The increasing adoption of artificial intelligence (AI) in health care has prompted a new study from Cornell University that focuses on integrating sustainability into these technological advancements. The research introduces a framework called Sustainably Advancing Health AI (SAHAI), which provides a strategy for optimizing energy consumption and reducing emissions associated with AI applications in medical settings.
This framework takes into consideration the greenhouse gas emissions resulting from AI-enabled patient messaging and the cooling water required for data centers. It also explores various scenarios that could influence the emissions profile of a significant health system deploying such tools.
Dr. Chethan Sarabu, director of clinical innovation for the Health Tech Hub at Cornell Tech, emphasized the importance of considering both the ethical implications and the environmental impact of AI in health care: “Our framework encourages health care organizations, and also technology developers, to think about these different levers and figure out how to balance the promise of AI in health care with being not only mindful of the ethical side, but also the environmental footprint of AI.”
The research, co-authored by Udit Gupta, assistant professor of electrical and computer engineering at Cornell Tech, was published on September 12, 2023, in NEJM Catalyst. The lead author, Dr. Anu Ramachandran, is an emergency medicine physician and postdoctoral fellow in medical informatics at Stanford University School of Medicine. Other contributors include Shomit Ghose from the University of California, Berkeley, and Dr. Vivian Lee, an executive fellow at Harvard Business School.
The U.S. health care sector is increasingly turning to AI solutions to alleviate pressure on a strained workforce. The market for automated patient interactions is projected to grow to a staggering $187 billion within the next five years. Despite the benefits of AI in reducing workloads for medical professionals, there are significant concerns regarding the energy demands of these technologies.
When evaluating the carbon footprint of AI tools, the researchers analyzed an AI-generated messaging application implemented in a large academic health system. They found that operating this AI-powered tool for one year, with 3,000 physicians responding to an average of 50 messages per day, would generate approximately 48,000 kilograms of carbon dioxide (CO2). This amount is roughly equivalent to the annual CO2 absorption of about 2,300 trees, with a “tree-year” defined as the volume of CO2 that one tree captures each year.
The researchers utilized a lightweight generative pretrained transformer (GPT) model for their calculations, which demands less computational power than larger models. While this model can efficiently direct patients to providers or answer general inquiries, more complex tasks would necessitate a more powerful model.
Dr. Sarabu noted the importance of weighing factors such as energy usage and water consumption when deciding on the deployment of AI systems. “If you’re responding to a patient about routine follow-ups, small differences in model accuracy, such as the difference between 83% accuracy and 85% accuracy, may not be noticeable,” he explained. “But if you generate double the amount of emissions with 85% accuracy, that’s probably not striking a good balance.”
The researchers advocate for a proactive approach to sustainability, stressing that considerations should be made during the system design phase rather than retrofitting after implementation. Dr. Sarabu remarked, “We’re really in the early days of AI being implemented in health care, and what happens in the next three years or so is going to get baked into the system. If we make energy-conscious decisions right now, we’ll have a more efficient system.”
Gupta added that prioritizing sustainability does not necessarily entail sacrifices in performance. He pointed out that data centers located on renewable energy grids can significantly reduce operational emissions. “Hospitals can make decisions on where they want to run these AI workloads, prioritizing data centers that operate on renewable energy,” Gupta stated.
The researchers concluded that there exists a crucial opportunity to align the large-scale integration of AI in health care with environmental considerations. “Although the impacts of climate change weigh most heavily on vulnerable, lower-resource patients and the health systems that serve them, emissions generation is driven largely by high-income countries and high-resource health systems, which must consider and mitigate their contributions,” they wrote.
This research serves as a reminder that while AI has the potential to transform health care, its environmental implications must be addressed to ensure a sustainable future for health systems worldwide.
