The dosing of vitamin K antagonists (VKA) is a complex and continuous process that is affected by many different variables. Dosing is traditionally carried out by our anticoagulation doctors and consultants with limited automatic decision support. We are investigating whether an algorithm can help them to tailor the dosage even more closely to the individual patient’s needs.
In 2018, Star-shl entered into a partnership with Pacmed, an Amsterdam-based company specialising in machine learning and developing artificial intelligence for healthcare. In collaboration with medical experts from Star-shl, they have developed a self-learning algorithm to predict the expected INR based on a patient’s history, medication, previous INRs and dosing regimens. This algorithm is capable of analysing millions of these variables. The algorithm is therefore trained to predict an appropriate dose for a specific patient.
In 2019, we reported that we were able to demonstrate that the algorithm can help with proper dosing advice. However, we missed some important variables, which did not always allow for the ‘prediction’ of the INR value. Examples of these variables include the patient’s compliance, diet and interaction with other medications. This created a ‘blind spot’ in the algorithm. Based on these findings, the algorithm was modified in 2020 with the help of Star-shl’s medical expertise.
This modified algorithm will be tested at the end of 2020. We will compare the advice of the Pacmed algorithm with the already available decision aids and the human advice of our anticoagulation doctors and dosage consultants. In this way, we are answering the question of whether data from millions of historical doses can help deliver even better anticoagulation care to our patients.
We expect to complete the study in the first quarter of 2021 and to be able to share our experiences with you then.
The Jan Schueler Foundation is supporting this initiative on the basis of its objective to make new laboratory diagnostics available for primary care.