FST JOURNAL

After the talks, the speakers joined a panel to discuss issues raised by the audience. These included: comparative advantage; the UK system, bias, regulation and standardisation.

AI is driven by applying enormous computing power to very large datasets and if the UK has any kind of comparative advantage it lies in our healthcare datasets. These will be of interest to the developers of AI.

The UK has substantial sets of longitudinal health records. It also has a large and very diverse population. However, much of this data is kept in legacy systems which may be difficult to access.

There are practical issues though. The data will already have some bias encoded, which has to be considered at the start of the data collection process. So for example, all the mammograms in the UK will come from women in a certain age range, because that's how we designed the screening programme.

The advantage the UK has is that we have a unified single system, a lifelong system, with an NHS number, patient identifier, and things can be tracked over time. However, the UK has managed to shoot itself in the foot several times over the issue of public trust. We did not engage with the public enough, before trying to implement these initiatives. So then there was a backlash with people opting out, withdrawing their consent for the data to be used for research. So public trust, communication, getting people on side is going to be the key to realising the potential of our health system as a data source.

There are major challenges about both regulation and standardisation for this field. However, the regulators are catching up and making major efforts to reconsider the way that they regard these technologies, how they classify them and how they evaluate them.

What the country does not have yet is the established pathway we have for drugs, for example. There are clear thresholds and standards for evaluating whether the extent to which a new drug works is worth the money that it costs. There is a process. There is no such process for medical technologies, including AI, although there is a lot of work taking place to develop this.

But there will always be early adopters. Standardising the adoption of technology comes later. So there has to be some flexibility where early adoption and implementation are piloted in certain places as testbeds.