Five Years That Transformed Patient Flow - Without Giving Radiologists More Time

A project that was more about the organisation and its users than about the technology

During this period, the researchers at the Norwegian Centre for E-health Research followed the work closely through interviews, observations and ongoing dialogue with staff. A clear finding was that much of the success hinged on everyday factors: good training, clear routines and confidence in using the system. It was also crucial that staff felt the solution fitted into their day-to-day work — both because it was well integrated with other systems and because it enabled better workflow and task allocation.

Project manager Gro-Hilde Severinsen at the Norwegian Centre for E-health Research points out that this is one of the most important lessons.

‘It is rarely the tool itself that determines whether it succeeds. It is about how it is put into use, the trust users have in it, and how the organisation adjusts along the way as it matures,’ she says.

Trust had to be built through local testing

One clear lesson was that the certification and documentation supplied by the vendor were not enough to instil confidence in the AI solution among clinical professionals. The solution had to be tested on their own images, using their own routines and equipment, to ensure the quality was high enough. Local validation therefore became a major part of the work — and for many, a prerequisite for being able to trust the results.

Severinsen describes it as essential reality-checking.

‘If you want the clinic on board, you have to show how this actually behaves in real life. CE marking and product descriptions are not enough,’ she notes.

The report concludes that the aim of saving time for radiologists was not achieved during the period in question. Nevertheless, the accompanying research highlights a range of positive outcomes. In urgent-care pathways, findings could be clarified more quickly, and patients could more often be moved on — either discharged home or referred for further follow-up — without unnecessary waiting.

‘When you focus solely on minutes saved in work processes, it is easy to miss the most important impact. For the patient, reduced waiting time and the reassurance of a clear outcome are often what matters most,’ says senior researcher Line Silsand, a co-author of the Norwegian Centre for E-health Research report.

The classic challenge: the cost in one place, the benefits in another

Another lesson was that much of the work and the costs of implementation largely sat within diagnostic imaging, while the benefits were most clearly felt in the emergency department and by patients. This can create imbalances when it comes to prioritising, funding and making the business case for solutions, and it highlights the need to view costs and benefits across today’s siloed health services.

What have we learned?

The accompanying research shows that implementing commercial AI in radiology is first and foremost about organisational change — not merely the introduction of new technology. It requires planning, local testing, close collaboration and sustained operational follow-up over time in order to realise benefits, as well as an organisation that is able to learn and adjust course along the way.

Perhaps that is precisely the most interesting finding from the accompanying research: not that AI made everything faster, but that even a straightforward, commercial AI solution triggers a wide range of knock-on effects across the organisation. This spans everything from legislation and data protection to workflow, task shifting and organisational change. Implementation also affects collaboration and funding models, competency requirements, ethics, and the need to think in new and innovative ways about what opportunities and value technology introductions can bring to health services.