Background: Physician Orders for Life-Sustaining Treatment (POLST) forms are used to document patient treatment preferences as medical orders. Prior research demonstrates that use of POLST alters medical treatments in a way that is consistent with the POLST orders. However, there are minimal data about the quality of POLST decisions, including whether they reflect the current preferences of well-informed patients. Objective: Evaluate the quality of POLST decisions. Design: Chart abstraction; interviews. Subjects: Nursing home residents and healthcare agents of incapacitated nursing home residents (n = 28). Measurements: Characteristics of the POLST conversation were assessed. Brief vignettes were used to assess knowledge about how POLST orders guide medical treatment. Current treatment preferences were elicited and compared with the patient's POLST orders to assess discordance. Results: A majority (59%) of participants recognized the POLST form. Participants were generally accurate in their knowledge of how POLST orders guide treatment concerning cardiopulmonary resuscitation (CPR) (68%), antibiotics (74%), and artificial nutrition (79%), but less so for medical interventions (50%). Current treatment preferences were initially discordant with one or more POLST orders for 64% (18/28) of participants, but half of these discordances were resolved with further discussion (e.g., participant agreed with the existing order). Discordance by treatment decision was as follows: CPR (7%), level of medical intervention (18%), antibiotics (21%), and artificial nutrition (11%). Conclusions: Discordance between current preferences and POLST orders is complex. Interventions are needed to support high-quality POLST decisions that are informed and concordant with current preferences.