Evaluation of Heidi AI Scribing Technology in Improving Efficiency and Documentation Quality in Frailty Services
Abstract
Introduction: Traditional clinical documentation methods were time-consuming and prone to inconsistency within the Frailty Team at City Health Care Partnership (CHCP), affecting productivity across workstreams including Urgent Community Response (UCR) and Comprehensive Geriatric Assessment (CGA). Heidi, an AI scribe developed by Heidi Health UK, generates clinical notes and structured care plans by interpreting patient-clinician discussions, reducing administrative burden and improving note quality.
Methods: York Health Economics Consortium (YHEC) conducted a mixed-methods evaluation of Heidi’s impact on time savings, documentation quality, staff experience, and resource use, adopting a cost-benefit analysis (CBA) approach. The evaluation included pre- and post-implementation staff surveys (n=34), user interviews, clinical audit data across 153 consultations, and analysis of usage metrics from the Heidi platform.
Results: Between 1st July and 31st August, 4,089 notes were created across 3,504 sessions. The average time writing consultation notes reduced from 29.7 minutes without Heidi to 18.3 minutes with Heidi, saving an average of 11.4 minutes per consultation. This reduced administrative time from 38% to 23% of patient management time. Clinical audits demonstrated improvements in documentation quality, including a decrease in spelling errors (from 56% to 18%) and more consistent use of professional language. Staff reported reduced cognitive load and enhanced patient engagement. The CBA showed a positive economic return, with annual time savings of 3,942 minutes valued at £313,484 and a cost-benefit ratio of 5.1 in Year 1.
Conclusions: Heidi has effectively embedded into Frailty Team workflows at CHCP, demonstrating significant time savings, improved note quality, and a positive economic return on investment. The technology frees up clinician capacity enabling Protected Time for Learning (PTL) sessions and reduces cognitive burden, supporting a more sustainable working pattern for staff.
Comments
Very interesting. Thanks for…
Very interesting. Thanks for this work.
Remote assesment
Did you use AVT for any remote consultations on the phone or over video ? We are involved in a proof of concept pilot of AVT in our Digital Care Hub and we are unfortunately coming across a few challenges in relation to accuracy and being able to gain consent.
Yes the AVT picks up both…
Yes the AVT picks up both sides of the consultation over the phone or video link so have been able to use it for all types of consultation. We followed a straightforward consenting process with patients and also other health care professionals, I'm only aware that consent was refused on one occasions. Sometimes hardware lets us down, older computers for instance with not as much processing power, connectivity can be an issue and occasionally causes issues when used as a mobile app however the software tends to work pretty well in most instances.
templates
Are your templates are available to view on heidi?
Excellent work.
Templates
not on the community in Heidi only within the team however I'm happy to share via email if you have an interest.
That would be great if i can…
That would be great if i can see your templates as this something I am interested in, in terms of service improvement.
email is andrewnoble@nhs.net…
email is andrewnoble@nhs.net if you wish to contact me.
Great work!
Interesting work! It will be interesting to see whether this could be translated into clinical practice, and its impact on geriatric practice and patients
AI concerns
Although I accept the potential for time saving, I have concerns about the potentially negative impact on clinical curation and decision-making. This GP eloquently outlines how the initial appeal of transciption led to negative outcomes: https://benngooch.substack.com/p/i-was-an-enthusiastic-early-adopter?ut…
I have read that but…
I have read that but certainly for me I don't agree. I felt the shift from Author to editor was an interesting transition but I find the checking and reviewing cathartic and actually helps me with my formulations and planning. I think you can equate it to moving from free text to structured templates which we seemed to do without thinking about it despite creating more presets and standardising the way notes were structured and taking some of the thought out of note taking. Personally I think it has improved my thinking and planning, I'm not sure there was a "Magic" in the way we wrote notes before, in some cases it was just a desperate dash to get it all down before forgotten and time ran out! So I think for me it has been better, others may experience it differently.
Integrating AI
This research provides great insight into the ways clinicians can adopt AI into clinical practice, showing how AI can improve the work of doctors when we work with it, rather than fearing replacement. Are there any other admin tasks that could integrate AI, for example would it be possible to use during ward rounds for better notes with adaptations?
Yes potentially it can be…
Yes potentially it can be used in all sorts of environments, in home visits etc... We probably need to be careful about just applying it to everything, I'm sure there will be some situations where it adds very little however the scope and potential for AI assistants is huge and very real. We are looking at how it can work in an MDT and creating summary care plans from multiple consultations on the same patient (i.e Physio, pharmacy, social workers, clincian etc,,) Will need careful checking and safeguards but potentially lots of admin removal for the team!
Confidentiality
This is very interesting! Are there any concerns around confidentiality/ sharing of patient's data when using AI scribes like Heidi?
No issues around…
No issues around confidentiality if the scribes are on the NHSE list of approved scribes then they must comply with NHS regulations. Heidi for instance all the data is processed in this country it is a class 1a medical device and has all the correct governance and documentation, your own IT teams will need to check these and create Clinical safety logs that are bespoke to your own use cases but we have SOPs in place so data is not held for more than 24 hours on the Heidi platform and you can set these relating to your organisations policies., Consent is sought from every patient.
Very Interesting!
Very interesting and likely the way forward. I wonder how challenging nhs IT / hardware / internet etc. will limit how widely AI tech will enable us to improve health care.
yes this is an issue, ageing…
yes this is an issue, ageing tech poor processing power poor reception in remote areas all cause problems with this tech. So much so that the companies involved are creating their own bits of hardware to work independently of the platforms to reduce all these issues. it won't be long before the AI actually exists within the device and does not need the internet!!