Clinical validation of a new algorithm for computerized dosing of vitamin K antagonist therapy: a retrospective simulation study

The number of patients on oral anticoagulant therapy has increased in recent years, and this trend is expected to continue. The increased workload for physicians has led to the development of computerized systems to make organizational workflow more efficient. These programs may include algorithms to propose a weekly dosage and timing for the following visit. Before introducing a new algorithm in clinical practice, its safety and efficacy must be validated. We undertook a retrospective simulation study to test a new algorithm for the TAOnet system. The main outcome was the percentage of concordant and discordant proposals between manual- and algorithm-based prescriptions. Pairs of computerized and physician prescriptions were assessed. They were categorized as 0.1–5, 5.1–10 and >10% if the dose was different, and assigned as “algorithm better” or “manual better” dependent upon the subsequent international normalized ratio value. In 61.0% of cases, the manual and computerized weekly dosage assignments were identical; in 15.3% of cases, the difference was between 0.1 and 5%; in 14.7 of cases, it was between 5.1 and 10%; and in 9.0% of cases, it was >10%. The algorithm did better in 43.9% of discordant pairs, generally due to less frequent under-dosing. In conclusion, the new algorithm proved to consistently overlap with the manual method. The algorithm is useful but must be tested in a multi-center, prospective, interventional study.