A Rubric for Algorithmic Fairness for Predictive Analytics in Healthcare

Predictive models, which are increasingly used in clinical medicine, raise concerns about introducing or amplifying health disparities. Different conceptions of fairness, however, may be mutually exclusive in predictive models. The investigators have shown how different fairness considerations are salient in situations involving allocation of scarce health care resources among patients compared to situations where the patient and physician make a shared clinical decision that is consistent with the patient’s values and preferences.