<p>Refusal of blood transfusion is often framed as a conflict between competent patients’ refusal and clinicians’ responsibility to prevent death or serious harm. In Japan, many hospitals have adopted relative non-transfusion policies, which seek to avoid transfusion where possible while reserving emergency transfusion when necessary. Existing debates have asked whether such policies are ethically or legally justified. This study examines how they structure access and responsibility as hospital policies. We conducted a document analysis of publicly available transfusion refusal policies from all designated advanced treatment hospitals in Japan (n = 88), using qualitative content analysis and ethical interpretation. Publicly available institution-wide policies were identified in 50 hospitals (56.8%). Three approaches were identified: structured relative non-transfusion policies specifying thresholds, emergency transfusion provisions, and the clinical or institutional locus of decision-making (41/50); transfer-first or standard-care prerequisite policies (5/50); and low-specification or deferred policies (4/50). Emergency transfusion override provisions appeared in 45/50 policies, thresholds in 47/50, decision-making locus in 46/50, elective–emergency distinctions in 25/50, and transfer or referral recommendations in 33/50. These findings suggest that relative non-transfusion policies do more than specify whether emergency transfusion may be undertaken despite refusal. They organize conditions under which patients access care, emergency transfusion may be reserved, referral or transfer may be recommended, and institutions retain, redirect, or leave under-specified responsibility. Ethical evaluation should consider meaningful access, transparent thresholds, feasible referral pathways, case-specific assessment, and institutional support for bedside judgment before emergencies arise.</p>