Introduction

Breast and cervical cancer are the leading causes of cancer morbidity and mortality in women in Sub-Saharan Africa (SSA). Breast cancer is the most commonly diagnosed cancer among women in SSA, whilst cervical cancer remains the leading cause of cancer death [1]. Most SSA countries do not have cervical or breast cancer screening programs and the majority of cancers (85–90%) present symptomatically and at an advanced stage [2].

Studies have shown that for symptomatic breast cancer, a shorter time between recognition of symptoms by women and first medical consultation is associated with early stage disease, and better breast cancer survival [3–6]. Further, a substantial proportion of advanced stage diagnosis of poorly differentiated breast cancer cases could be avoided if patients had presented within one month of detecting symptoms [4]. For both breast and cervical cancer there are often subtle but important symptoms in early stage disease, yet women may misinterpret these symptoms or wait until symptoms (and disease) progress before they seek medical attention [7–10]. Research outside SSA has indicated that cancer symptom awareness among the general population can be low [11–12], and that interventions to increase awareness lead to better outcomes. For example, in Malaysia, raising public awareness of symptoms has been associated with earler diagnosis of cervical cancer [13].

For people with potential symptoms of cancer, the pathway to cancer diagnosis is complex and influenced by many factors such as symptom and risk awareness, beliefs and barriers to accessing care [9,10,14–18]. Measuring these factors is important in developing and evaluating interventions to promote timely cancer diagnosis. A recent scoping study highlighted the need for local research arising from low-and middle-income countries to adequately address breast and cervical cancer prevention and control [19].

Although a few studies have reported on breast and cervical cancer symptom awareness, risk perception and help-seeking behavior in African settings [9,10,20,21], measurement and comparison is hampered by the lack of validated, culturally relevant measurement tools. The Cancer Awareness Measure (CAM) and the Awareness and Beliefs about Cancer (ABC) tools, developed and validated in the UK to reliably assess cancer awareness in the general public, are questionnaires that include generic and cancer site specific modules [22–26]. The CAM and ABC tools could be of value beyond the UK, but as they were developed in a very different cultural setting they need to be adapted and validated for use in SSA. A Kenyan research team adapted the Breast Cancer Awareness Measurement (BCAM) for local use and reported on validity testing of two of the tool domains (symptom awareness and barriers to screening) [27]. The team reported good content validity and internal reliability for the two domains, demonstrating that the BCAM could be adapted for use in African populations, but they also recommended further validation. The aim of the current study was to develop and validate an African Women’s Awareness of CANcer (AWACAN) tool to measure awareness of symptoms and risk factor for breast and cervical cancer, lay beliefs, confidence in appraising potential symptoms, help-seeking behavior and barriers to health care.

Methods

The four step processes of 1) generating tool items, 2) refining and 3) validating the English version of the AWACAN tool and, 4) developing local language versions of the tool are described below. The outcomes of Step 3 are presented in the results section. Ethics approval for the study was obtained from University of Cape Town, Faculty of Health Sciences Human Research Ethics Committee (HREC 544/2016), the Lacor Hospital Institutional Research Ethics Committee (LHIREC 027/11/2016) and the Ugandan National Council of Science and Technology (UNCST 194/212/01). Written informed consent was obtained from all participants. Permission to access health facilities was obtained from all relevant local health authorities. Permission to conduct interviews with non-medical university staff was obtained from the institutional human resource department.

Results

In total there were 139 community (SA n = 72, Uganda n = 67), 23 non-medical staff and 20 cancer expert participants who completed the validation studies (Step 3). Table 1 outlines the socio-demographic characteristics of the participant groups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Socio-demographic characteristics of participants.

https://doi.org/10.1371/journal.pone.0220545.t001

Test-retest reliability

There was no loss-to follow up between the first and second round of interviews. The percentage of agreement for responses to the socio-demographic questions was high (> 96% for all questions and total agreement for 8 of the 11 socio-demographic variables). Kappa scores for the socio-demographic questions ranged between 0.83 and 1.00.

Table 2 shows the agreement, the Brennan and Prediger prevalence-adjusted and bias-adjusted kappa and the unadjusted kappa scores in brackets for items relating to awareness, help-seeking behavior and confidence. For breast cancer, the adjusted kappa co-efficient ranged from 0.57 to 1.0 with one item (‘self-medicate’) showing moderate agreement and other items indicating substantial or almost perfect agreement. For cervical cancer, the adusted kappa coeeficients ranged from 0.65–1.0 i.e. all showed substantial or higher agreement. Unadjusted kappa co-efficient (shown in brackets) ranged from -0.06 to 1.0 for both cancers.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Test-retest reliability of awareness, help-seeking behavior and confidence.

https://doi.org/10.1371/journal.pone.0220545.t002

Test-retest reliability was moderate for knowledge of known risk factors, good for symptoms and excellent for lay beliefs of breast cancer (Table 3). For cervical cancer, test-retest reliability was moderate for known risk factors, lay beliefs and the symptom domain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Test-retest reliability of knowledge domains (N = 23).

https://doi.org/10.1371/journal.pone.0220545.t003

Construct validity

A high proportion of community participants had heard of breast and cervical cancer (90% and 84% respectively) and there were no statistically significant differences between experts and community members for these items (1-sided Fischer’s exact test p value = 0.184 and 0.222 respectively).

For breast cancer, community participant and expert scores were significantly different for knowledge of evidence-based risk factors and risk lay beliefs but similar for breast cancer symptoms (Table 4). There was clear discrimination between the community participants and cancer experts’ knowledge of evidence-based risk factors and symptoms of cervical cancer. Knowledge did not differ significantly for the 1 cervical cancer risk lay belief (p = 0.444).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Comparison of expert and community participant knowledge.

https://doi.org/10.1371/journal.pone.0220545.t004

Internal reliability

The internal reliability for the breast cancer knowledge questions was good with KR-20 values > 0.7 for the known risk factors, lay belief and symptom subscales (Table 5). For cervical cancer the KR-20 for the known risk factors subscale was lower (0.60). Cervical cancer risk factors items with a very low item-to-total correlation were: HPV (0.18), using birth control/family planning pills for more than 5 years (0.13), not going for regular screening (0.09). Internal reliability for the cervical cancer symptom questions was good (0.80).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Internal reliability of the AWACAN tool.

https://doi.org/10.1371/journal.pone.0220545.t005

Discussion

To our knowledge this is the first validated tool measuring breast and cervical cancer symptom and risk factor awareness, lay beliefs, confidence in appraising potential symptoms, help seeking behaviour and barriers to care in SSA populations. The AWACAN tool follows a similar format to the BCAM and CCAM [25,26]. Knowledge of risk factors and symptoms are first assessed with open-ended questions followed by a prompted checklist of known factors. Participants are not permitted to revise open-ended responses. This sequence of questioning allows respondents an opportunity to describe their own views without the possibility of guessing or being limited by the pre-existing options of the prompted questions that follow. Although responses to open-ended questions are more time-consuming to analyze compared to prompted questions, they can provide insights into terms and phrases commonly used to describe risk and symptoms which could assist in providing culturally relevant and easily understood language for interventions. Open-ended responses can also assist in identifying locally relevant risk and symptom lay beliefs.

Our AWACAN tool performed well in terms of test-retest reliability indicating that the questions are stable over time. Most socio-demographic, help-seeking behaviour and confidence items achieving substantial or higher adjusted kappa coefficients and all knowledge sub-scales showing moderate or higher reliability as measured by the ICC.

The AWACAN tool discriminated well between the expert and community level knowledge of breast cancer risk factors and risk lay beliefs. However, breast symptom knowledge was similar between these two groups. This could be a result of greater public awareness of breast symptoms, which is a focus of many breast awareness campaigns. Knowledge of cervical cancer items risk factors and symptoms was significantly higher in cancer experts compared to community participants, establishing construct validity for these subscales. At the stage of psychometric testing we only had one cervical cancer risk lay belief and although experts were better able to identify this lay belief the difference was not statistically different. Using responses from the open-ended questions we were able to identify additional lay belief items.

Internal reliability was good for our breast cancer questions, with all knowledge domains achieving a KR-20 above the recommend cut-off value of 0.7. Internal reliability was also high (KR-20 = 0.8) for the cervical cancer symptom domain, but lower (KR-20 = 0.6) for the subsection on risk factors. The three risk factor items contributing to this low internal reliability (HPV, use of birth control/family planning pills, not going for regular screening) were retained in the final tool on grounds of content validity.

After establishing good validity for the English version of the tool, it was translated into two 2 local languages (isiXhosa and Acholi), taking into account conceptual and cultural equivalence. In addition to these translations, there are some other important differences between our validated AWACAN tool and the BCAM and CCAM tools. Our tool includes additional prompted items on symptom and risk lay beliefs that were identified from the literature [27], and in responses to the open questions administered during the AWACAN development and validation process. These questions provide insights into lay beliefs that may need to be addressed when developing SSA-relevant and targeted interventions to improve timely diagnosis in these populations. In addition, the lay belief questions which are interspersed between known risk factors and symptoms serve as distractor items. For effectiveness, distractor items should represent incorrect responses that study participants could be expected to produce and avoid obviously implausible responses [41]. We were able to identify a number of breast and cervical cancer risk lay beliefs, but only one breast and one cervical cancer symptom lay belief. When the AWACAN tool is used in future studies, evaluation of open responses could assist in identifying additional locally relevant symptom distractor items.

Based on the recommendation of Wachira et al [27], we developed images for 3 breast cancer symptoms on dark-skinned women to be used with the tool. We believe that this is an important addition to cancer awareness measurement tools in our context. Further, these images can be used in awareness raising interventions. The AWACAN tool also includes simple explanations for terms that were not well understood by participants in our setting (e.g. menopause). Compared to the CAM tools our final tool does not include age-related risk factors as these were poorly understood by community participants and also received a low CVI score from experts.

The BCAM and CCAM tools make use of Likert scales, which call for a graded response to a series of statements about risk and confidence. Likert scales are widely used in surveys, however studies have reported challenges such as difficulty in understanding graded response options, skipping questions or bias toward either the midpoint or extreme options [42,43]. During cognitive testing we noted that participants in both SA and Uganda had difficulty understanding the Likert-type scales used in the CAM tools. Questions were thus modified in our tool to allow for simpler response options.

Understanding women’s help-seeking behaviour and potential barriers to care are critical to informing efforts to improve timely diagnosis of these cancers. The ABC and CAM tools includes measures of barriers to care and help-seeking behaviour relevant in high-income setting [23,25,26]. We included additional questions to reflect the different realities in the SSA context. For example, traditional healers are often consulted for health-related issues in SSA [44] and are included in our measures. We reduced the number of questions on barriers to seeking care in the final tool and made substantial changes to the wording of the remaining questions. Validity of this domain of our questionnaire will be further evaluated in future use of the tool and attention paid to context-specific variations.

Over the past few decades, there has been an increasing recognition that a holistic and comprehensive approach to women’s health is central to sustainable development [45,46]. We developed the AWACAN tool so that the two most common cancer in females in SSA can be addressed in one study to maximise data collection from women. However the tool can also be used for either breast or cervical cancer assessments alone.

A limitation of our study was the difference in mode of delivery to the cancer experts (online) versus other participants (face-to-face) during the validation process. We designed the online tool so that experts were unable to return to previous pages to change responses. However, we recognize that experts could have searched the internet or consulted elsewhere before completing the tool online. We used Classical Test Theory to assess reliability and validity of the AWACAN tool. An alternative would be to use Item Response Theory [47,48]. However this would have required an extremely large sample size given the number of items in the questionnaire. Classical Test Theory was used in the development of the CAM and is adequate for this developmental phase, especially in a context with limited resources. The isiXhosa and Acholi versions of the AWACAN tool have recently being used in a cross-sectional study and further psychometric testing with these local language versions of the tool are being undertaken. In addition we intend to administer and assess validity of the AWACAN tool in other SSA countries. Important and interesting differences are likely to be identified with this broader administration.

We have demonstrated that the breast and cervical AWACAN tool has evidence of reliability and validity and is ready to be used as a measure of risk factor and symptom awareness, lay beliefs, help seeking behavior and barriers to care for the two leading causes of cancer morbidity among women in SSA.

Supporting information

Acknowledgments

The authors thank the participants for sharing their experiences and participating in the validation processes; the providers and managers of the primary health care clinics and our students for support in the data collection process and the University of Cape Town human resources department for facilitating access to non-medical staff. This research is linked to the CanTest Collaborative, which is funded by Cancer Research UK [C8640/A23385], of which Fiona M Walter is Director.