Women’s knowledge and attitudes related to cervical cancer and cervical cancer screening in Isiolo and Tharaka Nithi counties, Kenya: a cross-sectional study

Isiolo and Tharaka Nithi counties have estimated populations of 143,294 and 365,330, respectively. The inhabitants of Isiolo county belong to several ethnic groups, the majority of whom are pastoralists. The population of Tharaka Nithi county are largely the Ameru and engage in mixed agricultural farming [12].

Isiolo county is considered a ‘hard-to-reach’ area. Accessibility is hampered by poor infrastructure, insecurity occasioned by conflicts among inhabitants (due to limited water supply, pasture and other reasons) and with neighbouring countries (Ethiopia and Somalia). This has contributed to economic instability, demonstrated by the uncharacteristically high poverty levels (63.0% for Isiolo county and 49.0% for Tharaka Nithi county versus the national level of 46.0%) [13].

Despite the government’s provision of free primary education and subsidised secondary education, literacy levels are low (59.8% in Isiolo county and 71.2% in Tharaka Nithi county versus the national average of 66.0% in 2013) [14]. The 2014 Kenyan Demographic Health Survey highlighted that in Isiolo county, 39.7% of women of reproductive age lacked formal education compared to 11.4% of their male counterparts. In Tharaka Nithi county, approximately one-third of women and men of reproductive age had some primary education [12]. The median age at first marriage in Isiolo county (18.5 years) is lower than the other counties of Kenya’s Eastern region (20.5 years) [12].

Isiolo county has one referral hospital and 27 health facilities with a physician-to-population ratio of 1:143,000. In 2014, two-thirds of women opted for a home delivery [8]. In contrast, Tharaka Nithi county has three district hospitals, one sub-district hospital and 84 health facilities. In 2014, the physician-to-population ratio was 1:21,000 [11] and 77.7% of women delivered in a health facility [12].

Females aged 18 years old and older at the time of enrolment, residing in the specified counties for at least six months prior to data collection, were considered eligible for participation in the study. Women with cervical cancer diagnosed before taking part in the study were considered ineligible for participation.

The sample size was calculated using the formula for estimating a population proportion n = p(1 − p)(1.96)² ÷ d² [15]. The expected proportion (p) of Kenyan women with adequate or comprehensive knowledge of cervical cancer was 50% (p = 0.5), with a desired precision of 7.0% (d = 0.07). Since this study utilised a multistage cluster sampling method, the sample size was multiplied by the design effect of 2. The sample size calculation also took into consideration a non-response rate of 10.0%. Therefore, the minimum sample size required was 431. Ultimately, a sample of 460 was used to gain sufficient statistical power to explore possible demographic factors associated with knowledge of cervical cancer.

Participants were selected using the multistage cluster sampling technique. This sampling method is effective in geographically dispersed populations. The method eliminates the need for a complete list of all units (households) in the population, and ensures that selected population units will be closer together, thus costs for personal interviews are reduced, and fieldwork simplified [16]. A proportional stratified sample was drawn from the respective counties. Random samples of 30 sub-locations were selected from each county. The households from each sub-location were selected by the systematic random sampling method [17]. The interviewers adhered to a predetermined sampling interval. Only one woman aged 18 years or older per household was interviewed. When an eligible respondent was not available during the first visit, an interviewer arranged alternative visits to complete data collection procedures.

There is no validated questionnaire to assess knowledge and attitudes related to cervical cancer specifically in Kenya. For the purpose of this study, questions to assess attitudes towards cervical cancer were adapted from other validated breast cancer questionnaires including Champion’s Health Belief Model Scale and Powe Fatalism Inventory (modified version) [18‐21]. Questions were chosen based on their relevance to the Kenyan cultural setting, considering the diversity of cultural and religious beliefs in Kenya. The questionnaire consisted of 8 closed-ended questions that assess knowledge of risk factors and 16 closed-ended questions that assess attitudes related to cervical cancer. The questionnaire was initially developed in English and then translated into the local language (Swahili).

The paper-based questionnaire contained sections to capture demographic characteristics, knowledge and attitudes towards cervical cancer and cervical cancer screening. Trained interviewers administered the questionnaire. In cases where the interviewer spoke the same local language as the respondent, questions were asked in the local language. One pilot session of the questionnaire was done in each of five ethnic communities to ensure women respondents were able to understand it and that questions were being interpreted as intended.

All of the questions used to assess knowledge of cervical cancer risk factors in the questionnaire were considered to be true. Knowledge scores for these questions were coded as ‘1’ for a correct response (“Yes”) and ‘0’ for an incorrect (“No”) or ‘not sure’ response. A composite score was derived for each of the 8 questions. A respondent who achieved a composite score greater than or equal to 4 (≥50%) was considered as knowledgeable (average and above), otherwise not [22, 23]. Attitude was assessed on a scale of 1 to 3 (yes / not sure / no, respectively). A negative response was assigned a score of ‘1’; not sure ‘2’; and a positive response ‘3’. An average score was calculated for each respondent from the sum total of 16 questions. The questionnaire is provided in Additional file 1.

The Kuder–Richardson Formula 20 (KR-20) [24] reliability coefficients and Cronbach’s alpha [25] coefficients were calculated for dichotomously scored variables and variables scored on a scale of 1 to 3, respectively. The KR-20 coefficient for the group of questions pertaining to knowledge of risk factors for cervical cancer was 0.71. Values greater than or equal to 0.70 were considered acceptable [26]. Similarly, the Cronbach’s alpha showed acceptable reliability for the group of questions pertaining to attitude assessment of cervical cancer, which was 0.75.

Data was captured in EpiData 3.1 [27] and exported to Stata 13.1 [28] for statistical analysis. Categorical variables are presented as frequency tables, and numerical variables as descriptive measures, expressed as median and range. The association between knowledge of cervical cancer (yes/no) and demographic characteristics was assessed using bivariate and multivariate logistic regression analysis. Odds ratios (OR) were used to test the association between binary variables and 95% confidence intervals (CI) that did not span unity were considered as thresholds of statistical significance. Adjusted odds ratios (aOR) were used in multivariate analysis.

Ethical clearance for this study was provided by the Isiolo and Tharaka Nithi County Health Departments (ethics committee reference: ICDH/NGO.1VOL.1/35) in January 2017. Participants provided written consent to participate in the study. Confidentiality was ensured throughout the process of data collection and analysis through the use of de-identified code numbers. Participants were not remunerated for participation.

A total of 451 women participated in the study, 29.5% from Isiolo county and 70.5% from Tharaka Nithi county, giving a total response rate of 98%. The median age of participants was 32 (ranging between 18 to 85 years) and approximately two-thirds (66.3%) were aged 18 to 39 years (Table 1). The majority were married (70.5%) and half (50.8%) were employed or self-employed. Thirty-five percent of the respondents had a primary level of education while 14.2% were non-literate. There were significant differences in the demographic characteristics between the study participants residing in the two counties (Table 1). The participants in Tharaka Nithi county were significantly older, 64.5% were over the age of 29 years of age compared to the participants in Isiolo county (51.1%) (OR = 1.7; 95% CI: 1.1–2.7). Over half of the participants in Tharaka Nithi (54.1%) had attained primary level of education compared to the participants in Isiolo (25.6%) (OR = 3.4; 95% CI: 2.1–5.5). A significant proportion of Tharaka Nithi women were employed or self-employed (62.6%) compared to the women in Isiolo (22.6%) (OR = 5.7; 95% CI: 3.5–9.5).

Overall, 79.8% (360/451) of the study participants were aware of cervical cancer, and 15.1% (68/451) had heard of HPV. Among those who were aware of cervical cancer, 83.6% (301/360) had heard of cervical cancer screening and 25.6% (92/360) had undergone a cervical cancer screening examination. Those who were aware of cervical cancer reported that their primary sources of information were from family or friends (45.0%, n = 162), a health care facility (40.3%, n = 145), radio/television (40.6%, n = 146), and less than 6.0% (n = 20) stated social media, newspaper or a non-governmental organisation. Fewer than two-thirds of those who had heard about cervical cancer gave the appropriate response to two of the eight questions on risk factors for cervical cancer; cervical cancer is preventable (61.9%, 223/360) and having many different sexual partners is a risk factor (61.1%, 220/360). One in six participants (16.9%, 61/360) knew that HPV is a risk factor for cervical cancer (Table 2).

As described in the methodology, using the composite score for knowledge, the results showed that fewer than half (44.4%) of the participants who were aware of cervical cancer had above-average knowledge of risk factors for cervical cancer. A significant association between the outcome variable (knowledgeable of risk factors for cervical cancer (yes/no)) and selected demographic variables (education, employment status and county of origin) was observed in the bivariate analysis. Only employment status and county of origin were significant predictors of knowledge when adjusted for all of the demographic variables in Table 3. Women who were employed were almost twice as likely to be knowledgeable of cervical cancer (aOR = 1.6; 95% CI: 1.0–2.6) compared to unemployed women, and women in Tharaka Nithi were almost three times more likely to be knowledgeable of cervical cancer (aOR = 2.8; 95% CI: 1.6–5.0) compared to women from Isiolo (Table 3).

Attitudes towards cervical cancer were assessed separately using 16 questions (Table 4). Almost all (89.2%) of those who had heard of cervical cancer categorised it as “scary”. Over half of the women responded that “cervical cancer would threaten a relationship with her husband, boyfriend or partner” (56.7%) and also preferred a female health worker to conduct a cervical examination (55.8%). Nearly two-thirds (61.4%) of respondents perceived the examinations to be positive and believed that “health care workers performing cervical examinations are not rude to women”. There was a marginal significant difference in the overall attitude assessment score towards cervical cancer between participants from Isiolo and Tharaka Nithi counties; the mean (SD) score was 2.13 (0.34) and 2.20 (0.30) respectively. The score was comparatively higher among participants residing in Tharaka Nithi (95% CI: 0.002–0.146; p = 0.043).

Our sample was predominantly rural and may not be generalisable to other populations. The inclusion of more women who were educated, working and married may limit the generalisability of findings to women in these areas who are less educated, unemployed and/ or single. We did not include questions around symptom awareness, which could have provided insights into potential associations with knowledge, attitudes and timely presentation. The self-report nature, although facilitated, might have caused bias and over- or under-estimation of certain variables such as lack of equivalent local terminology for medical words such as HPV. Understanding of the questions among people with first languages other than Swahili may have affected the responses provided, particularly in Isiolo county, where several other languages are spoken. We did not collect data on the languages in which each interview was conducted, but the interviewers were trained on study procedures and were fluent in Swahili and relevant local languages. The influence of religion on knowledge and attitudes was not assessed and could be another influencing factor. Isiolo county is predominantly Muslim while Tharaka Nithi county is predominantly Christian [17]. This variance might have influenced access to information and attitudes around cervical cancer however, previous research has not documented this association. Although the Cronbach’s alpha for the attitude scale was acceptable, we used unvalidated measures for the measurement of attitudes and knowledge, which may have affected the psychometric properties of the measure and is another limitation of our study.

Future studies that assess cervical cancer knowledge and attitudes should consider including questions around symptoms to explore ways for timely presentation at health services. Additional research to further understand and assess the effectiveness of different strategies to increase cervical cancer knowledge, improve attitudes and increase uptake of cervical cancer screening services is needed.