Impact of a risk based breast screening decision aid on understanding, acceptance and decision making

Internationally, population breast cancer screening is moving towards a risk-stratified approach and requires engagement and acceptance from current and future screening clients. A decision aid (www.defineau.org) was developed based on women’s views, values, and knowledge regarding risk-stratified breast cancer screening. This study aims to evaluate the impact of the decision aid on women’s knowledge, risk perception, acceptance of risk assessment and change of screening frequency, and decision-making. Here we report the results of a pre and post-survey in which women who are clients of BreastScreen Victoria were invited to complete an online questionnaire before and after viewing the decision aid. 3200 potential participants were invited, 242 responded with 127 participants completing both surveys. After reviewing the decision aid there was a significant change in knowledge, acceptance of risk-stratified breast cancer screening and of decreased frequency screening for lower risk. High levels of acceptance of risk stratification, genetic testing and broad support for tailored screening persisted pre and post review. The DEFINE decision aid has a positive impact on acceptance of lower frequency screening, a major barrier to the success of a risk-stratified program and may contribute to facilitating change to the population breast screening program in Australia.


INTRODUCTION
Since the introduction of population-based breast cancer screening significant inroads have been made in the understanding of breast cancer risk 1 particularly the role of factors such as mammographic density 2,3 and genetic factors and their impact on an individual's risk 4,5 .Potential harms of population screening such as overdiagnosis and false positives 6,7 have come to light as well as significant advancements in screening technology.Despite these changes only small shifts such as the 2014 inclusion of the family history policy 8 and breast density notification in some Australian states 9 , have altered to reflect this.Thirty years on, age and gender are still the predominant factors used to assess screening eligibility and breast cancer risk in the general population.
Risk-based screening, whereby recommendations are dependent on breast cancer risk category, the calculation of which would include mammographic density 10,11 , has the potential to further reduce breast cancer mortality by lowering rates of interval cancers in higher risk individuals 12,13 .Planning for the change of program structure is underway for breast cancer screening both internationally and in Australia 10,11,14,15 .Several international projects are underway in the UK 13,16,17 , Canada 18 , Europe 19 , Italy 20 and USA 21,22 to further understand the possible advantages of this approach including the role genomics may play in risk assessment.
Women who currently participate in routine population breast cancer screening are interested and supportive of a risk-stratified approach, including the use of genomics to assess risk 23,24 .However, there was less acceptance of decreased screening frequency if women are found to be lower than average risk 25,26 .Successful implementation of a risk-stratified approach to population breast cancer screening will depend on engagement and understanding from consumers, which will require comprehensive, and culturally sensitive communication made available in plain language.
A purpose built, online Decision aid (DA) called DEFINE (https:// www.defineau.org)was developed incorporating current breast screening participants' values, educational and decisional needs to support the process of choosing between current or risk-stratified screening 24,27 .An iterative process was undertaken with input from experts in a variety of fields as well as consumers who directed the content and improved readability.
The expansion in understanding of breast cancer risk, such as the impact of mammographic density and genetics, including polygenic risk information, has been shown to improve the assessment of an individual's risk.Incorporating these factors into existing risk prediction models significantly improves risk prediction [28][29][30] ; however, this benefit needs to be weighed up against feasibility and the increased expense and complexity of this personalised assessment especially collection of detailed family history.The inclusion of genetics requires careful consideration of the ethical, social, and psychological implications of genetic testing, and communication with consumers will play a critical role in managing these issues.
Decision making for an individual around screening is complex, personal, and likely to be based on emotion, experience or instinct rather than factual information 31 .There is an overwhelmingly positive opinion regarding the advantages of breast screening 32 and people tend to overestimate benefits and underestimate harms in population based screening 33 .Informed decision making requires the provision of information covering both the benefits and harm of an intervention 34 therefore, especially in the screening context where this has not always been the case, attention is needed to give education and communication with a balanced approach.
DAs, designed to assist people make health-related decisions that align with their personal values, can clarify decisional uncertainty, as well as weigh up the potential benefits and harms of a certain intervention.Overall, they improve understanding of clinical options, create more realistic expectations about outcomes and have been attributed to a higher likelihood of an informed value-based decision 35 .This has been a consistent finding throughout numerous studies of screening, including a metaanalysis on mammographic screening, confirming that DAs improve knowledge and promote informed decision making 36 .An online DA has the advantages over paper-based materials of accessibility, personalisation as well as being easily updated as information changes. 37.An online DA has the potential to be a powerful tool towards informed decision making for risk-stratified breast cancer screening.
This paper reports on the impact of DEFINE on knowledge of breast cancer risk and risk-stratified screening, acceptability of different screening frequencies as well as informed decision making for women participating in the current breast screening program in Victoria, Australia.

RESULTS
Participants 3200 email invitations were distributed, and 242 (8%) women completed the pre-DA review survey.127 (53%) women completed both the pre and post-surveys.Demographic data is summarised in Table 1.Participants in this study were of higher average age and had an overrepresentation of higher Index of Relative Socio-economic Advantage and Disadvantage (IRSAD) postcodes compared to BreastScreen participants which are evenly distributed amongst these quintiles 38 .A summary of missing results is presented in Supplementary Table 1.

Breast cancer and screening experience
The participants were well-experienced BreastScreen attendees with 197 of the 241 (82%) participants who responded to the question (one missing response) having had more than three mammograms.From the initial cohort of 242 respondents, 29% had a prior breast biopsy and 22% reported a family history of breast cancer.Table 2 lists participants' breast cancer and screening experience.

Risk perception
There was a normal distribution of responses on how likely a participant felt she was to develop breast cancer in her lifetime and no significant change between responses for pre and post DA review (scale from 0% "no chance of developing breast cancer" to 100% "will definitely develop breast cancer"-pre-review mean (M) = 41.0 (Standard deviation (SD) 22.1), post-review M = 44.1 (SD 22.6).The increase of scores indicates women perceived their breast cancer risk as slightly higher after reviewing the DA although this was not statistically significant (95% confidence interval (CI) −0.9 to 5.3, P value = 0.161).

Breast cancer worry and overall anxiety
Breast cancer worry was normally distributed with 50% of all respondents (n = 63 with no missing responses) choosing the middle response of rarely worrying about breast cancer, 24% (n = 30) often or sometimes worrying, and 27% (n = 34) never worrying.
Overall anxiety, measured by OASIS, was low (M = 2.6, range 0-25, SD 3.0).There was a weakly positive association between a participant's level of worry about developing breast cancer and their perceived risk of developing breast cancer (Spearman's rho = 0.28 P value < 0.001).

Knowledge of breast cancer risk factors
There was a significant increase in self-rated breast cancer risk factor knowledge pre (M = 65.5, SD = 20.0)compared to post (M = 75.4,SD = 16.8)reviewing the DA (t test 10.0; 95% CI 6.3-13.6;P < 0.001) and similarly for the risk factor knowledge test before (M = 4.6 SD = 2.0) and after (M = 6.0,SD = 2.0) (t test 1.4; 95% CI 1.0-1.7;P < 0.001).There was a weakly positive correlation between the post-test breast cancer risk factor knowledge test scores and self-reported knowledge on risk factors (Spearman's rho = 0.3; P < 0.001) suggesting participants had high levels of insight into their understanding of breast cancer risk factors.

Understanding of risk-stratified screening
Similarly, there was a statistically significant increase in how much women felt they knew about individualised breast screening with an increase mean score from 49.0 (SD = 29) to 74.4 (SD = 20.4,range 0-100) post reviewing the DA (95% CI 19.5-30.8,P < 0.001).
Figure 1 shows the improvement of subjective rating of riskstratified screening of the participants by contrasting the distribution of responses prior and after reviewing the website in a scatterplot.Open coding to assess the accuracy of responses to the free text question, "What do you know about individualised or personalised breast cancer screening?"., revealed eight categories.See Table 3 for categories and response examples.Given this was a compulsory question, five women in the pre and six women in the post-survey responded with non-sensical responses.There was a marked shift towards a more accurate response in the description of personalised screening after reviewing the website, as shown in Fig. 2.
Interest in risk-stratified screening overall Initial levels of interest in risk-stratified screening were very high.Of the 241 respondents who answered the question to the survey pre-DA review, 226 participants (93.7%) were definitely or probably interested in a risk-stratified approach to breast cancer screening.Analysis of the 127 participants who completed both pre and post-surveys saw a shift from 122 to 126 women into either the "Yes probably" or "Yes definitely" group, illustrated in Fig. 3. Grouping of responses into binary outcomes ("Yes probably" and "Yes definitely" into Interested and "Probably not" and "Definitely not" into Not interested) demonstrated no statistically significant change (P = 0.063).
There was a slightly higher rate of interest in the women who completed both pre and post DA review questionnaires compared to the 115 participants who did not complete the survey for after DA review, however this was not statistically significant (P = 0.779).
There was no association between levels of breast cancer worry and interest in risk-stratified breast screening (Spearman's rho = −0.159,P = 0.074) and similarly, there was no association between women's levels on anxiety as measure on the OASIS scale and their interest in risk-stratified breast screening (Spearman's rho = −0.084,P = 0.347).

Acceptability of varying screening intervals
In the pre-DA survey, there were very high levels of acceptance for having an annual mammogram if high risk (124 of the 125 respondents who answered the question, 99%) compared to the acceptance of having a 3-year screening interval if deemed lower than average risk (92 respondents with no missing responses, 72%).Figure 4 outlines the responses to all three questions of varying screening intervals.Of note, 35 women (28%) would not be accepting of a 3-year screening interval if lower than average risk prior to reviewing the DA but this reduced to 19 (15%) after reviewing the DA.Grouping the responses into binary categories resulted in a statistically significant shift towards acceptance of a 3-year screening interval if deemed lower than average risk (P = 0.005).
Levels of acceptance for annual mammograms if higher than average risk was exceptionally high with only one of the 125 respondents who answered the question (0.8%) not interested prior to reviewing the DA and two of the 127 respondents (1.6%) in the post review survey.On binary analysis there was no significant change between the two responses (P = 0.564).
In terms of a 5-year screening interval of much lower than average risk, this was acceptable to only 43% of women (54 of 127 respondents) pre-DA review, however this increased 54% (68 of 127 respondents) after reviewing the website.Grouping the outcomes for acceptance of a 5-year interval into binary categories there was a statistically significant shift towards acceptance (P = 0.004).Family history response allowed for more than one type of cancer accounting for the total being >100%.Genetic testing was open to interpretation-no specification was given to specific genes tested.
There was no relationship between breast cancer worry and accepting a five year interval for screening if lower risk (Spearman's rho = −0.1,P = 0.343).

Acceptability of genetic testing incorporated into screening and risk assessment
Levels of acceptance of genetic testing were very high with 120 of the 126 respondents who answered the question (95.2%) interested in the pre-DA survey and 117 of the 123 participants who answered the question (95.1%) interested in the post-DA survey.Similarly, in regard to accepting a risk assessment, this was very high in the pre-DA survey with 125 participants (98.4%) accepting this before and 124 (100%-3 missing responses) in the post-DA survey.This compares to 95% (109 of 115) of respondents who did not complete the post-DA survey who were accepting of a risk assessment.Figure 5 demonstrates responses to these questions.
Informed decision making 95 women (75% of respondents) gave adequate responses to the free text response describing risk-stratified screening after reviewing the DA indicating adequate knowledge.64 of those women had positive values and preferred individualised screening and three participants had negative values and preferred the current program.67 participants or 53% of the cohort have been classified as making an informed decision.Seven participants were unsure which option they prefer and 20 did not respond to the question of preference.

Satisfaction with decision
Satisfaction with decision was overall very high with 113 of the 125 respondents who answered the question (90%) rating a 0-6 high score.The remaining 12 respondents (10%) were in the intermediate 7-12 range and no participants had low levels of decision satisfaction.

Website feedback
Responses to specific questions around utility and usefulness of the website are summarised in Table 4. Interestingly, 114 (90%) of women spent 30 min or less and participants reported very little anxiety while reviewing the website.Most respondents reported the website content was new to them with 91 respondents (73%) reporting some of the information was new to them compared to 7 (6%) reporting none of it was new.

DISCUSSION
Our findings confirm that the DEFINE online DA (www.defineau.org)improves knowledge and in principle acceptance of a risk-stratified breast screening approach as well as  allowing for informed decision-making.These results are in keeping with assessments of other DAs which consistently report significant improvements in knowledge [39][40][41][42][43][44][45][46][47] and were confirmed in a Cochrane review of 105 studies reporting on DAs which have a significant impact on knowledge 48 .
In terms of acceptance of risk-stratified breast cancer screening, previous work has consistently reported strong levels of acceptance for risk assessment and a risk-stratified approach but a persisting reticence for accepting lower frequency screens for women who are deemed lower than average risk 23,25,26,[49][50][51][52] .A purpose-built DA could manage these hesitancies, and our work demonstrates that education plays a critical role in theoretical acceptance of less frequent screening.
While other studies have shown that a participant's perceived high risk influences acceptance of more frequent screening based on objective risk assessment 26 , this was not observed in our study.This discrepancy might be explained by the fact that the majority of our participants felt themselves at exceptionally high level of risk of developing breast cancer with a median perceived lifetime risk of developing breast cancer of 41%.
The reluctance to accept lower frequency screenings if lower than average risk is understandable from a cohort of women who  have been advised for two-yearly screening and they have been adherent to this for many years.82% of our cohort had more than three prior mammograms and only a small group of women (12 participants) had 1-3 prior mammograms.Given the disparate group sizes between less and more prior screening rounds, meaningful differences could not be ascertained in terms of acceptability of differing screening frequencies, or risk stratification more broadly.It is plausible a group of women with less experience in the current system would have different attitudes towards a lower frequency of screening and this would be an interesting future study.
The implications of our work for risk-stratified breast screening into an Australian context is reassuring in that we demonstrate women currently participating would not be dissuaded by a change in the program.We show women participating in our current program are interested in and motivated by this approach.There is broad acceptance of having risk assessed within our existing system as well as in the inclusion of genetic testing to this risk calculation.We have previously reported a proportion of current BreastScreen clients may have concerns about data security and privacy issues associated with genetic testing 24 ; however, our current results quantify those concerns to be small.
The first limitation of this study is the skewed sample of women which were recruited.By recruiting women already actively involved in population breast screening it is difficult to generalise these findings to the broader population including those women who have little or no experience with our current breast screening program.Our cohort is not a representative sample of the broader BreastScreen community given they were highly experienced with over 80% having had at least 3 prior mammograms and with an average age older than the average BreastScreen Victorian client.The impact of this experience and older age is challenging to quantify but is foreseeable as a confounding factor in risk perception and acceptability.
Secondly, with only an 8% uptake rate from the initial email invitation and an overrepresentation of higher socioeconomic postcodes 38 and an higher age than the average screening population, we may not have a representative sample of the overall screening population in Victoria.The other indicator we may have sampled a very well-informed group was that 73% of the cohort reported only 'some' of the website information was new to them.The use of email for recruitment will also have subselect participants with higher technological literacy which potentially biased our cohort.This method of recruitment was selected as it most closely aligned with our tool being web based; however, a paper or in person-based recruitment strategy may have uncovered different results.
Thirdly, we note just over half the participants completed both the pre and post-questionnaires.The group that completed only the pre-DA questionnaire was slightly younger with a marginally lower interest in risk-stratified screening and this further may limit the generalisability of our findings.
Another limitation is the hypothetical nature of the responses.There may be disparity between intention and behaviour and women in this cohort were not committing to an actual change or acceptance, rather indicating potential intentions to an approach not yet available to them.
We have custom built an effective decision aid which can support women's decision making around a risk-stratified model and have demonstrated that even brief intervention significantly improved understanding and acceptability of a risk-stratified approach to population breast screening in Australia.
International work from Canada, UK and Europe has consistently shown women are hesitant about reducing their screening interval 23,25,26,[49][50][51] but we have demonstrated this hesitancy can be lessened with brief educational and values assessment intervention; however, future work is needed assess whether this translates into real-world behaviour change with higher levels of uptake as well as the impact on a screening naïve population.
The transition from our current one size fits all population breast screening model to a tailored or personalised approach is well underway 10,14 .We recognise the importance of patient involvement, education, and communication in a new population health intervention.Future directions for this DA are planned addressing key stake holders perspectives on the implementation of DAs into a risk-stratified breast screening approach in Australia.

Design
Single arm mixed methods survey conducted pre and post-review of an online DA, DEFINE.How would you describe the website in terms of ease of understanding?Requests simpler language to be used 1 (1%) Demographic characteristics: Measured were: age, postcode (surrogate for socio-economic status), level of education, language spoken at home, experience with mammograms, prior breast biopsy, personal history of breast cancer and family history of breast cancer.Genetic testing was inquired about without specifying which genes had been tested.
Breast cancer risk factors knowledge: A knowledge measure of breast cancer risk factors was purpose-built using eight known breast cancer risk factors.Participants were required to "tick" as many of the factors they knew, and direct pre-post comparison was performed.The factors listed were -Mammographic density, family history, alcohol intake, body mass index, age, age of first pregnancy, age at menopause and being nulliparous 1 .Incorrect distractors (such as stress or the use of antiperspirant deodorants) were excluded to reduce the chance of confirmation bias reinforcing myths about breast cancer risk factors.Self-reflection of knowledge was assessed by a single question where participants rated their knowledge of breast cancer risk factors on a scale of 1-100 where 1 is poor and 100 is excellent.
Risk-stratified Breast screening knowledge: The only compulsory answer within the survey was a free text response to the question "What do you know about "individualised" or "personalised" breast screening?".The free text responses were interrogated using a structured qualitative content analysis approach.Discrete codes, 'accurate description', 'partially accurate' or 'value/belief about the program', were defined and responses were compared pre and post for a change in understanding.
Self-reported knowledge of risk-stratified breast screening was quantified and compared with a question asking participates to self-rate their knowledge of risk-stratified screening where 1 is poor and 100 is excellent.
Breast cancer worry and general anxiety: A single question was posed to assess overall worry about developing breast cancer "How much do you worry about getting breast cancer?" with the potential responses being 'never, rarely, sometimes, often, or most of the time' which was adapted from the UK Health Information National Trends Survey 55 .
General anxiety was measured using the Overall Anxiety Severity and Impairment Scale (OASIS) 56 .This five-item tool measures each response using a 5-point Likert scale with questions being equally weighted to give a total score of 0-25.A score of 0-5 indicates mild or no anxiety, 6-10 indicates moderate, 11-15 severe and a score of 16-25 indicates extreme anxiety.We were interested in examining the role anxiety plays in decision-making 57 , and although anxiety is only one of many factors which may impair decision making we chose to focus on this as it has been documented to have a particularly pertinent role in breast cancer screening decisions 58 .
Breast cancer risk perception: Risk perception was assessed using three validated questions using numeric, verbal comparative scales adapted from Gurmankin Levy et al. 59 .Risk perception can be a strong motivator for uptake of health behaviours such as cancer screening 60 .
Intention to participate: Two questions assessed intention to participate in a risk-stratified breast screening program.The first being: "If BreastScreen offered you the chance to take part in individualised breast screening in place of the current BreastScreen program, would you be:" of which the response options were: definitely interested, potentially interested, unlikely to be interested, or definitely not interested.The second question "Would you be interested in breast screening if each woman was managed differently depending on her personal breast cancer risk?" had potential responses of: yes definitely, yes probably, probably not, definitely not.
Acceptability of risk assessment and genetic testing: Two questions examining acceptability of risk assessment as well as acceptance of genetic testing were asked: "If you were offered an assessment of your personal breast cancer risk by BreastScreen at your next mammogram, would you be" and "If calculating your personal breast cancer risk meant having a blood test to look at your genes, would you be" with available responses on a four-point Likert scale: definitely interested, potentially interested, unlikely to be interested and definitely not interested.
Acceptability of change of screening: Three questions addressed the issue of acceptance of differing screening frequencies (for example, mammograms every one or three years dependent on level of risk).Prior to these questions, there was a descriptor explaining the current Australian recommendations for biennial mammography.These purpose built questions were aimed at assessing the acceptability of individualised screening and the impact of the DA on level of acceptability.The first question was "Hypothetically, if your personal breast cancer risk was lower than the average woman you could be advised to have a mammogram every 3 years.Would you be happy to have screening every 3 years if you were found to be low risk?".Available responses included yes definitely, yes probably, probably not, or definitely not.
The other two scenarios were gauging acceptance of a mammogram every 5 years if their risk were much lower than average and to have an annual mammogram if they were found to be higher risk.
Informed decision making: The multidimensional measure of informed choice (MMIC) acknowledges informed decision-making requires adequate knowledge, and alignment with values/ attitudes and intention.This has been validated in the context of antenatal screening for Down's syndrome 54,61,62 .In order to measure informed decision making we assessed three separate outcomes and combined three outputs of the post-website review survey: knowledge of risk-stratified screening from the free text response, attitudes of risk-stratified screening modified from the validated 6 questions developed by Dormandy 63 , and preference for current or risk-stratified screening.Participants were considered to have made an informed decision when they had adequate knowledge, positive values, and preferred risk-stratified screening or if they had adequate knowledge, negative values and preferred the current screening model.Adequate knowledge was defined as a response in the post-website review questionnaire to the openended question to "What do you know about "individualised" or "personalised" breast screening?"with 'accurate description' and 'partly accurate description' responses defined as adequate knowledge.
Satisfaction with decision-making: Satisfaction with decision scale is a validated six item scale modified from O'Connor's

Fig. 1
Fig. 1 Quantitative self reported knowledge of personalised breast screening.Scatterplots of responses to "How do you rate your knowledge of individualised or personalised breast screening on a scale of 1 to 100 where 1 is poor and 100 is excellent?"prior to reviewing the website above and after reviewing the website below.The blue dots represent each participant's response showing a substantial improvement in self-rated knowledge of individualised breast screening.n = 123 pre and 125 post (4 missing responses pre and 2 missing responses post).

Fig. 2
Fig. 2 Qualitative responses to knowledge of personalised breast screening.Open coding response frequencies pre and post DA review to "What do you know about individualised or personalised breast cancer screening?".n = 127 Figure 2 legend: (a) pre, (b) post.

Fig. 3
Fig. 3 Individualised screening interest.Responses to "Would you be interested in breast screening if each woman was managed differently depending on her personal breast cancer risk?".n = 127 Figure 3 legend (a) pre, (b) post, (i) Yes definitely, (ii) Yes probably, (iii) Probably not, (iv) Definitely not.

Table 5 .
Questionnaire domains.Pre Post Demographic data Risk factor knowledge/ confidence ➨ Risk factor knowledge/confidence Breast cancer worry and general anxiety Breast cancer risk perception ➨ Breast cancer risk perception Intention to participate ➨

Table 2 .
Breast cancer and screening experience.

Table 3 .
Themes and examples for open coding to "What do you know about individualised or personalised breast cancer screening?".
"Because my mother has had two bouts of breast cancer I am very particular about having my tests done" Nothing/not much/unsure "nil"Inaccurate description "I know the mammogram is checked and hopefully, if any cancer present it will be picked up but I know this is not always the case."Partlyaccuratedescription "Higher risks, more frequent screens" Accurate description "I think it means that instead of a blanket approach where everyone gets a mammogram over a certain age that instead it depends on your risk factors" More than before "Only what I have just read"Value or belief "Sounds like a good idea but I would still like the 'insurance' of a 2 year mammogram" "Enough to be interested"