Concordance of experienced-based maintenance warfarin dosing vs. algorithm-based dosing

Introduction

Warfarin maintenance dosing algorithms improve the time in therapeutic International Normalized Ratio (INR) range (TTR), a surrogate marker for clinical outcomes. Despite demonstrated benefit, many anticoagulation providers utilize experience-based dosing instead. This study assessed rates of concordance between experience-based and algorithm-based warfarin dosing at a single anticoagulation clinic.

Methods

Within University of Utah Health Thrombosis Service, patients on a maintenance dose of warfarin with an INR goal of 2.0–3.0 or 2.5–3.5 and who had INR results during November 2019 were included. Experienced-based approaches for out-of-range INRs were compared to a validated dosing algorithm to determine algorithm concordance rates as well as likelihood that algorithm concordance would return the INR into therapeutic range.

Results

During the one-month study period, there were 1120 out-of-range INRs in 770 patients included in this analysis. Providers’ decisions were 50.5% algorithm-concordant for dosing adjustments and 59.2% concordant for follow-up intervals. Algorithm-concordant dosing practices resulted in a significantly higher likelihood of returning the subsequent INR to the target range (odds ratio [OR] 1.33, 95% confidence interval [CI] 1.05–1.68), whereas algorithm-concordant follow-up intervals did not significantly impact return of INR to therapeutic range (OR 0.79, 95% CI 0.62–1.00). Baseline deviation from INR goal was determined to be significantly different between concordant and discordant study groups. Controlling for the deviation magnitude attenuated the significance of dosing concordance rates on return to INR target range (adjusted OR 1.16, 95% CI 0.91–1.48), while impact of follow-up concordance remained not statistically significant (adjusted OR 0.84, 95% CI 0.66–1.07). No provider characteristics were associated with the likelihood of return to INR goal.

Conclusion

Experience-based dosing was concordant with a validated dosing algorithm only half the time. Algorithm-concordant dosing increased the likelihood of returning the next INR to therapeutic range, though controlling for deviation magnitude attenuated the statistical significance of dosing concordance with return to INR goal rates. These findings support further research regarding implementing strategies that promote the use of a validated dosing algorithm among experienced anticoagulation providers.