Not All AI Is Generative: Finding ROI Faster—and with Less Risk
Generative artificial intelligence (Gen AI) holds strong potential to help unlock $1 trillion in healthcare savings, and the excitement for use cases in healthcare is palpable. However, Gen AI should be used with caution when it comes to making a diagnosis or creating a care plan, according to Jesse M. Ehrenfeld, MD, MPH, president of the American Medical Association (AMA).
Although probabilistic algorithms can be useful for diagnosing a “textbook” patient or a very narrow clinical question, they lack the judgment, nuance and critical thinking that a clinician brings to the diagnostic process, Ehrenfeld explains during an AMA podcast.
That is the reason the latest HTI-1 rule from ONC requires vendors to disclose full transparency regarding AI used within an EHR workflow. Information such as, what is the source of the data, how old is the algorithm, and is there a measure for bias. Much discussion has been centered around how this rule will be applied within a care setting for a provider to make a decision while they are delivering care.
AI in healthcare is also a key topic within the Biden administration’s executive order to create a bi-partisan task force on safeguards around AI.
While both policy and technology are being played out with generative AI models, there is a solution with an immediate return on investment today. One of AI’s biggest potentials for clinical care impact lies in intelligent data extraction—including from handwritten notes sent by fax—that puts critical structured data directly into clinicians’ workflows at the point of care that would otherwise need to be manually entered. It’s an innovation that offers tremendous value for clinical care at a time when 80% of healthcare data is unstructured, and many transitional points in care, like post-acute facilities, and providers that treat the underserved in supportive care settings don’t have access to an electronic health record.
Despite all the buzz about AI, many health systems are too resource-strapped to invest in potentially pricey AI solutions, and other care sites never received funding to implement EHRs, creating a system of healthcare technology haves and have-nots. More often than not, these health tech have-nots still exchange medical information by fax.
However, there are ways in which health systems can get immediate value from AI with minimal investment.
Leveraging Intelligent Data Extraction for Faster Clinical Care
In an industry rife with unstructured data, there’s still a significant need to enable communications between disparate care settings in effective ways.
Enter intelligent data extraction, an AI-based approach that transforms unstructured documents, such as faxes, scans, clinical notes, diagnostic images, medical charts and recordings, into structured, actionable data. Using natural language processing (NLP) and machine learning (ML), intelligent data extraction unveils valuable, insightful information that can help accelerate patient treatment across the continuum of care.
Unlike generative AI, which is designed to generate content based on statistically likely combinations, intelligent data extraction is about taking existing data from a document or another source of truth; extracting and structuring the information so that it’s useful faster, and more easily processable and sharable.
While a predominant data sharing standard makes sense, and the impetus to effectively share data between facilities with disparate technological capabilities is increasing as the healthcare industry drives adoption of the Fast Healthcare Interoperability Resources (FHIR) standard for exchanging healthcare information electronically—a focal point of the information-blocking provisions of the 21st Century Cures Act – it cannot be supported in all care settings.
No matter where you are along the healthcare continuum, at some point, most healthcare providers are going to need to communicate with a facility that’s a digital have-not. To do so, you’ll need a pragmatic solution for extracting and exchanging data.
With intelligent data extraction, providers can use AI in a variety of ways that provide a quick ROI. Once data is extracted it can be shared using HL7, FHIR, X12, and direct secure messaging to those endpoints that can consume a more advanced data structure.
Practical AI Applications in Healthcare
Although extracting information from documents may not be the most flashy application of AI, it is a real-world application with a lot of potential to reduce errors and costs. For example, 40% fewer hours are needed to process routine paperwork when even the most rudimentary AI-based extraction techniques are implemented.
By integrating digital cloud faxing with new technologies like NLP and ML, we can harness the power of intelligent data extraction to unlock vital information. That information can then be exchanged in an accepted format that both enables interoperability and promotes better health outcomes.
Non-Generative AI-powered tools also can help speed up provider referrals, expedite prior authorizations, improve treatment times and help eliminate congested workflows that bog down daily operations. By upgrading from obsolete paper fax machines to a digital fax platform and adding new AI technology in an integrated environment, post-acute care facilities can still meet interoperability standards while helping to streamline the communication process. It is an affordable approach to drive efficiency as well as improve satisfaction among patients and staff.
