Unsustainable costs and inefficiencies plague the healthcare industry. Expenditures have surged to levels that surpass GDP growth in certain developed nations. Beyond financial burdens, the sector faces physician and nurse burnout, resulting in elevated errors and a talent shortage. It’s estimated that waste in the U.S. healthcare system ranges from $760 billion to $935 billion, accounting for approximately 25% of healthcare spending. For example, when Mercy Health partnered with Microsoft Azure to create a smart data platform, they saw reduced patient stays, improved patient engagement, and streamlined administrative tasks – allowing staff to dedicate more time to patient care.
This inefficiency often stems from inadequate coordination among stakeholders, administrative complexity, and a payment model prioritizing capitation or fee-for-service rather than value-based care. It’s relatively easy to identify these issues, but another matter entirely to find workable solutions.
Among the more promising solutions to these challenges in healthcare is generative AI. In fact, generative AI could be as transformative for the industry as personalized medicine or GLP-1 drugs have been in the way they now mitigate obesity, addiction, and cardiovascular diseases, and more.
At the time of its entry into the market, ChatGPT had the fastest diffusion in history, reaching one million users just one week after launch and 100 million users within two months. The next year Instagram’s Threads would surpass this by achieving 100 million users in five days. ChatGPT’s rapid success, however, was particularly noteworthy considering the seemingly modest announcement made on what was then known as Twitter by Sam Altman, the OpenAI founder and CEO: “Today, we launched ChatGPT. Try talking with it here: chat.openai.com.” The ease of this announcement message underlines ChatGPT’s usability. It’s a significant advantage, the minimal learning curve needed to use the technology and the fact that users can interact with it through natural conversations. This accessibility is particularly helpful for adoption in healthcare, where technology can sometimes face resistance, for example, of time-consuming learning and onboarding processes.
Furthermore, generative AI medical skills are being tested globally, with promising results. In one blind study, patients evaluated responses from both human doctors and an AI chatbot on an online medical network. The patients felt that AI responses were more empathetic than those that came from the actual doctors, and notably, the chatbots provided more detailed explanations about medical conditions. While clinicians already use specialized tools like UpToDate – a commercial evidence-based medicine platform that offers peer-reviewed content across medical specialties – the real paradigm shift happening now is that patients can now access generative AI chatbots. The doctor-patient relationship is fundamentally changing.
For this type of transformation, strong leadership is essential.
Generative AI tools allow users to create custom GPTs, specialized versions designed for specific healthcare tasks. Medical professionals can use these to automatically code patient conversations into records, reducing paperwork while maintaining focus on patient care. For instance, a regional pilot of ambient artificial intelligence scribes implemented by The Permanente Medical Group demonstrated significant improvements, enabling real-time transcription of clinical encounters that reduced documentation burdens, enhanced patient-physician interaction, and consistently produced high-quality, structured clinical notes.
These tools can also improve Electronic Health Records (EHRs) by translating between the various coding languages used by different healthcare providers and countries, making records more standardized and shareable. Other potential applications include drafting equipment acquisition proposals, creating clinical presentations from medical reports, and developing integrated tools like medication dosage calculators that could offer cost-effective solutions for healthcare organizations worldwide.
In fact, a working paper from the National Bureau of Economic Research has found that wider adoption of AI could lead to savings of five to ten percent in US healthcare spending, approximately $200 billion to $360 billion annually. According to the researchers, the estimates are based on specific AI-enabled use cases that employ today’s technologies, are attainable within the next five years, and would not sacrifice quality or access.
Generative AI has great potential for practical solutions to healthcare’s pressing challenges, by automating administrative tasks, optimizing operational processes, or minimizing medical errors – all while improving quality of care. Implementing these solutions, however, presents its own set of challenges. As is the case with any disruptive technology, generative AI, particularly chatbots, may be received by healthcare professionals with resistance, skepticism, and apprehension. Success requires more than just technological advancements; it needs the right conditions for rapid, effective, and safe adoption. For this type of transformation, strong leadership is essential, to not only navigate through resistance but also distinguish between genuine expertise and self-proclaimed authorities in the sector.
We may question how generative AI will be integrated into healthcare organizations. As with any other technology developed thus far, generative AI will be adopted gradually by potential users, initially by the innovative and early adopters, following an epidemic-like pattern. In the healthcare sector, this adoption will inevitably be impeded by regulations that mandate evidence of its effectiveness and safety when the application pertains to clinical practice (i.e., testing, diagnosing, and treating patients). Healthcare organizations that are more engaged in research and clinical trials and possess more sophisticated health information technology will undoubtedly gain a clear advantage in integrating generative AI for clinical purposes.
Furthermore, generative AI diffusion will be more efficient for applications unrelated to clinical practice, such as reducing clinicians’ administrative burdens or scheduling personnel slots in the emergency room. These developments are not subject to regulation because they do not impact clinical work. Consequently, they can be developed and integrated more rapidly, even in smaller clinical facilities without research activities. This type of generative AI application has the potential to enhance healthcare efficiency. Additionally, the non-clinical applications of generative AI can serve as a catalyst to foster familiarity and receptiveness to adopting generative AI tools for clinical practice applications.
The adoption of generative AI will vary among medical specialties, as evidenced by the varying performance of AI chatbots across different disciplines. For instance, a recent study revealed that Gemini exhibits exceptional performance in Gastroenterology but poor performance in Cardiology.
Finally, it is imperative to acknowledge that millions of individuals worldwide utilize these tools continuously for a wide range of purposes. Among these users, we encounter both healthcare professionals and their future patients. The former continue employing Chat-GPT-like tools to augment their clinical tasks, fostering their willingness to adopt this technology. Patients have already demonstrated the adoption of generative AI to enhance their self-informed health status, interpretation of test results, validation of diagnoses, and determination of treatment appropriateness. Consequently, the increasing reliance on AI-empowered patients will inevitably foster the adoption of similar tools by doctors and nurses.
Ethical considerations around AI in healthcare have rightly come to the forefront. Critics argue that unclear training processes of these AI tools could lead to flaws and mistakes. However, it’s important to recognize that human clinicians are also vulnerable to biases and errors due to outdated knowledge or cognitive limitations. While some fear AI will replace clinicians, the reality of healthcare worker shortages and looming retirement trends suggest that AI will more likely complement and support the healthcare workforce rather than replace it.
Healthcare systems can start small when implementing changes with generative AI, for example by focusing first on areas where AI can lighten administrative burdens, improve record-keeping, and support clinical decisions. These first steps should be guided by doctors and nurses, who know first-hand which problems should go at the top of the list. Generative AI is the tool for developing tailored solutions and it will take us into a new age of healthcare. Adapting to technological advances is no longer an option, it is a necessity. Just as healthcare systems adapted to handle COVID-19, they must now evolve to embrace AI innovations. The pandemic showed us how quickly healthcare can transform when needed; this same adaptability is key as we navigate the integration of AI into medical practice.
© IE Insights.
