In a previous post, we delved into the technology of ChatGPT and its functionality. It is a cutting-edge AI technology that enables human-like conversation. It has taken the world by storm. It is not surprising that there has been much enthusiasm over its possible medical applications. In this second part of a 3-part series, we shall explore the potential applications of ChatGPT in healthcare.
One potential application is the automation of routine tasks such as answering frequently asked questions, scheduling appointments, managing bookings, composing notes and letters, as well as carrying out other low-value, repetitive administrative writing tasks. For example, one rheumatologist in Florida has used ChatGPT to generate first drafts of letters to insurance companies requesting pre-authorization of surgeries and procedures. The letter was not only completed with appropriate heading and formatting, but also included essential elements such as the explanation of treatment and correct use of references. Others have used ChatGPT to generate other documents such as patient-friendly descriptions of procedures or post-procedure instructions. ChatGPT can also be trained to understand medical codes, which enables it to create billing reports and claims automatically. Furthermore, by identifying missing or erroneous charges, ChatGPT can help with charge capture and correction. All the above-mentioned applications have the potential to free up many hours of a healthcare provider’s daily schedule and reduce their workload.
Another potential application is communication with patients through web portals outside of scheduled appointments. It can be burdensome for providers with busy clinic schedules as their patients increasingly demand real-time information about their conditions and explanation of the test results and imaging studies that patients can now access online 24/7. If programmed appropriately, ChatGPT could provide patients with immediate access to information about their health and conditions through chat or text while alleviating providers’ workload and stress.
ChatGPT excels in interpreting English-language texts. It is very likely that it could be programmed to assist providers to collect, summarize and analyze large volumes of health data including diagnostic test results, family and past history, symptoms and current medications from patient records. This, in turn, could allow healthcare providers to evaluate patient needs more efficiently by spotting early warning signs and trends. Ultimately, ChatGPT could function as a clinical decision support tool to recommend diagnostic procedures, medications and treatments to healthcare providers and assist them in delivering more personalized care.
In a similar vein, ChatGPT could potentially be designed to summarize and analyze thousands and thousands of long, detailed research papers expediently. It could be also used to sift through a multitude of clinical trials to assist providers and their patients in determining their eligibility.
Another promising future application of ChatGPT is the de novo design of drugs: the process of synthesizing protein molecules from scratch. The rationale behind this idea is based on the notion that amino acids, the chemical subunits of all proteins, could be represented by individual letters of the alphabet. The latter could, in turn, be used to characterize proteins as long sentence-like combinations, resulting in the creation of protein-language models. Such models could then be analyzed by a natural language processing model, such as ChatGPT, to predict the correct combinations and sequences of amino acids to create protein molecules with specific therapeutic properties. If this approach works as designed, the time required for the early stages of drug discovery would reduce dramatically, from years to months. It would also empower scientists to address diseases for which effective treatments have remained elusive.
These potential applications of ChatGPT offer a glimpse at the future of medicine where AI technologies could be deployed for patient care. However, it is important to consider the limitations and risks of this technology, which will be explored in our next post.