Contributors: Dr Sarah Gebauer, Dr Ellie Asgari, Dr Justin Graham, Dr Keith Grimes, Dr Janko Hrvoj, Adam Khimji, Mr Leon Lindsey, Dr Emma Mi, Arun Nadarasa, Dr Gagandeep Sachdeva, Dr Denis Vincent, Jasmine Balloch
Executive Summary
This white paper explores the evolving landscape of clinical documentation, contrasting the traditional use of human scribes with the innovative integration of digital scribes that harness artificial intelligence (AI). It aims to provide healthcare professionals, policymakers, and technology developers with a comprehensive understanding of the benefits and challenges associated with both human and AI scribes, drawing on current usage, clinician experiences, and emerging trends in clinical documentation.
Introduction
In the high-pressure environment of healthcare, efficient and accurate documentation is paramount. Clinicians have traditionally borne the bulk of the burden of writing notes and documenting patient visits. However, the advent of the AI scribe promises further advancements in efficiency and could transform clinical documentation practices. On February 12, 2024, TORTUS gathered clinicians from the UK and the US to discuss their experiences with approaches to AI clinical documentation. This paper examines the use of human scribes, the potential of digital scribes, and the considerations for integrating AI into healthcare environments.
The use of human scribes has emerged as a strategic solution to alleviate the administrative burden on clinicians, especially in the United States. Approximately 12-20% of doctors across the US, particularly in high-demand areas such as emergency departments, orthopaedic surgery, and primary care, have integrated human scribes into their practice. The financial investment in scribes ranges from $33,000 to $50,000 annually, with an initial training cost of around $6,500. This investment can save clinicians an average of 2 to 3 hours of work per day. It also enhances note detail that subsequently increases reimbursement by about 20%, and significantly improves satisfaction levels among both patients and doctors.
Studies have further underscored the impact of scribes on clinical practice, demonstrating a marked reduction in clinician documentation time – on average, by 3 minutes per patient and 36 minutes per session. This shift not only mitigates the disproportionate time spent on electronic health records (EHRs), estimated at two hours of documentation for every hour of patient care, but also refocuses the clinician’s role towards direct patient interaction and clinical decision-making. This ultimately helps to enhance the overall quality of healthcare delivery.
The status quo – when, where, and why are human scribes used?
This being said, human scribes are expensive. UK clinicians in attendance reported that they have not used human scribes before, except for one clinician whose secretary served that function before the electronic medical record was introduced. Another clinician had previously used Google Glass which streamed to a remote scribe. Most clinicians reported either manually typing or dictating their notes, however, some mentioned that doctors in training often informally took on the role of human scribes by noting conversations between patients and senior doctors during visits. While some clinicians felt this could be a valuable learning experience, they also questioned whether the high volume of scribing that doctors in training do is entirely necessary.
US doctors, who had used human scribes, noted that advantages included a person who could fill in for other roles in the office if needed, such as rooming patients or acting as chaperon or assistant for physical exams. They also described increased efficiency and clinician satisfaction. It was also mentioned that over time, human scribes have the ability to learn the clinician’s style and adjust their documentation accordingly, similar to how a surgical assistant learns how to work with each particular surgeon.
Clinicians had several concerns about human scribes including multilingualism, the ability to attract and retain high-quality scribes, and practical issues like small exam rooms not built to accommodate an additional person. Several clinicians mentioned the length of time required to train human scribes to perform the role efficiently and to the doctor’s standard. Issues of privacy and whether patients would be as forthcoming about personal issues with an additional person in the room were also raised.
What do clinicians want from a digital scribe?
Clinicians were generally interested in using digital scribes and stated that they want these technologies to be accurate and easy to use. They noted that ambient scribes only solve a fraction of the overall documentation burden which includes filling out forms, writing orders, and sending referrals. AI for clinical documentation was seen as being advantageous for patients and/or clinicians who speak languages other than English, their evolving potential to pick up medical diagnoses from voice, and the absence of turnover. It was mentioned that digital scribes could potentially act as technical support and experts in the use of EHRs.
There was an expectation that, by using digital scribes regularly, they would become personalised to an individual clinician’s needs and style thus improving performance. There was a suggestion that digital scribes are a complementary tool that clinicians could use to enrich their documentation, for example, to remind them of details they may have forgotten during the consultation. Interestingly, some patients had provided positive feedback on the output from a digital scribe – they felt their history was more accurately captured and reflected.
Downsides of AI clinical documentation were noted to be lack of ability to pick up on affect or emotional tones that might be especially important in fields like psychiatry and primary care. Additionally, some clinicians considered note-taking as an activity that allowed time for reflection and decision-making, which could be impacted by the use of scribes.
It was agreed upon that different stakeholders like clinicians, department managers, and trust CEOs have their own viewpoints when it comes to digital scribes and the use of AI for clinical documentation. However, the ultimate goal is to improve workflow and to benefit patients.
Would clinicians tolerate digital scribe errors?
The participants discussed their tolerance for errors in transcription tools. There were different levels of acceptance based on the personal experiences of clinicians, some expressed significant frustration with the rate of errors in AI-generated clinical documentation whereas others found the scribe they had used to be quite accurate. It was commented that our expectations from technology are often higher than humans with more acceptance for human error than errors of technology.
Clinicians have a responsibility to ensure the information documented in a patient’s medical record is accurate before it is shared with other healthcare practitioners or the patient themself. There were concerns from some clinicians regarding the safety of utilising AI for clinical documentation. Clinicians expected that as the use of these technologies will become widespread, regulatory frameworks addressing their use will become more clearly defined.
Conclusion
In summary, human scribes are expensive and used mainly in high throughput specialties in the US where the delivered efficiency and income justifies the cost. Clinicians were open to the idea of using AI for clinical documentation and thought that this ambient scribe technology will save them time and improve note-writing and workflow. They expect the technology to produce accurate results and be easy to use but accept responsibility for checking the output. Regulatory frameworks would help to clarify if these technologies are to be considered medical devices, and how AI clinical documentation can be safely adopted.
