OBJECTIVE: We sought to evaluate the knowledge of probable outcome by medical personnel for in-hospital and out-of-hospital cardiac arrests, and self-reported history of CPR training referrals for family members of cardiac patients. METHODS: One hundred people from each of three population lists were randomly selected at a large, urban school of medicine and affiliated medical center: (1) year III and IV medical students; (2) residents in family medicine, emergency medicine, internal medicine, anesthesia, and surgery; (3) attending physicians in the same departments. A questionnaire was distributed that elicited estimates of in-hospital and out-of-hospital cardiac arrest (IHCA and OHCA, respectively) survival rates, and CPR training referral history. Estimates were compared against published data for accuracy (IHCA: 5-20%; OHCA 1-10%) RESULTS: The overall response rate was 63%. Accurate in-hospital cardiac arrest estimates [% (95% CI)] of survival were provided by 51.1% (36.8-63.4%), 47.3% (35.9-58.7%), and 36.7% (23.2-50.2%) of students, residents, and attending physicians, respectively. Accurate out-of-hospital estimates of survival were provided by 51.1% (36.8-63.4%), 52.1% (40.6-63.5%), and 70.8% (57.9-83.7%), respectively. Most thought that family members of cardiac patients ought to be CPR trained (92.6%). However, few had referred any for training in the past year (16.5%). There was strong support across respondent groups for including death notification information in the ACLS training program, with 80.4% of all respondents in favor. CONCLUSIONS: This study demonstrates that medical experience is not associated with accurate estimates of cardiac arrest survival. Overwhelmingly, medical personnel believe family members should be trained to perform CPR, however, few refer family members for CPR training.