In a recent webinar, Carta Healthcare CEO Matt Hollingsworth shared how his health tech business is using AI for data mining to mobilize healthcare data to enable organizations to transform the patient experience. The company’s goal is to reduce the burden for healthcare organizations to pull together both structured and unstructured data in a standardized format so all their data can be used efficiently and consistently across the organization—to ultimately improve patient care.
Rachel Ford Hutman, founder of Ford Hutman Media, moderated the discussion.
Hollingsworth highlighted how patients with complex conditions, such as his mother, need to bring binders of their healthcare data to healthcare appointments because their healthcare history is not easily accessible by their physicians. His company seeks to transform the status quo in healthcare data usability with automation using natural language processing.
On the flip side, Hollingsworth noted that although natural language processing and automation are important tools that healthcare institutions are leveraging, each healthcare institution is different in the way they store and process data. This means that clinical expertise from healthcare clinicians is needed to balance the limitations of AI in discerning where the appropriate information is stored. Similarly, AI can supplement humans’ ability to process large amounts of information in a short time.
“Because the data is messy and complicated and the same data can live in multiple places and in different levels of completion, you don’t necessarily get all the data when humans alone are mining the data, because people don’t have an infinite amount of time to read through infinite documentation,” Hollingsworth said. “It’s always possible to miss things.”
“On the other hand, lacking clinical knowledge means that you can end up with noisy data. For instance, problem lists are notoriously inaccurate. They were often accurate when they were first captured, but there’s no one that goes along and sets resolution dates for things and figures out how long the condition was there. So you can’t necessarily rely on that. But that information is present in things like HMPs and progress notes. So again, being able to deal with the fact that the data lives in multiple places, you have to have someone be able to teach the system where to find the high reliability sources of information. If you don’t do that, you’re going to end up with very noisy data that is inaccurate.
The webinar also offers insights on:
- How AI methods such as machine learning and natural language processing can standardize healthcare data for clinical registry submission
- The strengths and limitations of AI
- Why a human+computer approach to data mining yields the best results
- How to make clinical registry data available in real time across hospital systems for patient care, quality initiatives, and other internal projects
To listen to the webinar, please fill out the form below.
Photo: ipopba, Getty Images