AI (Artificial Intelligence) Scribe has recently attracted the attention of several industries that are looking for ways to reduce the cost of documentation in their workflow, especially in Canada and the United States. Throughout the healthcare sector, from hospitals to small clinics, and in law firms, technologies are being developed that will enable them to transcribe stakeholders' conversations and extract meaningful information from them. Our focus in this article is on the application of AI Scribe in the healthcare and medical industry, its pros, and cons, as well as the potential future impact of AI Scribe on the healthcare industry.
AI Scribe, also known as Ambient Scribe, is a technology that converts a dialogue between different parties into text and uses machine learning and natural language processing to extract key data from that text. AI scribes are often used in healthcare to transcribe physician-patient conversations and to reduce the burden of documentation, which is one of the main causes of physician burnout today.
The AI Scribe or Ambient Scribe can be integrated with a variety of natural language processing modules, including text summarization and named entity recognition, to automate a variety of tasks, including:
Transcribing the physicians' and patients' conversations from voice to text.
Identifying and recording the symptoms and conditions of patients in the electronic medical record (EMR).
Automating the entry of patients' medication lists and dosages into the electronic medical record.
Adding prescribed medications to the electronic medical record of physicians.
Generating billing codes based on the conversations and the physician's diagnosis.
Creating a summary of the entire conversation, converting the transcription, and documenting the encounter.
With AI Scribe, physicians could fully engage with their patients, have a face-to-face conversation, reduce distractions and context switches with computers, and thus reduce their burnout from excessive documentation.
The short answer is no, not yet. AI or Ambient Scribe is unfeasible to implement in healthcare settings due to a few drawbacks:
The accuracy of physician dictation is very high, whereas transcription of physician-patient conversations can be very challenging. In general, informal discussions, such as those between a patient and her physician, do not follow a well-structured linguistic format. Filler words, incorrect logic, pauses, repetitions, interruptions, and simultaneous speaking can occur during a conversation.
Currently, the accuracy of AI and machine learning is insufficient to accurately label all conversations. Many patients are unable to recall accurate information, make mistakes, or ask irrelevant questions in response to the physician’s response. In this situation, data tagging becomes an error-prone process.
A medical student often validates AI Scribe's output as a result of its low accuracy, requiring a human scribe.
Due to the manual process involved, AI Scribe becomes an expensive solution, especially for family physicians and small clinics.
A number of people are opposed to the use of AI scribes, believing that physician manual documentation can help them structure their thoughts and think critically, which will lead to the achievement of better clinical outcomes.
AI Scribe licenses cost roughly $1500 CAD to $2500 CAD in Canada. As previously mentioned, since a fully automated AI Scribe does not exist at this time due to the low accuracy of the technology, a human scribe is required to be hired to evaluate the output of the AI.
As of right now, AI Scribe does not provide any diagnosis or recommendations based on the conversation.
Only a few companies have begun working on AI Scribe, and none have been able to fully automate the entire process. In the United States, DeepScribe and Nuance DAX are the two major players.
The AI Scribe offers a number of promises that can significantly reduce physician burnout in the healthcare industry. In Canada, at this time, a family physician or even larger clinics or hospitals cannot afford to implement the hybrid scribe systems due to the low accuracy and high operating costs. There have been some attempts in recent years to implement pilots of AI Scribe in the US, however, the outcome is unknown.
At Tali AI, we approach the documentation problem from a different perspective. Our AI medical dictation system captures training data for our research and future roadmap, but our focus is on providing a voice-enabled virtual assistant that only interacts with physicians and automates a lot of their day-to-day tasks.