THE VANDERBILT UNIVERSITY

An important potential enabling resource for Personalized Medicine is the combination of a DNA repository with Electronic Medical Record (EMR) systems sufficiently robust to provide excellence in clinical care and to serve as resources for analysis of disease susceptibility and therapeutic outcomes across patient populations. The Vanderbilt EMR is a state of the art clinical and research tool (that includes >1.7 million records), and is associated with BioVU, a DNA repository which has been in development for over 3 years; these are the key components of VESPA, the Vanderbilt Electronic Systems for Pharmacogenomic Assessment project proposed here. The BioVU model acquires DNA from discarded blood samples collected from routine patient care, and can link these to de-identified data extracted and readily updated from the EMR. The VESPA project will include a robust EHR system that includes point of care decision support; an extensive pharmacogenetic database linking genotypes to drug response phenotypes; and high-throughput, reliable, and increasingly inexpensive genotyping capabilities that allow dense genomic information to populate an EHR. Establishing platforms and approaches such as the one we propose here is a necessary first step to a genomic future in which these data are routinely incorporated into personalized, predictive, and preemptive healthcare. A fundamental question for progress in this area is whether Electronic Health Record (EHR) systems that are designed to provide excellence in clinical care can also serve as reliable resources for analysis of drug response phenotypes representing a range of therapeutic outcomes across patient populations. Multiple factors conspire to make drawing scientific conclusions from EHR data more challenging than data collected specifically to answer research hypotheses: examples are that sicker subjects tend to be monitored more closely and to receive different medications than healthier ones, or that clinical events are misclassified due to time pressure on clinicians or incentives to maximize reimbursement. Accordingly, we propose in Specific Aim 1 to validate BioVU, the Vanderbilt repository linking DNA to a de-identified EHR image, as a tool for pharmacogenomic research. Formal models are needed to represent the relative contribution and reliability of the classes of information elements present in EHRs, and their effects on the certainty of inferring associations between genotypes and heterogeneously documented clinical phenotypes. Therefore, the goal of Specific Aim 2 is to model the relative contributions of clinical data and genomic data to clinical event prediction.

Clarification of Codes