The emergency department (ED) is a unique clinical environment, with pressured patient loads, time constraints, and a high documentation burden. My experience has shown that using an AI scribe can provide considerable advantages in this setting. Personally, I've found it has increased my efficiency, allowing me to see more patients and discharge them sooner. It has also improved the accuracy and comprehensiveness of my clinical notes and reduced the cognitive load associated with often delayed transcription, which has honestly increased my enjoyment of the job.

The following are tips and best practices from my experience using this tool within the ED environment.

Strategic Device Deployment

The varied clinical settings within the ED—fast track, resus, acute and subacute, and waiting rooms —requires a flexible approach to recording device selection.

• Mobile Devices / Smartphone: These offer maximum mobility, facilitating use across all ED locations, including at the bedside and in areas inaccessible to larger workstations (e.g., Computers on Wheels - COWs). Using the LH phone app to listen and save the note, then accessing it through the website on a workstation before copying them over to the EMR is a few extra steps but ones I’m happy to make. This is made easier if there is tap on / tap off virtual desktops whereby the website remains open and logged in wherever you log in to a workstation. 
  

Critical Consideration: The deployment of personal or hospital-issued mobile devices mandates strict adherence to information governance, privacy, and security protocols. Robust institutional policies and technical safeguards are essential.

• Computers on Wheels (COWs): COWs are a functional compromise, offering both mobility and integrated (EMR) access. They are suitable for many ED environments but may present limitations in confined spaces or during high-intensity resuscitation efforts (and often just not available!)
• Fixed Workstations: While providing stability, fixed workstations inherently restrict AI scribe utilisation to designated rooms. This may be appropriate for dedicated consultation spaces but offers limited adaptability to the broader ED workflow.
• Lapel Microphones: I have heard of ED clinicians using a lapel mic, which could be connected to a workstation or COW.  This could potentially improve audio capture quality and preserve clinician mobility, and bypass the need for security clearance of personal devices.

Informed Consent Protocols

Securing and documenting patient consent prior to the activation of any recording device is a key ethical and legal obligation.

• Standard Clinical Encounters: For the majority of patients, a clear and concise explanation of the technology's purpose, followed by documented verbal consent, should be standard procedure. 
• High-Acuity and Resuscitation Scenarios: In critical situations where a patient lacks the capacity to consent (e.g., major trauma, cardiac arrest), the use of an AI scribe may be precluded. Subject to institutional policy, family members may be considered to provide consent if the clinical situation permits. 

Optimising Scribe Performance and Workflow Integration

• Interruption Management (Pause and Restart Functionality): ED consultations are subject to interruption or occur over multiple discussions. The capacity to pause recording and start again when obtaining further info —eg, to obtain collateral history from relatives, ambulance personnel, or nursing staff — is a feature that allows you to accumulate relevant info across space and time (and not just from the patient.) The scribe can collate all of this disparate info into the same document.
• Modifications to History-Taking Technique: Adjustments to conventional history-taking and examination practice can improve the completeness of AI-generated documentation:
  
  • Explicit Verbalisation of Discrete Data: Clearly articulating medication names, dosages, and frequencies (e.g., when reviewing a Webster pack or medication chart) permits the AI scribe to include this information within the note. Similarly, ‘playing back’ information to the patient regarding key parts of the history will allow the scribe to generate more structured and clinically accurate notes, which previously is something I would tend to do internally.
  • Articulation of Examination Findings: Verbalising physical examination findings as they are elicited (e.g., "Respiratory examination reveals bilateral clear air entry," ) enables the AI to draft this section, reducing subsequent manual data entry, though takes some getting used to.
  • Arti of the Clinical Plan: Articulating the assessment / impression and management plan directly to the patient will improve their comprehension but also provides content for the scribe to provide an impression and management plan draft.

Post-Capture Note Review and Augmentation

While AI scribes provide considerable efficiency, the initial output should be regarded as a sophisticated preliminary draft that needs thorough review and supplementation.

• Verification and Enhancement: Clinicians must review and edit the AI-generated note to ensure accuracy, completeness, and clinical appropriateness. In my experience, this has included
  
  • Validating and correcting medication lists.
  • Ensuring precise documentation of past medical history and comorbidities.
  • Incorporating or clarifying recent investigation results and their implications from prior EMR documentation
  • Correcting facts that the AI may misheard, or less likely hallucinated.
• Sustained Efficiency: Notwithstanding the necessity for review, I’ve found this process remains substantially more time-efficient and less cognitively demanding than traditional methods of retrospective recall and manual transcription.

Leveraging AI for Enhanced Clinical Communication

• Streamlined Discharge Summaries: A key benefit is the capacity for rapid generation of discharge letters, often achievable with a single command.
  

Refinement Strategy: Initial drafts may occasionally be verbose. Utilising the AI editing function or manual refinement can produce a concise more appropriate summary

• Patient Letter: The capability to generate a distinct, patient-appropriate summary of their ED visit is a valuable adjunct that I never had time for previously. Creating a dedicated patient note allows for tailored communication, differentiating from the more technical information in the discharge summary (previously I found myself striking a balance of clinical background, but a layman plan for the patient.) You can also use a feature to translate into other languages, however the local institutional medico-legal considerations and liabilities of this must be considered and may preclude this.

Infection Prevention and Control Measures

The introduction of additional hardware into clinical areas necessitates adherence to established infection control protocols, which as ED clinicians we aren’t known for.

• Routine Device Disinfection: All recording devices, particularly high-touch surfaces such as smartphones and COWs must be disinfected between each patient encounter, in strict accordance with institutional policy.

‍

More about Dr. Hassan Ahmad

Hassan is a Provisional Fellow in Emergency Medicine and a Certified Health Informatician. He leads research into the practical application of machine learning models within the ED to address operational and clinical challenges. His background includes experience in global health, NGOs, and social enterprise, with clinical interests spanning trauma, prehospital care, and rural and remote medicine.

Follow him on LinkedIn

‍