At Mt. Sinai Health System in New York, a combination of personalized medicine, natural language processing, and clever integration with electronic health record software is allowing clinicians to adjust medication selection and dosages based on patients' genomic differences.
Omri Gottesman, MD |
Personalized medicine is one of those technology topics that perpetually comes up in conversations about The Next Big Thing.
Think combining genomics data with population health, throw in some predictive analytics, and you've got the basic idea.
As a direct-to-consumer play, personalized medicine has run into some roadblocks, and at least one big setback. See the FDA's takedown of 23andMe's service that tested consumers' genomes and suggested correlations to particular predicted conditions or diagnoses.
But within the controlled environment of a health system, personalized medicine is making inroads.
At Mt. Sinai Health System in New York, a combination of personalized medicine, natural language processing, and clever integration with electronic health record software is allowing clinicians to adjust medication selection and dosages based on patients' genomic differences.
The clever integration, invented at Mt. Sinai, is the Clinical Implementation of Personalized Medicine through Electronic Health Records and Genomics platform, or CLIPMERGE for short.
Heading up development of CLIPMERGE was Omri Gottesman, MD, a UK-trained physician-scientist focused on the translation and implementation of genomic and data-driven medicine into clinical practice.
Gottesman's work focuses on developing and evaluating tools and best-practices that will facilitate the translation and implementation of genomic and data-driven medicine into clinical practice today.
Much of this involves the Institute for Personalized Medicine BioBank, a clinical cohort of more than 30,000 Mount Sinai Hospital patients who have donated their DNA to an electronic health record-linked DNA biorepository. These patients have consented to research on a de-identified image of their medical record and to being re-contacted for clinical trials.
Gottesman's work at the IPM has included clinical data representation and electronic phenotyping—creating case definition algorithms that leverage information from the electronic medical record to automatically assign case and control status for use in Genome Wide Association Studies (GWAS), replication studies and implementation trials.
"We wanted a way to be able to implement this kind of genomic medicine at the point of care, and so we built the CLIPMERGE system," Gottesman told me.
"It sits on top of the electronic health record and communicates with it in real time, and so when a consented patient is filling [a prescription] with a Mt. Sinai physician, if there is something relevant for us to say, then we respond and interject in the clinical workflow by firing clinical decision support which kind of pops up on the physician's screen and gives them advice that's tailored by the patient's data, be it genetic or otherwise."
CLIPMERGE alerts are already able to fire based on three paired relationships between genes and drugs. So far, the drugs are clopidogrel, simvastatin, and warfarin, and, soon, codeine will be added.
"If there is a pharmacogenomic reason why that patient should not receive that medication, then the alert fires on the provider's screen," Gottesman said.
"In the case of simvastatin, there is a genetic variant that predisposes a patient to a higher risk of a side effect of simvastatin called statin-induced myopathy, and if a patient has that genetic variant, and is prescribed certain doses of simvastatin, then an alert flashes on the provider's screen that says, the patient carries this genetic variant, which places them at higher risk of statin-induced myopathy. Consider prescribing a lower dose of simvastatin or a different statin."
Mt. Sinai's early outcomes aren't public yet. In fact, the studies are still accepting patients. The results will be the subject of peer-reviewed publications to start appearing at the end of 2014 or later, Gottesman told me.
There's also a role for another potential breakout technology, natural language processing (NLP), in the Mt. Sinai studies. The reason why this is has to do with the way doctors enter medication information into EHRs.
"What we were finding was that even though 40 mg of simvastatin were prescribed, actually the patient was only being prescribed 20 mg," Gottesman said. "The reason that was, that there was an option in the electronic medical record, if a medication is being re-prescribed, you could re-prescribe the existing dosage and modify it.
"What I mean by that is if a patient is taking say 40 mg of simvastatin, and the provider wanted to halve that dose, they could re-prescribe the simvastatin and put in the comment section, they could put take half a pill at bedtime."
The inverse of this is also true. If a patient was on a 40 mg dosage of simvastatin, and the care provider wanted to double that dose, they could re-prescribe 40 mg and put in the comment section, take two at bedtime.
"So what the CLIPMERGE platform was receiving for a patient that being prescribed 20 mg was 40 mg, so that was incorrect, and if a patient was being prescribed 80 mg, in those situations we were receiving 40 mg, which meant that the genome-informed clinical decision support that we were returning to the EMR was incorrect, because it was based on the wrong information."
The solution was an NLP technology, in this case from Clinithink, that turned the narrative text into something discrete that CLIPMERGE could recognize, to adjust recorded dosages accordingly, Gottesman said.
Where today there are four drug/gene pairs being processed by CLIPMERGE, in the future we will see dozens, hundreds, and eventually thousands. Some patients will balk at signing the consent forms that allow these alerts to fire. Others will welcome the information.
Physicians themselves will also have to adjust to a growing number of such alerts. Over time, it may become the standard of care.
It may be that each large medical institution will eventually have its own tech take on personalized medicine. Over time, expect the best of these ideas to either be licensed widely through the healthcare industry, or else be absorbed into existing EHR and analytics software, so patients everywhere can benefit from these ideas.
For the time being, savvy patients will need to shop around to see which providers are offering what flavors of personalized medicine.
Scott Mace is the former senior technology editor for HealthLeaders Media. He is now the senior editor, custom content at H3.Group.