Carbon Health Technologies Inc. today debuted a new software tool that uses OpenAI LP’s GPT-4 model to help physicians create medical charts faster.

Founded in 2015, Carbon Health operates a network of more than 100 primary and urgent care clinics in the U.S. It provides its clinics with access to an internally developed software platform, called Cloud EHR, that can be used to manage patients’ medical charts. The new AI tool Carbon Health detailed today is built into Cloud EHR.

Medical charts consist partly of notes that physicians take during patient visits. Historically, doctors had to write medical notes manually, which can take up a non-trivial portion of their workday. As a result, less time is left for patient visits.

According to Carbon Health, its new AI tool significantly reduces the amount of time necessary for medical charting. The company says that manually creating a medical chart usually requires 16 minutes. Its new AI tool promises to perform the same task in less than four.

After securing patient consent, medical professionals can activate the AI by clicking a single button in Cloud EHR. The software captures and transcribes audio using Amazon Web Services Inc.’s AWS Transcribe Medical service. Then, the tool’s machine learning algorithms combine the transcript with other patient information such as lab results.

Carbon Health says 88% of the text generated by its AI is accepted by medical professionals. That means doctors don’t have to invest a significant amount of effort in making edits, which frees up yet more time for patient visits. Additionally, it says, its AI tool makes it possible to create charts that are two and a half times more detailed than records produced using manual approaches.

The company has already piloted the technology at one of its clinics in San Francisco. According to Carbon Health, the clinic achieved a 30% increase in patient visit volume after adopting the AI tool. 

“It’s about eye contact over iPads,” said Carbon Health co-founder and Chief Executive Officer Eren Bali. “For too long, providers have had to choose between connecting with their patients and taking notes. That will no longer be the case at Carbon Health clinics.”

Carbon Health is not the only company applying large language models in the healthcare sector. In April, Google LLC detailed Med-PaLM 2, an internally-developed AI model designed to answer medical questions. It can generate responses in a natural language format.

During an internal test, Google had Med-PaLM 2 answer a series of questions written in the style of the U.S. Medical Licensing Examination. According to the search giant, the AI achieved a 85% score. At the time of the model’s introduction, Google stated that it was planning to make it available to early customers within a few weeks.

Photo: Carbon Health