JMIR Publications

ABSTRACT

Healthcare organizations increasingly procure artificial intelligence (AI) systems whose environmental impacts extend beyond the software itself to the energy, water, chemical, and community burdens of the supporting infrastructure, yet they lack a shared starting point for evaluating this information during procurement. Existing sustainability and procurement frameworks address parts of this problem, but we did not identify one that combines healthcare-specific, pre-contract use with disclosure expectations spanning energy, water, chemical, and community-siting dimensions. Drawing on a scan of existing frameworks, we propose a draft, discussion-stage question set covering 7 dimensions: disclosure quality, governance and accountability, energy and carbon, renewable energy, water, community burden, and lifecycle and chemical accountability. For each dimension, a request-for-information instrument records a vendor's disclosure status, verification, applicability, the responsible supply-chain actor, and documented effort to obtain missing information, rather than assigning a composite score; it characterizes disclosure, not environmental performance. Because smaller or downstream vendors often depend on upstream infrastructure providers, we distinguish a vendor's capacity to disclose from its documented effort to obtain unavailable information, an equity consideration central to the framework's design. Health systems and group purchasing organizations can use this instrument now as a discussion and negotiation tool, while standards bodies and policymakers develop complementary requirements for AI vendors and data-center operators. The instrument is unvalidated: further work is needed to establish inter-rater consistency, test applicability across vendor types, and clarify how disclosure relates to independently measured environmental outcomes.