HHS Section 1557 Rule on Clinical Algorithms

Section 1557 of the Affordable Care Act prohibits discrimination in covered health programs and activities on specified protected grounds. HHS’s 2024 final rule added 45 CFR 92.210 on nondiscrimination in the use of patient care decision support tools. The rule defines these tools broadly to include automated or non-automated methods used by a covered entity to support clinical decision-making, such as screening, risk prediction, diagnosis, treatment planning, health-care operations, and resource allocation as applied to patients. It therefore is not limited to machine-learning models. Covered entities must make reasonable efforts to identify patient care decision support tools that use input variables or factors measuring race, color, national origin, sex, age, or disability, and make reasonable efforts to mitigate the risk that their use discriminates on those bases. The rule does not categorically ban every tool that uses a protected characteristic. HHS states that a covered entity may routinely change tools and did not add a requirement to notify patients about each tool’s training data or assumptions. The legal status must be described carefully. HHS announced in June 2026 that a federal court had vacated specific provisions of the 2024 rule concerning expanded interpretations of sex discrimination tied to gender identity. HHS also said it would continue enforcement of protections involving race, color, national origin, age, disability, and unaffected aspects of sex discrimination. That notice does not say §92.210’s clinical decision-support provision was vacated. Covered entities should monitor court orders and HHS notices rather than assume the entire 2024 rule is either fully enforceable in all respects or entirely void. A covered provider, payer, or health program should inventory patient decision tools, identify which ones use factors related to protected characteristics, assess potential disparate effects, and document reasonable mitigation. Review vendor materials, validate tools for the intended population and setting, monitor performance, and retain a path to correct clinical decisions. Other civil-rights, privacy, device, and state laws may also apply.