Assessment of warfarin algorithms for hospitalized adults: searching for a safe dosing strategy

This study evaluates three warfarin dosing algorithms (Kimmel, Dawson, High Dose ≥ 2.5 mg) for hospitalized older adults. A random selection of 250 patients with overshoots (INR ≥ 5 after 48 h of hospitalization) and 250 patients without overshoots were accessed from a database of 12,107 inpatients ≥ 65 years treated with chronic warfarin during hospitalization between January 1, 2014 and June 30, 2016. Algorithms were retrospectively applied to patients 2 days prior to overshoots in the overshoot group, and 2 days prior to the maximum INR reached after 48 h of hospitalization in the non-overshoot group. Patients were categorized as overdosed or not overdosed and compared using descriptive statistics. Logistic regression modeling determined predictors for overshoots. There was no significant difference between overdose and non-overdose groups for progressing to overshoots by the Kimmel (51.0% vs. 48.7%, p = 0.67) or Dawson (48.5 vs. 57.9%, p = 0.19) algorithms. The Low Dose Group (≤ 2.5 mg) was significantly more likely to experience an overshoot than the High Dose Group (56.6% vs. 45.5%, p = 0.04). The Low Dose Group was more likely to be older (81.4% vs. 71.1%, p = 0.02), female (63.5% vs. 49.8%, p = 0.02), weigh less (71.3 ± 21.9 vs. 79 ± 23.1, p = 0.002), and be prescribed amiodarone (16.6% vs. 8.1%, p = 0.01). While none of the algorithms predicted overshoots in logistic regression modeling, weight over 70 kg and black race remained protective. The High Dose Algorithm revealed that providers appropriately gave lower doses to patients at highest risk for warfarin sensitivity. Future studies are needed to investigate tools for inpatient warfarin dosing in older adults.