Validity of criteria used to evaluate fingerstick devices that assess International Normalized Ratio

Background. Investigators commonly rely on unvalidated, mainly arithmetic criteria to predict if point-of-care fingerstick devices that assess International Normalized Ratio (INR) lead to the same warfarin dosing decisions as a standard measure. Methods. Criteria that predict warfarin dosing agreement between 2 INR measurements were evaluated using clinicians' actual dosing decisions as the standard. Bayesian hierarchical modeling was used to rank the criteria by the proportion of correct dosing predictions and the magnitude of difference between actual and predicted dosing agreement. Results. The prediction criteria misclassified dosing agreement for between 19% and 38% of paired INR values (x̄: 27%). The magnitude of misclassification varied inconsistently throughout the INR scale. Conclusion. The unvalidated criteria used to predict warfarin dosing agreement between 2 INR measurements are associated with large error. Warfarin dosing decisions should be measured directly in such assessments.