PII-052 - PRECISION MEDICINE IN PRACTICE: DEVELOPMENT OF AN AUTOMATED WARFARIN PHARMACOGENOMIC DOSING CALCULATOR EMBEDDED IN THE ELECTRONIC HEALTH RECORD.

K. Hess¹, J. Martin², S. Hansen¹, A. Radwan², A. Hansen¹, Y. Lee², D. Kao², E. Kudron², N. McDaniel², N. Rafaels², C. Swartz¹, A. Tanaka¹, T. Trujillo², S. Vandenberg¹, C. Aquilante²; ¹UCHealth, Aurora, CO, USA, ²University of Colorado Anschutz Medical Campus, Aurora, CO, USA.

Background: Evidence supports the use of pharmacogenetics (PGx) and clinical factors to estimate warfarin doses. However, operationalizing PGx-informed warfarin dosing in the electronic health record (EHR) is challenging, particularly if automated data entry is desired. As a result, many institutions use external websites, where PGx and clinical data are manually entered to calculate warfarin doses. Here, we describe the development of a novel, automated warfarin PGx dosing calculator embedded directly in a health system EHR.

Methods: We preemptively return clinical PGx results to the EHR for patients who have participated in a large population biobank. We return results for CYP2C9, VKORC1, CYP2C cluster, CYP4F2, and other PGx genes. We used portions of the EPIS (Exploration, Preparation, Implementation, Sustainment) process framework to guide the development of the warfarin PGx dosing calculator.

Results: During the Exploration phase, we reviewed validated warfarin PGx equations, CPIC guidelines, and publicly available warfarin PGx calculators. We chose the International Warfarin Pharmacogenetics Consortium (IWPC) algorithm, along with CPIC dose modifications for CYP2C9*5, *6, *8, *11 and rs12777823, for our dosing equation. We selected an interruptive alert to deliver calculator-derived warfarin doses to clinicians at the point of prescribing. During the Preparation phase, we custom-built the calculator using the native EHR functionality of patient score-based rules. The rules evaluated specific variables in the dosing equation and pulled data from discrete EHR fields. Input variables (and EHR sources) included: age/race (demographics); last recorded height/weight (flowsheets); amiodarone/inducer use (active medication lists); and PGx results (variant items). The calculated warfarin dose was output as a return message; SmartLinks within a SmartText were used to display the formatted return message in the alert user interface. We validated the calculator using a cohort of test patients, and then deployed the calculator in the EHR production environment.

Conclusion: We successfully designed and deployed an automated warfarin PGx dosing calculator in the EHR. The calculator overcomes traditional challenges associated with warfarin PGx implementation and advances the utility of PGx-informed warfarin dosing in practice.