Company reputation
It is worthwhile to begin by researching the company offering the AI scribe.
Software integration
There are many types of AI scribing tools available. Some electronic health records (EHR) platforms offer a built-in tool, others can communicate with your EHR via application programming interfaces (API), and some are standalone products. Those that are built into your EHR may offer ease of use, as EHRs are increasingly operating as comprehensive practice management solutions (e.g., offering billing, clinical record, virtual platform, AI scribing, etc.). Whereas AI scribes that are standalone or that integrate via API may offer you more choice to select among the many available options.
Functionality
Most AI scribes operate using an ambient listening mode (e.g., a provider enables the AI scribe to run in the background, essentially “listening” to a clinical appointment). Many AI scribes also have a dictation function. This option does not require the AI to “listen in” on a session. Rather a provider would either dictate or upload a written summary of the clinical appointment, and the AI scribe would turn that summary into a note of your choosing (e.g., intake report, progress note, etc.).
Privacy compliance
At a minimum, AI scribes are accessing PHI and must be compliant with HIPAA (or GDPR and/or other applicable privacy laws and standards in the jurisdiction(s) where you practice).
Technical security
Given the confidential data to which AI scribes have access, it is critical to understand how data is secured. This includes the processes and procedures that are in place to limit the possibility of a breach and in the event of a breach, to limit the amount of identifiable information available (see data handling section below).
Data handling
In addition to robust technical security, it is also important to assess how data are handled. The concept of “data minimization” is a key consideration, meaning that the least amount of data is stored for a given purpose. This can help to limit the amount of data that could be compromised in the event of a breach. You are encouraged to choose companies that allow you to customize settings to minimize the storage of data beyond what is necessary for your tasks or easily delete such data manually.
Large language models (LLMs)
This is the technology that powers AI scribes. Some companies have developed their own, proprietary models, and others contract with third-party AI companies to use their LLMs.
Informed and legal consent requirements
It is important to develop an informed consent process and documentation policy for the use of AI scribes in your practice.
Legality and liability considerations
You must review AI generated documentation for accuracy and edit as needed prior to signing and the documentation becoming part of the clinical record. While AI may assist with documentation, you remain responsible for all clinical decisions and documentation accuracy.
Documentation and periodic review
You are encouraged to document your review of any product that you select for use in your practice.
Needs of your practice
A decision about whether to use an AI scribe and if so, which tool to incorporate into your practice, is an individual decision based on one’s unique practice needs. However, these steps will help you gather the relevant information needed to make an informed decision.
