Technology
variously being used to: l Help prioritise which patients to see l Suggest diagnostic tests that should be organised
l Optimise care pathways (e.g. reduce the number of appointments needed)
l Flag risks and potential diagnoses l Highlight possible prescribing issues / conflicts
Secondary Care While generative AI and AI built into systems to help manage and triage patients are used by over half of the sample, there is also a smaller cohort of consultants using AI for very specific aspects of their practice in secondary care settings: l For assessing stroke patients l For administering local anaesthesia l For delineating organs at risk in radiation therapy planning
The frequency with which doctors use these AI systems varies, with some using it multiple times a day and others using it less routinely. These doctors often give very detailed explanations of the specific function of an AI tool within their practice. Unlike doctors using generative AI and/or AI embedded within primary care patient management systems, they tend to be using AI more to assist complex diagnoses and aid clinical decisions. As one Consultant Anaesthetist commented: “I do a lot of regional anaesthesia …injecting local anaesthetic around peripheral nerves to numb a body part for pain relief, or anaesthesia for surgical procedures and we do that with ultrasound…I have seen, and used…these systems, to help identify the relevant structures on the ultrasound machine screen.” Other clinicians were critical about NHS
adoption of technology in general and were pessimistic about the implementation of AI. One Consultant Radiologist commented that while there will be lots of clever ways that AI can be used in radiology, the general IT situation in the NHS is “so incredibly bad, that the idea that we’re going to spearhead this evolution is beyond laughable. Most of our computers don’t work properly…. we’ve still got stuff running on Windows XP. Even if we did have very good AI tools, I have absolutely no faith that the NHS would be able to implement effectively.”
Data bias Some doctors reference risks linked to bias in the datasets and/or range of datasets that generative AI is basing its outputs on. They believe that AI outputs (such as academic summaries, teaching materials etc.) may
not be fully or appropriately evidenced by AI. References therefore need to be carefully checked to make sure that the AI is not skewed towards a particular set of evidence and that any studies informing the AI output are robust and representative of the full relevant population. They explained that the data is not necessarily false, but that it is not always sufficiently comprehensive. Although rarely flagged spontaneously as a risk in relation to the AI systems they are using themselves, doctors acknowledged that AI for diagnostic and decision support more broadly may also be subject to bias in any data that the system is trained on. Care needs to be taken to ensure that no patient group is disadvantaged. One doctor gave the specific example of image classification tools for melanoma not being sufficiently
Serchem_Press
ad_C_120x168mm_AW.indd 1
04/08/2025 12:20 September 2025 I
www.clinicalservicesjournal.com 65
t
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80
