To determine the prevalence of cardiopulmonary resuscitation (CPR) and do-not-attempt-resuscitation (DNAR) orders, to define factors associated with CPR/DNAR orders and to explore how physicians make and document these decisions. We prospectively reviewed CPR/DNAR forms of 1,446 patients admitted to the General Internal Medicine Department of the Geneva University Hospitals, a tertiary-care teaching hospital in Switzerland. We additionally administered a face-to-face survey to residents in charge of 206 patients including DNAR and CPR orders, with or without patient inclusion.; 21.2% of the patients had a DNAR order, 61.7% a CPR order and 17.1% had neither. The two main factors associated with DNAR orders were a worse prognosis and/or a worse quality of life. Others factors were an older age, cancer and psychiatric diagnoses, and the absence of decision-making capacity. Residents gave four major justifications for DNAR orders: important comorbid conditions (34%), the patients' or their family's resuscitation preferences (18%), the patients' age (14.2%), and the absence of decision-making capacity (8%). Residents who wrote DNAR orders were more experienced. In many of the DNAR or CPR forms (19.8 and 16%, respectively), the order was written using a variety of formulations. For 24% of the residents, the distinction between the resuscitation order and the care objective was not clear. 38% of the residents found the resuscitation form useful.; Patients' prognosis and quality of life were the two main independent factors associated with CPR/DNAR orders. However, in the majority of cases, residents evaluated prognosis only intuitively, and quality of life without involving the patients. The distinction between CPR/DNAR orders and the care objectives was not always clear. Specific training regarding CPR/DNAR orders is necessary to improve the CPR/DNAR decision process used by physicians.