Objectives To compare actual treatments with preferences for starting or forgoing treatment of older adults at the end of life. Design Mortality follow-back study of relatives of deceased older adults. Preferences and actual treatment were studied for each of four treatments: starting or forgoing resuscitation (do not resuscitate), artificial nutrition and hydration (ANH), antibiotics, and artificial respiration. Setting Older adults in the Netherlands. Participants Proxies of deceased members (in 2006-2009) of two cohorts representative of the older Dutch population (n = 168) and of people with an advance directive (n = 184). Measurements Relationship between preferred and actual treatment. Results In most individuals who preferred receiving treatment, this preference was followed (n = 2/2, resuscitation; 23/26, ANH; 33/38, antibiotics; 23/24, AR). In approximately half of the individuals who preferred that a treatment be forgone, the preference was followed (n = 6/13, resuscitation; 11/18, ANH; 3/5, antibiotics), except for artificial respiration (n = 1/8). The majority of people for whom no preference was known received treatment (n = 5/9, resuscitation; 19/33, ANH; 15/20, antibiotics; 8/13, artificial respiration). People with a known preference for receiving a specific treatment had a seven times higher chance of preference being followed than people with a known preference for forgoing that treatment. People with a known preference for forgoing a treatment had a six times higher chance of treatment being forgone than people having no known preference. Conclusion Although concordance between preferred and actual treatment is high in older adults who prefer treatment and lower in people who prefer no treatment, making preferences for forgoing treatment known is useful because it increases the chance of treatments being forgone in those who wish so. © 2013, The American Geriatrics Society.