Physicians' influence over decisions to forego life support

Objective: To quantify the association between individual physicians and decisions to limit use of life supporting therapies for critically ill patients. Study design: Prospective, observational data collected 2002 - 2005 in the adult medical intensive care unit (ICU) of a publicly owned teaching hospital. Nine intensivists staff this closed-model ICU; in rotating 2 week blocks of time one intensivist is responsible for directing all care. In order to uniquely associate care with individual physicians, eligible patients were cared for by a single intensivist throughout their ICU stay. Life support decisions were identified as orders to withhold or withdraw any form of life supporting medical therapy, including cardiopulmonary resuscitation (CPR), defibrillation, invasive mechanical ventilation, vasoactive drugs, and renal replacement therapy. We used multivariable Cox modeling to identify variables associated with decisions to limit use of life support. The association with the individual physicians was assessed as the hazard ratios of indicator variables representing the individual physicians. Results: A decision to limit use of life support was made in 191 (14.0%) of 1363 ICU admissions. The hazard ratios associated with individual intensivists spanned a 15-fold range (0.069 - 1.042; p=0.0003). Decisions to limit life support were more strongly related to the identity of the intensivist than to comorbid conditions, acute diagnostic category, and the source of ICU admission. Conclusions: We have, for the first time, quantified the association between individual physicians and decisions made to limit life support for critically ill patients. More research is needed to understand the nature and implications of this association.