Site selection and recruitment

The region of Kédougou is divided into three medical districts: Kédougou, Saraya, and Salemata. Each of these health districts has a single health center in the district capital and multiple health posts in the rural surrounding communities. We selected nine sites in the Kédougou Medical Region through non-probability sampling including one health center and two rural health posts from each of the three districts comprising the region. In the Kédougou District we selected the Dalaba health post (population accessing this health post = 5995), Bandafassi (7189), and Dindefello (9370). In the Salemata District we selected the Salemata Health Center (7278) and the health posts of Dar Salaam (3084) and Dakately (3037). In the Saraya District, we selected the Saraya Health Center (5890) and the posts of Nafadji (3759) and Khossanto (3471). Each of the nine sites was mapped through OpenStreetMaps. [33] Printable maps were created (using FieldPapers [34]) to illustrate structures (assumed to be households), roads, and rivers in each site. Maps were divided into four sectors with approximately the same number of structures in each quadrant. Structures were numbered and Google’s random number generator was used to determine the starting point. Counting in increments of the limiting factor, each chosen structure was marked and recorded. This ensured a relatively even distribution of structures selected throughout each site. Twenty structures per site (n = 180) were selected and visited in order to assess for eligibility. Potential participants were recruited using the approved recruitment script. Households were selected if there was both an eligible woman and man who agreed to participate. An additional five women were recruited in each site from among the women’s group to strengthen the assessment at the community level. The target sample size of 225 (135 women and 90 men) was determined based on the need to have a heterogeneous sample across language and district. Our sample was not adequately powered to detect differences between screened and unscreened women.

Development of documents

The questionnaire included closed-ended, quantitative questions seeking information on participant’s perceptions of discrimination, cancer stigma, opinions, and attitudes. We included adapted questions from the Everyday Discrimination Scale [35] and the Cancer Stigma Scale. [36] The questionnaires were first created in English, translated into French and the local languages of Jakhanke/Malinke and Pula Fuuta (a dialect of Fula/Pulaar), and then back-translated for accuracy by certified Sénégalese translators. Questionnaires were field tested for comprehension prior to the initiation of the study. All IRB approved documents including study overview, recruitment scripts, and the informed consent were available in French, Malinke, and Pulaar.

Consent and data collection

All research assistant data collectors participated in a three-day training on the project protocol including data collection methodology facilitated by the lead investigator prior to field testing the instrument. After final institutional review board approval, research assistants attended an additional three-day training to review all data collection procedures. The study research assistants read the informed consent aloud, in the participants preferred language, and participants reviewed and signed the approved informed consent short form, written in French. In cases where the participant did not read French, a trusted contact was requested by the participant to witness the informed consent process, observe the signature of the participant, and then sign as a witness. After participants were consented, data collection was conducted immediately with a female research assistant collecting data from women and a male research assistant collecting data from men. All data collection activities were performed in a private setting. Data collection interviews occurred in the preferred language of the participant. All responses were recorded on hard copy interview forms with the name of the participant being the only item recorded on the final page. All data collection instruments were immediately handed over to the lead research assistant, who recorded the participant’s name on the participant code book, placed a unique identifier on page one of the data collection instrument, removed and destroyed the final page of each instrument, scanned all documents, and transmitted them through a secure portal to a research assistant in the United States.

Data analysis

Data were double-entered into an electronic spreadsheet by two research assistants, compared, approved and subsequently cleaned by the principal investigator. Bivariate analyses were conducted to 1) assess gender differences in the distribution of all variables including the adapted everyday discrimination scale, cancer stigma scale, and various cervical cancer attitude measures; 2) determine whether there were significant differences in beliefs and attitudes by screening behavior; and 3) explore the potential effect of educational attainment on gender differences in attitudes and beliefs. To accomplish the third analysis, we created a composite variable combining gender and education, categorizing those with Quranic school or no school as having “Low Education” and those who attended primary school, secondary school, and above as having “Higher Education.” Associations were tested using the Fisher’s Exact test statistic (we computed the p-value for analyses by way of the Two- Stage Fisher’s Exact Test using RStudio version 1.2.1578 through the Dplyr and Arsenal packages).

The conduct of responsible research and partnership

The University of Illinois at Chicago has an ongoing partnership affiliation agreement with the Institute of Health and Development at the University Cheikh Anta Diop and with the Medical Region of Kédougou. This partnership uses a participatory approach ensuring that all activities are well-aligned with the expressed priorities of the local health system. This human subjects research was approved by the University of Illinois at Chicago Institutional Review Board and the Institutional Review Board at the Ministry of Health and Social Action in Senegal. The Medical Region of Kédougou, the three health districts, and participating health posts granted researchers permission through signed letters of support to implement and conduct the data collection activities. Each investigator and U.S. based research assistant received the Collaborative Institutional Training Initiative (CITI) training certification prior to conducting the research. [37] All local research assistants were trained in research ethics through a locally approved ethics of human research training program.

Screened correlation

Subsequently, we correlated all variables with screening behavior among women. S1 Table lists these results. Among women who have been screened once, 92.3% strongly recommend that other women get screened compared to 44.0% of women never screened. (P = 0.006) Among women who have never been screened, 48.2% disagree and 4.7% strongly disagree that they would not feel comfortable around someone with cancer. However, among those who have been screened a single time, 15.4% disagree and 46.2% strongly disagree with this statement (P < 0.001). Among women who were screened one time, 61.5% strongly agreed that if a person has cancer it’s probably their fault compared to 22.4% of women who have never been screened (P = 0.003). In addition, 76.9% of women who have been screened one time strongly agreed that other women often state that they are worried about getting cancer compared to only 44.7% of women never screened (P = 0.019).

In correlating cancer attitudes with screening behavior we discovered that 91.7% of women screened a single time strongly agreed that they would want to know if they were diagnosed with cancer compared with 49.4% of women never screened (P = 0.039). However, results indicating desire for a family member to know about diagnosis were insignificant. It is notable that on this question nearly half of those who had been screened one time either disagreed (23.1%) or strongly disagreed (23.1%) that they would want a family member to know about their diagnosis compared to 17.6% and 8.2% respectively for women never screened. We found that among both women who have never been screened and those who have been screened one time, most prefer to know about the diagnosis of someone else in the family. Of those who have never been screened, 40.5% agree and 29.8% strongly agree that they would want to be informed. Among those screened once, 15.4% agree and 53.9% strongly agree that they would like to be informed of a family member’s diagnosis. (P = 0.041) Among women screened a single time, 61.5% strongly disagreed that cancer is fate and there is nothing they can do to change fate compared with 18.8% of never screened women who strongly disagreed with this statement. (P = 0.012) (Table not shown. See S1 Table)

Education effect on gender perceptions

In examining the effect of education on gender perceptions across all variables, it should be noted that 97% of the women in the study received no more than primary education, while 23.1% of men received some secondary schooling or above. We, therefore, compared the effect of Quranic school or no education to primary education or above across all variables. Through this analysis, we identified statistically significant findings for education level relative to perceptions of a woman’s role as caretaker (P < 0.001) and a man’s role as decision maker (P < 0.001). Among women with “Higher Education” 47.5% disagreed with the statement “A woman’s most important role is to take care of her home and cook for her family” and only 11.7% of women with “Low Education” disagreed while among men, 3.8% and 16.7% of those with “Low Education” and “Higher Education” respectively disagreed. Among women with “Higher Education” 37.5% disagreed with the statement “A man should have the final word about decisions in his home” and only 14.8% of women with “Low Education” disagreed while among men, 10.3% of those with “Higher Education” disagreed. No men classified as having “Low Education” disagreed with this statement. (See S2 Table)

Concerning the adapted everyday discrimination scale analyzed with the gender-education composite variable, we found significant results for a) ‘generally feel treated with less courtesy than others’ (P = 0.008), b) ‘feel treated with less respect by their spouse,’ (P = 0.010) c) ‘perceived as unintelligent by others’ (P < 0.001), d) ‘feel perceived as being dishonest’ (P = 0.008), and e) ‘feel threatened by others’ (P = 0.042). (See S3 Table) Regarding the adapted cancer stigma scale, we found significant results regarding a) ‘comfort with being around someone with cancer’ (P < 0.001), b) ‘those diagnosed with cancer are no longer normal’ (P < 0.001), c) ‘the healthcare needs of people with cancer should not be prioritized’ (P < 0.001), d) ‘having cancer is probably the fault of that person’ (P < 0.001), e) ‘cancer is more frightening than other diseases’ (P < 0.001), and f) ‘other women often state that they are worried about getting cancer’ (P < 0.001),. Illustrating this effect of education, we found that 59.0% of women with “Higher Education” disagreed that having cancer is the fault of that person, while 27.9% of women who had “Low Education” disagreed with this statement. Among men, we found that 50.0% and 34.5% of those with “Low” and “Higher” education respectively strongly disagreed with the statement. (See S4 Table) In addition, we found significant effect of education on gender perceptions on other cancer attitudes including: a) the value of cancer testing, even if unpleasant (P = 0.011), b) desire to be informed about one’s personal cancer diagnosis (P = 0.024), c) desire for family to be aware of one’s cancer diagnosis (P = 0.002), c) desire to know of cancer diagnosis for another family member (P < 0.001), and d) getting cancer is fate (P = 0.001). (See S5 Table) We did not find any significant impact of education on gender perceptions of recommendation for the cervical cancer screening test. (See S6 Table)

Structural gender disadvantages

Our data reveal some structural disadvantages for women within our study population. Women were less educated than men, and, in turn, were less likely to speak the national languages, Wolof and French. Education and language are both meaningful social determinants of health [38] and may be indicators of status and empowerment, both of which are key to accessing healthcare within this population. In addition, the social role for women in Southeastern Sénégal appears to be largely agreed upon between women and men. In this region, women generally work in the home and men conduct business and make decisions on the part of the family. Somewhat more women than men disagreed with these prescribed roles lending weight to the desire by some women for more autonomy.

Our data also indicate gender differences among variables in the adapted everyday discrimination scale within this population. Women were much more likely to state that they are treated with less courtesy or respect by their spouse on a frequent basis. In contrast, feeling general disrespect outside of the home are not significant. This finding may, therefore, be closely tied to the accepted social roles of women and men. Interestingly, men were somewhat more likely to feel perceived as being unintelligent. We have no explanatory mechanism, but this phenomenon may be related to men being much more likely to take on the role of leader in business interactions. In doing so, men are much more likely to travel and engage with others who have higher levels of education. They, therefore, may be comparing themselves to a different audience than women.

Attitudes and stigma

Our findings indicate that women are much more likely to personally stigmatize cancer. Women are significantly more likely than men to state that a) they would feel uncomfortable around someone with cancer, b) someone with a cancer diagnosis is never normal again, c) the health needs of people with cancer should not be prioritized, d) if a person has cancer it is probably their fault, e) I feel that cancer is more frightening than most other diseases, and f) other women often state that they are worried about getting cancer. Both women and men agreed on the importance of seeking screening or treatment for cancer, as well as the desire to know personally about a screening result. However, we found a significant difference between women and men concerning the desire for others to know about a personal cancer diagnosis. We found a similar pattern concerning the desire to know about a family member’s cancer diagnosis. This may indicate that women are also more likely to have a perceived stigma against cancer. These findings are intriguing and warrant further investigation. Higher rates of both personal and perceived cancer stigma among women has profound implications for how population and community level communication strategies for cancer prevention and control should be designed. Given that social norms play a critical role in the development of stigmas, we need a better understanding of the negative social influences that shape women’s and men’s knowledge, attitudes, and beliefs, as well as whether early positive social influences may be impactful in increasing early uptake for cancer prevention services.

Social influences

Our data show that women are much more likely to rely on the guidance and advice of their spouses or others within their immediate social network while men rely on the recommendations of health professionals when considering health care decisions. The large majority of women and men agree that it is the men who have the healthcare seeking decision making power at home. While the individual variables in our study (categorized as attitudes, discrimination, and stigma) are not altogether specific to cervical cancer, our data does indicate that both women and men would recommend the cervical cancer screening test to others. The perceived recommendation of the screening test from other women, however, is not as considerable.

Screening behavior

Women who have been screened are much more likely than non-screened women to recommend that other women get routine cervical cancer screening. In addition, it appears that they are less likely to display personal stigma toward cancer patients as evidenced by disagreement with having discomfort around a cancer patient. Screened women do note that other women are worried about getting cancer. If they were personally diagnosed or someone in their family were diagnosed, they would overwhelmingly want to know. However, on the question to inform family members of their own diagnosis, they are split. Further inquiry to explore this contradictory finding could pursue whether some women prefer others not be informed of a diagnosis because of shame, perceived stigma, stoicism, or other potential reasons.

Interestingly, screened women tended to strongly agree that a diagnosis of cancer is the fault of the person. Even though all questions were field tested, this may indicate confusion underlying the premise of the question. It is not possible to know if the women who responded in this way have responded specific to the context of cervical cancer and are indicating that they feel that screening is a way to prevent this illness, and, therefore, it is the responsibility of the individual to seek this service. Lending weight to this supposition is the fact that 61.5% of screened women strongly disagreed that cancer is fate. This significant finding illustrates the importance of ensuring that women feel empowered to prevent cervical cancer through screening by ensuring that stigma, attitudes, and beliefs are prioritized through outreach efforts aimed at uptake.

Effect of educational achievement on gender perceptions

Our findings exploring the effect of educational achievement on gender perceptions have considerable implications for future work. Concerning a woman’s role, there is a clear correlation with increased education among women and men and disagreement with the statement outlining a very traditional perspective. Likewise, increased levels of education indicate that both men and women are more likely to disagree that men are the sole decision maker in the family. We see similar attenuation of gender specific findings on the adapted everyday discrimination scale, the adapted cancer scale, and other cancer-specific attitudes and opinions. This brings to light critical questions to be addressed through future research, the results of which could impact the development of educational, behavioral change, and social mobilization focused interventions aimed at cancer and, specifically, cervical cancer prevention and control at various levels (individual, household, community, organizational, and policy).

Confronting barriers and context

Our findings that illustrate that gender roles in decision making, gender influences in discrimination and cancer stigmas, and other structural barriers such as educational attainment and language are meaningful social determinants of health related to cervical cancer screening uptake in Senegal. The existing literature exploring these themes is not robust but presents some insights that may be helpful in further interpreting our findings and guiding our next steps. As an example, cultural norms, gender roles, knowledge, and stigma were identified as socio-cultural factors influencing a woman’s decision to seek cervical cancer screening in disadvantaged communities in Cape Town, South Africa. [39] In addition, a study in Cusco, Peru linked underlying determinants such as fear, embarrassment, community conversations about cervical cancer, willingness to talk about cervical cancer, and gender dynamics, including spousal support, to health communications preferences. They found that cultural misconceptions and male perspectives were significant factors predicting screening uptake with an overwhelming need for interventions addressing sociocultural influences in order to address the underlying root causes. [40] These findings suggest that behavior interventions aimed at increasing the uptake of cervical cancer screening services should utilize strategies that go beyond simple health communication and education. In fact, one study examining factors related to breast and cervical cancer screening uptake among Cambodian and Thai women in Southern California identified similar structural barriers that these women were, in large part, unable to overcome such barriers without the assistance of a community navigator. [41] While it is critical to address knowledge gaps, our challenge will be to concurrently address underlying social determinants of cervical cancer screening uptake through positive social influence.

Limitations

There were some limitations in the methodology used to recruit households within each site. It was assumed that every structure identified on the map was a household and it was thus, included in the count because specific household information was not provided. In reality, some structures were businesses, vacant, or not present. When encountering this scenario, we progressed to the next marked structure on the list. In addition, the satellite images used to create the maps on OpenStreetMaps may have been outdated. In some cases data collection was attempted separately from the recruitment and consent process. This may have resulted in the inability to collect data from some of the selected households due to the participant being away during the day(s) that the research assistant was present. Considerable effort was made to select communities that are representative of the immediate and surrounding districts and regions of rural Sénégal. However, our sample may not be generalizable to other areas of Sénégal or other countries in West Africa. In addition, we must use caution in interpreting these results given the low numbers of women within our study sample who have been screened for cervical cancer. We will follow the trend of these indicators over time as women are exposed to the peer education program.