Development and implementation of a mathematical model for inpatient warfarin dosing: rationale and methods

Despite a narrow therapeutic range, need for frequent monitoring, and continuous dosing adjustment, vitamin K antagonists like warfarin remain widely prescribed worldwide owing to superiority over newer anticoagulants in key patient populations and low cost. Several models exist to guide warfarin dosing in stable outpatients, but none validated for inpatient warfarin dosing. Instead, clinicians use blanket pre-specified warfarin initiation strategies, and adjust the dose based on clinical gestalt. There is a need for warfarin dosing guidance during the inpatient period to improve time to achieve the therapeutic effect, potentially decrease adverse events related to over- or under-anticoagulation, and reduce hospitalization times. Warfarin response is characterized by inter-individual and inter-temporal variability. At the same time, modern electronic health record infrastructure has made relevant data easily accessible. A simplified pharmacokinetic/pharmacodynamic (PK/PD) model for inpatient warfarin dosing has been previously described, but no implementation has been disseminated. This methods paper is intended to introduce a stepwise derivation and an open-source implementation of this previously described simplified PK/PD model designed for warfarin dosing in the inpatient setting that can be used for subsequent large-scale validation. The model has the potential, after extensive validation and extension, to serve as the core of a clinical decision support system that can accurately provide dosing recommendations to clinicians.