The NHS’s clinical lead for primary care digital delivery in South East London talks from HLTH Europe about what’s going right, what’s going wrong and what’s on the horizon for AI in the NHS.
As an associate medical director and GP with a passion for integrating digital innovations into healthcare, Matea Deliu very much has skin in the game when it comes to the NHS’ continued move towards AI integration.
Part of her role as clinical lead for primary care digital delivery in South East London is to drive the adoption of advanced technology to enhance patient care and streamline services, so she’s perfectly placed to tell Healthcare Today about further government investment into the NHS app, AI accuracy and digital exclusion.
As part of a £50 million upgrade from the government, the NHS app will become the default method to send appointments, screening invitations and other important information to patients. Will this keep on top of demand?
The NHS app intends to be, effectively, a patient portal geared towards patient needs.
In its current format, it’s doing some things right. It’s capturing data, giving access to appointment bookings, prescription requests… things patients didn’t have before. But it’s still relatively clunky. The capabilities aren’t fully realised yet.
In reality, £50 million is probably not enough. The app needs integration with existing tools like the Joy app that already handles community services well, as well as apps from Age UK, VCSE (Voluntary, Community, and Social Enterprise) groups, social prescribing and so on. Right now, NHS app integration is painfully slow. When we piloted AccuRx SMS integration in South-East London, the process took far longer than it should have.
This isn’t necessarily the app developers’ fault, they’re resource-constrained. The question is whether £50 million can build an open platform with which others can integrate if they can’t develop everything in-house. It’s a good step, but we’ll need to prioritise strictly patient-centred improvements.
Does the continued turn towards digital tools increase the risk of exclusion of low-income and elderly groups?
This is the age-old question about digital exclusion. I categorise patients into three groups:
First, the digitally savvy who use technology daily. They’ll adopt AI naturally.
Second, the digitally enabled but not tech-savvy. They have devices but need training. Organisations like the Good Things Foundation are vital here in teaching digital literacy skills.
Then there are the digitally excluded, which splits further: those who consciously opt out – they love their landlines and always will – and those excluded by socioeconomic factors. For the latter, we need practical solutions such as loaner phones and data dongles. But we shouldn’t force tech on those who’ve deliberately avoided it.
The key is using AI to create efficiency gains that free up capacity for traditional access methods. As one clinician told me: “AI helps the tech-enabled group so we can better serve those who need face-to-face care”.
Concerns have been raised about AI accuracy. How do we ensure that all data is reported precisely?
It’s important whenever you implement an AI tool that you should also have a clinical safety officer to guide that implementation.
With every tool, regardless of its complexity, there are risks. A clinical safety officer can be useful for teams as you go through the system that’s being implemented and look at everything that could potentially go wrong.
From the AI misidentifying a patient or not effectively triaging to the right place, pinpointing all the risks that can happen from a patient safety perspective is of vital importance.
Risks happen with human clinicians, right? So, we can never minimise risk, we can only put in place mitigating factors.
Crystal ball time! Where do you see AI and digital transformation taking UK primary care in the next couple of years?
I see it very much integrated as part of our workflow, but I think it’s going to depend on which part of the workflow is primarily needed.
As a clinician, I do not need AI to guide my decision-making, to diagnose or prescribe for me. I’m a doctor at heart. I still want to be able to make those decisions myself.
Where I find AI helps is a lot of the admin burden of workload that I have to go through every single day. There are a lot of back-end processes that are low in terms of risk, but actually high reward. If I can save time by, let’s say, being able to summarise notes from patient records in two seconds, as opposed to trawling through pages and pages and dashboards and dashboards, that’s where I would like to see AI playing a strong part.
