ObjectiveTo examine prospectively the relationship between memories of intensive care (ICU) and levels of anxiety after ICU discharge, the stability of these memories with time, and their relationship to the development of acute posttraumatic stress disorder (PTSD)-related symptoms.DesignCase series cohort assessed by interview at 2 and 8 wks after ICU discharge.SettingDistrict general hospital (serving a population of 350,000) general intensive care unit.PatientsMemories of ICU and anxiety levels were studied in 45 patients after ICU discharge. Thirty patients were examined again at 8 wks to assess memory stability and development of acute PTSD-related symptoms.Measurements and Main ResultsStandardized interviews and questionnaires were used to assess memory for ICU, anxiety, and depression 2 wks after ICU discharge. In addition, PTSD-related symptoms and panic were assessed 8 wks after ICU discharge. A total of 33 of 45 patients had delusional memories from ICU at 2 wks; nine of the patients with delusional memories had no factual memories, and these patients had higher anxiety levels 2 wks after ICU discharge (p< .0001). Thirty patients had paired assessments at 2 and 8 wks. Those patients who had no factual recall of ICU but had delusional memories at 2 wks scored highly for PTSD-related symptoms and panic attacks at 8 wks (p= .023 and .014, respectively). The only predictors of possible acute PTSD-related symptoms at the 8-wk assessment were trait anxiety (p= .006) and having delusional memories without recall of factual events in the ICU at 2 wks (p< .0001). Only delusional memories were retained over time, whereas the recall of factual events in the ICU declined.ConclusionsWe propose that the development of acute PTSD-related symptoms may be related more to recall of delusions alone. This study suggests that even relatively unpleasant memories for real events during critical illness may give some protection from anxiety and the later development of PTSD-related symptoms when memories of delusions are prominent.