An ambitious, multifaceted intervention at ICUs in 12 hospitals, led by J. Randall Curtis, attempted to change intensivist behavior in approaching end-of-life and palliative care with patients and families. However, families of dying patients receiving the intervention reported the same satisfaction with care as controls. There were no differences in ICU length of stay or time to withdrawal of life support. Patients / families were assigned by cluster randomization (each ICU was its own control) to minimize institutional or other biases. (n=2300) AJRCCM 2011;183:348-355.