A new Lancet viewpoint by IMH Director Angela Woods, with Kirsten Ostther from Rice University and Eivind Engebretsen from the University of Oslo, calls for caution on the rapid adoption of ‘ambient scribes’ in healthcare.
They are a central element in the digitisation of healthcare strategy presented in NHS England’s 10-Year Plan and are also used widely in the US, despite mixed evidence on whether they improve either efficiency or the accuracy of clinical notes.
This new article suggests that their potential impact on patient care, trust, and how patients are understood requires more attention.
Drawing on insights from medical humanities scholarship, the authors argue that ambient scribes should be understood as "narrative technologies".
Rather than simply recording conversations, these systems select, filter, summarise and reformulate clinical encounters, actively shaping how patients are understood, treated and remembered. They create accounts that become part of the patient’s long-term medical record, influencing future clinical communication, treatment decisions and patient care.
The ambient scribe functions more similarly to another clinician in the room – e.g. exercising judgement, making selections, and importing assumptions – than as a stenographer.
The authors warn that the uncritical adoption of AI-scribes risks worsening existing problems in healthcare.
Patients’ stories contain more than reports of symptoms and biomedical facts. Patients communicate their emotions, uncertainty, personal identity, their social and material circumstances, and their own explanations of illness. These elements are often important to healthcare decisions but may be lost in AI-generated records that prioritise standardisation and biomedical relevance.
Existing inequities in healthcare may also be reinforced by AI systems trained on large, opaque datasets that are not representative of the communities they are designed to serve. Discomfort with surveillance, particularly for members of vulnerable groups and those with histories of discrimination, exclusion or misuse of health data, could increase medical mistrust and lead to patients withholding their full-lived experience during AI-monitored clinical consultations.
The authors identify five priorities for policy makers, developers, healthcare organisations and practitioners to ensure that the roll-out of ambient scribes does not exacerbate existing harms but instead helps to contribute to healthcare that is more humane, trustworthy and patient centred.
These include improving AI literacy among clinicians and patients, developing more transparent and representative technologies that are optimised for better health outcomes rather than efficiency, increased revenue and cost savings, placing patient perspectives at the centre of design, implementation and evaluation, ensuring meaningful consent and increased transparency, and conducting long-term research into how AI-generated records shape clinical and patient understandings of their experiences of health and illness over time.
/filters:quality(70)/prod01/prodbucket01/media/durham-university/research-/research-institutes/institute-for-medical-humanities/art-by-lonfeldt-711404-unsplash.jpg)