Part 2: Ambient Clinical Intelligence (Autonomous Clinical Documentation)

The Ambient Clinical Intelligence (ACI) technology is changing the landscape of autonomous clinical documentation at warp speed. Who doesn’t want real-time results, at a lower price? But be careful. Due to powerful new AI technology, there are many start up companies rolling out autonomous clinical documentation solutions. When performing your due diligence, it is critical to do your homework and take the time to deeply vet your vendor options. This will protect your investment.

Part 2 in our series will focus on understanding how to evaluate vendors, pre-evaluation questions, vendor evaluation checklist, and adoption best practices. These are key components for driving your business case and selection, as outlined in part 1.

Key Vendors in the Autonomous Clinical Documentation Industry

As described in Blog 1, there is currently a convergence of front-end speech, automated ambient scribe, and autonomous dictation to Autonomous Clinical Documentation software/workflow, with or without a human in the loop for quality review and editing of the note. In other words, AI with clinical intelligence to listen to clinician-patient encounters and generate near instant clinical notes.

Comparing Autonomous Clinical Documentation Vendors

The vendors below are the current major players in offering this technology As you review the table, consider the following:

• Category – First, for your personal ranking it’s important to understand if you want a vendor that offers Fully AI Automated (a note generated minutes following a conversation with no intervention from the vendor) OR a vendor that offers a Hybrid approach (fully Automated PLUS a human scribe/QA). Human in loop is a more expensive option with slower turnaround time to the clinician. However, this option may be a necessity for clinicians who demand stellar quality. A hybrid approach allows the customer flexibility in approach, based on the clinician current state and desire AND flexibility in pricing.

• Data – You want to ensure the vendor has access to large amounts of data for training their models. This will ensure a higher quality outcome, faster. This is a scenario where the saying “Data is King” applies. The more clinicians served; the more data used to train AI models. A library of dictated / transcribed documents provides a richer clinical vocabulary for fine tuning before the engine in even implemented. This level of medical specificity can be hard for clinicians to add into a patient-clinician dialogue. The vendors in the table are ranked based on the number of Estimated Clinicians column, which infers the amount of data used to train.

• Cost – Vendor cost is not outlined in the table, but average cost model falls in line with these offerings: No EHR Integration ~$100 ppm; Fully Integrated ~$250 ppm; Human in the Loop ~$450 ppm.

Autonomous Clinical Documentation Vendor Chart

There is also a subcategory of vendors that are Direct to Consumer. They target the physicians directly via social media with a low subscription fixed cost model (typically $100/month/clinician) and non-integrated approach. This is a popular solution in the physician world because they find time savings and typically skirt the formal approval process. Clinicians will cut and paste the information into the EHR since there is no integration. This is a great approach for a trial, but HIM and security should be closely vetting this approach for a permanent solution.

Pre-Evaluation Questionnaire

To narrow down the vendor scape you intend to officially engage for full evaluation, a pre-evaluation questionnaire should be distributed. The goal is to establish baseline understanding of the vendor’s capabilities and if their solution would meet your organization’s needs. This will allow you to weed out vendors and save your selection team valuable time.

Consider the following questions as you customize your ACI Vendor Pre-Evaluation Questionnaire:

• Number of active customers by category (hospitals, physician groups, urgent care, etc.)

• How many clinicians uses your technology?

• Outline your technology approach.

• What model do you use for your generative AI?

• Do you offer human in the loop service options? If so, describe options (ie. Scribe vs. transcriptionist vs. quality reviewer…)

• Average turnaround time of the note, post conversation?

• What is your average quality percentage out of the gate? 3 months?

• Current EHR integrations.

• Describe how you map clinical findings to the EHR data fields in various EHR instances (look for vendors who embed ambient workflow within the EHR for near real time results from conversational audio to mapping in the chart).

• Do you offer a pilot? If so, outline approach, average length, exit criteria, and cost.

• Outline the steps in your implementation process and time associated with each.

• Outline your recommendations for phasing in your technology and optimal adoption techniques.

• Describe how you evaluate the clinician’s current approach to documentation for optimal adoption.

• Describe training and support approach.

• Describe security and compliance measures.

• Outline proven multilingual languages (Language translation for multi-lingual conversations to documentation instantly).

• Describe any other adjunct autonomous offerings, such as autonomous CDI or coding solutions.

Vendor Evaluation Tool

In Blog 1 the key selection criteria was outlined. Now you have narrowed down the vendors you want to fully engage for the evaluation and selection process, based on the pre-evaluation questionnaire. The selection team has been formed and your meetings and demonstrations are underway. The tool below will ensure consistency and unity for your team’s assessment and selection. Customizing to your facility and documentation needs will be your final step in this process.

Adoption and Change Management Tips

Before planning adoption and change management strategies you must first consider the clinician’s current state approach for documenting clinical notes. If the team approaches as a “one shoe fits all” approach, there will be points of failure. Remember, there is a tradeoff between accuracy and delivery time with autonomous documentation: the ambient engine is very fast, but it is typically less accurate in the first draft than a human in the loop. It is important to give clinicians workflows that provide optionality for their involvement in the edit process.

For example, if a clinician in the current state has not had any involvement in their editing process due to a scribe, it will be a big leap to ask them to now edit and make corrections, even if the turnaround time is 2 minutes and the quality is “acceptable.” Whereas, if a clinician has been receiving their dictation back and editing at the end of the day, the immediate delivery/editing will make them very happy. Current state evaluation of each clinician with optionality is key to a successful adoption rate.

Your vendor should perform extensive discovery to understand current state, including documentation approaches, document samples, and other requests for interface and system configuration. A Go Live Action Plan, Testing Outline, and a Phase-In/Training Plan should be provided to the team to instill confidence in their support during this critical transition. It is important to ensure your vendor offers a multi-team, integrated approach to onboarding. They should be closely monitoring quality and TAT expectations, along with any need to change clinician preferences. This will require daily checkups with operations.

Other Tips:

Look for a user-friendly solution that requires minimal workflow changes and minimal training.

Start with a small group of champion level early adopter clinicians that are willing to provide feedback to refine the process and workflows. This will inform your change management plan for phasing in with a larger population.

As outlined above, the vendor’s level of support during implementation is critical and will significantly impact adoption.
Be careful choosing a solution that is purely web-based. This means there is no platform associated with the ambient solution. There are some web-based solutions that have acquired security best practices (although this should be carefully vetted), but they are not typically integrated into the EHR, let alone mapped to the clinical fields. Without this, clinicians will need to copy and paste, which will not help adoption or clinician satisfaction.