While the use of race in algorithms for clinical decision-making has received attention in legal scholarship, the use of sex, another protected class, has been largely ignored, despite being commonly used in many algorithms.
During the COVID-19 pandemic, men were more likely to be hospitalized and to die than women. This was true across geography and age. The difference became relevant tome when an epidemiologist from the UMB School of Medicine asked if I thought the Maryland Department of Health could legally use an algorithm preferencing men to allocate a scarce COVID treatment. Such a formula would implicate both the Fourteenth Amendment Equal Protection Clause and the Affordable Care Act’s antidiscrimination provision (Sec. 1557), but I was not sure of the answer.
That question led to a series of co-authored papers, first in Health Affairs, then in the Houston Journal of Health Law & Policy, and most recently in the New England Journal of Medicine (NEJM). In the Houston piece, we concluded that constitutionality under an intermediate scrutiny standard would depend upon whether courts view sex-based differences in COVID-19 outcomes as mediated primarily by biology, or by biological and socio-cultural factors.
Although we believe there is room for the courts to find that either basis is constitutionally valid, courts would be more likely to find it so if they judged that biological differences between males and females explain sex-based differences in COVID-19outcomes. We submit, however, that survival under strict scrutiny review (required under some state constitutions) and under the ACA, is far less certain.
Given the exigencies of a public health emergency and the limited relevant case law under Section 1557, we concluded that a motivated court could uphold the algorithm under the ACA. Without such a ruling, this uncertainty would have a chilling effect on state and healthcare system attempts to steward scarce lifesaving resources during public health emergencies and on other clinical algorithms that incorporate sex.
The NEJM article proposes a framework for determining when it is appropriate to include sex in algorithms for clinical decision-making, asking three questions:1) Is the use of the protected class necessary to achieve the government or hospital’s objective? 2) If yes, why is sex prognostically informative? Is it based on biology or unconscious biases in care delivery, sex-based stereotypes or assumptions, or historical medical underrepresentation? 3) If not biology, would the algorithm’s anticipated use “penalize” the disadvantaged sex from Question2? If so, sex’s inclusion is inappropriate if the algorithm is used in a nondiscretionary manner (i.e., to compel a decision rather than to prompt a discussion about risks).
We hope these questions will provide a starting point for healthcare providers considering the use of sex in algorithms for clinical decision-making and lead to a clinical consensus. Such a consensus could be highly relevant, and possibly persuasive, to courts and regulators adjudicating such questions.
Diane Hoffmann is the Jacob A. France Professor of Health Law and director of the Law & Health Care Program.