Using individual patient data to adjust for indirectness did not successfully remove the bias in this case of comparative test accuracy

DSpace/Manakin Repository