MetiStream Incorporated and Carilion Clinic today announced a partnership to create an Artificial Intelligence (AI) enabled solution – the Surgical Clinical Reviewer (SCR) – that will find the proverbial needle in a haystack – extracting meaningful insights from unstructured surgical quality data in electronic health records (EHRs).
“Carilion physicians know the ins and outs of reviewing surgical quality data. We are experts in Natural Language Processing (NLP) and AI,” said Chiny Driscoll, CEO and founder of MetiStream. “Together, we’re the right team to address the long-standing challenge of transforming and analyzing massive amounts of clinical text and documents.”
The search for usable insights from unstructured healthcare data within EHRs confounds clinicians across health systems every day. Clinical analysts search for those “needles in a haystack” to help clinicians construct more complete stories about their patients. The power of NLP and AI allows the data on hand to inform those stories better and rapidly improve patient outcomes.
“We understand how vital innovation is to the future of healthcare and improving patient safety and quality,” said Carilion President and Chief Executive Officer Nancy Howell Agee. “We’re excited about MetiStream’s solution and look forward to partnering with them to advance patient care.”
The gold standard for surgical quality is the National Surgical Quality Improvement Program (NSQIP). It was created by the American College of Surgeons and is used by more than 700 hospitals. NSQIP is the nationally validated, risk-adjusted, outcomes-based program that measures and improves surgical care quality. The NSQIP registry and database allow programs to track clinical outcomes.
For many organizations, the collection, analysis and reporting of surgical quality data to support NSQIP guidelines requires significant manual review and abstraction of clinical notes and charts by specially trained reviewers. The process is tedious, time-intensive and costly. It often results in only ~20% of surgical cases being analyzed each year, according to an analysis performed by Carilion.
Also, Carilion’s reviewers can spend, on average, 30 minutes to more than two hours reading and interpreting dense, highly unstructured clinical documentation like anesthesia records, patient histories and physicals, operative notes and more to complete just one NSQIP case submission. They often analyze and interpret over 180 clinical variables per patient. When in free text form requiring manual identification methods, important surgical quality data is often overlooked.
MetiStream and Carilion are collaborating to develop an AI-enabled solution – the Surgical Clinical Reviewer (SCR) – that extracts insights from the unstructured EHR data using NLP and AI. The Surgical Clinical Reviewer will improve the case review process and facilitate decision-making for leaders and clinicians who strive to improve surgical quality and patient care.
“Sometimes identifying quality improvements in a surgical setting is straightforward,” said Dr. Michael Nussbaum, chair of surgery for Carilion. “Most of the time, though, finding those areas for improvement means combing through an immense amount of unstructured physician notes. With MetiStream, we’re finding a way to analyze more data faster. That will enable us to iterate faster to improve outcomes for our patients.”
Carilion’s Surgical Quality Team has significant experience extracting quality data from patient charts and in-depth knowledge of each variable. MetiStream will build the application using its Healthcare Analytics Platform for Unstructured Data called Ember, powered by NLP and AI. The companies’ strategic collaboration will create a solution that replicates a Surgical Clinical Reviewer’s process and workflow to increase efficiency, accuracy and the number of quality variables to be analyzed.
“Our vision is to provide a virtual reviewer that analyzes clinical data and identifies the critical quality variables for each patient from the clinical notes data,” said Driscoll. “When complete, the solution will leverage expanded and enriched surgical data sets, enhance existing predictive analytics models, and drive the rapid development of new predictive analytics models across pre-operative, perioperative, and post-operative clinical scenarios – all to improve patient care.”
Already in Action
Carilion Clinic uses MetiStream’s NLP technology to track adenoma detection rates for our endoscopists. “We have been delighted with this model and the performance of the Ember platform to assist us with our overall quality of work,” stated Dr. Steve Morgan, Chief Medical Information Officer for Carilion. “We feel the platform will allow Carilion Clinic to expand into other future projects using NLP and AI to improve patient care.”
About Carilion Clinic
Carilion Clinic is a not-for-profit healthcare organization serving more than one million people in Virginia’s Blue Ridge and Southwest Virginia regions. Headquartered in Roanoke, Carilion’s comprehensive hospital network, primary and specialty physician practices, and other complementary services deliver high-quality, patient-centered care close to home. Carilion’s enduring commitment to the health of our communities has advanced over the last decade. Starting with the transformation to the clinic model of a physician-led, integrated health care system, Carilion has developed a robust partnership with Virginia Tech by creating the Virginia Tech Carilion School of Medicine and the Fralin Biomedical Research Institute at VTC. Carilion advances care through education and research elements to improve the health of the communities we serve.
MetiStream tries to solve the challenging unstructured data problem in the healthcare industry. Their mission is to help organizations access and gain value from all of their assets to develop a deeper understanding of the patient population to improve quality and outcomes. their Ember platform is a comprehensive healthcare analytics platform to transform unstructured data into clinical facts and evidence to deliver patient and population-level insights.