From December 2001 through September 2002, we mailed 1,500 questionnaires to a 10% random sample of OB-GYNs affiliated with the Brazilian Federation of Obstetricians and Gynecologists (FEBRASGO). At the time, FEBRASGO consisted of approximately 15,000 members. Our sample consisted of 10% of FEBRASGO members in each state, selected by assigning random digits to each member and then selecting a 10% random sample. In collaboration with FEBRASGO, we attempted to increase the response rate by using FEBRASGO stationery, publishing an advertisement in the FEBRASGO journal, sending reminder faxes, and offering a raffle. The Council's Institutional Review Board as well as FEBRASGO approved this study.
The anonymous questionnaire, which had undergone two rounds of pre-testing prior to fielding the study, contained 18 questions to assess respondents' knowledge of abortion law (specifically in cases of rape and life-threatening congenital malformations), their opinions of current abortion law in Brazil, their familiarity with various abortion procedures, and their experiences providing abortions. We also collected sociodemographic information, including sex, age, religion, and region of residence.
The primary outcome measure was opinion of abortion law. On the questionnaire, respondents were asked to choose one or more of the following circumstances in which he or she thought abortion should be legal in Brazil: life-threatening congenital malformation, rape, risk to a woman's life, risk to a woman's physical health, socioeconomic reasons, as an elective procedure, never legal, or under other circumstances. We classified respondents as conservative or liberal on abortion law based on their opinions of abortion law under the different circumstances. Conservative physicians were those who felt that abortion should only be legal under the circumstances codified in the current law (i.e., in rape cases or when the woman's life is in danger), or those physicians who thought the law should be more restrictive. Liberal physicians were those who felt that the current abortion law should be liberalized to allow for legal abortion under at least one other circumstance in addition to the two already permitted.
Using chi-square tests and crude logistic regression models, we determined which sociodemographic, knowledge-related or practice-related variables were associated with physician opinion. Following the bivariate analysis, we conducted a multivariate analysis of physicians' opinions of abortion law. In the initial multivariate analysis, we included variables that were associated with physician opinion of abortion law in the bivariate analysis at a p-value ≤ 0.10. We also included those variables that did not have a statistically significant relationship with opinion in the bivariate analysis, but were hypothesized to be associated with opinion based on the literature. For the final multivariate model, we used the Wald test to determine which variables were most highly associated with physician opinion of abortion law. In our final model, we include all variables that are significant at the p < 0.05 level as well as those that were marginally significant.
All analyses were conducted using SPSS statistical software version 10.0.6 and Stata, version 8.2.