New procurement guide for hospitals, payers takes aim at AI's environmental impact

Health Care Without Harm (HCWH) has unveiled a new United States Artificial Intelligence Procurement Guide to address the environmental consequences of AI adoption in healthcare. The organization's guide aims to empower healthcare stakeholders to leverage their purchasing power to demand transparency from AI vendors in terms of environmental impact.

HCWH collaborated with health systems and payers to develop the guide, which provides a framework for identifying AI vendors and encouraging them to become more environmentally responsible.

The guide highlights various environmental issues associated with AI, such as air pollution, energy disruptions, rising electricity prices, community displacement, reduced water availability, noise pollution, waste, and lack of green space access. HCWH emphasizes that healthcare organizations possess the potential to protect both the planet and human health by wielding their collective influence.

Key recommendations from the guide include urging AI vendors to measure and reduce greenhouse gas emissions, disclose their energy mix and increase renewable energy use. Additionally, the guide encourages vendors to be transparent about data center energy efficiency, water use, and backup generators, as well as the impact on community and public health. The guide also suggests that AI vendors should help healthcare organizations reduce the environmental footprint of their own AI use.

To facilitate this process, the guide provides ten environmental disclosure questions for AI vendors, along with boilerplate language for healthcare organizations to use when making purchases. The guide also offers scoring guidance for the responses, allowing vendors to earn points for committing to provide data within a 12-month timeframe. After the initial year, vendors must provide the requested data to maintain their points.

The guide draws upon the Sustainably Advancing Health AI (SAHAI) Framework, which offers strategies for scheduling nonurgent processing during off-peak hours, using mixed-quality models, and leveraging optional AI features that can be turned off to reduce emissions and costs. By implementing these measures, healthcare organizations can save money while simultaneously promoting environmental sustainability.

Despite the progress made by some healthcare organizations in achieving Science Based Targets initiative (SBTi) commitments, the guide acknowledges that AI companies, particularly smaller ones, have yet to demonstrate the same level of commitment. The guide aims to support AI companies in adopting sustainable practices and encourage collective action.

Bancroft, HCWH's CEO, emphasizes the importance of healthcare's mission and obligation to ensure health is not optional. She stresses that AI vendors must safeguard the environment and wellbeing if they hope to secure business from healthcare stakeholders.