The use of contraception and abortion are common, but numerous studies have shown that women have poor knowledge about the safety and risks of both contraception and abortion [1-10]. Two recent studies, one from Canada and one from the United States, have shown that attitudes about abortion, specifically the belief that abortion should be restricted, is associated with poorer knowledge about contraception and abortion [9,10]. Poor knowledge about contraceptive methods and abortion often involve the overestimation of risks with many respondents erroneously believing that both contraception and abortion are more dangerous than pregnancy and giving birth [5,6,8- 10]. There have been no study reports comparing different countries with respect to the association of attitude to abortion and knowledge of contraception and abortion.

There is robust evidence that abortions are safe procedures with a mortality rate of less than 1:100,000 and that abortions do not increase the risk of mental illness, breast cancer or infertility [11-16]. However, several studies reveal that women overestimate the risks of mental illness, breast cancer and infertility associated with abortion [8-10]. In a US study of 67 women returning to the clinic after having had an abortion, three-quarters of participants overestimated the health risks of a first trimester abortion compared to the risks of continuing a pregnancy and giving birth; 43% of participants overestimated the risk of mental health problems like depression after first trimester abortion for an unplanned pregnancy compared to continuing a pregnancy and 22% incorrectly answered that having one first trimester abortion will make it more difficult to get pregnant in the future [8]. A Canadian survey of 978 women having abortions found similar misconceptions with respondents overestimating the risks associated with abortion; with two-third indicating that the health risks of abortion were the same as or higher than birth; almost a third that the risks of depression were higher than after birth; 28% said there was an increased risk of infertility; and almost 1 in 10 that there was a higher risk of breast cancer. Additionally, one-third of the women in this study believed that abortion should be restricted [9]. Restrictive beliefs about abortion were associated with overestimation of the risks of abortion and greater anxiety after an abortion procedure [9,17]. An association between restrictive beliefs about abortion and poor knowledge about the safety and consequences of abortion and poor knowledge about contraception was also shown in an internet survey of reproductive aged men and women [10].

The purpose of this survey was to describe and compare knowledge and attitudes about contraception and abortion of women in Canada, US, UK, France and Australia and to examine the following questions:

1. Does the association between the attitude that abortion should be restricted and poor knowledge about the risks of contraception and abortion is similar across different countries?

2. Which countries and which demographic characteristics are associated with lower knowledge about contraception and abortion risk?

The more we understand about these prevalent misconceptions overestimating the risks of contraception and abortion, the better we will be able to educate women about the true risks, reduce their anxiety and decrease unintended pregnancies.

Surveys are an important source of collecting standardized information. Despite numerous strategies to increase the response rates, [18,19] there has been a steady downward trend in response rates to surveys [20]. The increase in use of mobile phones and Internet has introduced new opportunities and challenges to traditional methods of survey data collection. There are several online consultation platforms (registries of potential participants combined with ability to create and administer surveys) that are currently available to researchers. However, most have been established by private sector companies and are primarily focused on marketing research (such as: SurveyMonkey Audience–www. surveymonkey.com/mp/audience; Leger–www.leger360.com/canada_en/ legerwebpanel.asp; and Ipsos–www.ipsos.ca/en/products-tools/ipsospanels/ online-panels/). As such, they offer limited reliability, transparency and control over the entire research process, and their appropriateness for conducting scientifically sound research remains largely unknown. The second objective of this study was, therefore, to assess the usefulness and validity of online surveys for this type of research.

We used SurveyMonkey Audience that has a panel of about 3 million people in the US, about 500,000 in Canada, about 350,000 in the UK, and about 200,000 in France and Australia willing to participate in a variety of surveys. We purchased a panel of women aged 18-44 (at least 200 per country) in January 2013. There were no inclusion/exclusion criteria other than sex, age and country. The survey asked demographics (age, education, number of children and number of abortions), attitude to abortion and knowledge about risks of IUDs and abortion vs births. The attitude question was: ”Which ONE of the opinions best represents your view about abortion in the first 3 months of pregnancy?” followed by 4 possible responses, “Abortion should never be allowed, Abortion should be allowed ONLY if the life and health of the woman is in danger, Abortion should be allowed if, due to personal reasons, the woman would have difficulty in caring for the child, Abortion should be allowed for ANY reason, because no one should be forced to continue a pregnancy” This question had been used in previous research [6,7]. The knowledge questions were modified from previous studies [6,7]. We piloted these surveys in Canada, US, and Australia and had them reviewed by our colleagues in each country for appropriate wording. Approval was granted from University of British Coumbia Research Ethics Board H12-03384

Surveys were conducted via SurveyMonkey Audience. Survey takers are randomly selected by Survey Monkey from their pool of participants based on the demographic features requested. For each survey completed by a Survey participant, a $0.50 donation to the charity of their choice is made by SurveyMonkey, and the Survey taker is entered into lottery (by SurveyMonkey) for a $100 gift card. SurveyMonkey provided data in the form of survey responses as an SPSS (IBM Statistical Package for Social Scientists version 21) file. The data contained no identifying information about participants. The questions were adapted from previous research and customized or translated for each country. The knowledge questions were presented first and were mandatory (i.e., the women did not get credit for their survey if they did not answer these) and the demographic questions were last and were not mandatory. France does not allow ethnicity and immigrant status questions, so these were omitted for the French questionnaires. Data were analyzed using SPSS v 20. Women were asked to answer which statement best reflects their view about abortion. For the purpose of this study, women choosing the response, “Abortion should be allowed for any reason, because no one should be forced to continue a pregnancy” were categorized as “pro-choice” and those choosing one of other responses were categorized as “anti-choice”. A similar categorization technique for women’s attitudes about abortion was described previously [7]. Correct responses to the knowledge questions were tallied and compared with respect to country and attitude to abortion.

To assess the representativeness of the respondents, sociodemographic profiles of respondents as well as the responses to abortion-specific questions obtained from the SurveyMonkey Audience were compared with country-specific census data and the results of published surveys that used probability-based samples (e.g., using random digit dialing) and asked the same or similar questions.

Confidentiality

SurveyMonkey is an online survey company located in the USA, subject to the US Patriot Act, which allows authorities access to the records of internet service providers. The company servers record incoming IP addresses of the computer that subjects use to access the survey, but no linkage is made between the data and the computer’s IP address. The security and privacy policy for SurveyMonkey can be found at the following link: http://www.surveymonkey.com/mp/ policy/privacy-policy/.

Statistics

We chose our sample size on the basis of the question, “Do women categorized as “anti-choice” differ from the women categorized as “pro-choice” in their knowledge about health risks associated with abortion and contraception?” In our two previous studies, the 4 abortion knowledge questions were answered incorrectly by (1) 79% and 68%, (2) 46% and 31%, (3) 25% and 28%, (4) 6% and 9%. There were 34% and 52% who were classified as anti-choice. With a sample size of 200 women per country, we had sufficient statistical power to detect differences in proportion of incorrect answers of 14% or more (using two-tailed alpha of 0.05 and power of 80%). We used descriptive statistics to compare women across 5 countries with respect to knowledge of and attitudes to risks associated with abortion and IUDs. We used SSPS file to do chi-square tests on categorical variables and t-tests on continuous variables.

Email requests were sent by SurveyMonkey Audience to 20609, 13900, 50000, 10933, 16316 women in Canada, US, UK, France and Australia respectively in January 2013. Within two days, 1117 surveys were completed: 233 in Canada, 223 in the US, 230 in the UK, 221 in France and 210 in Australia. The demographics of our samples differed significantly between countries (Table 1). About half (47.1%) of the participants were classified as pro-choice because they indicated that women should be allowed to have an abortion for any reason in the first 3 months: 38.7% in Canada, 37.1% in USA, 42.0% in UK, 68.7% in France and 53.6% in Australia (p<.001; Table 2). There were many women who incorrectly believed abortion and contraception were more risky compared to births (Tables 3 and 4). Women classified as having anti-choice beliefs were more likely to provide incorrect answers to all 10 knowledge questions about abortion and contraception (p=<.001; Table 5). Specifically, women who believed abortion should be restricted were 12.5% more likely to provide incorrect answers to the question about abortion procedure risks, 14.4% about risks of breast cancer, 21.9% about risk of depression and 23.0% about infertility. When asked about contraception, women who believed abortion should be restricted were 10.8% more likely to provide incorrect answers to the question about risks of intrauterine devices, 11.6% about oral contraceptives and 11.8% about injectable contraceptives. There were few differences between the women of the different countries.

Table 1: Demographics of women surveyed in five countries.

Table 2: Attitude to abortion in women of five countries.

Table 3: Knowledge about risks of contraception (% correct answers).

Table 4: Knowledge about risks of abortion (% correct answers).

Table 5: Attitude to abortion and knowledge of risks of abortion (% correct answers).

We compared the demographics the five countries in (Table 6). For Canada we used the Statistics Canada census data and found that the online panel had a higher proportion of women with college degrees and women that reported no religion, were not married and had no children but a similar proportion of immigrants. For the US, our comparison data included US Census bureau and the National Vital Statistics Reports and the on-line panel had a higher proportion of women with college degrees and women who reported no religious affiliation, were not married and had no children but a similar proportion of immigrants. For the UK, our comparison data included the Office for National Statistics Census as well as the Northern Ireland Statistics and Research Agency data and the on-line panel had a higher proportion of women with college degrees and women who reported no religion and had no children but a similar proportion of immigrants and a higher proportion of married women. For France, our comparison data were from the Institut national de la statistique et des études économiques and the on-line panel had a lower proportion of women with college degrees but higher proportion of women who had no religion and had no children and a similar proportion of married women. For Australia, our comparison data was from the Australian Bureau of Statistics and the on-line panel had a lower proportion of women with college degrees but higher proportion of women who had no religion and had no children and a similar proportion of married women.

Table 6: Comparison of demographics in online sample and population samples.

We compared attitude data in (Tables 7a-e). For Canada we used an Angus Reid Public Opinion survey from 2013 and a telephone Forum Research survey from 2012 and found that the on-line panel had a lower proportion of women that we categorized as pro-choice about. For the US, our comparison data were a telephone survey by Gallop in 2011 and an on-line survey by Angus Reid in 2010 and the on-line panel had a higher proportion of women that we categorized as pro-choice about abortion. For the UK, our comparison data came from a face-to-face in-home interview study by Ipsos in 2011 and two on-line surveys by YouGov and Angus Reid in 2010 and 2011 and the on-line panel had both higher and lower proportions of women we categorized as pro-choice about abortion. For France, our comparison data were on-line surveys from the French Institute of Public Opinion (IFOP) in 2010 and 2014. The on-line panel had a lower proportion of women we categorized as pro-choice about abortion.For Australia, our comparison data were a telephone survey by Newspoll in 2013 and an online survey by Crespigny in 2008. The on-line panel had a lower proportion of women we categorized as pro-choice about abortion

Table 7a: Comparison of attitudes to abortion in surveys in five CANADA.

Table 7b: Comparison of attitudes to abortion in surveys in USA.

Table 7c: Comparison of attitudes to abortion in surveys in UK.

Table 7d: Comparison of attitudes to abortion in surveys in FRANCE.

Table 7e: Comparison of attitudes to abortion in surveys in AUSTRALIA.

Women who favored restrictions to abortion access also selected more incorrect answers and believed more exaggerated risks about abortion. This is consistent with the findings of two other surveys of women presenting for abortion in Canada and the US [5,7]. The lack of knowledge about the comparative risks of birth and contraception or abortion was pervasive through this sample and also is consistent with the finding from other studies [1-4]. These findings indicate that we need to improve education about the comparative risks of birth, abortion and contraception for all women. The women who believe abortion should be restricted may be more resistant to changing their beliefs about the risks of abortion and contraception and therefore education may not be as effective.

The advantages of using the survey panel are important; we were able to collect survey data from over 1000 reproductive aged women in five countries within a few days for less than US $9,000. All surveys have errors and bias in sampling and the issue here is whether the errors and bias in survey panels would invalidate our findings on knowledge and attitude about contraception and abortion. Although there were many differences in the demographics between our samples and the comparison samples, most of these differences were similar in size and quality to the differences between other surveys or samples in the same country. Each survey asked the knowledge and attitude questions differently and the national census asked marital status, immigration status and education level questions differently, making it difficult to make direct comparisons. These are problems that occur commonly when assessing the accuracy of survey data.

The first limitation of survey panels is the small percentage of the total population in these panels (between 0.5 and 1.5% in the five countries). These women had chosen to be on the panel to answer surveys, have access to computers and internet and may be different in numerous ways. We dealt with this problem by comparing our sample to other data from each country with respect to the factors we expected to be important for knowledge and attitude, for example, religiosity, marital status and education. The next limitation is the bias of who answered the questions within the panel. Since our survey was completed within days, only those panel members who answered the email request quickly were captured.

In more traditional surveying methods, there is also a bias towards people willing to answer surveys. In random digit dialing, there is a bias towards those who have land-based phone lines. It is impossible to have no bias in population based surveys.

Women from these 5 countries were similar in terms of their knowledge about the risks of abortion and contraception. The majority of women gave incorrect answers to the knowledge questions. Women who favored abortion restrictions to abortion access, in all five countries, were more likely to incorrectly overestimate the risks of both abortion and contraception. It is important that clinicians and educators are aware of this and work together to inform women. We found that on-line panels can be a useful method of assessing attitude and knowledge.

We thank our colleagues Sharon Cameron (UK), Elizabeth Aubeny (France) and David Grundman (Australia) for their advice.