Diversity MD
Source: Forbes
Written by Sahil Gupta
February 6, 2023– The process of becoming a physician is notoriously arduous, requiring years of specialized study and training. Before applying for a medical license in the U.S., aspiring physicians must pass the three-step United States Medical Licensing Examination, which covers topics including basic sciences, clinical knowledge and patient treatment and diagnosis. Most students take Step 1 at the end of their second year of medical school, Step 2 in their fourth year and Step 3 during their first or second year of residency.
According to a recent research experiment—which has not yet been peer-reviewed—ChatGPT, the artificial intelligence chatbot created by OpenAI, demonstrated that it was capable of passing all three parts of the USMLE without supplementary medical training. The USMLE’s passing threshold is approximately 60%. ChatGPT achieved over 50% accuracy across all examinations and surpassed 60% in most.
In late 2022, Google Research and AI company DeepMind published a paper outlining its work on Med-PaLM, an open-sourced large language model designed to generate accurate, useful answers in medical domains via multiple datasets. Med-PaLM has shown encouraging results, including 67.6% accuracy on MedQA (USMLE exam questions), but at the moment “remains inferior to clinicians.”
These are just two of the latest examples of how AI could continue to transform healthcare. I think that in the near future, AI will play an even more significant role in medicine, assisting physicians with clinical and administrative tasks to improve the quality, affordability and accessibility of care.
Decision Support
AI technologies are advancing rapidly, getting faster and more accurate at processing enormous volumes of data. While AI will not replace physicians, it can analyze large data sets and provide comprehensive answers to medical questions in a way that humans can’t.
Much of medicine is algorithm-based, taking into account factors such as patient medical history, risk factors, symptoms and test results to help make informed decisions about care. AI—while still imperfect and constantly improving—can be used as an effective decision-support system, providing physicians with valuable data to use in diagnosis and treatment plans. The human eye might miss a tiny abnormality on a CT scan, but AI can be trained to track the smallest details. Having an AI assistant could be like having an expert second opinion at your fingertips.
Creating greater access to this type of support may have considerable global health implications. There will be a projected shortage of 10 million healthcare workers by 2030, based on data from the World Health Organization, primarily in lower-income countries. AI could expand affordable access to quality healthcare in traditionally underserved parts of the world.
Administrative Support
Administrative work in healthcare is often repetitive, time-consuming and tedious. AI tools are capable of taking over many of these functions and allowing physicians to spend their time on patient-centered tasks.
A tool like ChatGPT could save physicians tremendous amounts of time in responding to insurance claims, for example. Instead of having to research and record documentation for each claim, justifying why a particular diagnosis requires an MRI or a blood test, a doctor could use an AI tool to draft a response and quickly revise it before sending.
Treatment Support
Treatment plans for many conditions are, by definition, algorithmic. There are clearly outlined steps physicians must follow, and if the first approach doesn’t work, they go on to the second. AI provides the opportunity to make treatment plans more personalized based on patient data.
In fertility treatment, for instance, there are standard dosing protocols to stimulate the growth of follicles in the ovaries. But the protocol can vary depending on factors including age, BMI or hormone levels. With the help of AI, physicians can analyze patient data more accurately to develop precise dosing for each individual. Fertility drugs are expensive, so this precision can save patients a considerable amount of money throughout the course of treatment.
We are on the verge of a fundamental change in healthcare. I believe that within the next five years, AI will offer diagnostic, treatment and administrative support that is far more accurate and efficient than what a single physician could provide. AI is creating the path forward, and we should be excited about where it will lead.
