Building AI tools into existing models is futile. It is time to go back to the drawing board, argues Orion Health’s Brad Porter

There are many reasons to doubt the future of our healthcare service. The number of staff vacancies remains a big concern for the NHS, with an estimated 125,572 posts currently unfilled. Emergency departments are filling up, waiting times are increasing to unacceptable levels, and nurses and doctors have taken strike action.

But despite this there is a cause for optimism. Healthcare and healthcare technology is rapidly evolving.

At the various healthcare conferences I attended last year there was one common thread of discussion – AI and the challenges and opportunities we face adopting it.

There are concerns that it’s a passing phase or gimmick. To avoid this, we need to embrace AI in a way that supports the workforce and solves some of the biggest delivery challenges facing the NHS.

The potential of AI is infinite, and this makes it hard to identify how to best harness it. Given its capabilities to sift through vast amounts of data, perhaps the best place to start is its application for analysing healthcare information. We should be using it to identify and predict population health trends, as well as improve individual healthcare outcomes for patients; and this all begins with shared care records.

Defining healthcare data

Today, healthcare data is bigger than we can comprehend. The NHS manages around 200 national data collections and on average, hospitals can produce 50 petabytes of data each year.

However, around 97% of the world’s healthcare data is going unused. England is further along on this journey, with the NHS referral programme and other national initiatives such as the National Record Locator, aiming to join up some of those critical datasets for the purpose of direct care.

But it’s a drop in the ocean. Simply put, there aren’t the resources to deal with the amount of data being produced, and even less resources to use it to its fullest potential.

Not broken, just disorganised

I hear that healthcare is broken everywhere I go but I don’t believe that – instead I like to think it’s not organised in the right way, quite yet.

It’s not functioning as it should, and that’s because the ‘back office’ is messy. There is unstructured data, a lack of interoperability and bits of important healthcare information appearing in ad hoc notes. A recent report titled ‘The clinician of the future’ surveyed 3000 practising doctors and nurses around the world. Almost two thirds reported that they were overwhelmed with the current volume of data. And they also predicted the use of widespread digital health technologies will become a burden to them in the future, without proper data organisation.

Healthcare is not broken. We have information, critical expertise and digital solutions available to us like never before. But the dots need to be connected to organise it in the right way.

We could and should be using AI to identify those patients that need to be sent to hospital before they get unwell. We should be using AI to ease the administrative burden, scheduling care pathways and referrals across the healthcare system. Predicted personalised medicine and patient monitoring should be the norm.

Until we tidy up the back office, we cannot get too giddy about the application of AI. In many cases, we are still in the starting blocks of transformative digital healthcare, and until this is addressed these ideas are just that – ideas.

Return to the drawing board

For the power of AI to be truly realised, the basic solid foundations still need to be established. We must have complete datasets. As an industry, we cannot gloss over the foundations that AI requires and try to fit a square peg into a round hole. Simply trying to build AI tools into existing models will prove impossible, it requires a return to the drawing board.

Take for example Ireland, as it looks to implement its National Shared Care Record in 2025. With no existing system in place, it is in the unique position of being a ‘blank canvas’ and has an opportunity to build an exemplary model that is primed for AI innovation.

A basic record might be cheaper or easier to implement and still provide lots of benefits. But an enhanced SCR, that has health intelligence capabilities could extract and organise data in a way that would mean AI tools could be easily implemented and provide unthinkable insights.

Shared care records enable AI

We need to build trust in AI and to do that, we need complete datasets. Shared care records are an AI enabler, as they provide a complete source of patient information – bringing together hospital visits, patient interactions across community settings, general practice and specialist settings.

AI is not something we should be afraid of. It’s a predictive engine and is completely reliant on the information we feed it. That’s why shared care records need to be elevated as part of the AI discussion. They are integral to laying the foundations of reliable data and will help to build trust, security and integrity across the system.

Technology might be outpacing people when it comes to processing capabilities but rest assured, healthcare workers will not be replaced by AI. We may find that a healthcare system enhanced completely by AI is still far out of reach but there are strides the UK and Ireland can take to unify data and improve healthcare delivery.

Brad PorterBrad Porter is chief executive officer at Orion Health.