This population-based study employing after-death interviews with proxies describes older persons' preferences regarding medical care at the end of life. Interviews were held with 270 proxy respondents of 342 deceased persons (age range 59-91) in the Netherlands, The deceased were respondents to the Longitudinal Aging Study Amsterdam. The prevalence of advance directives (ADs), preferences for medical decisions at the end of life (i.e. withholding treatment, physician-assisted suicide euthanasia) and preferences about the focus of treatment in the last week of life (i.e. comfort care versus extending life) were examined. Written ADs were present in 14% of the sample. A quarter had designated a surrogate decision-maker. Co-morbidity and perceived self-efficacy (PSE) were positively associated with ADs. About half the sample had expressed a preference in favour or against one or more medical decisions at the end of life. Predictors positively associated with expressing a preference were co-morbidity, dying from cancer, and PSE. Being religious was negatively associated with expressing a preference. The knowledge of the proxy regarding the older person's preference for the focus of treatment was dependent on the patient's symptom burden as perceived by the proxy. The majority of older persons had died without either an AD, or having expressed preferences for end-of-life care. Stimulating the formulation of ADs may help professionals who work with older people to understand these preferences better, especially in the case of non-cancer patients and those with low PSE.