Equity at the point of care: auditing AI-supported resource allocation in obstetric emergencies

Equity in artificial intelligence-supported obstetric emergency care should be assessed as a service outcome, not as a model property, because preventable harm is mediated through operational delay, escalation, and resource contention. This perspective synthesizes implementation-relevant literature and quality and safety principles to propose a practical approach for translating equity from an abstract aspiration into auditable operations. We argue for using “avoidable delay” as a shared denominator across emergencies (such as postpartum hemorrhage, hypertensive crises, and obstetric sepsis) and for evaluating equity across the full chain from risk detection to resource delivery. We propose a minimum fairness audit set that can be captured largely from routine timestamps and logs: consistency of triggering across comparable presentations; timeliness of first response and definitive treatment; readiness of critical resources (blood products at bedside, operating room access and anesthesia start, and monitored-bed availability); completion of escalation and transfer steps; and structured documentation of overrides, missing data, and exception reasons. We further outline governance requirements—clear cross-service accountability, change control with re-audit after threshold or workflow modifications, and patient-facing transparency—so that equity is treated as an accountable, measured, and managed risk item within routine quality improvement rather than a one-time publication metric. In this perspective, “AI-supported” is used as a pragmatic umbrella to encompass deployed algorithmic decision-support systems at the point of care, including static rules-based early warning triggers, machine-learning risk scores, and operational routing/queuing engines; the Minimum Fairness Audit Set (MFAS) audits the service-chain consequences of any such trigger when it is coupled to an executable pathway with auditable timestamps.