Guidance for healthcare and information governance professionals on the use of ambient scribe technology has launched, covering checks to be made before use, consent, transparency, checking for accuracy, the lawful basis for processing of data, record retention and storage, and more.
It notes that healthcare professionals should check whether the AI product they wish to use has been approved by their organisation, and what rules should be followed for its use to ensure legal requirements and regulatory guidelines are adhered to.
It is appropriate to rely on implied consent, rather than having to seek explicit consent from patients for the use of the technology, NHS England states, but professionals should still maintain transparency and share that the tech is in use. If that use is for individual care, patients and service users should be informed at the beginning of the session, it continues, sharing an example script detailing what the tool does and how the data it collects will be used.
The user of the technology is both responsible and accountable for the accuracy of the information added to patient records, according to NHSE, with healthcare professionals encouraged to check information and consider taking additional steps to validate accuracy. “When you use the content produced by an ambient scribe, it is good practice to indicate within the output that it has been produced with the help of an ambient scribe,” it adds.
For information governance professionals, NHSE emphasises the importance of organisations independently assessing tools for compliance with data protection and confidentiality, involving decision makers including CIOs, CCIOs, and SIROs, and establishing a lawful basis for the processing of data by an ambient scribe. For GDPR, the most likely lawful bases are article 6 (1) (e) public task, or article 9 (2) (h) health and care, it outlines, whereas for the common law duty of confidentiality the most likely lawful basis is implied consent.
As well as letting patients and service users know that the technology is in use at the beginning of their session, NHSE recommends updating privacy notices with details on how data is processed, exploring ways to promote understanding of the use of ambient scribes, such as with leaflets or posters in waiting areas, and ensuring transparency materials are appropriate for different audiences.
Once a summary produced by ambient scribe has been signed off, the original recording and transcript can be deleted, NHSE suggests, with the outputs then subject to retention periods in line with their use or eventual storage, such as those for EPR. Deleting records from other storage locations ensures information isn’t duplicated and reduces risk of being retained for longer than necessary, it concludes, noting: “If any data is stored with the provider of the tool at any time, you must give them clear instructions on storage and retention and ensure that they are able to comply with these.”
Information governance professionals must document who the controller and processor is for each processing activity, with the controller to be the relevant health and care organisation where used for individual care, NHSE confirms. This requires health and care organisations to determine the purpose of any processing, give clear instructions to the tool’s provider on data processing, storage, and retention, and ensure responsibilities are clearly documented. If an AI supplier determines a portion of the means or purposes of processing, it could then be considered a joint controller with shared responsibility for compliance with data protection legislation, NHSE outlines.
NHSE also explores the potential for ambient scribes to introduce security vulnerabilities both in their processing of data as well as their integration with systems, highlighting the need for organisations to engage with tech and security specialists to assess the risks prior to using the technology. Guidance on assessing security in AI is offered by the ICO, it notes.
Also of note are individuals’ rights to object to the processing of their data; to receive copies of information held about them including recordings or transcriptions from ambient scribe tools; to rectify inaccurate or incomplete data; to request erasure; or to request that the processing of their data be restricted.
Wider trend: Ambient Voice Technology
NHS England’s Transformation Directorate has published a list of 19 suppliers who have evidenced the criteria required to be part of its self-certified Ambient Voice Technology (AVT) registry – “a national capability to support safe and effective scaling and adoption of AVT across the health and care system”. The 19 suppliers include 33n, Accurx, Anathem, Aprobrium (Lexacom), Beam Up, Corti, Dictate IT, eConsult, HealthOrbit AI, Heidi Health, Lyrebird Health, Microsoft Dragon, Optum (EMIS), Pungo t/a Joy, Scribetech, Tandem, Tortus, T-Pro, and X-On Health.
University Hospitals of Leicester (UHL) and University Hospitals of Northamptonshire (UHN) have awarded a £1.9 million contract to Accurx for the provision of its Ambient Voice Technology solution. The award follows a competitive procurement that saw a total of five tenders evaluated, according to the trusts, seeking to find a supplier capable of implementing and deploying AVT to support both clinical and non-clinical documentation across multiple hospital sites. The solution will be used to capture consultations and draft documents such as clinical notes, summaries, and letters, to be reviewed by clinicians for accuracy before being sent out to patients.
A study exploring informed consent for ambient documentation using generative AI in outpatient care has highlighted nuances including that patients are more likely to self-censor when talking about mental and sexual health or illicit activity during consultations. The study, published in Jama Network Open, was conducted from March to December 2024 in ambulatory practices across specialities in a “large urban academic health centre”, involving 18 clinicians and 103 patients in an operational proof-of-concept. 74.9 percent of patients reported being comfortable or very comfortable with the use of ambient documentation, with this rising to 81.6 percent when provided with basic information about the technology. However, when participants were given information on AI features, data storage, and corporate involvement, this decreased to 55.3 percent.
