As the world stares down the barrel of climate change, the health care system — which is responsible for about 9% of all U.S. greenhouse gas emissions — is finally starting to take action.
Hospitals, in particular, are working to stem their impact, and have held up telehealth as a prime strategy to cut down on carbon, by eliminating millions of miles of travel to and from health care centers. But while digital health in many forms has significant potential to manage health care’s emissions, research into its impacts is still very nascent — and the measurements used often fall short.
“Digital health is late to the party,” said Enrico Coiera, a digital health researcher at Macquarie University who authored an editorial in the Journal of the Medical Informatics Association issuing a call to arms for the field to engage in climate change mitigation. “It’s very early days in terms of what the evidence says.”
Most research on the potential climate impacts of digital health has focused on the marginal benefits of trading one form of care for another: What happens when a patient sees their doctor over Zoom instead of going to the office? How much paper is saved when they check their records online instead of getting a printout? How much discarded personal protective equipment is avoided when a nurse takes a patient’s vital signs remotely?
One paper analyzing the records of a Spanish insurance company in 2020 estimated about 2,000 net tons of carbon dioxide emissions were avoided as a result of 640,000 digital appointments during the Covid-19 pandemic — after accounting for the increase in emissions associated with videoconferencing — and nearly 5,000 tons saved by shifts to digital report downloads. That’s the equivalent of about 8,000 round-trip, cross-country flights, thanks to telehealth.
“But actually, that’s only a small component of the digital health technology footprint,” said Zerina Lokmic-Tomkins, a nursing and informatics researcher at Monash University in Australia. “There’s this whole other life that nobody is aware of.”
Indeed, digital health’s carbon footprint stretches from electronic health records and e-prescriptions to online referrals and robotic surgery. “While people were looking at digital health technologies’ carbon emissions, they were doing it in a very fragmented way,” said Lokmic-Tomkins, who reviewed the methods used to measure digital health’s carbon impact in the JAMIA climate issue. “And that fragmented way seemed to be confined to things that are easy or easier to measure and estimate.”
The benefits of digital health adoption, Lokmic-Tomkins emphasized, should be balanced against their significant carbon impacts — which can be far more difficult to tabulate. There are upstream and downstream environmental costs associated with expanding digital care systems, from increased computing infrastructure and data storage to the mining and disposal of materials for a growing ecosystem of devices used to support remote care.
“I would take with a grain of salt any study that says any one thing is a solution,” said Coiera.
In particular, climate analyses of digital health need to consider the way a patient’s care trajectory changes when new technologies are introduced. “The easy thing to do is to imagine this little optimization, whether it’s a telehealth consult or something else, and then assume nothing else has changed,” said Coiera. But what happens to the time and money that patients and providers used to spend traveling to doctors’ offices, or to the technicians whose training is no longer necessary to interpret radiological images? One paper called these unknowns “rebound effects”: “If these alternate activities were high carbon generating, then there is a risk that much of the savings could be undermined.”
Consider a group of patients who received follow-up care via telehealth after being discharged from emergency rooms in Los Angeles. A recent analysis published in JAMA Network Open found they were more likely to return to the emergency department and to be hospitalized than those who had in-person visits. The initial telehealth follow-up may have cut carbon emissions — but overall, it led to more health care utilization, travel, and likely emissions.
That kind of careful unpacking is critical not just for careful deployment of remote care, but of a growing class of algorithmic clinical tools that require energy-intensive computing to develop. “One of the major issues facing digital health technology is actually data — the generation of data, how much of it is being stored, and for how long it’s being stored,” said Lokmic-Tomkins. “How do you redesign these infrastructures to be less carbon intensive?”
One recent paper found that substituting an ophthalmologist’s exam for diabetic retinopathy with an FDA-cleared autonomous AI screening could reduce greenhouse gas emissions by 80%. “Implementing autonomous AI in the health care setting is something that requires and creates emissions in its development,” said Risa Wolf, a pediatric endocrinologist at Johns Hopkins Hospital who published the paper with the AI company that built the tool. “But once it’s done, it’s highly scalable, and it also reduces the amount of provider and patient travel,” she added.
But that study didn’t include a full lifecycle analysis of the tool’s deployment. And not all clinical AI is created equal. Relatively few algorithms are purely autonomous; more common are “assistive” AIs that still require the oversight of a health care provider. Those, Wolf said, may in fact increase greenhouse gas emissions, because the technology requires additional resources without cutting down substantially on human power. “I think rigorous investigation of different systems in the future would be helpful,” said Wolf.
As artificial intelligence and other forms of digital health become a more established part of clinical care, climate advocates say it’s critical to focus on emissions from the earliest stages of development.
“If we do that, then we can definitely mitigate the digital health footprint overall,” said Lokmic-Tomkins. “If we don’t think about it, then there is always this need to create better, bigger, fancier, without that consideration of what it actually does down the track.”
Planning ahead could help the right people in digital health deploy tools that will simultaneously improve health outcomes and reduce carbon emissions. A recent survey of 85 health informaticists found that about three-quarters wanted to be involved in decarbonization — but nearly as many were unfamiliar with any tools that could help them address the impacts of climate change in their work.
And while focusing on carbon footprints could put a dent in businesses’ bottom line, Coiera emphasized that industry would be well-served by standing up and paying attention now. “There is plenty of opportunity here for people who get their act together,” said Coiera. “There’s no business in 20 year’s time if you don’t deal with climate change.”
This story is part of ongoing coverage of climate change and health, supported by a grant from The Commonwealth Fund.