Going forward
If the project succeeds in developing a good prediction model, the goal is to test it in a clinical setting in early 2024
The next step is to get the algorithm CE-certified and approved by the Norwegian Medicines Control Authority.
Clinical decision-support tools based on artificial intelligence are considered medical-technical equipment and must be approved for use in a clinical setting. An important distinction between ordinary medical-technical equipment and a decision-support tool based on artificial intelligence, is that the latter must be retrained regularly to prevent reductions in accuracy. The Norwegian Medicines Control Authority does not currently have a precedent for approving medical-technical equipment that requires retraining. Ahus has therefore been granted project funding from Live Science Growth House to explore, in consultation with DNV, the possibilities of CE-certifying an algorithm that requires retraining.
This sandbox project has highlighted a potential risk of the EKG AI discriminating against some patient groups that are currently under-represented in the algorithm’s data source. Ahus will conduct a clinical trial to determine whether the algorithm’s predictions are less accurate for patients with a different ethnic background or has the potential to discriminate on other grounds. The results will indicate whether corrective measures are required.
In the project period, we have discovered that there is no common, baseline method for identifying algorithmic bias. If we had had more time in the project, we would have developed our own method, based on experiences gained in the project period. In addition, it would have been interesting to dive even deeper into the ethical requirements for use of artificial intelligence in the health sector.