As the fundamental basis of all computerised tasks, algorithms are increasingly integral to healthcare. They perform functions at every level, from the research laboratory to the clinic; in rudimentary operating system processes to highly advanced analytical systems. Despite their prevalence, the precise definition of an algorithm is increasingly contentious, and this lack of clarity has ramifications for how algorithms in healthcare are developed and regulated.
Determining the ownership of an algorithm, who is liable if it fails, how patient data is handled and the patient’s right to explanation when looking at black box techniques are all major concerns, with unclear answers.
Working to address these issues, the PHG Foundation - a close collaborator with LML - hosted the first of two stakeholder workshops on Tuesday 20th March. 'Regulating algorithms in healthcare - the GDPR and IVDR in practice' was attended by Dr Kathy Liddell (LML Director), Dr Jeff Skopek (LML Deputy Director), Prof. Mateo Aboy, Dr Cristina Crespo, and Dr David Erdos. During this workshop, academics and regulators convened to consider such key issues as:
- Does the GDPR contain a right to explanation?
- Might counterfactual explanation satisfy a right to explanation?
- How does the IVDR (and MDR) intended purpose test for the definition of IVD compare with the FDA’s risk-based strategy?
- Is the intended purpose text flexible enough to adequately regulate mHealth and other algorithms that might be used for human health?
- Consider the difficulties of validation of software under the MDR/IVDR.
- Consider the difficulties of surveillance of software under the MDR/IVDR.
The PHG Foundation will publish a briefing note on 'what is an algorithm' in April 2018, and the next stakeholder workshop ('Regulating algorithms in healthcare - Liability and intellectual property') is scheduled for September 2018.