“For AI to add the most value and for patients and physicians to embrace it, it needs to support, not supplant, the patient-physician relationship … AI will be most effective when it enhances physicians’ ability to focus their full attention on the patient by shifting the physicians’ responsibilities away from transactional tasks toward personalized care that lies at the heart of human healing.”
Steven Lin, MD, Vice Chief of Technology Innovation, Stanford University
Overview
Generative AI like ChatGPT is truly exciting, and it’s easy to be seduced by the technology’s potential to produce, well, almost any sort of output. Be careful. The opportunity in generative AI is enormous but requires a thorough analysis of where the best applications lie. Healthcare, in particular, needs this assessment – this isn’t an industry known for fast change, and the risks of inappropriately deploying new technology can be huge. For instance, consider the hype around IBM’s Watson Health a few years ago; this AI was going to figure out complex cancers! It didn’t, and it was sold off cheaply in parts last year.
Few industries are as data-rich, text-heavy, and in critical demand for automation as healthcare. Beyond these attributes, there is sharp information asymmetry that exists both for patients and for clinicians; patients want to be better informed about their health, and clinical teams yearn for more timely, easily accessible insights on their patients and populations to better inform their delivery of healthcare.
In this context, generative AI models, particularly LLMs such as GPT-3.5 (and others such as Google’s PaLM) and derived technologies, such as chatGPT, can potentially transform healthcare. On the patient side, generative AI can generate rich, accurate medical advice and information (from multiple independent sources) to better inform and educate patients on their condition or symptoms. On the clinician side, generative AI can potentially reduce the administrative burden on clinicians, for example, by automating tasks such as writing referral letters, clinical coding, and the summarization of clinical consultations.
With a market opportunity worth a colossal $6 trillion, according to Morgan Stanley, it’s impossible to ignore the impact that this technology is set to have. Generative AI isn’t just a passing trend; it’s a rapidly evolving ecosystem of tools growing in popularity and showing tremendous potential to revolutionize healthcare in ways we’ve never seen before. This technology encompasses much more than just ChatGPT and can handle various data types, including text, images, audio, video, 3D modeling, and even coding. While some estimates suggest that it could even raise global GDP by 7% over 10 years, the potential applications of generative AI go well beyond just economic gains.
Exploring the potential use cases
There are several ways to segment the potential use cases for generative AI in healthcare. Boston Consulting Group (BCG) did one excellent approach in their June 2023 analysis of the state of the market.
For each segment—providers, pharmaceutical firms, payers, MedTech, services and operations, and public-health agencies—they grouped the options into three categories: validated solutions already on the market, early-stage or conceptual use cases, and potential future use cases not yet in development. You can read their analysis and approach here.
While I like the segmentation described in the BCG research, I think that most healthcare organizations will look to a more straightforward approach to deploying generative AI to solve pressing issues where they can realize immediate benefits. Generative AI is evolving at record speed while CEOs are still learning the technology’s business value and risks. Hospital CEOs tend to share three core business challenges: Negative margins, staff recruitment and retention, and staff burnout. What if generative AI could remove administrative and documentation burdens, help with the pre-op workflow, and assist with appealing claims denials? Many potential low-risk use cases could result in more time clinicians spend with patients. I offer some potential initial use cases to consider.
First, go after operational efficiencies – A study by the National Bureau of Economic Research found that access to a generative AI-based conversational assistant increases workers’ productivity by 14% on average, “as measured by issues resolved per hour.” Healthcare providers can reduce the administrative burden by implementing generative AI for various use cases, including enhancing member communications through digital channels and acting as a physician scribe. Using generative AI systems, physicians can automate the extraction of medically relevant information from discussion recordings, summarize the interaction and integrate notes into EHR systems. This can result in improved physician productivity and increased accuracy of patient data. Providers can also lower costs and offer improved member experience because generative AI enables more personalized and proactive communications through virtual agents (chatbots).
Generative AI use cases can impact capabilities across the payer value chain. As in the case of providers, communication through digital channels like chatbots can be significantly enhanced by tapping into the knowledge base of the payer, including policy documents. Generative AI can also be used to auto-generate approval and denial letters—this would encompass supporting responses to prior authorization requests and claim requests to improve speed and effectiveness.
Next, I’d look at clinical decision–making – Generative AI is already assisting doctors and medical professionals in making accurate and informed diagnoses. Generative AI can analyze data from a patient’s medical records, lab results, previous treatments, and medical imaging, such as MRIs and X-rays, to identify potential problem areas and suggest further testing or treatment options. One example proving this ability is Glass.Health who have created a generative AI tool capable of generating diagnoses and clinical plans based on the input of symptoms. By integrating data traditionally in the electronic health record (EHR) and data from outside the EHR, like social determinants of health and social networking data, generative AI algorithms could help identify chronic diseases earlier to improve health outcomes. This could help healthcare providers make more accurate and timely diagnoses, leading to earlier treatment and better patient outcomes.
Third, I’d look at personalized medication management and home care – The ability to provide customized care is essential in today’s healthcare landscape. Wearable devices can collect real-time, continuous data on a patient’s health indicators, including heart rate variability, blood oxygen, and blood glucose levels. The data can then be fed into generative AI algorithms, which can analyze and interpret the data and offer tailored recommendations and treatment options. In this way, AI would be deployed to manage diseases detected by wearables, like cardiovascular disease, for example. By leveraging wearables and at-home monitoring devices with generative AI, healthcare providers can move away from the traditional, reactive healthcare model to a proactive one.
“NHS England has previously reported there are 18 million GP appointments and 2.1 million visits to A&E every year for conditions that could be dealt with at home, costing £850 million. By empowering patients across the UK with verified information that would allow them to manage self-treatable conditions instead of going to a GP, millions of unnecessary NHS appointments could be released. Through the Healthwords platform, Clinova aims to effect this change and improve efficiency in the British healthcare system by relieving pressure on heavily burdened GPs and freeing up appointments for those who need them most.”
Sir Simon Burns, Government Liaison Officer for Clinova and former MP and Former Minister in the Department of Health
Finally, I’d explore risk prediction and pandemic preparedness – According to National Geographic, there are more viruses in existence than stars in the universe, and on average, approximately two new species of human viruses emerge yearly. As we have seen with the recent global pandemic, a new human-to-human virus without prior immunity could rapidly escalate into a pandemic, leading to millions of fatalities. Generative AI models have emerged as a vital source of insights for scientists studying the societal-scale effects of catastrophic events, such as modeling new pandemics and developing preventive measures. For instance, new generative AI models are being trained on large amounts of protein sequences to identify new antibodies which could address infectious diseases and support outbreak response.
What should we worry about? – While there are many potential benefits to using generative AI in healthcare, there are also some possible challenges and drawbacks. Some examples include:
Privacy and security: Patient privacy is strictly regulated. The use of generative AI in healthcare also raises concerns about protecting patient privacy, sensitive medical data, and the potential for misuse or unauthorized access to the healthcare data.
Bias and discrimination: Generative AI algorithms can be prone to bias and discrimination, especially if they are trained on healthcare data not representative of the population they are intended to serve. This can result in unfair or inaccurate medical diagnoses or treatment plans for underprivileged groups such as women or non-white races.
Misuse and over-reliance: If generative AI algorithms are not used properly, they can lead to incorrect or harmful medical decisions. In addition, there is a risk that healthcare providers may become overly reliant on these algorithms and lose the ability to make independent judgments.
Ethical considerations: Using generative AI in healthcare raises several ethical concerns, such as the potential impact on employment in the healthcare sector.
What questions should leaders ask? – In health systems considering generative AI, executives will want to quickly identify the parts of their business where the technology could have the most immediate impact and implement a mechanism to monitor it, given that it is expected to evolve quickly. A no-regrets move is to assemble a cross-functional team, including data science practitioners, legal experts, and functional system leaders, to think through essential questions such as these:
- Where might the technology aid or disrupt our industry and/or our system’s value chain?
- What are our policies and posture? For example, are we watchfully waiting to see how the technology evolves, investing in pilots, or looking to build a new service? Should the posture vary across areas of the system?
- Given the limitations of the models, what are our criteria for selecting use cases to target?
- How do we pursue building an effective ecosystem of partners, communities, and platforms?
- What legal and community standards should these models adhere to so we can maintain trust with our stakeholders?
Who’s applying generative AI in healthcare today? – The University of Kansas Health System is rolling out generative AI throughout its Kansas City-area medical centers. In one of the earliest large-scale uses of generative AI, the system is making Abridge AI Inc.’s application available to its 1,500 physicians and other clinicians. At UPMC, a minority investor in Abridge, a small cohort of clinicians uses technology from Abridge to document interactions with patients automatically. Abridge “listens” to conversations between a patient and their health care provider and extracts the critical points, like a change in medication or behavior, and creates notes for the patient and the electronic record (EHR). So far, both patients’ and clinicians’ reactions have been positive. UNC Health, meanwhile, is another early adopter. It has agreed to participate in a much smaller generative AI pilot with EHR giant Epic. The initial rollout will begin with five to 10 physicians at UNC Health using the technology to auto-draft responses to common patient questions that are time-consuming to answer. UC San Diego Health, UW Health, and Stanford Health Care also are participating in the pilot.
What’s next for generative AI – While the emergence of generative AI is exciting for many in the healthcare industry, it’s natural for others to feel nervous and uncertain of its future. Nevertheless, the potential for revolutionary progress within healthcare is undeniable. As such, the choices made by healthcare providers, practitioners, policymakers, and other stakeholders in the coming years will be critical in shaping the evolution of this technology.
As with any innovation, we must approach generative AI cautiously, acknowledging that its impact could be transformative, provided we adapt to its unique challenges and opportunities. Moments like this don’t come around often. A future marked by generative AI technology will usher healthcare into a new era of innovation, and those daring to experiment and lead in this space will help create opportunities for patients, providers, and healthcare institutions alike.