To transition from nascent to initial readiness and de-implement their first harmful race-based algorithm, institutions should cultivate engagement among key stakeholders and leaders, identify available resources, and begin developing foundational policies and data systems.

• Core knowledge and commitment: Facilitate discussions among key stakeholders and leaders to explore the potential harms of race-based clinical algorithms in practice.
• Education and engagement: Initiate provider education and patient engagement initiatives to address implicit bias and medical racism, and begin planning for enforcement and tracking.
• Resources and policies: Identify initial resources from established, broad health equity initiatives to begin addressing algorithmic bias at your institution.
• Implementation and integration: Begin developing informal work plans to identify harmful uses of race in healthcare decision-making at the institution and document the implications on patient care.
• Data and measurement: Establish consistent practices for collecting race and ethnicity data across the institution.

To progress from initial to active readiness, institutions should identify an executive champion, allocate dedicated resources, and formalize policies and work plans to enable the de-implementation of multiple harmful race-based algorithms.

• Core knowledge and commitment: Secure the commitment of an executive champion who will lead institutional efforts to remove racial bias from clinical algorithms.
• Education and engagement: Implement comprehensive provider education and patient engagement strategies that include specific enforcement and accountability mechanisms.
• Resources and policies: Dedicate resources and formalize policies that will support the de-implementation of harmful race-based clinical algorithms at the institution.
• Implementation and integration: Create a formal work plan for integrating race-conscious algorithms and more nuanced shared decision-making approaches into electronic health records and clinical workflows.
• Data and measurement: Standardize race and ethnicity data collection and invest in the infrastructure needed to consistently stratify patient outcomes by race, ethnicity, and social determinants of health.

To advance from active to sustained readiness, institutions must demonstrate their commitment to program improvement and collective action across the ecosystem, through advanced data practices and seamless workflow integrations that lead to measurable improvements in health equity outcomes, including restorative actions.

• Core knowledge and commitment: Promote a deep appreciation of the historical context and challenges of harmful race modifiers across all levels of the organization.
• Education and engagement: Make provider education on algorithmic bias and health equity mandatory and ongoing and formalize patient engagement through a standardized framework emphasizing restorative justice.
• Resources and policies: Dedicate staff, funding, and policies to ensure program sustainability and active participation in efforts to advance health equity initiatives across the healthcare ecosystem.
• Implementation and integration: Eliminate harmful race modifiers and replace them with race-conscious, shared decision-making approaches that are seamless, standardized, and guided by best practices.
• Data and measurement: Build a robust data infrastructure that incorporates comprehensive metrics, including race, ethnicity, insurance status, age, geography, zip code, language, and genomics.