It is well known that approximately 80% of patient data in EHRs consists of unstructured text. The need for automated systems that can actually use this crucial data has never been greater.
Dr. Tim O’Connell is co-founder and CEO of emtelligent (Booth 3185 at HIMSS24), a company offering medical AI platforms that provide actionable insights and verifiable answers. He said the platforms meet stringent requirements for high-quality patient data not only for physicians, but also for payers, life sciences and technology companies.
We interviewed O’Connell to discuss what he calls a next generation of clinical-grade AI, the differences between past and present natural language processing, the introduction of the next generation of his platform at HIMSS24, and examples of actionable insights that are adopted from this new type of clinical grade AI NLP.
Q: You say that a new generation of clinical-grade scalable AI takes a deep learning approach to in-context learning for large language models (LLMs). Please describe this to HIMSS24 participants and why this new generation is important.
A. Clinical data users across the healthcare spectrum are overwhelmed by a huge amount of clinical text, all too often leaving people searching for a needle in a haystack. Although text indexing technology is mature, it only finds what users search for, and without any context.
Medical NLP software can solve some of these problems, but can have a steep learning curve and is not the right solution for use cases that require a higher level of thinking as a post-processing step.
The current generation of large language models has opened the door to new possibilities for NLP to overcome historical challenges. However, the accuracy and hallucination problems with LLMs are well known, precluding the use of these models in clinical settings and potentially putting patients or organizations at risk.
Emtelligent has made it its mission to leverage new innovations to make NLP more secure and accurate at scale. Essentially, we brought AI to medical school.
Here at HIMSS24, emtelligent unveils the next generation of its medical AI platform, emtelliPro+, a collaboration between medical expertise and the best that artificial intelligence has to offer. The solution uses a custom, medically tuned LLM, which produces output that is resistant to hallucinations, and can be used for complex use cases that require higher level cognition to give users the data they need in a format that they can use easily.
Q. What have you noted as the differences between the latest generation of this NLP technology and the new generation you are working with today?
A. As a practicing physician, I understand how important it is for physicians to have accurate, reliable sources of information to make diagnoses and determinations. Previous AI solutions had problems with non-determinism and hallucinations, and had problems reliably referencing source data, to allow proper human assessment of the results.
The next generation of medical AI that comprises the emtelliPro+ platform combines a simple, intuitive interface with emtelligent’s medically-tuned LLM so that both clinical and business users can quickly get the actuarial, clinical and research insights they need about members, patients and population cohorts.
The key is to put the decision in the hands of the expert. The platform provides the information people need to make informed decisions, while also maintaining the auditable, verifiable evidence from the source data. With this approach, the model focuses on answering the question asked. The medical AI platform serves people in their work and supports them in applying their expertise to do their work.
Q: Please provide a few examples of actionable insights derived from this new type of clinical-grade AI NLP.
A. Making all that patient data actionable to create a complete health picture for patients, enrollees, and population cohorts can have a dramatic impact in healthcare and life sciences, even for the most complex use cases.
For example, the emtelliPro+ platform can provide both commercial and government payers with a holistic view of their members’ health and risk that was previously only accessible through manual card review. With an AI-powered conversational interface, it makes critical processes faster and more accurate, including underwriting, Medicare and ACA risk adjustment, and prior authorization approvals.
For pharmaceutical companies and clinical research organizations, this technology dramatically reduces the time and manual effort required to gather patients for clinical trials. Researchers can quickly process millions of patients and apply inclusion and/or exclusion criteria, allowing them to build their panels, monitor real-world evidence, and evaluate therapeutic efficacy.
Healthcare systems, EHRs and other technology platforms can use the platform to summarize patients’ medical histories and instantly gain accurate insights to accelerate care transitions, identify gaps in care, improve coding accuracy and more. This level of AI support helps staff work more efficiently and relieve them from manual tasks.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.