Machine Learning checks quality of GPMRI referrals

NEWS - eHealthNews.nz editor Rebecca McBeth

ProCare is using machine learning to check the quality of referrals for MRI scans being made by northern region GPs under a new national ACC programme.

The service allows GPs who have received MRI training to directly refer patients for a scan if they have knee, lumbar, or upper spine injures, rather than having to refer them to a specialist to make that decision.

The GPMRI programme was piloted with ProCare and Mercy Radiology from 2017 and has now been expanded nationwide with GPs in the upper North Island using ProCare’s inhouse built electronical referral platform Profusion.

Paul Roseman, ProCare’s general manager strategic development, says the electronic referrals system used by ProCare has allowed them to link the MRI report with the GP’s referral and apply machine learning technology to help determine the quality of the referral.

You’ve read this article for free, but good journalism takes time and resource to produce. Please consider supporting eHealthNews by becoming a member of HiNZ, for just $17 a month.

“This quality assurance is really important feature of this project because from ACC’s perspective they are taking a risk allowing GPs to refer for MRI,” he explains.

“Overseas evidence showed GPs are not always good users of MRI so we wanted to demonstrate that GPs are very good users of MRI and we believe that’s what we have done.”

The machine is now able to correctly identify around 70 percent of referrals as good quality. Of those that are not identified by the machine, half are determined to be consistent with the guidelines by a manual check. This means around 85 percent of GPMRI referrals are appropriate.

In the early phase of the project all referrals and reports were checked manually by Stephen Kara, associate clinical director for the GPMRI project, who still manually checks all referrals that cannot be identified as ‘good’ by the machine.

In these cases, the machine may not have had enough information or sometimes the system turns up really good uses of MRI that are not part of the guidelines.

“Then we contemplate changing the guidelines to acknowledge that circumstance and that’s where we get decision support around pathway building of our forms,” says Kara.

He also does random checks on referrals the machine has identified as good and says it is performing at around 98 percent accuracy.

Altis Consulting developed the machine learning programme with ProCare. Alex Gray, Altis regional manager, says this involved taking what Kara was doing via his manual checks and teaching the machine to look for the same words.

The machine was piloted in 2020 before going live around mid-2021 for the whole northern region.

“As we get more results those feed back in and the machine algorithms learn and develop over time, so it is getting more accurate,” explains Gray.

He emphasises that the machine is not making any clinical decisions, but rather is used to ensure referrals are consistent with the guidelines provided to GPs.

Kara gives individual feedback and advice to GPs where referrals do not follow the guidelines and says GPs are consistently improving their use of MRI.

“It’s an instructive process, rather than waving a big stick at them,” he says.

“It’s reduced my workload significantly: we could never have scaled up to cover the whole northern region without the use of this technology.”

To comment on or discuss this news story, go to the eHealthNews category on the HiNZ eHealth Forum

Read more Analytics news

Return to eHealthNews.nz home page