Healthcare experts welcome the launch of ChatGPT Health for consultations, despite concerns about accuracy and data security.
Healthcare professionals are showing support for the recently launched ChatGPT Health tool, designed to facilitate health consultations, despite the inherent risks of inaccuracies linked to AI technology. According to reports, many users are already opting for this service over traditional symptom searches.
Dr. Sina Bari, a surgeon and AI initiatives leader at iMerit, recounted a patient experience where ChatGPT suggested a 45% risk of pulmonary artery thromboembolism after a medication recommendation. Upon investigation, Dr. Bari found that this statistic originated from a study focused on a specific subgroup of tuberculosis patients, making it irrelevant for the patient in question.
Despite these potential pitfalls, Dr. Bari remains optimistic about ChatGPT Health. He believes that the platform can enable discussions about health matters in a secure, private environment, stating, “I think it's great...formalizing the process to protect patient information will make it more effective for patients.”
Users can enhance their interactions by integrating personal medical records and syncing the service with health apps, although this raises concerns about the security of sensitive information. Itai Schwartz, co-founder of MIND, pointed out that this data transition from HIPAA-compliant entities to less regulated providers poses significant regulatory challenges.
Weekly, over 230 million individuals are consulting ChatGPT about their health, often bypassing traditional search engines for advice. Schwartz emphasized this trend, noting that creating a specialized, secure version of the chatbot for health inquiries is a logical next step.
The Hallucination Issue
The subject of AI hallucinations remains a prominent concern, particularly within healthcare applications. A study indicated that OpenAI's GPT-5 has a higher tendency to generate inaccuracies than other AI models from competitors like Google and Anthropic. However, Dr. Nigam Shah from Stanford University argues that accessibility to healthcare providers is a more pressing issue than the risk of misleading information from AI. He highlighted the long wait times for primary care appointments, suggesting many might prefer immediate AI support over postponing care by months.
Administrative burdens limit the capacity of doctors to see patients, with Dr. Shah leading efforts at Stanford to enhance efficiency through their ChatEHR software, aimed at streamlining electronic medical record management. Dr. Sneha Jain, a tester of ChatEHR, noted that creating a user-friendly interface would enable doctors to spend less time on administrative tasks.
In a related development, Anthropic has introduced Claude for Healthcare, a toolkit designed for both healthcare providers and patients, signifying a growing trend towards integrating AI in medical practice.
