ChatGPT and Generative AI are Transforming Healthcare
Welcome to this week’s edition of Vitamin Z, a newsletter about health and wellness with 3,000+ subscribers. If you are new, you can join here.
In this issue:
ChatGPT and other forms of generative AI are gaining steam in healthcare
Generative AI is powering new solutions in telehealth, diagnosis, patient engagement, and personalized medicine
The technology is expected to improve quality of care, reduce costs and expand access to care
I’ve spent the last few weeks diving into the world of ChatGPT and generative AI. Why? It’s shaping the future of healthcare.
The New York Times this week reported doctors are using ChatGPT to be more empathetic when dealing with patients. Google announced a partnership with the Mayo Clinic to use generative AI to improve healthcare.
Generative AI, the technology behind ChatGPT, is projected to grow faster in healthcare than any other industry, with a compound annual growth rate of 85% through 2027, according to BCG.
In this piece, I’ll explore 3 ways generative AI is changing healthcare:
Improving quality of care
Expanding access to care
Saving money
1. Improving Quality of Care
As with any new technology, it’s worth asking: What problems can generative AI help solve?
Healthcare is an industry rife with challenges. It’s weighed down by massive amounts of data, complex medical and scientific jargon, and challenging patient communication.
Generative AI excels in a few areas where healthcare typically struggles: a) Interpreting unstructured data; b) Explaining data in a coherent way; c) Engaging people in conversation, and d) Generating new ideas.
I made this diagram to illustrate some of the use cases for Generative AI in healthcare, rated by their maturity and market potential. (These are rough estimates based on my conversations with industry leaders.)
Diagnosis
One challenge in healthcare is dealing with huge volumes of patient data: lab tests, imaging studies, genetic tests, clinical notes. Generative AI tools can digest these and build a holistic view of a patient’s condition.
They can help doctors predict disease onset by analyzing a patient’s healthcare records (imaging, lab tests, genetic tests) and detecting early indicators that may lead to a specific disease.
The following diagram (from a recent article in Nature) illustrates how generative AI can help doctors make better, data-driven decisions:
In this example, consider a young medical resident who is treating a crashing patient, and turns to the GPT app on her smartphone for help. The technology analyzes the patient’s data and provides guidance for management—in this case, to administer insulin.
A start-up called Glass Health is using generative AI to help clinicians draft clinical plans and generate differential diagnosis (DDx) based on patient symptoms. In a study, users rated 84% of the system’s DDx as helpful and 71% as accurate.
Until they get FDA approval, AI-powered tools will not provide definitive diagnoses or create treatment plans for patients. Experts believe AI-powered diagnostic tools should be used as an aid, not a replacement for human judgment.
Personalized treatment
Tailoring medical treatment to a patient’s unique profile and disease state may become easier with ChatGPT.
Generative AI algorithms can analyze a patient’s medical history, genetic information, and lifestyle factors to create a personalized treatment plan.
For a cancer patient, generative AI can analyze the patient’s tumor DNA and identify the genetic mutations driving the cancer. Based on this information, the algorithm can recommend a personalized treatment plan that targets specific genetic mutations.
A few other ways generative AI can improve care:
Psychiatry: Generative AI can deliver cognitive behavior therapy via chatbots that mimic a human therapist. It can diagnose specific forms of depression and recommend best-in-class treatments tailored to that diagnosis.
Surgery: The algorithms can be helpful during surgical procedures by analyzing intraoperative data and providing real-time guidance to surgeons, thus improving the accuracy and safety of surgical procedures.
Specialist referrals: Generative AI can determine the most appropriate type of care for an individual patient: What services are suitable for this patient? What kind of visit, in-person or virtual? At what location? With which clinicians? (Source: Out-of-Pocket)
Investors are pouring money into tech startups building generative AI tools to support doctors, nurses and pharmacists:
Atropos Health, a startup based in Palo Alto, is developing a digital consult service powered by real-world data.
InpharmD uses GPT to summarize medical literature and provide answers to drug queries.
2. Expanding Access to Care
One silver lining of Covid was a boom in telehealth. More healthcare visits are now conducted virtually.
ChatGPT and AI health chatbots could power a new wave of telehealth solutions, making healthcare cheaper and more accessible.
These chatbots can help solve a key challenge in healthcare: Engaging people in their own care. The technology may be helpful in obtaining screening data (e.g. do you feel safe at home?) and providing talk therapy for low-acuity behavioral health issues.
As the New York Times has reported, doctors are using ChatGPT to find the right words to break bad news and express concerns about a patient’s suffering. The chatbots can assemble talking points for doctors, explain the pros and cons of treatment options, or translate them into Spanish.
Here are some other examples of how generative AI can expand access to healthcare:
Addiction treatment: A GPT-powered service called You Are Accountable provides peer support services, care coordination and social reinforcement to encourage recovery for patients in addiction treatment.
Finding medical information: A startup called Jeti AI is helping healthcare providers integrate ChatGPT into their websites. It can “help patients find contact info for your doctor, how to pay your bill, schedule an appointment, answer medical questions in any language,” said Jeti AI CEO Marc Fischer.
Lab tests / Imaging: Need to get a blood test or CT scan, but aren’t sure where to go? Generative AI tools could show you a list of locations near you and compare prices for lab tests and imaging studies, making it easier to make informed decisions about your healthcare.
3. Saving Money
The most near-term opportunity for generative AI in health: Reducing administrative burdens for healthcare providers and saving money for health systems.
Routine tasks like documentation, charting, and messaging take time away from patient care. Administrative burdens are a key driver of burnout and a contributing factor to our nation’s shortage of primary care providers.
The more of this stuff we can automate with generative AI, the more we can improve healthcare efficiency, reduce costs and enable stronger doctor-patient relationships.
Medical Charting
A company called Carbon Health last week launched a product for hands-free medical charting—an AI-enabled notes assistant. The software uses ambient voice technology to capture audio of a patient-doctor visit, letting doctors focus more time on care and less on typing.
Other companies in this space include Abridge AI, which helps doctors digitize virtual primary-care conversations; and Syntegra, which uses generative AI to create realistic copies of patient data for research.
Electronic Health Records (EHRs)
Physicians spend up to six hours a day logging notes in their EHR, which results in less time spent with patients and contributes to burnout. Generative AI can help make sense of the ocean of data in EHRs by extracting data automatically, organizing it, and identifying trends.
Microsoft and Epic, the market leader in EHRs, are working together to use GPT to automatically draft replies to patient messages. Generative AI can simplify explanation of benefits notices and writing prior authorization request forms.
Billing
As anyone who’s ever received a medical bill can tell you, trying to understand medical billing is like trying to read hieroglyphics. Generative AI can break down healthcare bills and explain payments and appeals for patients.
Other Opportunities in Healthcare AI
Medical education: Generative AI can be used as a training tool for medical students. It can take cases residents treated from the EHR and let them analyze the case afterward—go through the whole patient journey, look at the decisions and outcomes, and compare to historical data and other cases.
Drug design: The algorithms can help biomedical scientists generate new molecular structures optimized for certain outcomes, and test them in simulations to find the most suitable ones for a particular treatment.
Clinical trial design: AI tools can generate synthetic data for clinical trials, accelerating an expensive process by creating more data for simulation and hypothesis testing, covering diverse patient populations, treatment outcomes and adverse events.
Risks of Generative AI
Healthcare leaders say generative AI tools must overcome a few hurdles.
Accuracy / Accountability
A chatbot called Tessa, developed by the National Eating Disorders Association (NEDA), was recently taken down after users complained the tool was dispensing diet tips that were potentially harmful.
Generative AI tools should have transparency and help users interpret the output, so they can make informed decisions based on the recommendations.
Bias
If an algorithm is trained on biased data, or its results are applied in biased ways, it will reflect and perpetuate those biases. This could lead to disparities in healthcare outcomes.
Regulation
The FDA in early 2023 announced a new approach to regulating medical devices that incorporate AI and machine-learning software. The approach would enable the agency to track and evaluate a software product from pre-market development to post-market performance.
AI will either save the world or destroy humanity, depending on whom you ask.
In healthcare AI, there’s reason for optimism. Investors are betting on health startups focused on ChatGPT and generative AI. More people are learning how to deploy large language models for important healthcare use cases.
We need intelligent medical technology to meet our growing health challenges. The hope is these tools can save lives and improve healthcare for all of us.
Thank you for reading this week’s edition of Vitamin Z.
Until next time,
By Daniel Zahler
Hi there and thanks for reading. If you stumble on my newsletter, you will notice that I write about health and wellness, and ways to optimize cognitive, physical and emotional health. I work with the world’s leading healthcare and life sciences companies to develop innovative new solutions to improve health globally. I was trained as a research scientist at Harvard, and I serve as a GLG council member where I advise global business leaders on healthcare and technology innovation.
Enjoyed your reading experience?
Follow me on Twitter.
Hit reply with your feedback and ideas :)
Share this post with others.
This is actually encouraging, Dan. I just hope ChatGPT and its brethren are familiar with Asimov's Three Laws of Robotics.