This cross-sectional Web-based survey was embedded in a large Canadian research initiative called the PEGASUS project (Personalized Genomics for Prenatal Aneuploidy Screening Using Maternal Blood) aiming to validate the performance and utility of the NIPT in the general population. In this large initiative, our overarching aim was to inform the future implementation of a DA to foster SDM in the context of prenatal screening for Down syndrome. Ethics approval was obtained from the research ethics boards of the Centre intégré universitaire de santé et de services sociaux de la Capitale-Nationale (#2013-2014-29), the Centre intégré de santé et de services sociaux de Chaudière-Appalaches (CER-1415-910), and the CHU de Québec (#B14-02-1929). We used the CHERRIES (Checklist for Reporting Results of Internet E-Surveys) checklist to guide the reporting of our Web-based survey [58].

The theoretical framework of this study was the Theory of Planned Behavior (TPB), which is one of the highest-performing and commonly used theories for identifying the determinants of intention [59]. According to the TPB, the direct determinants of intention are attitude (perceived advantages of adopting the behavior), subjective norm (the perceived social pressure from significant others to perform the behavior), and perceived behavioral control (perceived control over performing the targeted behavior). These direct determinants of intention are respectively associated with indirect constructs: (1) attitudinal beliefs (perceived advantages and disadvantages of using a DA for prenatal screening for Down syndrome during the course of a prenatal care visit during a subsequent pregnancy); (2) normative beliefs (a woman’s perceptions of to what extent partner, parents, or friends want her to perform the behavior); and (3) control beliefs (perceived barriers and facilitators of engaging in the behavior) [25]. The identification of beliefs that are associated with the intention to perform the behavior allows the specification of precise targets in future interventions. The extended behavior change model used in this study includes the main determinants of intention according to the TPB (attitude, subjective norm, and perceived control), as well as additional constructs known to improve the predictive capacity of the TPB: (1) anticipated regret, or an estimation of the regret that would result from not adopting the behavior; (2) descriptive norm, or the perceived prevalence of the practice; and (3) moral norm, or the moral principles involved [60-63] (see Figure 1).

From September 16, 2015, to October 8, 2015, we recruited eligible pregnant women (Table 1). Eligibility criteria included the following: being at least 18 years old, not less than 16 weeks pregnant, not presenting a high-risk pregnancy (eg, preeclampsia, gestational diabetes, multiple pregnancy), and having already decided about prenatal screening for the current pregnancy. Women who participated in a previous phase of the research were excluded. A private company specialized in polling was mandated to recruit eligible pregnant women in the province of Quebec (Canada) using a Web panel of willing participants in Internet surveys. Canada’s health care system consists of 13 (10 provincial and 3 territorial) independent health care systems. In this study, we focused on the province of Quebec, which is the second most populous Canadian province. First, the survey company sent an email invitation to all women on the panel aged from 18 to 44 years. After 2 weeks, to enhance recruitment, the same email invitation was sent to men on the panel aged from 25 to 44 years, as in this age range their partner was more likely to meet our eligibility criteria. The email invitation included the following information: (1) Subject: Research in health services led by Université Laval; (2) Financial compensation: 25 Canadian dollars; (3) Time to answer: next 5 business days; and (4) Personalized link to survey. Nonrespondents received a reminder every 2 weeks until the survey was closed. All interested persons who clicked on the personalized survey link were directed to the closed survey (password-protected) and asked to answer preliminary eligibility questions. Special filters allowed selection of the female partners of recruited men on the panel. Once eligibility criteria were confirmed, eligible women started the voluntary survey.

Participating pregnant women completed the Web-based survey through 39 Web pages that included up to 7 items, always appearing in the same order (see Multimedia Appendix 1). Clear preliminary statements provided information about the study and instructions and allowed participants to confirm their consent. To foster participation, the survey enabled pregnant women to stop the survey at any time and to restart it as long as the personalized survey link was active. No completeness check was possible before submitting the questionnaire. Once an item was answered, the answer could not be changed, as many items were similar and we wanted to test if participants’ responses were consistent. Once data collection was completed, the contracted company sent us the data anonymously, which were then stored on our secure network (password-protected).

In earlier research on factors that influence health-related behavior change [64,65], we observed that it is helpful to give participants a vicarious experience of the behavior of interest in order for them to understand it better [66]. As the pregnant women were not expected to have experienced the use of a DA, to help them understand the behavior of interest (action: use; target: a DA for prenatal screening for Down syndrome; context: during the course of prenatal care visits during a subsequent pregnancy; time: not specified), we asked them to watch a 10-minute video first. The video depicted a prenatal care follow-up during which a pregnant woman, her partner, and a health professional used a DA to decide about prenatal screening for Down syndrome. Production of this video had followed a validated process and had proved successful for communicating the behavior of interest [67]. The DA is available in Multimedia Appendix 2. After watching the whole video, eligible women answered the Web-based questionnaire based on the TPB but which included additional psychosocial factors known to influence the uptake of a new behavior [57,60-63]. In a previous step of the project, we had conducted a pilot study to validate this questionnaire [68]. We also measured underlying salient beliefs related to the direct constructs as elicited in a previous qualitative study (Leiva Portocarrero, M. Sc., personal written communication, February 2016). Using 52 closed items scored on a 5-point Likert-type scale, we measured intention, attitude, subjective norm, perceived behavioral control, anticipated regret, descriptive norm, moral norm, attitudinal beliefs, normative beliefs, and control beliefs. Except for attitude and anticipated regret, all direct constructs were assessed with multi-item measures. Anticipated regret was measured with 2 items and attitude with 6 items using bipolar adjective pairs assessing cognitive and affective dimensions of women’s attitudes. Cronbach alphas indicated good reliability of multi-items measuring each construct (alpha range .67-.94, Table 2). The questionnaire was developed following Ajzen's guidelines [69] and referred to using a DA to decide about prenatal screening for Down syndrome. All in all, the questionnaire included 52 psychosocial items, 9 sociodemographic items, and 50 health literacy items, for a total of 111 items. It was available in French and English.

To assess health literacy, after consulting with experts in the field [30,48,54,70] and reviewing multiple systematic reviews [71-75], we chose to use both subjective and objective scales. While objective scales measure competencies, subjective scales measure the perception of competencies and have been shown to reduce burden of participants [76-78]. We thus assessed pregnant women’s levels of health literacy using the following 4 complementary scales that measure health literacy and numeracy both objectively and subjectively: (1) the short version of the Test of Functional Health Literacy in Adults (S-TOFHLA), the literacy part only, which comprises 36 blank spaces and 4 choices of words to fill the blanks [79,80]; (2) a total of 3 self-administered health literacy questions (3HLQ, 5-point Likert scale: range 0 to 4, final score range 0 to 12) [81]; (3) a total of 3 numeracy questions (3NQ, 3 items, correct answers range 0 to 3) [82]; and (4) the Subjective Numeracy Scale (SNS, 2 subscales of 4 closed questions scored on a 5-point Likert scale, mean score range 1 to 5 for both subscales and complete scale) [76]. Finally, we assessed sociodemographics such as age, clinician in charge of monitoring, mother tongue, ethnicity, civil status, employment status, annual family income, education, and pregnancy number.

Details on flow of participants are depicted in Figure 2. On the basis of the CHERRIES statement [58], we considered all potentially eligible participants who clicked on the personalized link to visit the survey as unique survey visitors. As the Web survey did not include a middle stage between visiting the website and visiting the first survey page, we adapted the CHERRIES criteria to calculate the study's response rates [58]. Accordingly, view rate (ratio of unique survey visitors/unique receiver of survey invitation), participation rate (ratio of users who agreed to participate/unique survey visitors), and completion rate (ratio of users who finished the survey/users who agreed to participate) were respectively 15.09% (16,943/112,257), 2.26% (383/16,943), and 91.4% (350/383) (Figure 2).

Time of completion was not kept for analysis as participating women could stop and restart the survey. No data were missing as the Web-based questionnaire did not accept unanswered items. Mean age of pregnant women was 31 years. Out of 346 women retained for data analysis, 319 (92.2%) were white, 318 (91.9%) were French-speaking, and 168 (48.6%) had a university degree (see Table 1).

Intention to use the DA for deciding about prenatal Down syndrome screening during a subsequent pregnancy, and factors of this intention, showed generally high scores. Among the 346 pregnant women, 109 (31.5%) had an intention score of < 4 out of 5, a total of 92 (26.6%) had an intention score of 4 out of 5, and 145 (41.9%) had an intention score of >4 out of 5. All the direct determinants of intention showed similar scores (≥ 3.00 out of 5, see Table 2 and Multimedia Appendix 3 for details), except anticipated regret, which showed a median score of 2.00. The level of health literacy was generally high. A median score of 36.00 out of 36 was obtained for objective literacy (S-TOFHLA), 10.00 out of 12 for subjective literacy (3HLQ), 3.00 out of 3 for objective numeracy (3NQ), and 3.88 out of 5 for subjective numeracy (SNS total score; Table 1).

With our bivariate analysis, we found that all sociocognitive factors were significantly associated with intention (P<.001 for all, Table 2).

No sociodemographic and health literacy variable was significantly associated with intention (Multimedia Appendices 4 and 5). In exploring correlations between health literacy variables and sociocognitive constructs, we found the most frequent association was with perceived control, which showed significant associations with all health literacy variables except objective numeracy (Multimedia Appendix 6). In addition, we observed that all health literacy scales were correlated among themselves (rho range .14-.89, P value range .007-.001; see Multimedia Appendix 6).

Finally, we identified the most significant factors in women’s intention to use the DA. In the first multivariate model, including only TPB variables, attitude (odds ratio, OR, 13.38, 95% CI 6.40-27.90), subjective norm (OR 3.64, 95% CI 2.33-5.70), and perceived control (OR 2.36, 95% CI 1.43-3.90) were significant factors of pregnant women’s intention (Table 3). No sociodemographic variable was added to the model.

In the second multivariate model, still based on the TPB but including the additional variables of anticipated regret, descriptive norms, and moral norms, we found that attitude (OR 9.16, 95% CI 4.02-20.85), moral norm (OR 7.97, 95% CI 4.49-14.14), descriptive norm (OR 2.83, 95% CI 1.63-4.92), and anticipated regret (OR 2.43, 95% CI 1.71-3.46) were significant factors of pregnant women’s intention (Table 3). Comparison of deviance showed that the model that included additional sociocognitive variables better explained pregnant women’s intention (∆ deviance=41.33, P>.05, Table 3).

To investigate whether health literacy predicted pregnant women’s intention in our theoretical model, we sequentially added each health literacy variable to the ordinal logistic regression model and observed that the pregnant women’s intention was not affected by health literacy (Table 3). In addition, no interaction term was identified between health literacy variables and direct constructs determining intention. We performed structural equation modeling in parallel with our stepwise regression model and observed the same results (data not shown but available from authors).

To identify significant underlying beliefs, we performed another ordinal logistic regression model with beliefs related to attitude, as attitude was the only significant construct with underlying beliefs. We found 3 significant beliefs related specifically to the attitude construct, namely, that the use of a DA (1) would reassure pregnant women (OR 2.55, 95% CI 1.73-4.01), (2) would facilitate their reflection with their spouse (OR 1.55, 95% CI 1.05-2.29), and (3) would let them know about the advantages of doing or not doing a prenatal screening test for Down syndrome (OR 1.53, 95% CI 1.05-2.24; Table 4).

In this theory-based Web survey, we sought to identify psychosocial factors influencing pregnant women’s intention to use a DA for prenatal screening for Down syndrome and assessed whether health literacy added to the predictive power of this model. There are no data specifying the profile of pregnant women in the province of Quebec, but our sample compared well to that of women in the province overall, except for education and health literacy levels, which were higher in our sample [85-87]. Overall, we found that pregnant women showed high levels of intention to use a DA for prenatal screening for Down syndrome. Also, we observed that, in order of importance, attitude, moral and descriptive norms, and anticipated regret were the factors that explained most of their behavioral intention. In other words, the perception of the advantages of using a DA (attitude), the possible regret foreseen if the DA is not used (anticipated regret), the perception that it is a common practice (descriptive norm), and the feeling that using a DA for this decision would be in agreement with their moral values (moral norm) were significantly associated with a strong intention to use the DA. In addition, we identified 3 attitudinal beliefs significantly associated with women’s intention: perceiving that using a DA (1) would reassure them, (2) would facilitate their reflection with their spouse, and (3) would let them know about the advantages of doing or not doing a prenatal screening test for Down syndrome. On the other hand, our findings showed that neither health literacy levels nor individual sociodemographic characteristics had any influence on the behavioral intention of interest, suggesting that, regardless of their health literacy levels and sociodemographic characteristics, all women are under the influence of the same sociocognitive factors regarding whether or not they intend to use a DA for prenatal screening for Down syndrome. These findings lead us to make 5 main points with regards to pregnant women’s intention, the direct determinants of their intention, their underlying beliefs, the influence of health literacy, and the next steps.

First, to the best of our knowledge, this study is among the first to adopt an all-encompassing theory-based approach to identifying factors, including health literacy, influencing the intention to use a DA in prenatal care. Our results support earlier research on SDM implementation indicating that women showed a strong intention to engage in SDM regarding prenatal screening for Down syndrome [26]. Moreover, this high level of intention may reflect a need felt by pregnant women facing prenatal screening choices to become more skilled in discussing screening tests with their health care provider, which is congruent with the literature on pregnant women’s decision-making needs [4,6,9]. This strong intention suggests that future efforts to increase DA use and SDM among clinicians for prenatal screening for Down syndrome would find a favorable response in pregnant women.

Second, we observed that the following factors, in order of importance, influenced pregnant women’s intention: attitude, moral norm, descriptive norm, and anticipated regret. These findings are congruent with earlier research on SDM implementation in this context [26], which showed that attitude, subjective norm, self-efficacy, and moral norm were determinants of pregnant women’s intention to engage in SDM. Although the variables “descriptive norm” and “anticipated regret” were not investigated in the earlier study, the influence of social pressure came out through the subjective norm variable, which refers to the influence of significant individuals in women’s entourage. Contrary to the findings of our study, and despite the similarity of the samples, however, the previous study showed self-efficacy as a determining factor among women without postsecondary education (although not among women with a higher level of education), whereas in our study sample perceived control was not a significant factor. Self-efficacy and perceived control are not the same constructs, but they are closely related as they both refer to a person’s evaluation of the degree of difficulty of adopting a given behavior. This difference in findings could reflect the fact that the earlier study considered intention to engage in SDM while our study asked women about their intention to use a DA, a practice that constitutes one specific behavior in the overall SDM process. Pregnant women with low education could have more confidence (self-efficacy) about using a DA than about engaging in SDM in general, belying the common myth that the use of DAs is equivalent to the behavior of engaging in SDM [12,88].

Third, significant salient beliefs underlying attitude were, in order of importance, (1) the women’s belief that the use of a DA would reassure them; (2) the belief that it would facilitate their reflection with their spouse; and (3) the belief that it would let them know about the advantages of doing or not doing the prenatal screening test for Down syndrome. These results are congruent with earlier research on decisional needs among pregnant women facing prenatal screening for Down syndrome, which showed that the main difficulties perceived by pregnant women were pressure from others, emotions, and lack of information [4]. Our findings provide information about modifiable attitudinal beliefs regarding DA use that could facilitate design of implementation strategies to increase their use by pregnant women in clinical practice. According to the Intervention Mapping approach, efficient interventions should “contain specific messages that target selected beliefs within the determinants of interest, and require specific translation to practical applications to reach optimal fit” [23]. In practice, a public health communication campaign combined with interventions mediated by health professionals could reinforce the influence of women’s attitude to DA use by targeting its 3 identified underlying beliefs (reassurance, reflection with spouse, awareness of advantages of each choice). Key statements regarding these 3 salient beliefs could also be added to the DA to increase women’s intention to use it and help clinicians to explain it.

Fourth, health literacy was not a factor that influenced women’s intention in our study, although many studies have shown associations between health literacy and related notions, such as patient involvement in decision making [27,38,40-43,47,68]. Hence, a single population based DA implementation program would benefit any pregnant women in the province of Quebec to help them make informed values-congruent decision about prenatal screening for Down syndrome. To the best of our knowledge, this study is among the first to investigate the role of health literacy within the context of a theory-driven study using a behavior change model. Our findings suggest that regardless of their perceived or actual capacity to understand complex information, pregnant women would like to use a DA to decide about prenatal screening for Down syndrome. In addition, in our bivariate analyses we did identify some interesting associations between health literacy and perceived control. Although perceived control was not a predictive factor in our study, this association suggests that in contexts where it does play a predictive role, such as intention to engage in healthy eating behaviors [89], intention may be affected by health literacy. Similar studies that include populations with low health literacy should be conducted before decisive claims can be made. However, some have argued that it would be more efficient to work at clearer health communication and fostering participation among all patients rather than screening them for health literacy [90]. Our results should not be used to minimize the importance of improving patients’ understanding. It is of primary importance to communicate clearly with every pregnant woman about Down syndrome screening and invite them to ask questions, whatever their health literacy level.

Finally, if the findings of this study are valid, identifying the 4 most significant determinants of pregnant women’s intention to use a DA to decide about Down syndrome screening will be useful for the design of interventions to promote uptake. Each of the 4 determinants can be mapped to the Behavior Change Wheel [91], a method developed to inform the design of behavior change interventions. This generates a set of recommended interventions (such as education, training programs, persuasion, modelling, etc), each with its appropriate related methods. For example, clinicians could be trained to introduce the advantages of using the DA, and present the use of DAs as a common practice. Indeed, while it has been shown that lack of training was an important barrier to DA implementation [17], pregnant women indicated that clinicians had a key role in delivering such information [92]. An implementation intervention could also consist of a Web-based application coupled with a DA, which would enrich the current bank of online decision support tools [93]. In terms of moral norm, our findings confirm that there is a significant ethical dimension to the decision about prenatal screening [94-96]. No psychological techniques have yet been formally identified for considering the influence of the moral norm in an intervention [97]. Also, it has been shown that interventions that aim to manipulate moral norms can be counterproductive because of a “boomerang effect” that arises when a person perceives that his or her freedom is threatened by social pressure [98,99]. Health professionals could thus simply be invited to explain to women why moral values are at stake in the decision, so that they can subsequently discuss this and clarify the decision in light of the woman’s moral principles.

This study has limitations. First, although we suspected that the lack of variability in our health literacy findings could have been due to lack of scale discriminating capacity in our sample, except for the objective health literacy scale (S-TOFHLA), the administered scales covered many dimensions of health literacy and correlated together, indicating a convergence of the results (see Multimedia Appendix 7). In addition, this study will enable further validation of new French versions of the 3 scales.

Second, it is possible that the video used to present the behavior of using a DA mediated pregnant women’s intent. However, we felt it was more important to ensure that respondents understood the nature of the behavior being studied than to avoid any risk of mediated intent by not using a video at all.

Third, use of the TPB framework could in itself have shaped our findings. For example, the assumption that agents and their actions are rational may neglect the role of nonrational factors (such as emotion or experience) in human action and reasoning. Also, the approach underestimates the singularity of each agent, as it considers that similar agents are influenced by similar factors [100]. Despite these theoretical limitations, from the wider perspective of developing potentially effective implementation strategies, a TPB-based approach promotes the collection of comprehensive, consistent and valid information, and is still one of the most frequently applied theories in the domain of behavior change.

Finally, we cannot assume that the results can be extrapolated widely without further research. Pregnant women in the sample were mostly white, with French as their mother tongue. This may weaken the external validity of our sample, but not with regards to the general population of the province of Quebec, which was the aim of this study. Because Canadian health care services are organized into 10 provincial and 3 territorial health care systems, each requires its own population-based prenatal screening program. Also, pregnant women in our sample were highly educated compared with women in the province of Quebec overall, where less than a quarter of women aged from 15 to 44 years old have a university education [87]. Likewise, health literacy levels were very high overall, far above that of the general population of the province of Quebec but also of the rest of Canada, where more than half the population has inadequate levels [85,86]. Pregnant women in our sample were recruited from a large Web panel and, as participants willingly subscribed to the panel, their literacy levels, including health literacy and eHealth literacy, might be higher than that of the general population [101]. Our study sample may thus lack representativeness with regard to vulnerable and less literate populations.

Our recruitment methods reflected our main objective (psychosocial determinants of women’s intention), but methods in future studies looking at health literacy should be informed by the specificities of the studied population with respect to education levels, Internet use and eHealth literacy.

This study, based on a theoretical approach to behavior change, indicated which factors will need to be addressed to design an effective implementation intervention for the use of DAs in the context of prenatal screening for Down syndrome. Our findings indicate that women’s intention to use a DA in this context was determined by the consequent pros and cons they perceived (attitude), its compatibility with their moral values (moral norm), their perception of how much other women use it (descriptive norm), and the regret they perceive they might feel if they do not use it (anticipated regret). This study provides valuable and specific guidance for designing an intervention to implement the use of a DA and ultimately to foster SDM in prenatal care.