European Journal of Oncology Nursing 20 (2016) 42e48
Contents lists available at ScienceDirect
European Journal of Oncology Nursing journal homepage: www.elsevier.com/locate/ejon
Breast Cancer Screening Beliefs Questionnaire: Psychometric properties assessment of the Arabic version Cannas Kwok a, *, Gihane Endrawes a, Chun Fan Lee b a b
School of Nursing and Midwifery, University of Western Sydney, Sydney, New South Wales, Australia Department of Biostatistics, Singapore Clinical Research Institute, Singapore
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 June 2015 Received in revised form 7 August 2015 Accepted 13 August 2015
Purpose: The aim of the study was to report the psychometric properties of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire (BCSBQ). Methods: A convenience sample of 251 Arabic-Australian women was recruited from a number of Arabic community organizations. Construct validity was examined by Cuzick's non-parametric test while Cronbach a was used to assess internal consistency reliability. Explanatory factor analysis was conducted to study the factor structure. Results: The results indicated that the Arabic version of the BCSBQ had satisfactory validity and internal consistency. The Cronbach's alpha of the three subscales ranged between 0.810 and 0.93. The frequency of breast cancer screening practices (breast awareness, clinical breast-examination and mammography) were significantly associated with attitudes towards general health check-up and perceived barriers to mammographic screening. Exploratory factor analysis showed a similar fit for the hypothesized threefactor structure with our data set. Conclusions: The Arabic version of the BCBSQ is a culturally appropriate, valid and reliable instrument for assessing the beliefs, knowledge and attitudes to breast cancer and breast cancer screening practices among Arabic-Australian women. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Arabic women Breast cancer screening Validation Cultural sensitivity Psychometric properties
1. Introduction The global burden of breast cancer is growing. This is particularly the case in the Middle Eastern countries where, over the last two decades, the incidence of breast cancer among Arab women has increased significantly compared to that of women in developed countries such as the US and Australia (Mittra, 2011). Arab women moreover, develop breast cancer at an early median age of 49e52 years (El Saghir et al., 2007) compared to the 61e63 year median among women in more developed countries (American Cancer Society, 2013; Cancer Australia, 2015). Arab women also face a significantly higher risk of mortality due to the advanced stage of the cancer at diagnosis (El Saghir et al., 2007). In Western countries, significant improvements in breast cancer survival rates over the last 20 years have been largely due to the
* Corresponding author. School of Nursing and Midwifery, University of Western Sydney, Locked bag 1797, Penrith, 2751, New South Wales, Australia. E-mail address: [email protected] (C. Kwok). http://dx.doi.org/10.1016/j.ejon.2015.08.003 1462-3889/© 2015 Elsevier Ltd. All rights reserved.
practice of early detection measures, particularly breast selfexamination (BSE), clinical breast examination (CBE) and mammography (American Cancer Society, 2015; Cancer Australia, 2015). That achievement in turn can be attributed to the considerable efforts made by both governments and organisations to promote breast cancer screening practices. For example, women aged 50 to 74 are offered a free mammogram every two years in Australia. While this has benefited Caucasian women, immigrant women from culturally and linguistically diverse (CALD) groups living in Western countries have been consistently reported as having lower participation rates in screening practices. This phenomenon has been demonstrated by studies in Australia (Kwok et al., 2012), the USA (Choi et al., 2010) and the UK (Jack et al., 2014). Similar studies have indicated that Arab women living in their home countries (Elobaid et al., 2014; Othman et al., 2012) or after immigration to Western countries (Schwartz et al., 2008) have relatively low participation rates in breast cancer screening. This conclusion has been reinforced in a study by Petro-Nustas et al. (2012) which found that only 24.7% of Arab women in the USA had practised monthly BSE over the previous 12 months. In their
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
study of Arab women in Qatar, Donnelly et al. (2013) reported that only 31.3% and 26.9% of participants have presented themselves for CBE and mammographic screening respectively. While there has been no investigation of screening rates among women from Arabic background in Australia, it is likely that they may also have low screening rates. Since evidence suggests that screening rates among women from CALD backgrounds are low, extensive research has been conducted internationally to examine the breast cancer screening behaviours of CALD women (Crawford et al., 2015; Lee et al., 2012; Robb, 2010). By now, a body of evidence has been built up which conclusively demonstrates that religious beliefs and culturallybased attitudes play a vital role in determining women's screening behaviours (Ahmad et al., 2012; Padela et al., 2014; Wang et al., 2009). Some of these beliefs and attitudes are common across diverse cultural groups. For example, the belief among many Arab women that health and illness are due to the will of God “Allah” and that no one can escape their fate (Azaiza and Cohen, 2008; Donnelly et al., 2013) is similar to the belief of many Chinese women that if an individual is destined to have cancer, nothing can change that and therefore screening measures are pointless (Gonazlez et al., 2014; Shang et al., 2014). Robb (2010) concluded that some Arab women were concerned that performing BSE might threaten their religious values because it could be construed as a challenge to the will of Allah. A cultural barrier to BSE among many Chinese women is that touching the body, particularly the breasts, is a taboo (Kwok et al., 2006). Although the performance of screening measures by asymptomatic individuals is a well-established health concept in the Western health paradigm (American Cancer Society, 2013; Cancer Australia, 2015), this concept is unknown among some CALD groups. Many Arab women seek help for a health problem only when there is an obvious need for treatment and thus, when they are asymptomatic, they see no need for screening measures (Mamdouh et al., 2015) Studies have demonstrated that this is also the case among Chinese- (Kwok et al., 2012) and Korean-American women (Lee et al., 2015). Overseas studies indicate that for Arab women, further barriers to breast cancer screening include feelings of embarrassment, lack of language proficiency, lack of medical knowledge, fear of pain and discomfort during the procedure, economic difficulties and also difficulty with transportation which makes it difficult for them to access screening sites (Elobaid et al., 2014; Montazeri et al., 2008; Radi, 2013). In Australia, Arabic-speaking people represent 1.3% of the entire population and constitute the third largest immigrant language group after those who speak Mandarin and Italian at home (Australian Bureau of Statistics, 2012). Nevertheless, research into breast cancer screening practices within the Arabic community is very limited. To understand the factors associated with women's screening behaviours, a valid and reliable instrument is essential. The aim of the present study was to assess and report on the psychometric properties of the ‘Breast Cancer Screening Beliefs Questionnaire’ (BCSBQ) in an Arabic community in Australia. The BCSBQ was originally designed as a culturally sensitive instrument to assess the knowledge and attitudes of CALD women with regard to breast cancer and screening practices. It was first developed and validated by Kwok et al. (2010) for Chinese-Australian women and recently when validated in studies among Korean (manuscript under-review) and Indian communities (Kwok et al., 2015) in Australia, it was shown to have a high degree of reliability. 2. Methods This is a questionnaire.
cross-sectional
study
using
a
self-reported
43
2.1. Participants A convenient sample of 251 Arabic-Australian women were recruited through organizations such as churches and community centres in Sydney. The eligibility criteria were that the participant had migrated to Australia, was over 18 years old who self-identified from Arabic background and had no history of breast cancer. 2.2. Data collection Approval from the Ethics Committee of the relevant University was obtained prior to the data collection phase of this project. As a first step, the second author, who is from an Arabic-Australian background, made contact with leaders in the Arabic community and provided them with the details of the study before asking them to help with the recruitment of participants. After gaining the support of the leaders, women members of these organizations were invited to participate in the study. Those who agreed to do so received a participant information sheet and questionnaire in either English or Arabic according to their language preference. The second author was also available to provide further information to participants, who were invited to complete the questionnaire and place it in a secure container located in the common area of these organizations. Filling in the questionnaire (which took approximately 20 min) and returning it was taken as an implied consent to participate in the study. 2.3. Instrument The BCSBQ is a 13-item instrument composed of three subscales: 1) attitudes towards general health check-ups with a subscale of four items designed to ascertain whether a women had general health check-ups in the absence of signs and symptoms; 2) knowledge and perceptions about breast cancer (four items) which explored cultural beliefs relating to breast cancer and 3) barriers to mammographic screening practices (five items) which covered personal and practical issues perceived by women to hinder their participation in breast cancer screening. All of the items were rated on a five point Likert scale ranging from ‘strongly agree’ (score of 1) to “strongly disagree” (score of 5). Lower scores indicated the lowest attitudes, least knowledge or greatest barriers. A brief description of the items is illustrated in Fig. 1. The instrument collected demographic information such as age, length of time in Australia, English language proficiency and education levels. Participants were then asked if they had heard of the concept of breast awareness, (in terms of which women are encouraged to become familiar with their breasts and which in Australia, has replaced the narrower the term ‘breast self-examination’), and whether and how often they undertook CBE and mammograms. 2.4. Translation of the instrument The original English version of BCSBQ was translated by the second author who is fluent in English and Arabic. Back translation was conducted by an independent bilingual translator to ensure lexical equivalence and that the meaning is not lost in translation (Behling and Law, 2000). The BCBSQ was next piloted with ten Arabic-Australian women with diverse demographic backgrounds such as age, English proficiency, length of stay in Australia and educational level. Women were asked to comment on the clarity, comprehensibility and readability of the questionnaire. No problematic sentences or wording were reported.
44
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
Q1. If I feel well, it is not necessary to have a health check-up Q2. If I follow a healthy lifestyle such as a balanced diet and regular exercise, I don’t feel it is necessary to have a regular health check-up Q3. I see a doctor or have my health check-up only when I have a health problem Q4. If I feel healthy, I do not need to see the doctor
0.88 0.86 0.89
Attitudes towards general health check-ups
0.89
0.10 Q5. Breast cancer is like a death sentence; if you get it, you will surely die from it
0.82
Q6. Breast cancer cannot be cured; you can only prolong the suffering
0.92
Q7. Even if breast cancer is detected early, there is very little a woman can do to reduce the chances of dying from it Q8. If a woman is fated to get breast cancer, she will get breast cancer; there is nothing she can do to change fate
0.87
Knowledge and perceptions about breast cancer
0.38
0.72
0.22
0.29
Q9. I’m worried that having a mammogram will hurt my breasts
0.42
Q10. It would be difficult to arrange transportation for getting a mammogram
0.51
Q11. I don’t want to have a mammogram because I can’t speak English
0.55
Q12. I don’t want to go for a mammogram because I would need to take off my clothes and expose my breasts”
0.98
Q13. Having a mammogram is embarrassing
Barriers to mammographic screening
0.92
Fig. 1. Path diagram of a confirmatory factor analysis of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire. The values correspond to the standardized estimates.
2.5. Sample size While there is no specific or agreed minimum sample size for factor analysis, Streiner and Norman (2003) recommended the use of 5e10 participants per item. The authors aimed at recruiting the maximum number of ten participants for each of the 13 items of the BCSBQ scale and on this score the results were eminently satisfactory. Of the 350 women invited to participate, 268 returned the questionnaire, a response rate of 76%. Five of these were excluded because they had history of breast cancer while 12 failed to complete the questionnaire. Thus the final number of participants was 251, an acceptable sample size for the confirmatory factor analysis. 2.6. Statistical analysis The three subscale scores of the Arabic version of the BCSBQ were computed in the same way as those of the Chinese version reported elsewhere (Kwok et al., 2010). If at least half of the items in that subscale were answered and valid, missing values would be imputed by the mean of the responses in the same subscale, i.e. by the half-rule. Participants' demographic characteristics and the distribution of the subscale scores of the instrument were summarized using descriptive statistics. Floor and ceiling effects were examined. Substantial floor and ceiling effects suggested that a 5point Likert scale might not be sufficient clearly to distinguish the responses at the two extremes. Item performance was first assessed. Internal consistency
reliability was evaluated by the Cronbach alpha. A good Cronbach alpha should be between 0.7 and 0.9 since a low value indicates a low degree of homogeneity while a too high value indicates item redundancy (Streiner and Norman, 2003). Corrected item-total correlations (rcorr) were also computed to evaluate the convergent-divergent validity. Construct validity was examined by testing three hypotheses regarding the association of the subscales with the frequency of screening practices and education level: (1) those who reported more frequent practice of breast awareness, CBE and mammographic examinations, were likely to have more open attitudes towards breast cancer screening and this would be reflected in higher scores in the Attitude subscale; (2) those who had attained higher educational levels were likely to be more knowledgeable about breast cancer screening and thus obtain a higher scores in the Knowledge subscale; and (3) the results of those who performed screening practices more frequently and were associated with fewer barriers to screening, would be reflected in higher scores in the Barriers subscale. As the frequency of mammographic examinations and education levels were of ordinal-type data, the trend was subjected to Cuzick's nonparametric test (Cuzick, 1985). To study whether the Arabic BCSBQ had the originally designed 3-factor structure, a confirmatory factor analysis (CFA) was performed. Goodness-of-fit of the factor model was assessed by the fit indices, including the root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), comparative fit index (CFI), and non-normed fit index (NNFI). These
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
indices addressed the parsimony correction, absolute fit, and comparative or incremental fit, as recommended (Hu and Bentler, 1998). We followed the common criteria in considering the adequacy of the factor model, i.e., RMSEA 0.06, SRMR 0.08, CFI 0.95 and NNFI 0.95 (Hu and Bentler, 1999). Addition of covariance between items was made based on the largest modification index if there were doubts about inadequate fit (Brown, 2006). If the factor structure could not be confirmed, an exploratory factor analysis (EFA) would be conducted. The number of factors would be determined by the scree plot. Factor loadings after a varimax rotation would be computed. The final EFA structure would then be compared with the originally hypothesized 3-factor structure.
3. Results A total of 251 women completed the questionnaire, and their demographic characteristics are summarized in Table 1. The agerange of the cohort was between 18 and 70 years with a mean (standard deviation) of 39.3 (13.0) years. The participants had lived
Table 1 Demographic characteristics of the 251 participants. Characteristic
N (%)
Age (year) (Mean: 39.3, SD: 13.0) >20 11 (4.4) 20e29 63 (25.1) 30e39 56 (22.3) 40e49 54 (21.5) 50e59 49 (19.5) 60e69 16 (6.4) 70 or above 2 (0.8) Country of birth Egypt 171 (68.1) Lebanon 33 (13.1) Jordan 8 (3.2) Iraq 10 (4.0) Others 29 (11.6) Language spoken at home Arabic 217 (86.5) English 32 (12.7) Others 2 (0.8) Length of stay in Australia (year) (Mean: 10.8, SD: 12.3, missing: N ¼ 4) 0e5 142 (57.5) 6e10 6 (2.4) 11e15 17 (6.9) 16e20 30 (12.1) 21e25 22 (8.9) 26 or above 30 (12.1) Marital status Single 48 (19.1) Married/living together 178 (70.9) Divorced/separated 12 (4.8) Widowed 13 (5.2) Education level Primary school 28 (11.2) Secondary school 67 (26.7) TAFE/college 48 (19.1) Tertiary or above 108 (43.0) Current employment status Employed, full time 69 (27.5) Employed, part time 53 (21.1) Unemployed 102 (40.6) Retired 27 (10.8) English proficiency Not at all 42 (16.7) Little 30 (12.0) Average 59 (23.5) Good 55 (21.9) Very good 65 (25.9)
45
in Australia for a mean of 10.8 (12.3) years, most were married (70.9%), had been born in Egypt (68.1%) and spoke Arabic at home (86.5%). Imputation for missing values was not required because all participants answered all 13 items. The distributions of the three subscales are summarized in Table 2. All three subscales had a range from 0 to 100. The Attitude subscale showed a floor effect with 22.3% of the participants answering 0 for all items, while the Knowledge subscale showed a ceiling effect with 15.9% answering 5 for all items. The Cronbach alpha of the three subscales ranged from 0.81 to 0.93 (Table 3). For the Attitude and Knowledge subscales, items correlated strongly with their own subscale (all rcorr > 0.7) but only weakly to moderately with other subscales (all rcorr 0.5). Comparatively, rcorr between the Barriers subscale and its items were smaller, ranging from 0.44 to 0.83, but item Q11 had a correlation of 0.50 with the Attitude subscale. The mean scores of the three subscales stratified by the frequency of screening practices and education level are shown in Table 4. For both Attitude and Barrier subscales, the scores were significantly higher among those who more frequently practised breast awareness, CBE and mammographic screening (all pvalues < 0.001). Women with higher educational levels also had significantly higher scores in the all three subscales (all pvalues < 0.001). In other words, the three hypothesized associations were all significant. The CFA of the hypothesized 3-factor structure of this version of the BCSBQ resulted in a chi-square statistic ¼ 260.0 (degrees of freedom ¼ 62, p-value < 0.001), RMSEA ¼ 0.12 (95% confidence interval (CI) ¼ 0.11 to 0.14), SRMR ¼ 0.11, CFI ¼ 0.90 and NNFI ¼ 0.88. After examining the modification index, a covariance between Q9 and Q10 was added to the factor model and resulted in a chi-square statistic ¼ 258.2 (degrees of freedom ¼ 61, pvalue < 0.001), RMSEA ¼ 0.11 (95% CI ¼ 0.10 to 0.13), SRMR ¼ 0.10, CFI ¼ 0.92 and NNFI ¼ 0.89. The final CFA model is shown in Fig. 1. Since the pre-specified criteria were not satisfied, an EFA was performed. The eigenvalues for the first five factors were 4.89, 3.07, 1.64, 0.80 and 0.56. Having examined the scree plot (Fig. 2), a 3factor model was identified which explained 73.9% of the total variance. Table 5 shows the rotated factor loadings after varimax rotation. Items having a rotated factor loading with magnitude 0.4 within a particular factor were considered to be its major component, and are highlighted in Table 5. All items of the Attitude and Knowledge subscales loaded more heavily on the first and second factors respectively. While the items in the Barriers subscale loaded on the third factor, the item Q11 also had a loading of 0.51 on the first factor. If the three factors respectively represent the Attitude, Knowledge and Barriers subscales, the model would be the same as the originally hypothesized 3-factor structure tested in the CFA. 4. Discussion To improve breast cancer screening participation among ArabicAustralian women, it is imperative to have a valid instrument to explore the factors associated with their screening behaviours. The present study evidenced that the Arabic version of the BCSBQ has appropriate psychometric properties which can provide insights into tailor-made strategies designed to address the needs of this cultural group. A notable feature of this version of the BCBSQ was the severe floor and ceiling effects in the subscales, suggesting that the 5-point Likert scale may not be sufficient to distinguish responses at the lower extreme of the Attitude subscale and also the higher extreme of the Knowledge subscale. Nevertheless, the instrument appeared to have good internal consistency reliability under each of the three subscales ranging from 0.81 to 0.93 with no indication of overlap
46
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
Table 2 Distribution of the subscale scores of the 13-item Arabic version of the Breast Cancer Screening Beliefs Questionnaire. Subscale
Mean
Standard deviation
Median
Minimum
Maximum
% at floor
% at ceiling
Attitudes towards general health check-ups Knowledge and perceptions about breast cancer Barriers to mammographic screening
32.9 63.1 60.5
29.0 27.1 25.5
25.0 68.8 65.0
0 0 0
100 100 100
22.3 2.0 1.2
4.0 15.9 9.2
Table 3 Cronbach's alpha and corrected item-total correlation for the subscales of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire. Item
Attitudes towards general health check-ups
Cronbach's alpha 0.93 Attitudes towards general health check-ups Q1 0.84 Q2 0.83 Q3 0.85 Q4 0.84 Knowledge and perceptions about breast cancer Q5 0.04 Q6 0.10 Q7 0.08 Q8 0.10 Barriers to mammographic screening Q9 0.16 Q10 0.18 Q11 0.50 Q12 0.34 Q13 0.39
Knowledge and perceptions about breast cancer
Barriers to mammographic screening
0.90
0.81
0.08 0.05 0.10 0.12
0.42 0.40 0.33 0.41
0.75 0.83 0.82 0.69
0.29 0.32 0.30 0.26
0.38 0.40 0.18 0.19 0.16
0.44 0.57 0.47 0.83 0.76
Table 4 Construct validity of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire. N (%)
Breast awareness (missing: N ¼ 6) At least once a month Once every few months Once a year Never P-value for trend Clinical breast examination A year or less More than a year and less than two years Two to three years More than three years Never had one P-value for trend Mammogram Once a year Once every two years Once every three years or more Never had one P-value for trend Education level Never attended school Primary school Secondary school TAFE/college P-value for trend
Attitudes towards general health check-ups
Knowledge and perceptions about breast cancer
Barriers to mammographic screening
Mean (SD)
Mean (SD)
Mean (SD)
19 (7.8) 30 (12.2) 66 (26.9) 130 (53.1)
53.6 (35.5) 47.9 (33.3) 39.0 (30.5) 22.8 (21.5)
Contents lists available at ScienceDirect
European Journal of Oncology Nursing journal homepage: www.elsevier.com/locate/ejon
Breast Cancer Screening Beliefs Questionnaire: Psychometric properties assessment of the Arabic version Cannas Kwok a, *, Gihane Endrawes a, Chun Fan Lee b a b
School of Nursing and Midwifery, University of Western Sydney, Sydney, New South Wales, Australia Department of Biostatistics, Singapore Clinical Research Institute, Singapore
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 June 2015 Received in revised form 7 August 2015 Accepted 13 August 2015
Purpose: The aim of the study was to report the psychometric properties of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire (BCSBQ). Methods: A convenience sample of 251 Arabic-Australian women was recruited from a number of Arabic community organizations. Construct validity was examined by Cuzick's non-parametric test while Cronbach a was used to assess internal consistency reliability. Explanatory factor analysis was conducted to study the factor structure. Results: The results indicated that the Arabic version of the BCSBQ had satisfactory validity and internal consistency. The Cronbach's alpha of the three subscales ranged between 0.810 and 0.93. The frequency of breast cancer screening practices (breast awareness, clinical breast-examination and mammography) were significantly associated with attitudes towards general health check-up and perceived barriers to mammographic screening. Exploratory factor analysis showed a similar fit for the hypothesized threefactor structure with our data set. Conclusions: The Arabic version of the BCBSQ is a culturally appropriate, valid and reliable instrument for assessing the beliefs, knowledge and attitudes to breast cancer and breast cancer screening practices among Arabic-Australian women. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Arabic women Breast cancer screening Validation Cultural sensitivity Psychometric properties
1. Introduction The global burden of breast cancer is growing. This is particularly the case in the Middle Eastern countries where, over the last two decades, the incidence of breast cancer among Arab women has increased significantly compared to that of women in developed countries such as the US and Australia (Mittra, 2011). Arab women moreover, develop breast cancer at an early median age of 49e52 years (El Saghir et al., 2007) compared to the 61e63 year median among women in more developed countries (American Cancer Society, 2013; Cancer Australia, 2015). Arab women also face a significantly higher risk of mortality due to the advanced stage of the cancer at diagnosis (El Saghir et al., 2007). In Western countries, significant improvements in breast cancer survival rates over the last 20 years have been largely due to the
* Corresponding author. School of Nursing and Midwifery, University of Western Sydney, Locked bag 1797, Penrith, 2751, New South Wales, Australia. E-mail address: [email protected] (C. Kwok). http://dx.doi.org/10.1016/j.ejon.2015.08.003 1462-3889/© 2015 Elsevier Ltd. All rights reserved.
practice of early detection measures, particularly breast selfexamination (BSE), clinical breast examination (CBE) and mammography (American Cancer Society, 2015; Cancer Australia, 2015). That achievement in turn can be attributed to the considerable efforts made by both governments and organisations to promote breast cancer screening practices. For example, women aged 50 to 74 are offered a free mammogram every two years in Australia. While this has benefited Caucasian women, immigrant women from culturally and linguistically diverse (CALD) groups living in Western countries have been consistently reported as having lower participation rates in screening practices. This phenomenon has been demonstrated by studies in Australia (Kwok et al., 2012), the USA (Choi et al., 2010) and the UK (Jack et al., 2014). Similar studies have indicated that Arab women living in their home countries (Elobaid et al., 2014; Othman et al., 2012) or after immigration to Western countries (Schwartz et al., 2008) have relatively low participation rates in breast cancer screening. This conclusion has been reinforced in a study by Petro-Nustas et al. (2012) which found that only 24.7% of Arab women in the USA had practised monthly BSE over the previous 12 months. In their
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
study of Arab women in Qatar, Donnelly et al. (2013) reported that only 31.3% and 26.9% of participants have presented themselves for CBE and mammographic screening respectively. While there has been no investigation of screening rates among women from Arabic background in Australia, it is likely that they may also have low screening rates. Since evidence suggests that screening rates among women from CALD backgrounds are low, extensive research has been conducted internationally to examine the breast cancer screening behaviours of CALD women (Crawford et al., 2015; Lee et al., 2012; Robb, 2010). By now, a body of evidence has been built up which conclusively demonstrates that religious beliefs and culturallybased attitudes play a vital role in determining women's screening behaviours (Ahmad et al., 2012; Padela et al., 2014; Wang et al., 2009). Some of these beliefs and attitudes are common across diverse cultural groups. For example, the belief among many Arab women that health and illness are due to the will of God “Allah” and that no one can escape their fate (Azaiza and Cohen, 2008; Donnelly et al., 2013) is similar to the belief of many Chinese women that if an individual is destined to have cancer, nothing can change that and therefore screening measures are pointless (Gonazlez et al., 2014; Shang et al., 2014). Robb (2010) concluded that some Arab women were concerned that performing BSE might threaten their religious values because it could be construed as a challenge to the will of Allah. A cultural barrier to BSE among many Chinese women is that touching the body, particularly the breasts, is a taboo (Kwok et al., 2006). Although the performance of screening measures by asymptomatic individuals is a well-established health concept in the Western health paradigm (American Cancer Society, 2013; Cancer Australia, 2015), this concept is unknown among some CALD groups. Many Arab women seek help for a health problem only when there is an obvious need for treatment and thus, when they are asymptomatic, they see no need for screening measures (Mamdouh et al., 2015) Studies have demonstrated that this is also the case among Chinese- (Kwok et al., 2012) and Korean-American women (Lee et al., 2015). Overseas studies indicate that for Arab women, further barriers to breast cancer screening include feelings of embarrassment, lack of language proficiency, lack of medical knowledge, fear of pain and discomfort during the procedure, economic difficulties and also difficulty with transportation which makes it difficult for them to access screening sites (Elobaid et al., 2014; Montazeri et al., 2008; Radi, 2013). In Australia, Arabic-speaking people represent 1.3% of the entire population and constitute the third largest immigrant language group after those who speak Mandarin and Italian at home (Australian Bureau of Statistics, 2012). Nevertheless, research into breast cancer screening practices within the Arabic community is very limited. To understand the factors associated with women's screening behaviours, a valid and reliable instrument is essential. The aim of the present study was to assess and report on the psychometric properties of the ‘Breast Cancer Screening Beliefs Questionnaire’ (BCSBQ) in an Arabic community in Australia. The BCSBQ was originally designed as a culturally sensitive instrument to assess the knowledge and attitudes of CALD women with regard to breast cancer and screening practices. It was first developed and validated by Kwok et al. (2010) for Chinese-Australian women and recently when validated in studies among Korean (manuscript under-review) and Indian communities (Kwok et al., 2015) in Australia, it was shown to have a high degree of reliability. 2. Methods This is a questionnaire.
cross-sectional
study
using
a
self-reported
43
2.1. Participants A convenient sample of 251 Arabic-Australian women were recruited through organizations such as churches and community centres in Sydney. The eligibility criteria were that the participant had migrated to Australia, was over 18 years old who self-identified from Arabic background and had no history of breast cancer. 2.2. Data collection Approval from the Ethics Committee of the relevant University was obtained prior to the data collection phase of this project. As a first step, the second author, who is from an Arabic-Australian background, made contact with leaders in the Arabic community and provided them with the details of the study before asking them to help with the recruitment of participants. After gaining the support of the leaders, women members of these organizations were invited to participate in the study. Those who agreed to do so received a participant information sheet and questionnaire in either English or Arabic according to their language preference. The second author was also available to provide further information to participants, who were invited to complete the questionnaire and place it in a secure container located in the common area of these organizations. Filling in the questionnaire (which took approximately 20 min) and returning it was taken as an implied consent to participate in the study. 2.3. Instrument The BCSBQ is a 13-item instrument composed of three subscales: 1) attitudes towards general health check-ups with a subscale of four items designed to ascertain whether a women had general health check-ups in the absence of signs and symptoms; 2) knowledge and perceptions about breast cancer (four items) which explored cultural beliefs relating to breast cancer and 3) barriers to mammographic screening practices (five items) which covered personal and practical issues perceived by women to hinder their participation in breast cancer screening. All of the items were rated on a five point Likert scale ranging from ‘strongly agree’ (score of 1) to “strongly disagree” (score of 5). Lower scores indicated the lowest attitudes, least knowledge or greatest barriers. A brief description of the items is illustrated in Fig. 1. The instrument collected demographic information such as age, length of time in Australia, English language proficiency and education levels. Participants were then asked if they had heard of the concept of breast awareness, (in terms of which women are encouraged to become familiar with their breasts and which in Australia, has replaced the narrower the term ‘breast self-examination’), and whether and how often they undertook CBE and mammograms. 2.4. Translation of the instrument The original English version of BCSBQ was translated by the second author who is fluent in English and Arabic. Back translation was conducted by an independent bilingual translator to ensure lexical equivalence and that the meaning is not lost in translation (Behling and Law, 2000). The BCBSQ was next piloted with ten Arabic-Australian women with diverse demographic backgrounds such as age, English proficiency, length of stay in Australia and educational level. Women were asked to comment on the clarity, comprehensibility and readability of the questionnaire. No problematic sentences or wording were reported.
44
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
Q1. If I feel well, it is not necessary to have a health check-up Q2. If I follow a healthy lifestyle such as a balanced diet and regular exercise, I don’t feel it is necessary to have a regular health check-up Q3. I see a doctor or have my health check-up only when I have a health problem Q4. If I feel healthy, I do not need to see the doctor
0.88 0.86 0.89
Attitudes towards general health check-ups
0.89
0.10 Q5. Breast cancer is like a death sentence; if you get it, you will surely die from it
0.82
Q6. Breast cancer cannot be cured; you can only prolong the suffering
0.92
Q7. Even if breast cancer is detected early, there is very little a woman can do to reduce the chances of dying from it Q8. If a woman is fated to get breast cancer, she will get breast cancer; there is nothing she can do to change fate
0.87
Knowledge and perceptions about breast cancer
0.38
0.72
0.22
0.29
Q9. I’m worried that having a mammogram will hurt my breasts
0.42
Q10. It would be difficult to arrange transportation for getting a mammogram
0.51
Q11. I don’t want to have a mammogram because I can’t speak English
0.55
Q12. I don’t want to go for a mammogram because I would need to take off my clothes and expose my breasts”
0.98
Q13. Having a mammogram is embarrassing
Barriers to mammographic screening
0.92
Fig. 1. Path diagram of a confirmatory factor analysis of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire. The values correspond to the standardized estimates.
2.5. Sample size While there is no specific or agreed minimum sample size for factor analysis, Streiner and Norman (2003) recommended the use of 5e10 participants per item. The authors aimed at recruiting the maximum number of ten participants for each of the 13 items of the BCSBQ scale and on this score the results were eminently satisfactory. Of the 350 women invited to participate, 268 returned the questionnaire, a response rate of 76%. Five of these were excluded because they had history of breast cancer while 12 failed to complete the questionnaire. Thus the final number of participants was 251, an acceptable sample size for the confirmatory factor analysis. 2.6. Statistical analysis The three subscale scores of the Arabic version of the BCSBQ were computed in the same way as those of the Chinese version reported elsewhere (Kwok et al., 2010). If at least half of the items in that subscale were answered and valid, missing values would be imputed by the mean of the responses in the same subscale, i.e. by the half-rule. Participants' demographic characteristics and the distribution of the subscale scores of the instrument were summarized using descriptive statistics. Floor and ceiling effects were examined. Substantial floor and ceiling effects suggested that a 5point Likert scale might not be sufficient clearly to distinguish the responses at the two extremes. Item performance was first assessed. Internal consistency
reliability was evaluated by the Cronbach alpha. A good Cronbach alpha should be between 0.7 and 0.9 since a low value indicates a low degree of homogeneity while a too high value indicates item redundancy (Streiner and Norman, 2003). Corrected item-total correlations (rcorr) were also computed to evaluate the convergent-divergent validity. Construct validity was examined by testing three hypotheses regarding the association of the subscales with the frequency of screening practices and education level: (1) those who reported more frequent practice of breast awareness, CBE and mammographic examinations, were likely to have more open attitudes towards breast cancer screening and this would be reflected in higher scores in the Attitude subscale; (2) those who had attained higher educational levels were likely to be more knowledgeable about breast cancer screening and thus obtain a higher scores in the Knowledge subscale; and (3) the results of those who performed screening practices more frequently and were associated with fewer barriers to screening, would be reflected in higher scores in the Barriers subscale. As the frequency of mammographic examinations and education levels were of ordinal-type data, the trend was subjected to Cuzick's nonparametric test (Cuzick, 1985). To study whether the Arabic BCSBQ had the originally designed 3-factor structure, a confirmatory factor analysis (CFA) was performed. Goodness-of-fit of the factor model was assessed by the fit indices, including the root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), comparative fit index (CFI), and non-normed fit index (NNFI). These
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
indices addressed the parsimony correction, absolute fit, and comparative or incremental fit, as recommended (Hu and Bentler, 1998). We followed the common criteria in considering the adequacy of the factor model, i.e., RMSEA 0.06, SRMR 0.08, CFI 0.95 and NNFI 0.95 (Hu and Bentler, 1999). Addition of covariance between items was made based on the largest modification index if there were doubts about inadequate fit (Brown, 2006). If the factor structure could not be confirmed, an exploratory factor analysis (EFA) would be conducted. The number of factors would be determined by the scree plot. Factor loadings after a varimax rotation would be computed. The final EFA structure would then be compared with the originally hypothesized 3-factor structure.
3. Results A total of 251 women completed the questionnaire, and their demographic characteristics are summarized in Table 1. The agerange of the cohort was between 18 and 70 years with a mean (standard deviation) of 39.3 (13.0) years. The participants had lived
Table 1 Demographic characteristics of the 251 participants. Characteristic
N (%)
Age (year) (Mean: 39.3, SD: 13.0) >20 11 (4.4) 20e29 63 (25.1) 30e39 56 (22.3) 40e49 54 (21.5) 50e59 49 (19.5) 60e69 16 (6.4) 70 or above 2 (0.8) Country of birth Egypt 171 (68.1) Lebanon 33 (13.1) Jordan 8 (3.2) Iraq 10 (4.0) Others 29 (11.6) Language spoken at home Arabic 217 (86.5) English 32 (12.7) Others 2 (0.8) Length of stay in Australia (year) (Mean: 10.8, SD: 12.3, missing: N ¼ 4) 0e5 142 (57.5) 6e10 6 (2.4) 11e15 17 (6.9) 16e20 30 (12.1) 21e25 22 (8.9) 26 or above 30 (12.1) Marital status Single 48 (19.1) Married/living together 178 (70.9) Divorced/separated 12 (4.8) Widowed 13 (5.2) Education level Primary school 28 (11.2) Secondary school 67 (26.7) TAFE/college 48 (19.1) Tertiary or above 108 (43.0) Current employment status Employed, full time 69 (27.5) Employed, part time 53 (21.1) Unemployed 102 (40.6) Retired 27 (10.8) English proficiency Not at all 42 (16.7) Little 30 (12.0) Average 59 (23.5) Good 55 (21.9) Very good 65 (25.9)
45
in Australia for a mean of 10.8 (12.3) years, most were married (70.9%), had been born in Egypt (68.1%) and spoke Arabic at home (86.5%). Imputation for missing values was not required because all participants answered all 13 items. The distributions of the three subscales are summarized in Table 2. All three subscales had a range from 0 to 100. The Attitude subscale showed a floor effect with 22.3% of the participants answering 0 for all items, while the Knowledge subscale showed a ceiling effect with 15.9% answering 5 for all items. The Cronbach alpha of the three subscales ranged from 0.81 to 0.93 (Table 3). For the Attitude and Knowledge subscales, items correlated strongly with their own subscale (all rcorr > 0.7) but only weakly to moderately with other subscales (all rcorr 0.5). Comparatively, rcorr between the Barriers subscale and its items were smaller, ranging from 0.44 to 0.83, but item Q11 had a correlation of 0.50 with the Attitude subscale. The mean scores of the three subscales stratified by the frequency of screening practices and education level are shown in Table 4. For both Attitude and Barrier subscales, the scores were significantly higher among those who more frequently practised breast awareness, CBE and mammographic screening (all pvalues < 0.001). Women with higher educational levels also had significantly higher scores in the all three subscales (all pvalues < 0.001). In other words, the three hypothesized associations were all significant. The CFA of the hypothesized 3-factor structure of this version of the BCSBQ resulted in a chi-square statistic ¼ 260.0 (degrees of freedom ¼ 62, p-value < 0.001), RMSEA ¼ 0.12 (95% confidence interval (CI) ¼ 0.11 to 0.14), SRMR ¼ 0.11, CFI ¼ 0.90 and NNFI ¼ 0.88. After examining the modification index, a covariance between Q9 and Q10 was added to the factor model and resulted in a chi-square statistic ¼ 258.2 (degrees of freedom ¼ 61, pvalue < 0.001), RMSEA ¼ 0.11 (95% CI ¼ 0.10 to 0.13), SRMR ¼ 0.10, CFI ¼ 0.92 and NNFI ¼ 0.89. The final CFA model is shown in Fig. 1. Since the pre-specified criteria were not satisfied, an EFA was performed. The eigenvalues for the first five factors were 4.89, 3.07, 1.64, 0.80 and 0.56. Having examined the scree plot (Fig. 2), a 3factor model was identified which explained 73.9% of the total variance. Table 5 shows the rotated factor loadings after varimax rotation. Items having a rotated factor loading with magnitude 0.4 within a particular factor were considered to be its major component, and are highlighted in Table 5. All items of the Attitude and Knowledge subscales loaded more heavily on the first and second factors respectively. While the items in the Barriers subscale loaded on the third factor, the item Q11 also had a loading of 0.51 on the first factor. If the three factors respectively represent the Attitude, Knowledge and Barriers subscales, the model would be the same as the originally hypothesized 3-factor structure tested in the CFA. 4. Discussion To improve breast cancer screening participation among ArabicAustralian women, it is imperative to have a valid instrument to explore the factors associated with their screening behaviours. The present study evidenced that the Arabic version of the BCSBQ has appropriate psychometric properties which can provide insights into tailor-made strategies designed to address the needs of this cultural group. A notable feature of this version of the BCBSQ was the severe floor and ceiling effects in the subscales, suggesting that the 5-point Likert scale may not be sufficient to distinguish responses at the lower extreme of the Attitude subscale and also the higher extreme of the Knowledge subscale. Nevertheless, the instrument appeared to have good internal consistency reliability under each of the three subscales ranging from 0.81 to 0.93 with no indication of overlap
46
C. Kwok et al. / European Journal of Oncology Nursing 20 (2016) 42e48
Table 2 Distribution of the subscale scores of the 13-item Arabic version of the Breast Cancer Screening Beliefs Questionnaire. Subscale
Mean
Standard deviation
Median
Minimum
Maximum
% at floor
% at ceiling
Attitudes towards general health check-ups Knowledge and perceptions about breast cancer Barriers to mammographic screening
32.9 63.1 60.5
29.0 27.1 25.5
25.0 68.8 65.0
0 0 0
100 100 100
22.3 2.0 1.2
4.0 15.9 9.2
Table 3 Cronbach's alpha and corrected item-total correlation for the subscales of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire. Item
Attitudes towards general health check-ups
Cronbach's alpha 0.93 Attitudes towards general health check-ups Q1 0.84 Q2 0.83 Q3 0.85 Q4 0.84 Knowledge and perceptions about breast cancer Q5 0.04 Q6 0.10 Q7 0.08 Q8 0.10 Barriers to mammographic screening Q9 0.16 Q10 0.18 Q11 0.50 Q12 0.34 Q13 0.39
Knowledge and perceptions about breast cancer
Barriers to mammographic screening
0.90
0.81
0.08 0.05 0.10 0.12
0.42 0.40 0.33 0.41
0.75 0.83 0.82 0.69
0.29 0.32 0.30 0.26
0.38 0.40 0.18 0.19 0.16
0.44 0.57 0.47 0.83 0.76
Table 4 Construct validity of the Arabic version of the Breast Cancer Screening Beliefs Questionnaire. N (%)
Breast awareness (missing: N ¼ 6) At least once a month Once every few months Once a year Never P-value for trend Clinical breast examination A year or less More than a year and less than two years Two to three years More than three years Never had one P-value for trend Mammogram Once a year Once every two years Once every three years or more Never had one P-value for trend Education level Never attended school Primary school Secondary school TAFE/college P-value for trend
Attitudes towards general health check-ups
Knowledge and perceptions about breast cancer
Barriers to mammographic screening
Mean (SD)
Mean (SD)
Mean (SD)
19 (7.8) 30 (12.2) 66 (26.9) 130 (53.1)
53.6 (35.5) 47.9 (33.3) 39.0 (30.5) 22.8 (21.5)
