I very much enjoyed the presentation by Dr. Greene recently at the University of Colorado’s AB Nexus event on AI, Advanced Computing, Quantum & Human Health.
His take on AI: it is a way to engineer serendipity. Using pharmacogenomics (drug-gene interactions) as an inspiration, he described how information can be introduced into care pathways to engineer serendipitous moments. We can now isolate DNA from a patient’s blood, saliva, or other samples and detect genomic variants. Many of these variants affect an individual patient’s response to certain medications.
The trouble is that while many research biobanks can know a lot about a patient, getting this in front of patients and their physicians, APP’s and other clinicians at just the right moment, is the moment of engineered serendipity that makes a difference in improving care.
A common strategy in pharmacogenomics is testing by free-standing genomic companies that will run similar genomic analyses, and then generate a PDF or printout or even a wallet card and say “Be sure to show this to your doctor the next time you receive a prescription. Who knows, it may affect which drug you should take!” This is often impractical. Can you image a patient bringing her/his wallet card or PDF to every future doctor visit (urgent care, ER, clinic, hospital, virtual visit) on the off chance that their genomic variant and the associated tens to hundreds of affected drugs, is one that she or he will be prescribed in the future?
In our UCHealth system, we have incorporated key genomic variants, generated for greater than 30,000 patients in our biobank project, into our Epic Electronic Health Record. In total, this includes more than 200,000 actionable genetic data points. Now, Just-In-Time, if a prescriber plans to prescribe Plavix (for example), a common anti-platelet medication, to someone with a variant for which it is contraindicated, the EHR will put up an alert: “CAUTION: Pharmacogenomic alert. This patient’s genomic variant means that this medication may not be effective. Consider Ticagrelor instead. Click here to switch.”
Another example of serendipity is our ability to detect a confluence of a hospital patient’s symptoms to spot sepsis. This blood infection is subtle and deadly if not treated quickly enough. We use AI to spot patterns and notify the right person at the right time. As a result, we have been able to save 800 more lives from sepsis and related deteriorations in-hospital.
CMIO’ take? I like the idea that AI, or pharmacogenomics, is not just a tool, it is a way to engineer serendipity.
