said Bevey Miner, Executive Vice President, Healthcare Strategy and Policy. consensus cloud solutionIn , we share how extraction artificial intelligence (AI) can help train machine learning models to recognize and extract information from documents and provide confidence scores for structured data.
This approach to extracting data is presented as a solution to address challenges in the healthcare sector and can help organizations comply with evolving standards even when resources are limited, Miner explains. .
transcript
What role will emerging technologies such as AI large-scale language models play in shaping digital health strategies?
Looking at this movement to consume, send, and receive structured data primarily using HL7 FHIR; [Fast Healthcare Interoperability Resources] A standard that is modern HL7 messaging, or X12unstructured data can be thought of as documents because it is a method that has existed forever for sending, receiving, and sharing data.
So when you look at the fax, it becomes a PDF document. Documents like faxes, scanned images, medical settings are full of scanned images. And even his TIFF images where the photos were taken, everything is unstructured. So how can we structure it?
What I'm referring to with structured data is that the various important fields are all available as separate fields. Start with patient demographics: first name, last name, address, insurance number, etc. This is not structured data. Therefore, in order to attach these documents to a patient's medical record, someone must data-enter them.
There's a workforce shortage, burnout is everywhere, and we're asking nurses to sit down and do a lot of data entry before they can start treating patients. This is a major challenge in today's healthcare. I've heard the CIO say, [chief information officers] Those who have been in the industry for 20 years say they have never seen the situation so bad that they don't have the right kind of staff to treat patients, many of whom have to do data entry. not.
So what role does artificial intelligence play? And I want to take a step back a little bit here and talk about the differences between artificial intelligence. There's a lot of talk right now about generative AI. What the generation AI does is generate something that didn't exist before you used the prompt question. Generative AI has many potential applications, and there are many concerns in the industry about the use of Generative AI and Chat GPT.
The type of intelligence I'm talking about is intelligent document extraction.This allows the AI to look at the document and, depending on the context, tell you that John Smith is the name and that this is ICD-10 [International Classification of Diseases, 10th Revision] The code is this diagnosis and indicates that there is handwriting on the prior authorization. emergency The handwriting is written in a recognizable place. It's machine learning trained to understand forms and can extrapolate those forms into structured data. Generative AI incorporates machine learning rather than large-scale language models.
Another important thing to note is that with extraction AI, you always have a source of truth. You can go back to the original documentation, solutions like ours, clear solutions. [Consensus Clarity] indicates the confidence score that it is John Smith. 100% we know it's John, we know it's Smith, we know that's the name. Therefore, you can examine all orphaned fields and check the confidence score.
As a system, we can say, “I want you to estimate the data with a 90% confidence score. Even if the urgency is 85% or 90%, I want you to remove the urgency so I can decide if I need it.” Masu. If humans are going to intervene, we want them to see more of it. ” But you can always return to the source of truth.
With generative AI, there is no source of truth. These large language models are processing large amounts of content, and generative AI has a so-called psychedelic effect, meaning that all that data is actually making up something that hasn't been proven by any source. There are concerns that this may be the case. So the type of AI I'm talking about is extractive AI, and this is how AI is going to move into a structured type of format and provide all kinds of organizations with the ability to meet the new FIHR standards. It's a way you can make a difference.Even if there is no HR, it will appear [human resources] Or money to invest in expensive technology.
