To determine the incidence of amniotic fluid embolism and the incidence as modified by cesarean section and age. Data are from the National Hospital Discharge Survey. We examined the number of patients discharged from short-stay nonfederal hospitals throughout the United States, from 1980 through 2005, with a diagnostic code for amniotic fluid embolism. From 1980 through 2005, there were 112,712,000 deliveries, of which 12,000 patients (11/100,000) had amniotic fluid embolism. The incidence of amniotic fluid embolism was lower in 1993-2005 than in 1980-1992 (9/100,000 vs. 12/100,000)(p < 0.0001). The incidence of amniotic fluid embolism was higher with cesarean section, 5,000 of 22,937,000 (22/100,000) than with vaginal delivery, 7,000 of 89,775,000 (8/100,000) (relative risk 2.80, 95% CI 2.70-2.90) (p < 0.0001). The incidence was also higher in women aged 30-39 years, 6,000 of 35,039,000 (17/100,000) than in women aged 15-29 years, 6,000 of 77,673,000 (8/100,000) (relative risk 2.22, 95% CI 2.14-2.30) (p < 0.0001). The incidence of amniotic fluid embolism has decreased since the early 1990s. The risk is higher with cesarean section and higher in women aged > or =30 years.