Discussing the new AI health app

THANK YOU. KELLY ANN FIVE ON YOUR HEALTH THIS MORNING, A NEW AI TOOL THAT CAN PERSONALIZE ANSWERS FOR YOUR HEALTH QUESTIONS IS CALLED CHATGPT HEALTH. AND TO USE IT, YOU’LL HAVE TO UPLOAD YOUR MEDICAL FILES, INCLUDING LAB TESTS, BLOOD AND MEDICATIONS. BUT IS THIS A GOOD IDEA? JOINING US NOW IS DOCTOR DANIELLE BITTERMAN, THE CLINICAL LEAD FOR DATA SCIENCE AND AI AT MASS GENERAL BRIGHAM. DOCTOR, THANK YOU SO MUCH FOR JOINING US THIS MORNING. WE’RE SO CURIOUS ABOUT THIS TOPIC. HOW NEW IS THIS? AND IS THIS AN ACTUAL TOOL THAT CAN REPLACE YOUR DOCTOR’S ADVICE IN ANY WAY? THANKS SO MUCH FOR HAVING ME. SO IT’S REALLY NEW. CHATGPT HEALTH CAME OUT A LITTLE BIT MORE THAN A WEEK AGO, AND IN TERMS OF, IS THIS A TOOL THAT CAN REPLACE YOUR DOCTOR? NOT ENTIRELY. YOU KNOW, THESE AI CHATBOTS DON’T HAVE THE CLINICAL REASONING AND EXPERIENCE AND HUMAN TOUCH THAT YOUR DOCTOR HAS. SO IT’S NOT A SUBSTITUTE. I VIEW IT MORE AS A IN ITS BEST FORM AS AN ASSISTANT. MILLIONS OF PEOPLE ARE USING AI CHATBOTS TO ASK QUESTIONS ABOUT THEIR HEALTH INFORMATION. SO CLEARLY THERE’S A NEED AND DEMAND FOR NEW AND BETTER WAYS TO ACCESS HEALTH INFORMATION. AND SO IN ITS BEST FORM, WE CAN CONTINUE WITH CLINICAL CARE. AND PATIENTS CAN ACCESS THESE CHATBOTS TO LEARN MORE ABOUT THEIR HEALTH, BE MORE ENGAGED AND EMPOWERED, AND MORE ACTIVE IN THEIR HEALTHCARE DECISIONS. IN GENERAL. I MEAN, THIS IS KIND OF SCARY. ANYTIME YOU TYPE SOMETHING IN LIKE WHAT THE ANSWER COULD BE, DO YOU FIND THAT IF YOU HAD A MEDICAL QUESTION, YOU COULD GET AN ACCURATE ANSWER THROUGH AI? LIKE THIS. SO THE AI CHATBOTS, THEY ARE VERY QUIET. THEY’RE VERY GOOD, BUT THEY’RE NOT PERFECT. AND SO IT’S ALWAYS IMPORTANT TO STAY AWARE THAT WHILE THE ANSWER MIGHT SEEM LIKE IT’S REALLY CORRECT AND SEEM QUITE CONVINCING IN SOME CASES, THESE CHATBOTS DO SOMETHING CALLED HALLUCINATE, WHICH IS WHERE THEY PROVIDE OUTPUT THAT MAKES A LOT OF SENSE SUPERFICIALLY, BUT ACTUALLY IS INCORRECT. MY LAB HAS DONE RESEARCH SHOWING THAT THESE CHATBOTS HAVE A TENDENCY TO BE OVERLY HELPFUL AT THE EXPENSE OF ACCURACY, WHICH MEANS THAT EVEN IF THEY DON’T KNOW THE ANSWER, THEY’LL REALLY WANT TO GIVE A RESPONSE, AND THAT OPENS A WINDOW FOR THEM TO GIVE YOU INCORRECT MEDICAL INFORMATION. ON THE FLIP SIDE, YOU KNOW, ACCURACY ITSELF IS NOT THE ONLY BEHAVIOR. THESE CHATBOTS THAT CAN BE HELPFUL FOR PATIENTS. AND THERE’S A LOT OF POTENTIAL IN USING THEIR ABILITY TO TURN REALLY COMPLEX LANGUAGE IN YOUR MEDICAL RECORDS AND IN THE MEDICAL LITERATURE INTO LANGUAGE. THAT’S MUCH MORE UNDERSTANDABLE FOR PATIENTS. SO, FOR EXAMPLE, WE DID A STUDY SHOWING THAT WHEN WE GAVE PATIENTS WHO ARE ENROLLING INTO A CLINICAL TRIAL AN AI GENERATED SUMMARY OF THAT TRIAL, THE MAJORITY OF PATIENTS ACTUALLY SAID THAT IT IMPROVED THEIR UNDERSTANDING OF THE TRIAL. AND THAT’S A REALLY PROMISING USE CASE. AND IT’S NOT MEDICAL ADVICE. OKAY, I CAN SEE THAT. WHAT ADVICE WOULD YOU GIVE YOUR PATIENTS, AND ARE THERE ANY CONCERNS ABOUT PRIVACY? I THINK WHEN PEOPLE THINK OF UPLOADING ALL OF THEIR MEDICAL INFORMATION TO THE INTERNET, THEY THINK, WHERE IS IT GOING TO GO? SO ARE THERE ASSURANCES OF PRIVACY HERE TO. ABSOLUTELY. SO I TELL MY PATIENTS NOT TO USE AI CHATBOTS IN A MEDICAL EMERGENCY, AND NOT TO USE IT AS THE ONLY SOURCE TO BASE YOUR MEDICAL DECISIONS ON. ALWAYS GO TO YOUR DOCTOR AND REVIEW WHAT THE CHATBOT TOLD YOU, AND HAVE A CONVERSATION ABOUT WHAT THE BEST CHOICES FOR YOUR FOR YOUR DECISIONS. ONE OF MY FAVORITE USES OF AN AI CHATBOT FOR PATIENTS IS TO USE IT TO PREPARE IN ADVANCE OF SEEING YOUR DOCTOR, SO USE IT TO BRAINSTORM QUESTIONS YOU SHOULD ASK TO MAKE SURE THAT ALL THE TOPICS THAT ARE IMPORTANT TO YOU GET GET COVERED IN THAT SHORT PERIOD OF TIME THAT YOU DO HAVE WITH YOUR DOCTOR. IN TERMS OF PRIVACY, IT’S REALLY IMPORTANT TO THINK THROUGH. PATIENTS SHOULD BE REALLY THOUGHTFUL BEFORE UPLOADING THEIR SENSITIVE MEDICAL INFORMATION INTO AN AI CHATBOT. KEEP IN MIND, THESE CHATBOTS AREN’T HELD TO THE SAME HIPAA PRIVACY LAWS THAT YOUR DOCTOR IS, OR THAT HOSPITALS ARE. AND WE DON’T KNOW ENTIRELY WHAT WILL HAPPEN TO YOUR DATA ONCE IT’S UPLOADED INTO THOSE SYSTEMS. SO BE THOUGHTFUL. EVERYONE’S GOING TO HAVE THEIR OWN PERSONAL OPINION ABOUT THEIR RISK TOLERANCE FOR PRIVACY, BUT KEEP IN MIND THAT IT MAY NOT BE AS SECURE AS WHEN YOU TELL YOUR HEALTH INFORMATION TO YOUR DOCTOR. OKAY, THAT IS SO HELPFUL. THANK YOU FOR KIND OF NARROWING IT ALL DOWN FOR US, DOCTOR DANIEL BITTERMAN, THE LEAD CLINICAL LEAD CLINICAL LEAD FOR DATA SCIENCE AND AI AT MASS GENERAL BRIGHAM, THANKS FOR BEING WITH US THIS MORNING.