Doctors should think twice about using ChatGPT for medical advice, here’s why
In 2023, artificial intelligence has been a huge story.
This was the year large language models like OpenAI’s ChatGPT program were unleashed on the masses, giving people access to AI programs that can help with everything from figuring out what to eat for dinner tonight to explaining complex topics theories.
But ChatGPT, and every other large language model (LLM), is only as good as the information which feeds it. And a new study suggests that artificial intelligence still has much to learn about the medical field.
Over a 16-month period between 2022 and 2023, Sara Grossman, PharmD, Associate Professor of Pharmacy Practice at Long Island University, and her team challenged ChatGPT with real medical questions posed to Long Island University’s College of Pharmacy drug information service.
The study’s conclusion?
Healthcare professionals and patients should be cautious about using ChatGPT as an authoritative source for medication-related information, Grossman. Anyone who uses ChatGPT for medication-related information should verify the information using trusted sources.
Pharmacists involved in the study first researched and answered 45 queries and each answer was reviewed by a second investigator. The investigators then asked ChatGPT the same questions — minus six questions because there was a lack of literature to provide a data-driven response.
Of the 39 questions that the team asked the free version of ChatGPT, only 10 of the responses were judged to be satisfactory.
In one instance, ChatGPT was asked whether a drug interaction exists between Covid-19 treatment Paxlovid and blood pressure lowering medication verapamil. Chat GPT said that there were no interactions that had been reported for that combination of drugs.
In reality, these medications have the potential to interact with one another, and combined use may result in excessive lowering of blood pressure, Grossman said. Without knowledge of this interaction, a patient may suffer from an unwanted and preventable side effect.
For the 29 questions ChatGPT missed, researchers determined that the program did not directly answer 11 of them while 10 were inaccurate and/or 12 were incomplete.
This study raises important concerns about the reliability of using ChatGPT for medical advice. While artificial intelligence has made great strides in various fields, the medical field requires a specific and accurate knowledge base that may not be adequately represented in these language models.
The potential risks associated with relying solely on ChatGPT’s responses for medication-related information cannot be overlooked. Patients and healthcare professionals should exercise caution and consider verifying the information provided by the AI with reliable and trusted sources.
As technology continues to advance, it is crucial that AI models, especially those in the medical field, undergo rigorous testing and improvement to ensure their reliability and accuracy. While ChatGPT may have its uses in assisting with general queries or providing explanations on a wide range of topics, it is evident that it should not be relied upon as an authoritative source for medication-related information.
The study conducted by Sara Grossman and her team serves as a reminder that AI systems should be viewed as tools to augment human expertise, not replace it. The importance of human medical professionals and their knowledge cannot be understated, particularly when it comes to matters of health and medication.
In conclusion, while ChatGPT and other large language models have their benefits, healthcare professionals and patients should exercise caution and seek information from trusted sources when it comes to medical advice. AI has the potential to revolutionize the field of medicine, but it is essential to ensure that these systems are reliable, accurate, and backed by comprehensive knowledge.
