What is ambient clinical intelligence?

How does ambient clinical intelligence work?

Ambient clinical intelligence uses AI to automate the detailed documentation required during an average doctor's visit. Normally, the clinician takes notes, gathers data, and then transcribes that information immediately after the visit or later in the day. ACI automates this time-consuming documentation to boost the quantity and quality of doctor-patient interactions.

A typical ACI system operates in the following manner:

• Recording. Using a common device, such as a secure smartphone or tablet, ACI systems capture and record natural conversations between the clinician and the patient. Recordings are compressed and saved to storage media such as a disk drive.
• Transcription. The ACI system transcribes the audio into text. Speech recognition technology differentiates the clinician's voice from the patient's, establishing a clear conversational exchange that improves its overall analysis. Transcription happens in real time. The clinician then reviews the transcription when able. Transcriptions saved as digital files are associated with the corresponding audio recording. If any portion of the audio is unclear or cannot be properly transcribed, ACI systems may alert the clinician during the visit.
• Analysis. The ACI system analyzes the transcribed text using NLP techniques. Analysis establishes the patient's context or condition, identifies medical details obtained during the interaction and extracts other important information from the conversation.
• Coding. The analyzed transcription is processed for standardized clinical coding, such as LOINC codes or SNOMED CT codes. Do not confuse these with coding systems used in medical billing, including ICD-10 or CPT codes. Standardized coding facilitates data entry into the patient's EHR, as well as the unfettered exchange of data between authorized providers and health data systems.
• Notation. Finally, the ACI system compiles complete clinical notes outlining and detailing the visit. The clinician reviews, edits and approves these notes, which are then added to the patient's EHR. In the event of errors or further questions, clinical notes correlated to the transcription and audio recording provide traceability through the ACI process.

 Although current ACI systems focus on assisting clinicians with timely, accurate and complete documentation, advanced ACI platforms may take a more active role in the clinical process.

For example, a future ACI system with approved access to patient medical records knows the patient's prescribed medications. If the patient mentions taking a dietary supplement while describing their symptoms, an advanced ACI system may compare the supplement's contents to current medications, then flag potentially dangerous interactions for the clinician to discuss.

As another example, future AI-driven analytics may identify additional patient issues during regular clinician inquiries. Detecting vocal stress or hesitation in an audio response, the advanced ACI system alerts the clinician to pause and ask more probing questions to elicit patient details.

Such capabilities demand vastly improved technologies and training data sets, but the AI building blocks are in place.

What can ambient clinical intelligence do right now in healthcare?

Pilot ACI programs, despite shortcomings, have produced positive results, from strong notational accuracy to better doctor-patient engagement. Current ACI systems deliver limited capabilities, including the following:

• Real-time recording. ACI captures real time audio of conversations between clinicians and patients.
• AI transcription. Speech recognition and NLP convert the audio recording into a text transcript.
• Clinical coding. ACI extracts medical details, places them in context and applies suitable coding to relevant details.
• Clinical notes. ACI outlines relevant elements of the doctor-patient conversation and creates suitable clinical notes.
• Records integration. ACI links with EHR systems to convey coding and notes.

 Of course, clinicians must retain full review over ACI systems. The provider checks, edits and approves ACI content before conveying it to an EHR system. And patients, perhaps concerned about data privacy or security, have the right to opt out of ACI support. When this occurs, the clinician generates notes in the traditional way.

What are the drawbacks of ambient clinical intelligence?

ACI's benefits are compelling, yet several drawbacks must be considered before adopting an ACI system, including:

• Accuracy. ACI systems frequently misinterpret slang and non-lexical sounds – a muttered "uh huh" instead of a clearly enunciated "yes," for example, is challenging – and cannot process non-verbal communication, such as a shrug of the shoulders to mean either a non-committal "sort of" or "not really." In addition, ACI systems sometimes misidentify medical terms – especially when coupled with an accent or imprecise pronunciation – leading to notational errors. In all cases, clinicians must review and approve ACI-generated notes and codes for accuracy and completeness, editing content if necessary.
• Security. ACI systems collect, store, process and produce enormous amounts of data, including protected or personal health information. This demands data security to guard patient privacy. To protect data collections against misuse and breaches, medical organizations deploy comprehensive breach notification and remediation technologies, plus encryption and data anonymization to mitigate security concerns further.
• Ethics. ACI systems lack transparency in how they process data, transcribe it and extract its elements for coding and notation. Such "black box" behavior often leaves clinicians skeptical of ACI technologies. Conversely, successful ACI adoption sometimes leads to overdependence on ACI: Human judgment is displaced by ACI output, termed automation bias. Bias also occurs in training data sets, leading to improper or discriminatory results. Finally, proprietary concerns stretch beyond healthcare delivery. For example, using ACI data for training or research purposes or selling anonymized ACI data creates ethical issues.
• Deployment. Initial investments in resources—financial, technological and training—are costly for medical organizations, where those resources are often already in critically short supply. ACI systems demand strong interoperability with existing EHR systems, a resource-sapping organizational task.

How will ambient clinical intelligence change modern healthcare?

There is a shortage of physicians today, and it's expected to continue. Talented professionals enter healthcare fields with a genuine desire to help people, ease patients' suffering and improve their quality of life. Yet during too many health visits, physicians spend more time entering notes at computer terminals and far less interacting with their patients.

Accurate notes and coding are critical to proper care and ensure regulatory and legal compliance. However, these tasks demand long hours of administrative work with little relation to medicine, often resulting in physician dissatisfaction, frustration and burnout. This means fewer capable clinicians are available, and those who remain must manage even more patients.

ACI technologies promise to remove some mundane but necessary administrative tasks of modern clinicians, such as taking accurate and complete notes. Automating these tasks lets a physician focus on asking in-depth questions, listening intently to a patient's responses and, in turn, delivering better patient outcomes.