The 2014 Healthcare Informatics Innovator Awards: Third Place Winner: Vanderbilt University Medical Center | Healthcare Informatics Magazine | Health IT | Information Technology Skip to content Skip to navigation

The 2014 Healthcare Informatics Innovator Awards: Third Place Winner: Vanderbilt University Medical Center

January 28, 2014
by John DeGaspari
| Reprints
An enterprise-wide implementation of genotyping and clinical decision support is in the early steps toward personalized medicine in a clinical setting

As the gateway to personalized medicine, genetic information offers the possibility of ensuring that patients receive the “right” medications that are tailored to them, according to the individual’s genetic makeup. In Nashville, Tenn., Vanderbilt University Medical Center has established itself as a leader in the discipline of pharmacogenomics that is making this possible by applying genetic information that offers powerful tools to clinicians to help them choose the correct dose for the right drug, and to avoid unwanted side effects from taking certain medications.

The Vanderbilt team has distinguished itself by bringing the exciting potential of personalized medicine to a real-world clinical setting. Its novel Pharmacogenomic Resource for Enhanced Decision in Care and Treatment (PREDICT) project is an enterprise-wide implementation of genotyping and clinical decision support (CDS) designed to prevent adverse reactions to commonly prescribed medications. As noted by the Vanderbilt team, gene variants are increasingly recognized as an important factor behind variability in patients’ drug responses. PREDICT uses advanced electronic health record (EHR) and CDS tools to help clinicians make genome-informed prescribing decisions, averting adverse drug outcomes in patients with high genetic risk. This has been a team effort, in which Vanderbilt’s pharmacists have been a resource to help clinicians make drug decisions as well as understand the meaning of genetic variances in their patients.


Vanderbilt’s interest in personalized medicine—specifically, using DNA information to tailor patient care—began several years ago as a research project, according to Josh F. Peterson, M.D., assistant professor in Vanderbilt’s Department of Medicine and Department of Biomedical Informatics and Erica Bowton, Ph.D., program manager, personalized medicine, at the Vanderbilt Institute for Clinical and Translational Research. That program, called bioVU, was launched in 2009 and led to exciting genetics discoveries, Bowton says, but they weren’t necessarily being applied to clinical care. “One of the goals of PREDICT was to take those exciting genetic findings that we were finding in bioVU and also in the literature outside of Vanderbilt, and actually move that into clinical care,” she says.

PREDICT was launched in 2010 as a quality improvement initiative, and Vanderbilt has since genotyped more than 14,200 patients as part of routine clinical care. Patients are identified as candidates for genotyping if they could benefit from it right away, or if they will probably need genotyping in the next three to five years. With the latter group of patients, PREDICT looks at their diseases to determine the likelihood that the patients will be prescribed medications that are in the PREDICT program, and which can be tailored to the patient’s genotype. If so, the patients’ genotype information is stored.

Vanderbilt University Medical Center’s PREDICT team: Marc Beller, information services consultant, Office of Research Informatics; Jennifer Mitchell, former PREDICT operations and implementation project manager; Josh F. Peterson, M.D., assistant professor, Department of Biomedical Informatics and Department of Medicine; Erica Bowton, Ph.D., program manager, personalized medicine, Vanderbilt Institute for Clinical and Translational Research; and Josh C. Denny, M.D., associate professor, Department of Biomedical Informatics and Department of Medicine

As part of the patient identification process, PREDICT uses a risk prediction algorithm to recommend genotyping only for patients most likely to benefit from genotype-tailored therapies. The prognostic model was developed using historical data from thousands of Vanderbilt patients to predict the likelihood of a patient being prescribed clopidogrel, warfarin, or a statin over a three-year timeframe. Using diagnostic information contained in the EHR, the PREDICT prognostic model presents the clinician with a risk flag during an outpatient visit, indicating that his or her patient is at high risk of being prescribed one of the target medications, and offers the option of ordering the PREDICT test.

Patients who are potential candidates for genotyping are identified in Vanderbilt’s EHR. For patients who are tested, the EHR lists information such as drug interactions, allergies and the medication list. Vanderbilt’s laboratory manages high-volume microarray genotyping to measure 34 genes and 184 single nucleotide polymorphisms (SNPs) that are relevant to drug metabolism, excretion or efficacy. Vanderbilt has developed a sophisticated IT infrastructure to store and interpret the patient’s genotype information.


Peterson explains that one of the major challenges of genomics is that it generates large volumes of data. Some of that information is useful at the present time, and a subset of that information is very useful, he says. That data has to be managed over a long period of time, during which it is stored and its privacy is maintained. “For the clinicians you are serving, you have to pick out the relevant parts and push it to their attention at the right time in their clinical workflow. That is the real informatics challenge of translating genomics to clinical practice,” he says.

Bowton notes that DNA information does not change over time, which has implications for the genetic testing process and the storage of the information. Peterson explains that it makes economic sense to run genetic testing on a patient up front once, and then use the genetic information repeatedly over time as it is need. That’s also the rationale for storing the data and having mechanisms in place to push it to the clinician’s attention when they need it, he says. Once it’s in the patient’s record it’s treated like any other biomedical data, he says.

The genetic data from the lab falls into two groups. One set is actionable data, as determined by Vanderbilt’s Pharmacy and Therapeutics Committee, which is added to the EHR. The other set is sequestered data, which will be stored, and will be added to the EHR when the Pharmacy and Therapeutics Committee determines there is sufficient evidence to make the data actionable. Those decisions are based on guidance and recommendations from the FDA as well as other groups that release guidelines and information on pharmacogenomics, Bowton explains.

A sophisticated IT infrastructure manages the data. When genotype data from the lab is entered into the system, a translational layer converts the genotype into a phenotype, or how the patient’s scores respond to a certain drug based on the patient’s specific genotype. Those phenotypes are distributed to clinical applications through a services layer, and drive clinical decision support within the inpatient computerized physician order entry (CPOE) system and the outpatient e-prescribing system called RXStar. The information also populates a patient-friendly report within Vanderbilt’s online patient portal.

Marc Beller, information services consultant at Vanderbilt’s Office of Research Informatics, notes that many of Vanderbilt’s core clinical systems, including RXStar and the patient portal as well as its middleware, were developed in-house, a fact that has facilitated the integration of the information into the workflow.


So far, Vanderbilt has developed care models for seven drugs—clipidogrel, warfarin, simvastatin, tacrolimus, azathioprine, 6-mercatopurine, and thioguanine. Those drugs were selected because they were some of the best examples in the literature for using genomic data to tailor how the drug is prescribed, according to Peterson. “We are constantly monitoring the evidence to see at what point a drug-genome interaction reaches a threshold, where we think it is ready to be translated,” he says. Among the criteria for selection is replication of results, and whether there is an alternative course of action that can neutralize the risk, he says.

The Vanderbilt leaders maintain that the PREDICT model is unique in the scale and breadth of its implementation. So far, over 370 clinicians have ordered PREDICT testing across dozens of clinics. And the organization’s leaders report that they have identified a genetic risk in 88.3 percent of patients for the seven drugs currently included. In follow-up surveys, clinicians involved in the program have reported that CDS was vital to interpreting and making use of the genetic data.

Vanderbilt works with research consortia to solve informatics and implementation challenges related to pharmacogenomic medicine. “We are highly collaborative,” says Josh C. Denny, M.D., associate professor in the Department of Biomedical Informatics and Department of Medicine. Vanderbilt serves as the coordinating center for one such group, eMERGE, with the role of sharing knowledge about clinical decision support, workflows and other issues with other member healthcare institutions.

While acknowledging that there are still challenges, at least in the intermediate term, in determining how genes affect drug outcomes, Peterson anticipates that hundreds or even thousands of pharmacogenomic associations will be discovered over the next decade. Vanderbilt also notes that the PREDICT program creates prospects for saving money by averting adverse drug events over time. It is a powerful tool for health systems to proactively manage patient risks and translate evidence into practices.

The Health IT Summits gather 250+ healthcare leaders in cities across the U.S. to present important new insights, collaborate on ideas, and to have a little fun - Find a Summit Near You!


See more on

betebet sohbet hattı betebet bahis siteleringsbahis