Innovation Category
An AI Assistant for Documentation is an AI-powered voice-enabled tool that executes documentation tasks relieving burden and improving clinical documentation. It uses the power of AI to allow physicians to step away from their EHRs and focus on patient care.
AI Assistants as a category are pure software solutions that do not require “humans in the loop” making them much more affordable, scalable and broadly adoptable. While innovators will use humans to develop and continuously improve their language models, the solution does not require other humans in the documentation workflow significantly lowering costs. This allows AI Assistants to be much more affordable and scalable than Human in the loop (HITL) solutions including scribes, virtual scribes, and tech-enabled human solutions that may have natural language processing and use AI supporting dictated and ambient documentation but have HITL at some point.
Collective Voice
It’s been a breakthrough for my practice.
The anxiety of being rushed has significantly decreased.
I feel more at ease during longer visits because I know I will be able to make up the time in most cases through quicker charting.
The natural flow is to be more engaged with my patients, not just surviving and getting room to room, but actually connecting.
I think it’s going to be more sustainable and more enjoyable for physicians as long as we do not use the margin created to overload our schedules.
​
Characteristics
AI-Powered Foundation
It uses artificial intelligence and the analysis of large amounts of labeled training data, where the system identifies correlations and patterns in medical documentation, to understand context, increase the accuracy of speech recognition, and automate tasks for more streamlined and efficient processes.
​
Voice-Enabled Assistance
Physicians can simply use their voice to interact with the AI-powered assistant, allowing them to step away from their EHRs and the tedious interactions. Whether dictating notes, asking questions, issuing commands, or ambiently listening to the visit, the system understands natural language and assists in documentation.
​
Pure Software: Do not require "human-in-the-loop"
​
AI Assistants as a category are pure software solutions that do not require “humans in the loop” making them much more affordable, scalable and broadly adoptable. While innovators will use humans to develop and continuously improve their language models, the solution does not require other humans in the documentation workflow significantly lowering costs. This allows AI Assistants to be much more affordable and scalable than Human in the loop (HITL) solutions including scribes, virtual scribes, and tech-enabled human solutions that may have natural language processing and use AI supporting dictated and ambient documentation but have HITL at some point.
Assists on Tasks
​
-
Notes: More easily create clinical notes.
-
Diagnosis Code Selection: Selecting accurate diagnosis codes.
-
Patient Data Retrieval: Retrieve patient information from EHRs.
-
Answering Questions: Get quick answers to medical queries.
-
Automating Administrative Tasks
​
Streamlines Documentation
The assistant automates various documentation tasks, from creating clinical notes and selecting diagnosis codes to retrieving patient information, saves time and improves the quality and accuracy of medical records.
​
Mobility
AI Assistants are most powerful when they are mobile ensuring that physicians can document patient care wherever they are and whenever they want. This mobility enhances flexibility and convenience by untethering them from their EHRs. It allows them to document in the hallway, in their cars and while washing dishes. They are most convenient and affordable when available on the physician’s own smartphone.
​
EHR Integration
AI Assistants are most effective the deeper and more seamlessly they are integrated with EHRs. They can read from patient records and write to them in a safe, secure manner that complies with healthcare regulations. The integration is optimized when the AI assistant is able to create notes that are in the native format of the EHR, meaning that all the functionalities or 3rd-party applications that work with notes can be accessed by the user. This allows physicians to have their AI Assistant to “interact” with the EHR while they focus on their patients.
​
Expanding Capabilities
AI Assistants are based on models and systems designed to continuously learn and evolve with additional skills, improved intent detection, and enhanced data presentation. Medicine is dynamic and ever evolving and AI Assistants must also be to remain assistive.
How it works (by GPT4)
An AI assistant for physician documentation can be thought of as a super-smart transcriptionist and secretary rolled into one, but with the added capability to understand medical language. Here's a breakdown:
​
-
Listening to Conversations: Imagine a doctor, Dr. Smith, talking to her patient, Mr. Jones, about his recent headaches. As they speak, the AI is listening, similar to how Siri or Alexa listens when you talk to them. Dr. Smith: "How often have you been having these headaches, Mr. Jones?" Mr. Jones: "Almost every day for the past two weeks."
-
Understanding Medical Context: The AI isn't just passively transcribing. It understands medical terminology and context. So, when Mr. Jones talks about his headaches, the AI knows he's discussing a symptom and not just any regular topic.
-
Transcription and Drafting: After their conversation, Dr. Smith would typically have to write or type up her notes about Mr. Jones' headaches. But with the AI assistant, it's already creating a draft of the documentation based on their conversation. Drafted Note: "Patient reports experiencing headaches daily for the past two weeks."
-
Assistance and Suggestions: If Dr. Smith forgets to ask about something important related to headaches, like whether Mr. Jones is also experiencing dizziness, the AI might prompt her with a reminder or a suggestion.
AI Prompt: "Consider asking about associated symptoms such as dizziness or blurred vision."
-
Review and Edit: Once the consultation is over, Dr. Smith can review the AI-generated notes. She can make edits, add additional details, or delete anything that's not necessary. This saves her a lot of time compared to writing or typing everything from scratch.
-
Storing and Retrieving Information: The AI can also help store these notes in Mr. Jones' electronic health record. Later, if Dr. Smith needs to recall details about Mr. Jones' headaches, she can easily retrieve the information or ask the AI to summarize it for her.
-
Continuous Learning: Over time, as the AI interacts with Dr. Smith more and more, it learns her preferences and habits. Maybe Dr. Smith always likes to ask about medication history. The AI will recognize this pattern and ensure this is always included in the notes.
​
Products
There is a growing number of companies now saying they are providing AI Assistants to physicians either as products integrated on EHR or EHR’s now offering embedded voice powered assistants, “copilots” or “autopilots”. We intend to develop tools to help physicians to evaluate offerings in this category and make a decision informed by the in-practice experience of “physicians like them”. Here is list of companies with products that may be in this category:
​
​
Lab Partner: Suki
The AAFP Innovation Lab assessed the companies in this category to find a partner for a lab. Suki was chosen for several reasons: The Suki solution combines proven voice and AI technologies to provide a representative solution of this new category. Suki's reason for being aligned well with the goals of the lab (i.e., helping physicians primarily care for their patients).
This brand and physician focus has helped drive innovation and adoption. Suki is actively and successfully selling to family medicine and primary care clinicians. The solution is readily adoptable, software only, not requiring any new hardware. Physicians just download the application and sign up. The version of Suki studied is available on iPhone and Android mobile devices and on the web via the Chrome browser. Suki assists best when integrated with EHRs and is compatible with popular EHRs, including athenaOne, Epic, Cerner, Elation, and eClinicalWorks.