As the shift to value-based care accelerates, healthcare organizations are increasingly turning to artificial intelligence to help them improve patient outcomes and enhance provider efficiency.
During a May 4 webinar hosted by Becker's Hospital Review and sponsored by Apixio, industry experts discussed the value of a patient phenotype, and AI's role in building out a phenotype program.
The panelists were:
- Terry Ward, senior vice president of solutions at Apixio
- Bryan Lee, vice president of solutions at Apixio
Five takeaways:
1. Outside of patient phenotyping, healthcare information isn't easy to interpret. "Information is key, but it's really diverse and it's scattered across the healthcare ecosystem," Mr. Ward said. "Much of this data and the clinical value that it holds is still locked within that unstructured data," such as medical chart data that's locked within a PDF, for example. "We estimate about 80 percent of clinical data is still locked as unstructured," he said. Patient phenotyping solves for that, allowing providers to draw insights from otherwise disparate data.
2. Patient phenotypes offer a complete longitudinal picture of an individual's health. It's an evolving profile that encomasses everything from traditional information like medical conditions, lab results, medications, past procedures, to newer data such as social determinants of health. "Those are unique characteristics and access to some data that we've never had before," Mr. Ward said. "Social determinants of health have now become part of our infrastructure."
3. AI streamlines the process of both generating a patient phenotype and drawing out meaning for providers. Collecting all of the data required to power a patient phenotype, such as EHRs, lab results and prescriptions, is a demanding task, with AI dramatically simplifying the initial generation, the panelists explained. Once it's created, "it's then [about] applying the learnings that your computers have to be able to nurture out an insight," Mr. Lee said. It's about "looking at the specific characteristics of that individual member … and then driving said actions based upon the characteristics."
4. Evidence-based actions with patient phenotyping help avoid provider fatigue. "By basing [an action] on evidence, which is what a patient phenotype is going to do, it's going to drive the 'why' behind it," Mr. Lee said, adding that most providers don't just take the word of an alert without evidence.
5. Allow for the input of feedback back into the system. It's important for the system to know what action the provider takes at the point of care, Mr. Lee explained. "That's going to help inform and improve this system." If a patient phenotype shows a person has a reaction to a certain drug, for example, "we want to make sure we're not pushing a certain drug to that provider on a regular basis and incorporate that into the machine learning process … so that the machine continues to learn and alleviate that potential error or abrasion that might occur at the point of care."
To watch the full webinar, click here.