A cross-sectional survey of women aged 50 years and above from five countries (Belgium, France, Italy, Spain, and the United Kingdom [UK]) was performed in January 2021. Based on the age distribution of women aged 50 years old and above in the five countries in 2020, with a 95% confidence interval, the required minimum sample size was estimated to be 1125 women in total, with a minimum of 225 women per country.
A total of 1180 participants responded to the survey from five countries (Belgium, France, Italy, Spain, and the UK). Participants per country ranged from 228 (19.3% of total sample) to 239 (20.3% of total sample). Age of participants was reported only in the categories of aged 50–59 years old and 60 years and older. For the total sample, 55.1% were aged 50–59 years old and 44.9% were aged 60 years and older. As a proxy of socio-economic status, household income per annum was asked to participants. A total of 947 participants provided a response. The most frequent range of household income per annum for the total sample was €20,000 - €39,999, which was reported by 388 participants (32.9% of the total sample). The least frequent option for household income per annum for the total sample was €80,000+, which was reported by 29 participants (2.5% of the total sample). 233 participants (19.7% of sample) declined to provide data on household income per annum. The characteristics of the participants are represented in Table 1.
An online survey was developed in English and translated into the national languages of the five countries surveyed. To inform the questionnaire items a conceptual model was developed based on the Theory of Planned Behaviour (TPB) and the Health Belief Model (HBM) (Fig. 1). The model includes individual characteristics of participants, several cognitive variables informed by the models of behaviour.
Individual characteristics were concerned with age (reported in dichotomous categories of age 50–59 years old, and age 60 years old and above), country of residence (Belgium, France, Italy, Spain, and the UK), and level of household income per annum (< 19,000€; 20,000–39,999€; 40,000–59,999€; 60,000–79,999€; >80,000€) of participants. For the UK, the household income reported in Pound Sterling (GBP) was converted to Euros. the respondents’ history of participation in breast cancer screening. Participants were also asked if they had ever participated in mammography screening, with an option to define if they had been invited by the mammography screening programme or referred by a health professional. This was later dichotomized for the analysis as participated or not participated. Knowledge of the benefits and harms of breast cancer screening was assessed by asking participants whether they could currently identify that mammography came with benefits and risks, or if they believed it came with benefits but no risks, or its goal was to prevent cancer. The outcome was dichotomised to either correctly identifying breast cancer screening has benefit and harms or not
Cognitive variables measured by the survey were perceived social norms, perceived behavioural control, perceived susceptibility, and perceived barriers to screening. The items were informed by the Champion Health Belief Model Scale (CHBMS) used previously in studies on attitudes to breast cancer screening (24). Perceived social norms was measured by asking participants whether they believed that ‘most people who are important to me think I should have my breasts screened’; perceived susceptibility by asking participants whether they believed that ‘my chances of getting breast cancer in the next few years are great’; and perceived barriers by asking them whether they believed that ‘I have other problems more important than getting a mammogram’. For all three these items, a 4-point Likert scale was used with options ‘strongly disagree’, ‘disagree’, ‘agree’, and ‘strongly agree’, or to answer, ‘I am not sure’. Perceived behavioural control was measured by asking participants whether they believed that ‘Keeping my appointment for breast cancer screening will be … ‘, using a 4-point Likert scale ranging from very easy to easy, difficult, or very difficult, with an option to answer, ‘I am not sure’. For each variable, items were dichotomized in the subsequent analysis to the categories of ‘agree/disagree’ or ‘easy/difficult’.
Health literacy was measured using the 6-item version of the European Health Literacy Survey Questionnaire (HLS-EU-Q6) (25), using a 4-point Likert scale per item (very difficult, difficult, easy, very easy). Answers were coded on a scale from 1 to 4 (‘very difficult’ scoring 1; ‘very easy’ scoring 4). The Health Literacy score is then calculated as a mean of the scores of the completed items in the HLS-EU-Q6 Questionnaire (sum of answers/number of items). This presents a mean score that can range from 1 to 4. Three levels for the scale have been defined and validated against the more extensive 47 item version of the European Health Literacy Survey Questionnaire: Inadequate Health Literacy (≤ 2); Limited Health Literacy (> 2 and ≤ 3); Sufficient Health Literacy (> 3) (22). Cronbach’s alpha for the HLS-EU-Q6 score was calculated to check internal consistency on the data of each participating country separately.
Intention to be screened was operationalised by asking participants, after being presented with the correct statement that breast cancer screening carries both benefits and harms, whether this made them more likely to participate in screening, less likely to participate, or neither more nor less likely.
A convenience sample of ten women pre-tested the survey for intelligibility prior to translation, which revealed no problems in the construction of the questionnaire. The questionnaire was put into a web-based survey platform administered by Panelbase UK, which is a research consultancy that performs online surveys to an established panel drawn from the general population who have previously provided consent to be included in such research. Eligible potential participants were contacted via email with a link to the questionnaire, which was incorporated into the routine online omnibus surveys administered by Panelbase. The survey remained open until the minimal sample size per country (n = 225) was exceeded. Incomplete responses or responses with missing values were excluded.
Descriptive statistics were used to present absolute and relative frequencies of the dichotomised variables. Correlation analyses were performed to inspect the association between intention to screen and the antecedent variables included in the conceptual model.
Logistic regression analysis was applied to test two models explaining the intention to be screened for breast cancer: a first model testing the influence of cognitive variables, plus screening history and health knowledge on screening intention, and a second model adjusting the first model for age and household income per annum. Adjusted odds ratios (ORs) are reported with 95% confidence intervals (CIs), with significance set at p < 0.05. A mediation analysis using the Baron-Kenney method and bootstrapping was performed to examine the influence of health literacy (measured via HLS-EU-Q6 mean score) on the relationship between age, household income (as a proxy of socio-economic status), screening history, and health knowledge (independent variables) and intention to screen (dependent variable). PROCESS v3.5 using the Hayes method was calculated for the multi-categorical variable of household income per annum. Data were analysed using IBM SPSS Statistics for Windows, V.27.0 (IBM).