a move it said could result in formal disciplinary action.

Most expect use to grow substantially

Many respondents considered AI's biggest potential to be reducing time spent on clinical documentation, meaning more time with patients, alongside reduced clinician burnout and workload.

"We're using it for ... speeding up efficiencies and processes, and that ultimately delivers better patient care," Norwell said.

"I do feel now it will go more widespread across practices, rather than just individual clinicians."

She hoped one day it would be able to help predict diseases and outbreaks before they happened - "but we're not quite there yet".