By Jeff Butcher, PM, IC, MS
The latest version of ChatGPT 4 Omni has sparked widespread excitement and curiosity, particularly within the EMS community. With its advanced capabilities, many are pondering the potential roles it could play in revolutionizing emergency medical services.
As a dedicated EMS professional and technology enthusiast, I decided to explore whether this cutting-edge AI could be a reliable tool for identifying pills in the field.
Through hands-on testing and real-world applications, I’ve uncovered fascinating ways ChatGPT can support and enhance our work, offering promising solutions to some of our most pressing challenges.
Testing ChatGPT’s Pill Identification Capabilities
One afternoon, I spread all the pills in our house across the kitchen counter, with my son snapping pictures using the ChatGPT app on my iPhone. It was quite a sight when my wife walked in with grocery bags, only to see us engaged in this experiment.
Before diving into the results, it’s essential to emphasize two warnings about this type of “testing”:
- This experiment was purely exploratory. I am not suggesting ChatGPT is ready for field use or recommending anyone try this at work.
- Testing AI on various pills at home, especially with a child involved, can be hazardous, especially if your wife catches you doing it!
During the test, we laid out a variety of medications outside their prescription bottles and took pictures using the ChatGPT app. Surprisingly, it correctly identified all eight of the pills.
It’s important to note that the goal was to see if the AI could identify the pill type, not to delve deeper into its capabilities.
Practical Applications and Limitations
So, where could this technology be useful? Imagine arriving at the scene of an overdose, finding various pills scattered across the table, and only a couple of identifiable pill bottles.
With the right resources and time, ChatGPT could help quickly identify the pills the patient ingested. Additionally, having a photo of the pills could be useful when briefing the doctor in the emergency room.
However, there are significant disadvantages to using AI for pill identification on the scene. Most overdose calls where pills cannot be easily identified also mean the patient is already affected, and immediate medical attention is crucial.
While knowing the medication can help prepare for potential complications, emergency medical providers use a wealth of other information to form a complete picture.
Over-reliance on an application not designed for this specific use case could be detrimental. Furthermore, using personal phones for patient care introduces privacy concerns and potential legal implications if the phone becomes evidence in a criminal investigation.
Integrating ChatGPT into Field Operations
After my initial testing, I envisioned how this technology could be practically integrated into field operations. First, I would seek approval for a dedicated phone with the ChatGPT app, provided and paid for by my organization. My primary use case for the application would be as a language translator.
Second, I would consider its use in specific scenarios: when there are extra first responders on scene or when identifying one or two pills.
In such cases, an extra responder familiar with the app could be tasked with identifying the pills, allowing the primary medical team to focus on patient care.
Finally, findings should always be cross-checked with a quick internet search to ensure accuracy. Does the identified pill make sense in the context of the patient’s history?
This additional step can help confirm the AI’s identification and provide a layer of verification that is crucial in emergency situations.
Training: The Most Promising Use Case
Where ChatGPT 4 Omni shines the most is in training scenarios. A great drill would involve setting up an overdose lab with a variety of over-the-counter medications and some candy resembling pills.
This setup allows participants to engage in a controlled, yet realistic, environment. Participants could complete worksheets identifying pills, matching them to their bottles, and determining side effects and quantities. These exercises not only enhance their ability to recognize different medications but also improve their understanding of pharmacology and toxicology.
Additionally, they could connect overdose scenarios to patient profiles with detailed vital signs. These profiles would include information such as blood pressure, pulse, breathing rate, pulse oximetry, and cardiac rhythm, providing a comprehensive picture of the patient’s condition.
By working through these scenarios, participants can practice critical thinking and decision-making skills, preparing them for real-life emergencies.
Utilizing ChatGPT in these training exercises adds an element of technological integration, demonstrating how AI can assist in emergency response and education.
To help you set up this training experience, I have created a prompt that you can use with your favorite language model (LLM).
Simply take this prompt, customize it with your specifics, and input it into the text box of any LLM. If you don’t have access to one, consider using Bing Chat, which is free and a great place to start experimenting with AI for training.
Overdose Training Prompt
Create a detailed scenario for an overdose drill for EMS training, utilizing the following medications: [list medications here]. The scenario should include:
- A worksheet for participants to identify pills, match them to their bottles, identify side effects, and determine the amount taken based on varying quantities.
- Profiles for a group of patients, including vital signs and symptoms that match the overdose scenario. Each profile should contain:
- Blood pressure
- Pulse
- Breathing rate
- Pulse oximetry
- Cardiac rhythm
- SMART training objectives, ensuring the goals are Specific, Measurable, Achievable, Relevant, and Time-bound.
- Fields for entering the length of time for the training and the instructor’s name.
- Instructions for dividing participants into three groups: one using a pill identification book, another using any available resources, and the third utilizing ChatGPT.
- Ensure the scenario promotes discussions and offers diverse actions for participants.
Suggestions for Enhanced Training
- Include diverse demographics and medical histories for patient profiles to add complexity.
- Provide descriptions, common uses, and potential side effects for each medication.
- Outline key discussion topics for debriefing.
- Ensure vital signs and symptoms accurately reflect the overdose scenarios.
- Develop clear and concise SMART objectives for the training.
Questions for Customization
- What specific medications should be included in the scenario?
- Are there particular patient demographics or medical histories to feature?
- Do you have a preferred format for the worksheet and patient profiles?
- How many patient profiles are needed?
- Should the scenario include unexpected complications or variables?
- Are there specific overdose symptoms to highlight?
- What are the desired SMART objectives for the training?
- What is the length of the training?
- Who is the instructor for the session?
Conclusion: Harnessing AI for the Future of EMS
Testing out ChatGPT 4 Omni’s ability to identify pills has opened a dialogue on the potential uses of AI in emergency medical services.
While our initial tests show promise, particularly in training scenarios, it’s crucial to approach this technology with caution and a clear understanding of its limitations.
In the field, our primary focus is always on providing the best possible care for our patients. AI can be a fantastic sidekick, helping us identify unknown medications quickly in those rare, head-scratching situations.
However, we must remain cautious and aware of its limitations. The last thing we want is to waste precious time fiddling with an app when a patient needs our immediate attention.
On the flip side, the training potential of ChatGPT 4 Omni is nothing short of amazing. By simulating real-world scenarios and providing a platform for interactive learning, it can enhance the preparedness of EMS providers, ensuring that we’re always ready to respond swiftly and accurately.
As technology continues to evolve, so too must our methods and strategies in EMS. By staying informed and open to innovative solutions, we can improve our practices and ultimately provide better care to those we serve.
The journey of integrating AI into EMS is just beginning, and with careful consideration and strategic implementation, ChatGPT 4 Omni could become an invaluable asset in our mission to save lives.
About the Author
Jeff Butcher, paramedic, firefighter, is a seasoned EMS professional with extensive experience in emergency services management and public safety education. He is currently a senior education partner at Waldorf University and has a strong background in developing and directing EMS training programs. His work focuses on integrating innovative technologies into EMS operations to enhance training and field effectiveness.
