Opinions of the UK general public in using artificial intelligence and “opt-out” models of consent in medical research
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Due to its complexity, Artificial Intelligence often requires large, confidential clinical datasets. 20-30% of the general public remain sceptical of Artificial Intelligence in healthcare due to concerns of data security, patient-practitioner communication, and commercialisation of data/models to third parties. A better understanding of public concerns of Artificial Intelligence is therefore needed, especially in the context of stroke research.
Aims
We aimed to evaluate the opinion of patients and the public in acquiring large clinical datasets using an “opt-out” consent model, in order to train an AI-based tool to predict the future risk of stroke from routine healthcare data. This was in the context of our project ABSTRACT, a UK Medical Research Council study which aims to use AI to predict future risk of stroke from routine hospital data.
Methods
Opinions were gathered from those with lived experience of stroke/TIA, caregivers, and the general public through an online survey, semi-structured focus groups, and 1:1 interviews. Participants were asked about their perceived importance of the project, the acceptability of handling deidentified routine healthcare data without explicit consent, and the acceptability of acquiring these data via an opt-out model of consent model by members within and outside of the routine clinical care team.
Results
Of the 83 that participated, 34% of which had a history of stroke/TIA. Nearly all (99%) supported the project’s aims in using AI to predict stroke risk, acquiring data via an opt-out consent model, and the handling of pseudonymized data by members within and outside of the routine clinical care team.
Conclusion
Both the general public and those with lived experience of stroke/TIA are generally supportive of using large, de-identified medical datasets to train AI models for stroke risk prediction under an opt-out consent model, provided the research is transparent, ethically sound, and beneficial to public health.
