Unplanned or unwanted? A randomized study of national estimates of pregnancy intentions

Objective: To evaluate the effect of question wording on national estimates of pregnancy intentions. Design: Data drawn from a national probability survey. Setting: The FECOND study in France in 2010. Patient(s): Five thousand two hundred and seventy-two women and 3,373 men who reported 11,603 pregnancies. Intervention(s): Participants randomly assigned to answer 1 of 2 questions on whether they had planned or wanted each of their pregnancies. Main Outcomes Measure(s): Generalized estimated equation regression models used to test for differences in pregnancy intentions by question wording. Result(s): The use of different wording yielded a 6%point difference in estimates: 33.5%pregnancies were "unplanned," and 27.4%were "unwanted." The addition of information on reasons for not using contraception at the time of conception lead to significant recoding, which resulted in a significant reduction in the wording effect: 23.7% (95% CI 22.4-25.0) of pregnancies were unplanned, and 21.2% (95% CI 19.9-22.5) were unwanted. Results from the multivariate analysis confirm the greater chance of reporting an unplanned as compared with an unwanted pregnancy (relative risk 1.25 [95%CI 1.17-1.33]), even after recoding (relative risk 1.15 [95%CI 1.06-1.24]). Conclusion(s): This study shows the strong effect of question wording on estimates of pregnancy intentions. The results also support the value of adding information on reasons for nonuse of contraception when assessing pregnancy intentions.