Objective To investigate whether self-rated health (SRH) in pregnancy can predict childbirth complications, adverse birth outcomes, and maternal health problems up to 3 years after delivery. Methods A retrospective analysis was performed of data obtained in a prospective longitudinal population-based birth cohort study. Pregnant women resident in the Brno or Znojmo regions in the Czech Republic were included if they were expected to deliver between March 1991 and June 1992. SRH data were collected between 1991 and 1995 via pen-and-paper questionnaires administered in mid-pregnancy, and at 6 months, 18 months, and 3 years after delivery. Medical records were reviewed for pregnancy complications, childbirth complications, and birth outcomes. Multivariate regression analysis was performed. Results Overall, 4811 women were included. Better SRH in pregnancy predicted fewer childbirth complications (b =–0.03; P = 0.036); lower odds of cesarean delivery (odds ratio 0.81; P = 0.003); and fewer maternal health problems at 6 months (b =–0.32; P < 0.001), 18 months (b =–0.28; P < 0.001), and 3 years after delivery (b =–0.30; P < 0.001). The effects of SRH were independent of diagnosed complications and self-reported health problems in pregnancy. Conclusion SRH in pregnancy has predictive value for subsequent health outcomes, and might be an additional tool for assessment of pregnant women's health.