Mitigating clinical algorithmic discrimination (13 December 2024)

Summary

This article in JAMA Health Forum examines how the growing use of clinical algorithms exacerbates health disparities through perpetuating discrimination. Anirban Basu discusses five critical issues that should be considered when developing and using clinical algorithms to minimise algorithmic discrimination:

1. Should race belong in a clinical algorithm?
2. Do we then have to develop two separate algorithms for decision-making invoking compensation vs reward?
3. What happens if there is differential mismeasurement of race?
4. What happens when there are differential measurement errors in other predictors, including biological variables?
5. What happens when differential measurement errors exist in the clinical outcomes for which the algorithm is being developed?