The emerging role of electronic medical records in pharmacogenomics

Abstract

Health-care information technology and genotyping technology are both advancing rapidly, creating new opportunities for medical and scientific discovery. The convergence of these two technologies is now facilitating genetic association studies of unprecedented size within the context of routine clinical care. As a result, the medical community will soon be presented with a number of novel opportunities to bring functional genomics to the bedside in the area of pharmacotherapy. By linking biological material to comprehensive medical records, large multi-institutional biobanks are now poised to advance the field of pharmacogenomics through three distinct mechanisms: (i) retrospective assessment of previously known findings in a clinical practice-based setting, (ii) discovery of new associations in huge observational cohorts, and (iii) prospective application in a setting capable of providing real-time decision support. This review explores each of these translational mechanisms within a historical framework.

Figure 1. One approach to the construction of a biobank for pharmacogenomic research

Electronic medical records (EMRs) typically contain a combination of unstructured text reports and structured data. Structured data includes most laboratory values, vital signs, and such data as computerized provider order entry (CPOE) records. In addition, administrative billing codes (ICD9, CPT) form valuable components for electronic phenotyping. These data can then be de-identified using algorithms to remove personal health identifiers from text through a combination of statistical and pattern-matching techniques [ref 13]. Finally, de-identified medical records are linked to DNA samples using research unique identifiers, which can be generated using a one-way hash algorithm that prevents discovery of the input number (e.g., a medical record number).

Figure 2. Quantifying drug toxicity. Therapeutic Index (TI) = TD₅₀/ED₅₀

ED₅₀ = dose of a drug observed to yield half-maximal efficacy. TD₅₀ = dose of a drug observed to yield half-maximal toxicity.

Figure 3. Structured and unstructured data generate high-quality phenotypes

Upper Left: Recent advances in Natural Language Processing (NLP) allow extremely accurate reconstruction of comprehensive medication histories. Upper Right: Structured medication data generated by computerized provider order entry software (e.g. name-value pairs, such as “medication = tamoxifen”) can be easier to collate and analyze. However, structured data must be normalized across diverse systems of care. The National Library of Medicine (NLM) has developed a terminology called RxNorm [http://www.nlm.nih.gov/research/umls/rxnorm/], linking drug names (dose, ingredient, and formulation) with drug vocabularies commonly used in pharmacy management systems (e.g., First Databank, Micromedex, MediSpan, MediSpan, Gold Standard Alchemy, Multum). Bottom: Structured and unstructured data can be merged to yield high-quality drug exposure phenotypes that facilitate pharmacogenomic studies using EMRs.

Figure 4. Quantifying drug efficacy within large populations, using EMRs

4A. Dose-response for atorvastatin. LDL cholesterol plotted by dose. 4B. Gender-stratified distribution for E_max within an EMR biobank