Name: Erwin Bottinger
Title: Director of Charles R. Bronfman Institute for Personalized Medicine at Mount Sinai Hospital
Experience: Associate professor of medicine and molecular genetics, Albert Einstein College of Medicine;
Attending physician, Montefiore Medical Center and Jacobi Medical Center;
Visiting associate, National Cancer Institute and visiting staff physician in the Kidney Disease Section at the Warren Grant Magnuson Clinical Research Center of the National Institutes of Health
Education: MD, Friedrich-Alexander University School of Medicine in Erlangen, Germany
Mount Sinai Hospital is looking to become one of the first hospitals to make patients' genomic sequence linkable to their electronic medical records for clinical use.
Erwin Bottinger, head of the Institute for Personalized Medicine at Mt. Sinai, is spearheading a pilot project that is designing the necessary infrastructure, including a physician interface, which it plans to begin testing in January and to roll out to patients next spring or summer.
Initially, the team will do targeted sequencing, based on a pharmacogenomics panel, for several thousand patients in the institute's biobank. If successful, it plans to scale up to exome and eventually whole-genome sequencing for these patients.
While the project will become one of the first examples of making patient sequence data accessible through electronic medical records, others are also seeking to do the same. The National Human Genome Research Institute is investing $40 million over four years in its newly created Clinical Sequencing Exploratory Research Projects program. Projects funded under that program will look at, among other things, how genomic information can be integrated into electronic medical records (CSN 12/7/2011).
Additionally, the Faroe Islands are planning a project to offer sequencing to all 50,000 of its residents. The data could be queried by the patient's physician only for a specific question, such as testing carrier status of a particular disease (CSN 11/2/2011).
Recently, Bottinger spoke with Clinical Sequencing News about how he is implementing this infrastructure at Mt. Sinai and how it could impact patient care.
Can you explain your plans to link patient sequence data to electronic medical records and where you are in the project?
We have developed an approach to have sequence data housed in a separate external database and extract the clinically relevant information from the sequence data, which is then ... translated into clinical decision support for patients participating in this study. Sequence data is shared with the patient's provider and the patient as part of a consented research project.
The clinical decision support and underlying supporting genomic evidence is displayed in the electronic medical records through a standardized interface that we are developing. We're currently working on generating the [clinical decision support module] and the interface. We'll have our first testing happen in January and then I think we'll be ready to go live sometime in March.
Can you explain what you mean by a clinical decision support module?
What is not likely to be informative and comprehensible for clinical practitioners is raw sequence information. So the allelic information will need to be explained and presented in the context of what the clinical implications are based on the existing evidence, and what potential actions the physician might take with this information in hand. That's what is generally called clinical decision support, and here we would call it genome-informed decision support.
How many patients do you have enrolled?
In our Institute for Personalized Medicine biobank, we have a total of about 90,000 patients. We plan to re-contact those patients and re-consent them to participate in this study. We don't know yet exactly what our rate of enrollment will be, but we anticipate we'll have 1,000 to 5,000 patients.
For this initial pilot, what kind of sequencing will you be doing — whole-genome, exome, targeted panels?
There will be a pharmacogenomics panel. That's in the works. What we also plan beyond that, in a second phase, is a pilot to use expanded exome sequence, and eventually we would want to use whole-genome sequence. But for that we will need to obviously expand on our data-management capabilities.
In January, we will do testing of our interface, and then we would want to probably go live by mid-spring of next year.
What sequencing platform or platforms will you be using?
We have not committed to exactly what platform will be used. It's still in the works. It will depend on what platforms are established in a CLIA-approved environment at our genomics institute.
How are you testing the interface?
It’s a complicated back and forth between the electronic medical record and the external database that is used to house the genomic data and then used to generate the clinical decision support.
Such a link between a genomics-driven clinical decision support tool and electronic medical records currently does not exist. So we need to create that from scratch.
It will be essentially an IT task, testing that the right information gets to the right record at the right time, to the correct provider. These are standard procedures in clinical informatics that would need to be verified for accuracy.
That will be the testing [that we start in January], and not part of the study involving research subjects. It will be essentially infrastructure testing, and then we anticipate [rolling] that out for clinical implementation studies in late spring, early summer.
You said that genomic information will be housed separately from patients' electronic medical records. What is the importance of keeping them separate? Why not include the patient's genomic information directly with the medical record?
That's a question that needs to be asked of the vendors of electronic medical records, but it gives us flexibility. The reality is that in the health information technology market, vendors produce electronic medical records, and I think there are different factors driving the decisions of developers of electronic medical records to respond to various needs.
If we have control over how we are processing whole-genome sequence information, how we are managing it, and then, as needed, to provide that information through an interface to an electronic medical record, that gives us more flexibility than actually having to work with the developers of electronic medical records on responding to our needs.
When you talk about incorporating whole-genome sequence data such that it actually resides in an electronic medical record, that is not something we envision at this point. We envision that we have whole-genome sequence data in a separate data environment that is linkable with the electronic medical records at the point of care.
Does that also help ensure patient privacy?
That's obviously a key aspect, yes. We still have to assure that, and that's part of the complexity of this interface tool that we need to test carefully in January. The tool has to provide access to those data only to authorized providers and the patient. So we need to make sure that unauthorized individuals do not have access. That's part of the functionality of this interface tool. So from that perspective, I think you could say [that] at this stage we feel more comfortable having the [genomic sequence data] reside in a separate database with a tightly controlled interface.
What have been some of the main hurdles and challenges and what do anticipate will be the main hurdles going forward?
At this point, we just need to take this step to have a serious approach to the clinical implementation of genome-based decision support. This is going to take, obviously, the appropriate infrastructure, what I've just described to you, that needs to be created at our institution, and then it will require the participation of physicians and patients in these studies.
To what extent we will be able to motivate patients and physicians to contribute to this research will be an important factor.
But I think the time is now to really do this and move forward. [It's time to bring] this type of genomic information and decision support to an environment where it can be used in patient-physician conversation through an electronic medical record context.
How will data be returned to the patients? Will only specific questions of the data be asked and interpreted, or would all the genomic data be interpreted?
This is part of the infrastructure that I referred to that has a level of complexity that will require careful testing. For instance, [consider] a pharmacogenomics panel. Say a physician considers prescribing medication … for which we have sequence data of the locus that has been linked to affect either the response to this medication or gives us an indication of risk for adverse event. The physician wanting to order this medication would trigger an alert in the electronic medical record based on the patient's sequence information for the relevant genes.
If there's a high likelihood that the drug will not be effective or there's a high likelihood that the drug has a risk for an adverse event, the physician might want to consider an alternative drug, or an adjusted dose for that medication.
So, it will be a point-of-care approach, where an intent or action by the physician will trigger the information to be provided to inform the physician's decision making. We don't foresee a global sharing of genomic information with the physician and the patient, like a direct-to-consumer mode. We don't foresee that at this point.
How do you think having genomic sequence data linked to electronic medical records will impact patient care?
What we really need is to generate an evidence base that provides us with real data in a quantitative and statistical sense about how physicians take up genomic information. How are they incorporating genome-based decision support into their clinical practice and decision making? How does it affect their prescribing patterns? On the patient side, how does that affect utilization of medication? And in the longer term, over years, how does that affect outcomes?
That's obviously the final goal, to assess how this information affects typical outcomes. Does it reduce adverse events? Does it reduce response failure for certain medications? So, that will need to be assessed in formal clinical research studies because beyond anecdotal information, there's no evidence base of a scientific nature in the clinical care arena.
Do you think that linking genomic information with electronic medical records will become more common in the future?
I do think that certainly this will be very common in clinical practice in the long term. The way I foresee this is [that] there will be the emergence of a new type of software and health information technology tool that is dedicated to managing whole-genome sequence information — to extracting the clinically relevant information for the individual based on the individual's genomic variation and that can then make that information available to the physicians and patients. It can be linked to by various electronic medical records for personal health charts.
I think this will be more likely the scenario that's going to evolve. There will be an emergence of new tools that will provide this kind of service in the clinical care context.
Is there anything else you'd like to add?
I think there's probably one misconception that is widely held. Clinical decision making, in my view, with the exception of some clear cut, well-documented cases based on evidence, will unlikely be based solely on sequence information. The genomic information will be part of an overall spectrum of different data that will be computed into a personal health care advice and recommendation. These data will include other data types, including patient's clinical information, laboratory data, age, sex — all those more traditional factors that determine risk for disease or response to treatment. They will certainly be enhanced by genomic information, but not replaced by genomic information.