The Ottawa Hospital doctors to try AI for patient notes

The Ottawa Hospital (TOH) is launching a pilot project this June that will use a Microsoft artificial intelligence tool to record conversations between patients and doctors and transcribe them into medical notes.

Hospital chief of staff Dr. Virginia Roth said a select number of doctors in various departments, including the emergency room, have volunteered to take part in the yearlong trial of DAX Copilot.

With the consent of the patient, physicians will use AI to record, transcribe and translate the appointment into a draft medical note for patient's chart. That includes information shared by the patient, the diagnosis and what the care plan is.

The doctors will review the notes and patients will be able to access them.

Roth said this will reduce the 10 hours a week or more doctors spend on paperwork. 

"This really is not why most physicians went to medical school, to enter things into a computer," she said. "We really want to interact with our patients and their families and ensure that they're getting good care."

The hospital has already tested the tool, which Roth said is "surprisingly accurate."

For patients concerned about privacy, she said all recordings are uploaded directly to a secure cloud like any other electronically stored record.

Still, the doctor acknowledged AI is "an evolving tool" with "upgrades that need to be made."

She said the software can have difficulty correctly transcribing accents. It also only recognizes English and Spanish, with Microsoft being a U.S.-based company.

It has yet to include French, but Roth said she's hopeful that will soon change.

AI to assist, not replace doctors

TOH joins a growing number of medical institutions incorporating artificial intelligence.

A former Canada Research Chair for AI and medical imaging said it's helping physicians triage, diagnose patients with more accuracy and recommend treatment plans.

Dr. Alex Wong said AI can be a tool that helps a health-care system dealing with finite resources and increasing doctor burn-out.

"There [are] more and more people and hospitals are not necessarily gaining proportional resources to help with that, he said.

"That's where artificial intelligence can really come in, to help doctors become more efficient."

AI will never replace doctors, said Wong, but it can assist them to help people faster and more accurately.

Bias concerns

The vice-president of data science and advanced analytics at Unity Health Toronto said there are considerable benefits, "but the investment in time and resources to be able to develop and deploy these solutions in a safe, effective and responsible manner should not be underestimated."

"These algorithms, in many cases, they tend to learn as you go," said Dr. Muhammad Mamdani.

That means AI is constantly evaluating and learning from a doctor's behaviour, which can include any conscious or unconscious biases.

"The AI algorithms may learn these biases and what they end up doing is magnifying these biases if they're not assessed carefully," he said.

Medicine is also advancing all the time, so what treatment might be approved today can differ from what could be approved a month from now. 

"How do you keep a track of that and monitor how it's changed over time? It's incredibly challenging," he said.