Jeff Rush, BSCSE
May 3, 2023
"It's not about the tool, it's about the problem it's solving." - Matt Heinz
ChatGPT operates upon text. For any areas of healthcare where blocks of text are present, it can offer:
Exploration of artificial intelligence has been going on since the 1950s, weathering seasons of heady successes and disappointing setbacks. So why does it seem so different this time with ChatGPT?
Notification letter templates and educational materials are a rich source of static, non-patient text in healthcare. ChatGPT can assist in maintaining them.
1. Summarization, Translation, Comparison, Perspective, and Reading Level
For example, ChatGPT can summarize multi-page documents into abstracts according to your preferred tone, reading level, or human language. These can be attached to the documents and help with cataloguing them or finding the appropriate one to use. Documents in a non-English language can be summarized into English to help a health librarian who doesn't speak that language manage them.
Parallel maintenance of multi-language documents can be eased by having ChatGPT generate non-English versions from an original in English, which makes quality checks by a multi-lingual document specialist more efficient. Sometimes, such document libraries experience quality drift over time, so ChatGPT can compare the multi-language versions against an original to insure they are still equivalent in meaning.
ChatGPT can also be used to transform these documents in useful ways to adjust the reading level for optimal comprehension and change the wording point of view from that of a patient (2nd person perspective) to their caregiver (3rd parson perspective).
2. Customization, Emphasis and Content Examination
ChatGPT can also play a role in automated document customization for each patient by being told to omit the section about a topic that is not relevant or to emphasize sections that are critical.
Beyond summarization, ChatGPT can save staff time. It eliminates the need to read a trove of documents that must be quality checked by pointing a document for conversation and then allowing content librarians to ask questions that examine it for desirable properties. Examples of such questions include but are not limited to:
Templates for assessments, which are basic questions for patients/members that also offer a list of multi-choice values, can be operated upon by ChatGPT. Use of a content management system can provide the previously mentioned translation and reading level transformations.
Summarization can be obtained across a set of such questions either statically for a fixed assessment or dynamically across an assessment assembled from a collection selected based on a patient's diagnosis.
Although often challenging, due to different styles and cryptic writing styles, there are opportunities for ChatGPT to explain, summarize, and extract information from doctor notes. These can be used to elevate such notes into automatic facts about a patient that a care management system can use more effectively.
ChatGPT has the interesting ability to generate imaginary roles, data or situations and then interact as if they were real. This can be useful for care management training systems to help them role play particular types of patients and verify the care staffs’ actions.
Prompt: "Imagine there is a patient reporting to the emergency room who is 24 years old, male, diabetic, and experiencing headaches and high blood pressure. His name is John. He seems confused and a little hostile. Role play John and let me ask him questions."
It's still very early in the widespread adoption of systems like ChatGPT, but I hope you can see the potential uses in healthcare and how it can improve the efficiency of care management systems and staff.
Learn more about what makes ChatGPT unique compared to other AI.