In this second part of the blog, we will explore the potential role of ChatGPT in the medical field, its impact on the advancement of digital health, and important considerations and limitations. We will also examine how AI models can be integrated into clinical practice, healthcare, and research and how they can shape the future of medicine. Finally, we will discuss the implications of these latest technology developments for health professionals and their practices.
ChatGPT Won’t Do Your Job For You, But It Can Certainly Help
The possible applications of ChatGPT are so far-reaching we have yet to uncover them all and probably never will as the technology continues to improve. It’s a compound effect: the more users engage with the chatbot, the more information it processes and can learn from, becoming increasingly more sophisticated over time.
ChatGPT has proven effective as an automation tool, reducing the burden of routine tasks from patient scheduling to note compilation. Its ability to process large amounts of data and identify key trends makes it a powerful research instrument. But the real excitement is in its potential to transform healthcare delivery and revolutionize the future of medicine.
A Better Practice Automation Tool
ChatGPT is a practice automation tool that elevates the patient and provider experience. With seamless data integration, AI technology can answer frequently-occurring questions, schedule appointments, manage bookings, and carry out simple tasks. At the same time, providers can leverage the chatbot’s unique fluency to help alleviate a rising concern among health professionals – responding to a growing number of patient portal messages.
- Rote Daily Tasks
Clinicians have already begun testing ChatGPT’s ability to ease their daily tasks, including repetitive, low-value writing, creating macros and bulk text items, as well as simple formulaic work.
Anobel Odisho, MD, MPH, physician, practice manager, and researcher from the University of California San Francisco, provided examples of how he uses it in practice in a recent MedPage Today article.
“Odisho said he has used the program to generate first drafts of letters that he then reviewed and edited. He also used it to generate more patient-friendly descriptions of procedures or post-procedure instructions. He even used it to create a draft of an on-call schedule.”
Ultimately both simple and more complex tasks can be automated, becoming the domain of ChatGPT, freeing up many hours on a practitioner’s weekly schedule and reducing workload.
- Responding To Portal Messages
A current and pressing burden on physicians is the extent to which patients communicate through patient portals outside of scheduled appointments. Now that most patients have online access to their medical information, they are increasingly seeking real-time information and explanations of their clinical results. Not only are these messages adding to an already overburdened clinician’s workload, but they also pose a billing challenge. Reports are appearing noting that many healthcare facilities are seeking to bill patients for these interactions, which could, in many cases, be handled by a large language model like ChatGPT. The AI chatbot could help alleviate stress for both patients and providers if programmed appropriately and responds accurately.
- Summary and Analysis of Data and Research
Another potential application of ChatGPT is summarizing and analyzing research papers and datasets. ChatGPT can summarize a long, detailed research paper or list the keywords in an abstract. It can also be used to analyze large amounts of medical data to identify trends and parse through data from millions of studies. In clinical practice, the AI model could help physicians aggregate patient information expediently, condensing medical history results, diagnoses, allergies, previous visit notes, and results into a short summary. The technology would provide physicians with a comprehensive understanding of their patients quickly and efficiently, allowing more time to be spent caring for patients.
- Novel Drug Development
Now, for a promising future application of ChatGPT, the natural language processing model can be used to analyze and synthesize proteins to assist with discovering and developing new drugs. Utilizing the AI technique that powers ChatGPT, researchers can create protein-language models to study biological data.
“The models encode what might be called the grammar of proteins—the rules that govern which amino acid combinations yield specific therapeutic properties—to predict the sequences of letters that could become the basis of new drug molecules,” an article published on News Update explains. “As a result, the time required for the early stages of drug discovery could shrink from years to months.”
While the technology is in the very early stages, researchers and manufacturers are leveraging protein-language models to optimize established molecules, including improving the efficacy of existing drugs and drug candidates.
- Healthcare Delivery
Another possible benefit of ChatGPT and AI technology, in general, highlights its potential to revolutionize the future of medicine. By employing artificial intelligence to create intelligent processes and workflows, current healthcare delivery models can be restructured to be more efficient, effective, accessible, and affordable.
If allowed, ChatGPT can run in the background at all times, sensing patterns and missed opportunities, offering suggestions, and identifying trends. This will ultimately result in more personalized patient care and less strain on providers and the already overburdened healthcare system.
As the model becomes more advanced, it may eventually be able to diagnose patients and recommend treatments on its own. By leveraging the power of AI, healthcare providers would be able to better serve patients and significantly improve not only care delivery but also population health outcomes.
Interestingly, Google’s AI division, DeepMind, announced the release of MedPaLM a few short weeks after ChatGPT launched. Their 540-billion parameter large language model is specifically designed to answer healthcare-related questions and is based on open-question answering datasets covering professional medical exams, research, and consumer questions.
While MedPaLM could overtake ChatGPT in its clinical applications (following substantial refinement), DeepMind concludes that the tool “performs encouragingly, but remains inferior to clinicians” in a paper published last month.
Limitations And Their Implications
The AI chatbot will remain inferior to clinicians for the foreseeable future. Due to the intricate, complex nature and high stakes of healthcare, the deployment of artificial intelligence models for patient care carries significant risks.
As is usually the case, caution is highly advised when dealing with novel technologies in the industry due to the potential for fatal outcomes.
The legal considerations of using artificial intelligence for clinical purposes are many and remain to be fully elucidated as the technology’s applications in clinical practice expand.
- Un-Humanness
Dr. Sahil Mehta, a former interviewer for prominent medical schools, investigated the chatbot’s performance as a potential candidate by conducting an interview with the model. If ChatGPT were a human student, they would have passed the majority of examination questions with flying colors. Yet the chatbot’s inability to display empathy, recall memories, or possess other innately human characteristics made personal interview questions impossible to answer.
While ChatGPT cannot provide sufficient evidence of its “humanness” as it lacks personal experiences and feelings, it excels in answering difficult questions, presenting thoughtful arguments, and having complex conversations.
A report published earlier this month suggests the technology can write convincing fake scientific abstracts – something that does not require emotion. The findings of a recent study conducted by Northwestern Medicine researchers revealed that reviewers of a pool of scientific abstracts could only detect fake abstracts 68% of the time. This suggests the potential of ChatGPT’s ability to “outsmart” humans and undermines its universal access if used inappropriately.
- Accuracy
OpenAI, the model’s developer, acknowledges that ChatGPT is not connected to the internet and can therefore produce incorrect answers. Limiting its accuracy even further is the fact that ChatGPT has limited knowledge of “events after 2021” or the latest version of data it was trained on.
Furthermore, the AI model is rewarded for recognizing and predicting patterns in its training data. It then uses that information during run-time, but in the medical setting, especially in patient note summarization, it can inaccurately predict events and treatments that never happened.
- Patient Privacy
Another critical limitation of ChatGPT is that it does not support services covered under the Health Insurance Portability and Accountability Act (HIPAA) through accessing protected health information (PHI). Therefore, using the model in healthcare workflows in which OpenAI can access protected patient data, for example, clinical note analysis and summary, violates the Act and terms of use.
- Fear
Despite the immense potential of these novel NLP techniques, companies like OpenAI may be hesitant to implement them in healthcare, given the recent failure of AI with IBM Watson Health. Instead, innovation in NLP for healthcare will likely come from smaller startup companies rather than big tech.
It is important to note that ChatGPT is still a machine-learning model and is not yet capable of replacing human doctors or nurses. It is a tool that can be used to support and enhance the work of healthcare professionals.
The Future Is Already Here
The healthcare industry is on the cusp of a major revolution spurred by advancements in natural language processing (NLP) techniques and other artificial intelligence technologies. With the launch of ChatGPT and the latest developments in healthcare-related AI, we are being offered a glimpse at the future of medicine.
The abundance of healthcare data that is becoming available will revolutionize how we live and approach healthcare, putting the power of exponential technologies in the hands of patients. With ubiquitous access to wide-ranging data, researchers aided by AI will dramatically accelerate longevity science, amplifying the human healthspan as the medical care paradigm shifts from acute symptom management to long-term, continuous monitoring.
However, the question remains, are we ready for this revolution?
As we continue to push the boundaries of AI and NLP in healthcare, it is essential to ensure that we are prepared to harness the power of AI and data to improve the health and well-being of the population. It is also crucial that we are prepared for the ethical, privacy, security, misuse, misconduct, and societal implications that come with this technology. Finally, it is increasingly evident that in order to stay relevant and competitive in the healthcare industry, it is crucial to adapt to these changes and embrace the opportunities that AI presents.
