Research | Open Access | Volume 9 (3): Article 146 | Published: 16 Sep 2026
Menu, Tables and Figures
Table 1: Socio-demographic characteristics of participating caregivers, overall and by ward (N = 204)
Table 2: HPV vaccination uptake among adolescent girls aged 15–18 years, overall and by ward (N = 204)
Table 3: Reasons for not intending to vaccinate, among caregivers of unvaccinated girls (n = 123, multiple responses permitted)
Table 4: Item-level distribution of correct responses on the 15-item knowledge scale (N = 204)
Table 5: Composite knowledge score and knowledge category, overall and by ward (N = 204)
Table 6: Item-level distribution of favourable responses on the 18-item perceptions and attitudes scale (N = 204)
Table 7: Bivariate associations between caregiver characteristics and HPV vaccination uptake (N = 204)
Table 8: Crude and adjusted prevalence ratios for HPV vaccination uptake: modified Poisson regression with robust variance (N = 204)
Sabla Galgalo Kubi1,2, Mary Joy Kaimuri1, Jane Rutto1
1Department of Public Health, School of Health Sciences, Meru University of Science and Technology, Meru, Kenya, 2Isiolo County Health Department, Isiolo, Kenya
&Corresponding author: Sabla Galgalo Kubi, Department of Public Health, School of Health Sciences, Meru University of Science and Technology, P.O. Box 972-60200, Meru, Kenya, Email: sablagalgalo@gmail.com ORCID: https://orcid.org/0009-0009-8713-4530
Received: 24 Jun 2026, Accepted: 31 Aug 2026, Published: 16 Sep 2026
Domain: Vaccine Preventable Diseases, Cancer Epidemiology
Keywords: HPV vaccination, pastoralist, Kenya, caregivers, knowledge, attitudes, mixed methods, vaccine hesitancy, Health Belief Model
©Sabla Galgalo Kubi et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Sabla Galgalo Kubi et al. Caregiver factors associated with HPV vaccination uptake in a pastoralist community in Kenya. Journal of Interventional Epidemiology and Public Health. 2026; 9(3):146. https://doi.org/10.37432/jieph-d-26-00194
Introduction: Cervical cancer remains the leading cause of cancer death among women in Kenya, and HPV vaccination coverage is lowest in mobile pastoralist populations. This study examined the level of, and factors associated with, HPV vaccination uptake among caregivers of adolescent girls aged 15–18 years.
Methods: A concurrent mixed-methods study among 204 caregivers selected by multi-stage sampling across the three wards of the sub-county. Knowledge was measured with a 15-item composite score, and perceptions and attitudes with an 18-item scale. Proportions are reported with Wilson 95% confidence intervals (CI), and binary logistic regression exhibited complete separation; adjusted prevalence ratios (aPR) were estimated using modified Poisson regression with robust variance. Two focus group discussions and ten key informant interviews were analysed thematically using the Braun and Clarke framework.
Results: Uptake of at least one dose of HPV was 38.7% (95% CI: 32.3–45.6), and completion of the two-dose schedule was 11.3% (95% CI: 7.6–16.3); only 29.1% (95% CI: 20.3–39.9) of those who initiated completed the schedule. Ward differences were not statistically significant (P=0.121). Knowledge was poor in 65.2% (95% CI: 58.4–71.4) of caregivers, and attitudes were poor in 82.8% (95% CI: 77.1–87.4). Among the 125 caregivers of unvaccinated girls, 61.6% expressed no intention to vaccinate; the most frequently reported reasons were fear of infertility (62.6%) and doubts about vaccine safety (59.3%). In the adjusted model, good knowledge (aPR: 2.32, 95% CI: 1.40–3.84) and fair knowledge (aPR: 1.74, 95% CI: 1.12–2.70) relative to poor knowledge, and good attitudes (aPR: 1.72, 95% CI: 1.16–2.56) relative to poor attitudes, remained independently associated with uptake. Occupation remained an independent association, with livestock-dependent caregivers less likely (aPR: 0.41, 95% CI: 0.20–0.81) and petty traders more likely (aPR: 1.81, 95% CI: 1.12–2.92) to have vaccinated daughters than housewives. Qualitative accounts converged on fertility anxiety, moral framing of adolescent sexuality, paternal veto over girls’ health decisions, weak initial mobilisation and the exclusion of out-of-school and mobile girls from school-based delivery.
Conclusion: HPV vaccination uptake in this pastoralist setting is far below national and global targets and is associated most strongly with caregiver knowledge and attitudes, with an additional independent contribution from livelihood. Interventions should prioritise community-embedded communication delivered through trusted local voices, deliberate engagement of male household heads, and outreach adapted to mobility, alongside consolidation of Kenya’s November 2025 transition to a single-dose schedule.
Human papillomavirus (HPV) is the most common sexually transmitted infection worldwide and is responsible for a wide range of pathological conditions in women and men, including precancerous lesions that may progress to cancer [1,2]. Most infections are asymptomatic and clear spontaneously within two years, but persistent infection with high-risk genotypes may lead to cervical intraepithelial neoplasia and, if untreated, to invasive cervical cancer [3,4]. HPV-16 and HPV-18 account for approximately 70% of cervical cancer cases globally, with types 31, 33, 45, 52 and 58 contributing a further substantial share [5,6]. Current global estimates place cervical cancer as the fourth most common cancer among women, with roughly 660,000 new cases and 350,000 deaths each year, some 90% of which occur in low- and middle-income countries, and sub-Saharan Africa carries a disproportionate share of both HPV prevalence and cervical cancer mortality [2,7].
Vaccination and screening are the two pillars of the World Health Organization strategy for cervical cancer elimination. HPV vaccines are safe and highly effective when administered before first sexual exposure, and universal vaccination of girls is the single most effective means of reducing HPV-attributable cervical cancer [8,9]. The global elimination targets require that 90% of girls be fully vaccinated by age 15, that 70% of women be screened with a high-performance test, and that 90% of women with cervical disease receive appropriate treatment [2].
Kenya introduced HPV vaccination into its routine immunisation programme on 18 October 2019, following a demonstration project conducted in Kitui County between 2013 and 2015. The programme initially offered a two-dose schedule to 10-year-old girls, with the intention of extending to girls aged 9 to 14 years. In November 2025 Kenya transitioned to a single-dose schedule in line with revised WHO recommendations, and on 15 January 2026 the Ministry of Health launched the country’s first National Cervical Cancer Elimination Action Plan 2026–2030, a costed five-year roadmap aligned to the 90-70-90 targets that prioritises free HPV vaccination for girls aged 10 to 14 years, expanded HPV DNA screening and guaranteed treatment [10,11]. The data reported here were collected while the two-dose schedule was still in force, and the two-dose completion figures should be read in that light.
The publicly available policy brief for the National Cervical Cancer Elimination Action Plan does not identify pastoralist communities as a separate target group. It nevertheless establishes equity as a core value and includes operational objectives to expand equitable, timely and integrated HPV vaccine delivery, strengthen vaccination workforce and data systems, promote multisectoral coordination, and intensify behaviour-change communication. These broad commitments provide a policy basis for adapting vaccination delivery to mobile, out-of-school and geographically isolated girls, although the brief does not specify pastoralist-focused implementation mechanisms [10].
Pastoralist communities face structural and socio-cultural barriers that directly affect vaccination uptake and continuity of care. Seasonal migration removes households from fixed health facilities, making school-based and facility-centred delivery difficult to sustain [12,13]. In Kenya, Pertet and colleagues [14] reported that 60% of children in nomadic pastoralist households had never received any vaccination, compared with 7.2% in settled populations. Adolescent girls in these settings are more likely to be out of school because of early marriage, domestic labour and recurrent mobility, which limits their inclusion in school-based campaigns [15,16]. Decision-making authority is frequently concentrated among male household heads, yet vaccination communication rarely engages men directly [17,18], and weak formal communication systems increase reliance on informal networks through which misinformation about safety, infertility and sexual morality circulates rapidly [19,20].
These conditions converge in Garbatulla Sub-County, Isiolo County, where routine data placed HPV vaccination coverage among girls aged 15–18 years at 14.13%, among the lowest in the country [21]. Existing Kenyan evidence on HPV vaccination has concentrated on urban, peri-urban and school-attending populations [20,22], and the small number of studies among mobile populations have addressed Maasai and coastal communities rather than the Borana. No study to date has combined caregiver-level quantitative measurement with health-system and community qualitative accounts in a Borana pastoralist setting, and none has examined the caregivers of the older, largely out-of-school cohort of girls aged 15–18 years who were age-eligible at national introduction but were never reached. That specific combination is the contribution of the present study. The objectives were to determine the level of HPV vaccination uptake among adolescent girls aged 15–18 years, to assess caregiver knowledge of HPV and HPV vaccination, to examine caregiver perceptions and attitudes, and to identify factors independently associated with uptake.
Study design and setting
This was a concurrent mixed-methods study comprising a quantitative analytical cross-sectional study and a qualitative exploratory study. The two strands were conducted in parallel, analysed independently, and integrated at the interpretation stage. The cross-sectional label applies to the quantitative strand only; the qualitative strand followed an exploratory design and is not described in cross-sectional terms. The study was conducted in Garbatulla Sub-County, Isiolo County, Kenya, a predominantly Borana pastoralist area characterised by nomadic and semi-nomadic livelihoods, weak health infrastructure, poor road networks, low literacy and low school enrolment among girls. The sub-county comprises three administrative wards: Garbatulla, Kinna and Sericho.
Design and analysis framework
The study was guided by the Health Belief Model (HBM), which explains preventive health behaviour through perceived susceptibility, perceived severity, perceived benefits, perceived barriers, cues to action and self-efficacy [6,23]. These constructs informed selection of questionnaire items and provided an organising framework for interpretation of the mixed-methods findings. Perceived susceptibility was represented by beliefs about personal or adolescent risk of HPV infection; perceived severity by understanding of the health consequences of HPV and its relationship with cervical cancer; perceived benefits by beliefs about the protective value of HPV vaccination; perceived barriers by concerns about safety, side effects, fertility, sexual morality, paternal refusal and access; and cues to action by information or encouragement from health workers, media, teachers, Community Health Promoters and community forums. Knowledge was treated as a contextual informational variable relevant to these beliefs rather than as a formal HBM construct.
Application of the Health Belief Model in analysis
The HBM was not tested as a separate statistical or latent-variable model. In the quantitative strand, HBM-aligned knowledge and perceptions and attitudes measures were analysed as observed variables, and their associations with HPV vaccination uptake were estimated alongside socio-demographic covariates. In the qualitative strand, HBM constructs were used as sensitising concepts when codes and candidate themes were reviewed, while explanations that emerged inductively from participants were retained even when they did not fit a single construct. At the mixed-methods integration stage, limited understanding of risk and cervical cancer was interpreted in relation to susceptibility and severity; beliefs about protection in relation to benefits; infertility, safety, moral concerns, paternal refusal and mobility in relation to barriers; and health workers, teachers, media, Community Health Promoters and local leaders in relation to cues to action. Self-efficacy was not scored as a separate quantitative construct; qualitative accounts of decision authority and the ability to act on vaccination advice were used to contextualise caregiver agency. Thus, the HBM organised measurement, qualitative interpretation and integration, but was not treated as a causal model requiring a separate model-fit test.
Operational definitions
HPV vaccination uptake, the primary outcome, was defined as caregiver report that the eligible adolescent girl had received at least one dose of HPV vaccine before the interview. Completion was defined as receipt of two doses, consistent with the national schedule in force when the data were collected. Caregiver knowledge was measured from 15 dichotomously scored items (correct=1, incorrect or not known=0), producing a score of 0-15; poor knowledge was defined as 0-7, fair knowledge as 8-12 and good knowledge as 13-15. Perceptions and attitudes were measured from 18 items scored so that favourable responses received 1 and unfavourable responses 0, with negatively worded items reverse-scored; poor perceptions and attitudes were defined as 0-8, moderate as 9-14 and good as 15-18. Intention to vaccinate was assessed only among caregivers whose daughters had received no HPV vaccine dose and was based on the caregiver’s reported intention. Socio-demographic variables were analysed using questionnaire categories; education was collapsed to no formal education, primary education, and secondary or higher for inferential analysis because the upper education categories contained sparse observations.
Study population and sample size
The study population comprised caregivers of adolescent girls aged 15–18 years resident in Garbatulla Sub-County who, by age, should have completed the two-dose HPV vaccination schedule. Ministry of Health estimates for 2024 place approximately 20,489 girls aged 15–18 years in Isiolo County, of whom approximately 7,601 reside in Garbatulla Sub-County. One caregiver was interviewed per eligible household to preserve independence of observations.
The sample size for the quantitative strand was calculated using the Cochran formula for a single proportion, n = Z²p(1−p)/d², with Z = 1.96, d = 0.05 and p = 0.141, the latter derived from Kenya Health Information System records for the sub-county (2,393 vaccinated of 16,937 eligible girls). This yielded 186, which was adjusted by 10% for anticipated non-response to give a target of 204 caregivers.
Sampling procedure, quantitative strand
A multi-stage sampling approach was used. In the first stage, the sub-county was stratified into its three administrative wards, and the target sample was allocated proportionally to ward population, giving Garbatulla 113 (55.4%), Kinna 28 (13.7%), and Sericho 63 (30.9%). In the second stage, we listed community health units within each ward and selected units with probability proportional to the number of registered households. In the third stage, households containing an eligible girl were identified from community health unit registers with the assistance of Community Health Promoters, and the required number of households was drawn from each sampling frame using a random number generator application. In the fourth stage, where a household contained more than one eligible girl, one girl and her principal caregiver were selected at random.
A total of 204 eligible caregivers were approached, all of whom consented and completed the interview, yielding a response rate of 100%. No eligible caregiver declined participation, and none could not be reached during the household visits.
Sampling procedure
Participants for the qualitative strand were selected purposively rather than at random, because the aim was to obtain information-rich accounts of community beliefs and of health system implementation rather than a statistically representative sample. Focus group participants were caregivers of unvaccinated daughters who had not taken part in the quantitative survey, so that the two strands drew on separate respondents. Key informants were selected on the basis of direct operational involvement in HPV vaccination delivery in the sub-county.
Two focus group discussions were conducted with a total of 16 caregivers of unvaccinated adolescent girls, recruited from the community through Community Health Promoters and community health units in Garbatulla Sub-County. The discussions were separated by sex, with one group comprising eight female caregivers and the other comprising eight male caregivers. The discussions were conducted in Borana, lasted approximately 60 minutes each, and were moderated by the principal investigator, a public health professional trained in qualitative interviewing, with assistance from a trained research assistant who took field notes and managed the audio recordings. Informational sufficiency was considered to have been achieved when participants’ accounts became repetitive and no substantively new beliefs, barriers or explanations emerged from the second discussion and subsequent key informant interviews.
Ten key informant interviews were conducted with personnel directly involved in HPV vaccination planning, delivery and community mobilisation. The key informants comprised two Sub-County Public Health Officers, two community health strategy coordinators, two facility in-charges, two nursing officers and two Community Health Promoters. They were purposively recruited from the Garbatulla Sub-County Health Management Team, selected health facilities and community health units across Garbatulla, Kinna and Sericho wards. The interviews were conducted in English or Kiswahili, according to the participant’s preference, in private locations within the participants’ respective workplaces to protect confidentiality. Each interview lasted approximately 40 to 60 minutes.
Data collection instruments and procedures
Quantitative data were collected using an interviewer-administered structured questionnaire developed from the study objectives and the Health Belief Model constructs. The instrument covered four domains: socio-demographic characteristics; HPV vaccination status of the daughter; caregiver knowledge of HPV, cervical cancer and HPV vaccination; and perceptions and attitudes toward HPV vaccination.
Knowledge was scored across 15 items, each scored 1 for a correct response and 0 otherwise, giving a range of 0 to 15. Items not applicable to a respondent because of a filter question, for example, the meaning of a Pap smear among those who had never heard of it, were scored 0 because they represent an absence of the knowledge in question rather than missing data. Perceptions and attitudes were scored across 18 items, each scored 1 for a favourable response and 0 otherwise, giving a range of 0 to 18, with negatively worded items reverse-scored so that a higher score always denotes a more favourable orientation.
Both composites were categorised for reporting and analysis using pre-specified cut-points. Knowledge was categorised as poor (0-7), fair (8-12) and good (13-15), while perceptions and attitudes were categorised as poor (0-8), moderate (9-14) and good (15-18). The categories were specified before analysis and were not derived from the observed distribution, consistent with categorical approaches used in comparable HPV knowledge, attitude and practice studies in sub-Saharan Africa [24-26]. To assess whether the findings were dependent on categorisation, associations between each composite and vaccination uptake were also examined using the continuous scores; the direction and statistical significance of the associations were unchanged. The questionnaire was pretested with 38 caregivers at Badana location in Sericho ward, an area with comparable demographic characteristics that was not included in the main study.
Following the pretest, items concerning the eligible age for HPV vaccination, the relationship between HPV and cervical cancer, the mode of HPV transmission, the need for cervical cancer screening after vaccination, and concerns about vaccine safety and infertility were reworded to improve clarity and reduce ambiguity. Response options for sources of HPV vaccination information, reasons for non-vaccination and reasons for unwillingness to vaccinate were expanded to include locally relevant responses such as paternal refusal, religious or cultural concerns, fear of infertility, doubts about vaccine safety, lack of transport and uncertainty about where the vaccine was available. No items were removed, as all the items were considered relevant to the study objectives and the Health Belief Model constructs.
The questionnaire and focus group discussion guide were administered in English, Kiswahili or Borana (Afaan Oromo), depending on each caregiver’s language preference, while the key informant interview guide was administered in English or Kiswahili. The original instruments were developed in English and forward-translated into Kiswahili and Borana by bilingual public health professionals familiar with the local cultural context. Independent bilingual translators, who had not reviewed the original English instruments, then back-translated the translated versions into English. The original and back-translated versions were compared by the research team, and differences in wording, meaning and cultural interpretation were discussed and reconciled before the final instruments were used in the main study.
Internal consistency of the composite scales was acceptable, with a Cronbach alpha of 0.832. Qualitative data were collected through audio-recorded focus group discussions and key informant interviews using semi-structured guides with open-ended questions and probes. Focus group guides explored community perceptions, beliefs and social norms around HPV vaccination; key informant guides explored health system perspectives, service delivery gaps, outreach challenges and frontline implementation realities. Recordings were transcribed verbatim and checked against the audio for accuracy.
Data management and analysis
Quantitative analysis
Data were entered into EpiData version 3.1 and analysed in SPSS version 26, with confirmatory analysis of the regression models conducted independently. Categorical variables are summarised as frequencies and percentages with Wilson 95% confidence intervals, which are preferred to the normal approximation where subgroup counts are small, or proportions approach 0 or 1. Normality of continuous variables was assessed with the Shapiro–Wilk test supported by visual inspection of histograms and normal quantile plots. Age, number of children, the knowledge score and the attitude score all departed significantly from normality (Shapiro–Wilk P<0.001 in each case) and are therefore reported as medians with interquartile ranges; means with standard deviations are given alongside for comparability with earlier literature, but inference is based on the non-parametric summaries. Group comparisons of continuous variables used the Kruskal–Wallis test.
Bivariate associations between candidate variables and vaccination uptake were tested with the Pearson chi-square test. Where any expected cell count fell below 5, the chi-square result was verified with a Monte-Carlo exact test based on 20,000 permutations, which extends the Fisher exact approach to tables larger than 2×2; both P values are reported. Education was collapsed to three categories (no formal education, primary, secondary or higher) before testing because the tertiary and university categories contained 9 and 1 observations, respectively.
Because vaccination uptake was common at 38.7%, the odds ratio would materially overstate the prevalence ratio, and binary logistic regression was therefore not used for the primary analysis. Adjusted prevalence ratios with robust 95% confidence intervals were estimated using modified Poisson regression, that is, a generalised linear model with a Poisson family, log link and Huber–White sandwich variance estimator. A log-binomial model was attempted first and failed to converge. All variables associated with uptake at P<0.25 in bivariate analysis were entered simultaneously into the model; this liberal screening threshold is used to avoid excluding variables that are non-significant in isolation but become important after adjustment, and it was applied consistently, so that age group (P=0.180), marital status (P=0.186) and ward (P=0.121) were all retained in the model alongside education, occupation, knowledge and attitude. Religion (P=0.881) and number of children (P=0.390) did not meet the threshold and were excluded. All categorical predictors were entered as indicator (dummy) variables with an explicit reference category, so that education contributes two coefficients, occupation five, marital status three, ward two, and each composite category variable two.
Multicollinearity was assessed using variance inflation factors computed on the model design matrix. Model fit was assessed by the omnibus likelihood ratio test and, for the logistic sensitivity model, by the Hosmer–Lemeshow test. Statistical significance was set at two-sided P<0.05, and P values are reported to three decimal places with a leading zero. No formal adjustment for multiple comparisons was applied, and the analysis of subgroups is therefore treated as exploratory.
Qualitative analysis
Transcripts were analysed using the six-phase reflexive thematic analysis framework described by Braun and Clarke [27]: familiarisation with the data, generation of initial codes, searching for themes, reviewing themes against coded extracts and the full data set, defining and naming themes, and producing the report. Focus group and key informant material was coded within a common framework so that community and health-system accounts could be compared.
Qualitative analysis was conducted using NVivo 14. Before coding, transcripts were checked against the audio recordings for accuracy. The principal investigator read each transcript repeatedly, assigned descriptive and interpretive codes to relevant text segments, and maintained a coding framework that was refined as analysis progressed. Related codes were grouped into candidate themes, compared across the focus group and key informant datasets, reviewed against the original transcripts, and revised until they represented coherent patterns relevant to the study objectives and the HBM. Coding was undertaken by the principal investigator; therefore, no inter-coder agreement statistic was calculated.
Validation of the qualitative findings used complementary trustworthiness procedures. Credibility was strengthened through data-source triangulation by comparing accounts from the female and male caregiver focus groups with one another and with accounts from health-system key informants, and through mixed-methods triangulation by comparing the resulting themes with the quantitative distributions and associations at the interpretation stage. Dependability was supported by consistent application of the six analytic phases and maintenance of a coding framework linking raw extracts, codes, candidate themes, final themes and illustrative quotations. Confirmability was strengthened by reviewing candidate themes against the original transcripts and grounding each reported theme in verbatim extracts, rather than relying on the investigator’s interpretation alone. Transferability was supported through detailed description of the study setting, participant groups and service-delivery context. Because the principal investigator conducted the interviews, transcribed the recordings and led the analysis, this dual role was explicitly treated as a potential source of interpretive influence and is acknowledged in the limitations; credibility therefore rests on transparent analytic procedures and triangulation rather than inter-coder agreement.
Ethical considerations
Ethical approval was obtained from the Meru University of Science and Technology Institutional Ethics Review Committee, and a research licence was issued by the National Commission for Science, Technology and Innovation. Administrative approvals were secured from the Isiolo County Government and the Garbatulla Sub-County administration.
This was an observational study and involved no experimental procedures. It was conducted in accordance with the ethical principles for research involving human participants set out in the Declaration of Helsinki as revised in 2013, and with the Kenyan National Guidelines for Ethical Conduct of Biomedical Research Involving Human Participants.
Given the low level of formal education in the study population, particular attention was paid to the consent process. Information sheets and consent forms were prepared in English and translated into the local language, and the study purpose, procedures, risks, benefits, voluntary nature of participation and right to withdraw at any point without consequence were explained verbally in the participant’s preferred language by a fieldworker fluent in that language. Participants were given the opportunity to ask questions before deciding. Literate participants gave written consent by signature. Participants who could not read or write gave consent by thumbprint after the information had been read aloud in full, in the presence of an impartial literate witness who was not a member of the research team and who countersigned the form to attest that the information had been accurately conveyed and that consent was freely given.
Participants received no financial inducement. Audio recordings and completed questionnaires were stored on password-protected devices accessible only to the research team, identifiers were removed at the point of data entry, and no individual is identifiable in any output. Quotations are attributed by respondent number only.
Characteristics of participants
A total of 204 caregivers were interviewed, distributed as Garbatulla 113, Kinna 28 and Sericho 63 in accordance with the proportional allocation. Almost all participants were Muslim (198, 97.1%), and three-quarters were currently married (155, 76.0%). Median caregiver age was 38 years (interquartile range 35-44), and the median number of children was 5 (IQR: 4 – 6); neither variable differed significantly across wards (Kruskal–Wallis P=0.257 and P=0.120, respectively). Educational attainment was low, with 96 caregivers (47.1%) having no formal education and a further 73 (35.8%) having completed only primary education; just 35 (17.2%) had secondary or higher education. The most common livelihoods were housekeeping (68, 33.3%) and livestock-related activity (49, 24.0%), and only 11 caregivers (5.4%) held salaried formal employment (Table 1).
HPV vaccination uptake
Seventy-nine of the 204 adolescent girls had received at least one dose of HPV vaccine, giving an uptake of 38.7% (95% CI: 32.3 – 45.6). Completion of the two-dose schedule was substantially lower at 11.3% (95% CI: 7.6 – 16.3), and 56 girls (27.5%, 95% CI: 21.8 – 33.9) had received one dose only. Among the 79 girls who initiated vaccination, only 29.1% (95% CI: 20.3 – 39.9) went on to complete the schedule. One hundred and twenty-five girls (61.3%, 95% CI: 54.4 – 67.7) had received no dose at the time of data collection (Table 2).
Uptake ranged from 21.4% in Kinna to 42.5% in Garbatulla, but the confidence intervals overlapped substantially, and the difference across wards was not statistically significant (χ²=4.225, df=2, P=0.121; Monte-Carlo exact P=0.119). The ward-level estimates for Kinna in particular are imprecise because only 28 caregivers were sampled there, and the apparent gradient should not be over-interpreted.
Among the 79 girls who had been vaccinated, 39 (49.4%) received the vaccine at school, 33 (41.8%) at a health facility and 7 (8.9%) through community health outreach. The reason most frequently given for vaccination was a mass vaccination drive at school (63, 79.7%), followed by a recommendation from a health care provider (38, 48.1%), a wish to protect the daughter from HPV (26, 32.9%) and encouragement from a friend (18, 22.8%).
Intention to vaccinate and reasons for non-vaccination
Among the 125 caregivers whose daughters were unvaccinated, 77 (61.6%) reported no intention to have the daughter vaccinated, 33 (26.4%) were not sure, 8 (6.4%) said they might do so, and 5 (4.0%) intended to vaccinate; 2 caregivers (1.6%) did not answer the item. Reasons offered for the daughter not having been vaccinated, which allowed multiple responses and were answered by 116 caregivers, were dominated by safety concerns about the vaccine (85, 73.3%) and by simply not knowing about the vaccine (65, 56.0%); paternal refusal was reported by 28 caregivers (24.1%) and religious or cultural objection by 27 (23.3%), while logistical reasons were comparatively rare, with 11 (9.5%) citing lack of transport and 8 (6.9%) not knowing where to obtain the vaccine.
Among the 123 caregivers who gave reasons for not intending to vaccinate, the belief that the vaccine could cause infertility was the single most frequently reported reason (77, 62.6%, 95% CI: 53.8 – 70.6), closely followed by the belief that the vaccine may not be safe (73, 59.3%, 95% CI: 50.5 – 67.6). Cost was reported by only one caregiver (Table 3).
Caregiver knowledge of HPV and HPV vaccination
Item-level responses show a consistent pattern of high surface familiarity coupled with low functional understanding (Table 4). Most caregivers had heard of cervical cancer (170, 83.3%), of HPV (161, 78.9%) and of the HPV vaccine (151, 74.0%), and 143 (70.1%) were aware that the vaccine is offered to young girls. Only 72 caregivers (35.3%) correctly identified sexual intercourse as the mode of transmission, 77 (37.7%) correctly identified what HPV infection causes, and 45 (22.1%) correctly identified who is at risk of infection. Knowledge of the vaccination programme itself was weaker still: only 39 caregivers (19.1%) could correctly state the eligible age range, and only 29 (14.2%) knew that cervical cancer screening remains necessary after vaccination. Familiarity with Pap smear was low, at 40 caregivers (19.6%).
The composite knowledge score had a median of 6 out of 15 (IQR: 3 – 9; mean 6.15, SD 3.75). Applying the pre-specified cut-points, 133 caregivers (65.2%, 95% CI 58.4 to 71.4) had poor knowledge, 56 (27.5%, 95% CI: 21.8 – 33.9) had fair knowledge, and only 15 (7.4%, 95% CI: 4.5 – 11.8) had good knowledge. Scores did not differ significantly across wards (Kruskal–Wallis H=2.112, P=0.348), indicating that the deficit is uniform across the sub-county rather than concentrated in the more remote wards (Table 5).
Caregiver perceptions and attitudes
Favourable responses were infrequent across the 18 attitude items and fell below one-third on 14 of them (Table 6). The two items attracting the most favourable responses were rejection of the claim that vaccination encourages sexual promiscuity (92, 45.1%) and agreement that vaccination is cheaper than treating disease (90, 44.1%); even on these, a majority of caregivers responded unfavourably. Perceived susceptibility was particularly weak: only 53 caregivers (26.0%) believed a child could be at risk of HPV infection and only 36 (17.6%) considered themselves likely to contract HPV. Willingness to recommend vaccination was limited, with 64 caregivers (31.4%) willing to recommend it for their own child or a relative and 63 (30.9%) willing to recommend it for girls generally. The least frequently favourable response was to the statement that it is important for males to receive the HPV vaccine, agreed by 9 caregivers (4.4%).
The composite attitude score had a median of 3 out of 18 (IQR: 2 – 7; mean 4.51, SD 4.09). Applying the pre-specified cut-points, 169 caregivers (82.8%, 95% CI: 77.1 – 87.4) had poor attitudes, 26 (12.7%, 95% CI: 8.8 – 18.0) moderate attitudes and 9 (4.4%, 95% CI: 2.3 – 8.2) good attitudes. The distribution is strongly right-skewed with a marked floor effect: 23 caregivers (11.3%) scored zero and 111 (54.4%) scored 3 or below. Unlike knowledge, attitude scores differed significantly across wards (Kruskal–Wallis H=14.405, df=2, P=0.001), with a higher median in Garbatulla (4, IQR: 2 – 8) than in Kinna (2, IQR: 1 – 6) or Sericho (2, IQR: 1 – 6). Knowledge and attitude scores were strongly correlated (Spearman ρ=0.76, P<0.001).
Bivariate associations with vaccination uptake
Education, occupation, knowledge category and attitude category were significantly associated with vaccination uptake at the bivariate level (Table 7). Uptake rose from 32.3% among caregivers with no formal education to 62.9% among those with secondary or higher education (χ²=10.562, df=2, P=0.005). Uptake varied more than fivefold across occupational groups, from 14.3% among caregivers dependent on livestock activity to 100% among the 11 caregivers in formal employment (χ²=41.501, df=5, P<0.001). The two composite measures showed the strongest gradients: uptake was 23.3% among caregivers with poor knowledge, 58.9% among those with fair knowledge and 100% among the 15 with good knowledge (χ²=46.689, df=2, P<0.001), and 29.6%, 76.9% and 100% across the poor, moderate and good attitude categories respectively (χ²=36.177, df=2, P<0.001).
Three tables contained cells with expected counts below 5, namely religion, occupation and attitude category, and the chi-square results for these were confirmed by Monte-Carlo exact tests, which gave P=0.685, P<0.001 and P<0.001 respectively. The complete absence of unvaccinated girls in the formal employment, good knowledge and good attitude categories is a genuine feature of the data rather than a coding artefact, but it rests on 11, 15 and 9 observations respectively, and the corresponding confidence intervals are correspondingly wide.
Age group (P=0.180), marital status (P=0.186) and ward (P=0.121) did not reach conventional significance but met the P<0.25 screening threshold and were carried forward. Religion (P=0.881) and number of children (P=0.390) did not and were excluded.
Factors independently associated with vaccination uptake
A binary logistic regression model with the same specification was fitted first and proved unusable: the complete absence of unvaccinated girls in the formal employment, good knowledge and good attitude categories produced complete separation, with three coefficients exceeding 20 in absolute value and standard errors above 10³, and correspondingly meaningless odds ratios with infinite confidence limits. Because the outcome is also common at 38.7%, the odds ratio would in any case overstate the prevalence ratio. The primary analysis therefore uses modified Poisson regression with robust variance, which estimates prevalence ratios directly, is not affected by separation in the same way, and gives interpretable estimates for every category (Table 8).
After adjustment, caregiver knowledge remained the strongest independent correlate of uptake. Relative to caregivers with poor knowledge, those with fair knowledge were 1.74 times as likely to have a vaccinated daughter (95% CI: 1.12 – 2.70, P=0.013) and those with good knowledge 2.32 times as likely (95% CI: 1.40 – 3.84, P=0.001). Good attitudes were also independently associated with uptake (aPR: 1.72, 95% CI: 1.16 – 2.56, P=0.007), while the estimate for moderate attitudes was of similar magnitude but did not reach significance (aPR: 1.44, 95% CI: 0.99 – 2.09, P=0.059).
Contrary to the interpretation offered in the earlier analysis, occupation did not lose its association after adjustment. Caregivers dependent on livestock activity remained substantially less likely to have vaccinated daughters than housewives (aPR: 0.41, 95% CI: 0.20 – 0.81, P=0.011), and petty traders remained more likely (aPR: 1.81, 95% CI: 1.12 – 2.92, P=0.015). The crude association for formal employment was strong (cPR: 2.43, 95% CI: 1.83 – 3.23) but attenuated to non-significance after adjustment (aPR: 1.22, 95% CI: 0.65 – 2.29), consistent with that association operating largely through knowledge and attitudes; the estimate rests on 11 caregivers and is imprecise.
Education showed no consistent independent association: the adjusted estimate for secondary or higher education was close to unity (aPR: 1.10, 95% CI: 0.58 – 2.09) despite a strong crude association (cPR: 1.95, 95% CI: 1.32 – 2.86), and the estimate for primary education was below unity and borderline (aPR: 0.67, 95% CI 0.44 – 1.01, P=0.053). The most plausible reading is that the crude educational gradient operates through knowledge and attitudes rather than independently of them; the borderline inverse estimate for primary education is not readily interpretable and, in the absence of a plausible mechanism, is best regarded as unstable. Widowhood was associated with higher uptake (aPR: 1.89, 95% CI: 1.04 – 3.41, P=0.036), an estimate based on 11 caregivers that should be treated as hypothesis-generating. Neither age group nor ward was independently associated with uptake.
Variance inflation factors for all model terms ranged from 1.09 to 2.08, well below the conventional threshold of 5, indicating that multicollinearity among education, occupation, knowledge and attitude did not materially affect the estimates despite the strong bivariate correlation between the knowledge and attitude scores. The omnibus likelihood ratio test was highly significant (χ²=97.86, df=16, P<0.001), and the Hosmer–Lemeshow test on the logistic sensitivity model indicated adequate calibration (χ²=4.76, df=8, P=0.782).
Qualitative findings
Five themes were developed from the focus group discussions and key informant interviews. They are presented here as findings; their relationship to the quantitative results is considered in the Discussion.
Theme 1: Health system communication gaps and frontline capacity. Key informants located the knowledge deficit in the health system rather than solely in the community, pointing to staff turnover and the absence of refresher training.
“Awareness within caregivers is not 100 per cent because most staff are newly employed” (KII 8)
“Lack of proper training among new health workers and lack of continuous health education among CHPs” (KII 8)
Uncertainty about eligibility was described as a specific and consequential gap.
“Some caregivers have knowledge deficit… others are not aware of what age it should be given” (KII 2)
Focus group participants described the resulting confusion in similar terms.
“Most of the participants didn’t know about the HPV vaccine… some said they had heard it on radio or television, but they don’t understand how HPV and cervical cancer are related” (FGD 2)
Theme 2: Fertility anxiety as the organising fear. Concern that the vaccine would compromise a girl’s future fertility was described not as an isolated rumour but as an established community understanding.
“mothers and the general community perceive that HPV vaccine could cause infertility for young girls in future” (KII 7)
“a way of family planning that prevents them from getting pregnant during their entire life” (FGD 1)
“they fear the infertility… and believe that the vaccine is not safe” (FGD 2)
Theme 3: Moral framing of adolescent sexuality. Because HPV is sexually transmitted, vaccination was widely read as an implicit endorsement of premarital sexual activity, and in some accounts as an instrument of external population control.
“some community members suspected it could be a government ploy aimed at population control” (KII 4)
“they also believe that it is associated with immoral behaviour” (KII 4)
“We don’t want our daughters to think that being vaccinated means they can do anything” (FGD 1)
Theme 4: Paternal authority over girls’ health decisions. Fathers were consistently described as holding final authority over decisions affecting daughters, including vaccination, while programme communication was directed almost exclusively at mothers.
“the husband is the one who decides… if he says no, that is the end of the matter” (FGD 1)
“Consent of the parent, especially fathers, is often a significant hurdle” (FGD 2)
Theme 5: Weak initial mobilisation and the limits of school-based delivery. Informants traced current mistrust to inadequate community preparation at the point of programme introduction, and identified trusted local intermediaries as the channel through which acceptance was subsequently achieved.
“When the program began, there was poor mobilisation and advocacy; thus the community awareness was low, leading to poor uptake” (KII 2)
“Poor attitude by caregivers, inadequate support from County and National government for community awareness” (KII 1)
“Community members tend to trust and comply more when messages come from Community Health Promoters and local leaders” (KII 3)
This study found that fewer than two in five adolescent girls aged 15–18 years in Garbatulla Sub-County had received any dose of HPV vaccine and that barely one in nine had completed the two-dose schedule in force at the time of data collection. Both figures fall far below the 90% coverage target set by the global elimination strategy and adopted in Kenya’s National Cervical Cancer Elimination Action Plan 2026–2030. The uptake observed here is nevertheless higher than the 14.13% recorded for this age group in routine sub-county data, which may reflect a household survey capturing doses administered outside the sub-county or through campaigns that were incompletely recorded, or may reflect caregiver recall error in the absence of card verification. That discrepancy is itself a finding of programmatic relevance and argues for strengthening routine documentation.
The most consistent quantitative finding was the independent association between caregiver knowledge and uptake, which survived adjustment for education, occupation, marital status, age and ward. Item-level analysis clarifies what this means in practice. Awareness in the conventional sense was not the binding constraint: four in five caregivers had heard of HPV and of cervical cancer, and three in four had heard of the vaccine. What was missing was the operative content, namely who is at risk, how transmission occurs, which ages are eligible, and whether screening remains necessary afterwards. Fewer than one in five caregivers could state the eligible age range. This pattern, in which terminology is familiar but understanding is too thin to support a decision, corresponds to what Oketch and colleagues [28] describe as superficial knowledge, and has been documented in Ethiopia [24,25], Uganda [8,29] and across sub-Saharan Africa [26,30]. It has a direct programmatic implication: awareness-raising campaigns that stop at naming the vaccine will not shift uptake, because the population has already crossed that threshold.
The qualitative strand explains why the knowledge deficit persists despite a functioning immunisation programme. Key informants attributed it to staff turnover, absent refresher training and discontinued community health education, rather than to community indifference. This matters because it identifies a supply-side determinant of a demand-side barrier. Where frontline workers cannot answer questions confidently, the informational vacuum is filled by community narrative, and the caregivers least connected to services are the least likely to encounter authoritative information at all. Gammino and colleagues [13] identify precisely this self-reinforcing exclusion as a structural feature of pastoralist health systems.
Attitudes were independently associated with uptake and were, on the evidence of the item-level data, extremely unfavourable, with a pronounced floor effect. More than half of caregivers scored 3 or below out of 18 and one in nine scored zero. Two readings of a floor effect this severe are possible and cannot be distinguished with these data. The first is that unfavourable orientation is genuinely near-universal in this community, which the qualitative accounts support. The second is that the instrument, which scored filtered items as unfavourable for caregivers who had never heard of the vaccine, conflates absence of belief with negative belief; this measurement decision is defensible for a scale intended to capture readiness to vaccinate, but it compresses the lower end of the distribution and will exaggerate the apparent depth of negative attitude among the unaware. The strong correlation between the knowledge and attitude scores (ρ=0.76) is consistent with that conflation, although variance inflation factors confirm that it did not destabilise the regression estimates. Future instruments should separate the two more cleanly.
The qualitative themes give substance to the attitude scores and identify the specific beliefs at issue. Fertility anxiety was the single most frequently reported reason for not intending to vaccinate, and the focus group accounts show it operating not as a discrete rumour but as a coherent interpretation of the vaccine as a covert contraceptive. In a Borana pastoralist context, where fertility bears directly on marriage prospects, family continuity and a woman’s social standing, refusing vaccination on these grounds is framed as parental protection rather than neglect, which is why corrective information alone rarely dislodges it. Comparable narratives have been documented across sub-Saharan Africa [4,31]. The moral framing identified in Theme 3 mirrors findings from Kisumu and from Kenyan parental studies more broadly [18,20], and supports the recommendation of Wong and colleagues [32] that messaging be anchored in cancer prevention and family wellbeing rather than in sexual transmission. The paternal veto described in Theme 4, reported as a direct reason for non-vaccination by one in four caregivers of unvaccinated girls, aligns with the analysis of reproductive health decision-making among north-eastern Kenyan pastoralists by Kenny and colleagues [17], and points to a specific and correctable design flaw, namely that communication is delivered almost exclusively to mothers while decision authority rests with fathers.
Viewed through the HBM, the integrated findings show how individual beliefs and contextual constraints operate together. Limited understanding of who is at risk and of the link between HPV and cervical cancer weakens perceived susceptibility and severity, while recognition of cancer prevention represents the principal perceived benefit. Fertility anxiety, safety concerns, moral objections, paternal refusal and mobility function as perceived or practical barriers, whereas Community Health Promoters, health workers and local leaders provide credible cues to action. The qualitative accounts of paternal decision authority further show that a caregiver may understand and value vaccination yet have limited ability to act, which contextualises self-efficacy in this setting. The HBM therefore helps organise the pattern linking knowledge and attitudes with uptake, while the mobility findings show that belief-focused interventions must be combined with delivery strategies that address structural access barriers.
Two findings depart from the pattern reported in the earlier analysis of these data and warrant emphasis. First, occupation retained an independent association with uptake after adjustment. Caregivers dependent on livestock activity were less than half as likely as housewives to have vaccinated daughters, and petty traders were substantially more likely. This is unlikely to be a matter of income, since cost was reported by a single caregiver in the entire sample; the more plausible mechanism is physical presence. Livestock-dependent households move with the herds and are structurally least available at the moments when school-based and facility-based vaccination occurs, whereas petty traders are anchored to market centres and to the social networks that form around them. Read alongside the finding that four in five vaccinated girls were reached through a school campaign, this indicates that the delivery model itself, rather than caregiver disposition alone, is generating part of the coverage gap. Second, education did not show a coherent independent association once knowledge and attitude were in the model, and the crude educational gradient is best understood as operating through them. This is consistent with evidence from Uganda and Ethiopia that socioeconomic disadvantage affects vaccination behaviour indirectly through information access and belief formation [4,19], and it is encouraging for programme design, because communication deficits are more tractable in the short term than educational attainment.
The gap between initiation and completion deserves separate comment. Fewer than a third of girls who received a first dose went on to receive a second, a shortfall consistent with findings from northern Uganda [33] and from low-resource settings generally [34]. For a population whose mobility makes scheduled follow-up structurally difficult, this was the strongest argument for a single-dose schedule, supported by the demonstration of non-inferior protection from a single dose in a randomised trial conducted in Kenya [35] and by modelling of the resulting reduction in the completion gap [36,37]. Kenya adopted a single-dose schedule in November 2025, after these data were collected. The present findings should therefore be read not as an argument for that change, which has already been made, but as an indication of how much of the coverage deficit it can be expected to close and how much it cannot. The completion gap documented here, some 27.5% of all girls having received one dose only, should convert directly into full coverage under the new schedule. The 61.3% who received no dose at all will be unaffected by it, because their non-uptake is driven by the knowledge, belief and access barriers described above, none of which a change in schedule addresses.
Strengths and limitations
The main strengths of this study are its focus on a population that is consistently absent from the Kenyan HPV literature, its concurrent mixed-methods design, which allowed community and health system accounts to be set against measured uptake, its probability-based multi-stage sampling within the sub-county, and the use of an analytical approach appropriate to a common binary outcome, which yields prevalence ratios that can be interpreted directly rather than odds ratios that would overstate the associations.
Several limitations qualify the findings. The design is cross-sectional, so the associations reported here cannot establish that knowledge or attitudes cause vaccination uptake; reverse causation is entirely plausible, since caregivers whose daughters were vaccinated will have encountered information and health workers in the course of that experience. Throughout this paper the estimates are therefore described as factors associated with uptake rather than as determinants or predictors.
Vaccination status was based on caregiver report and was not verified against vaccination cards or the electronic immunisation register. This introduces recall and social desirability bias of unknown direction, and it is the most likely explanation for the divergence between the uptake measured here and routine coverage data for the same age group. Future studies in this setting should include card or register verification.
Several subgroup estimates rest on very small numbers. The complete absence of unvaccinated girls among caregivers in formal employment, those with good knowledge and those with good attitudes is based on 11, 15 and 9 observations, respectively, and while the confidence intervals reported throughout make that imprecision explicit, these estimates should not be relied upon for programme targeting. The same applies to the ward-level estimates for Kinna, where only 28 caregivers were sampled, and to the association observed for widowhood. No adjustment was made for multiple comparisons, so subgroup findings are exploratory.
The attitude instrument scored items that were filtered out for caregivers unaware of the vaccine as unfavourable, which conflates absence of belief with negative belief and probably exaggerates the depth of the attitude deficit, as discussed above. The knowledge and attitude cut-points, although pre-specified and conventional, remain arbitrary in the sense that any categorisation of a continuous score is arbitrary; results were unchanged when the scores were analysed continuously.
The study was conducted in a single sub-county of a single county. Findings should generalise reasonably to Borana pastoralist communities with similar service configurations, but should not be extended to Kenyan pastoralist populations generally, whose social organisation, religious composition and health-system exposure differ considerably. Finally, the qualitative strand comprised two focus group discussions and ten key informant interviews, all conducted, transcribed and coded by the principal investigator. Comparison of caregiver and health-system accounts and integration with the quantitative findings strengthened contextual interpretation, but the limited number of discussions and the absence of independent coding may have constrained thematic breadth and increased the possibility of interpretive bias. The qualitative findings should therefore be treated as context-specific explanations rather than exhaustive representations of all views in the community.
HPV vaccination uptake among adolescent girls aged 15–18 years in Garbatulla Sub-County is far below national and global targets, at 38.7% for a first dose and 11.3% for completion of the two-dose schedule then in force. Caregiver knowledge and attitudes were the factors most strongly and independently associated with uptake, with livelihood making a further independent contribution: caregivers dependent on livestock activity were less than half as likely to have vaccinated daughters as those based at home. Because the design is cross-sectional, these are associations and not causes.
The pattern of results points to three programmatic priorities. Communication should move beyond awareness-raising, which has already achieved its purpose in this community, to the specific operative content that caregivers lack, namely eligibility, transmission, the continuing need for screening, and direct engagement with the fertility and moral concerns that the qualitative data identify as the organising barriers; this is likely to be effective only when delivered through Community Health Promoters and local leaders, whom informants identified as the trusted channel. Male household heads should be addressed directly, since they hold the decision and are currently outside the communication strategy. Delivery should be adapted to mobility, because the reliance on school-based campaigns systematically excludes the out-of-school and migrating girls who are at greatest risk. Kenya’s transition to a single-dose schedule in November 2025 should close much of the completion gap documented here, but it will not by itself reach the three in five girls who received no dose at all.
What is already known about the topic
What this study adds
Conceptualization: Sabla Galgalo Kubi
Data curation: Sabla Galgalo Kubi
Formal analysis: Sabla Galgalo Kubi
Methodology: Sabla Galgalo Kubi
Resources: Sabla Galgalo Kubi
Supervision: Mary Joy Kaimuri, Jane Jemeli Ruto
Writing – original draft: Sabla Galgalo Kubi
Writing – editing & review: Sabla Galgalo Kubi, Mary Joy Kaimuri, Jane Jemeli Ruto
| Characteristic | Category | Overall n (%) N=204 | Garbatulla n=113 | Kinna n=28 | Sericho n=63 |
|---|---|---|---|---|---|
| Marital status | Married | 155 (76.0) | 89 (78.8) | 20 (71.4) | 46 (73.0) |
| Widowed | 11 (5.4) | 8 (7.1) | 1 (3.6) | 2 (3.2) | |
| Divorced or separated | 19 (9.3) | 7 (6.2) | 5 (17.9) | 7 (11.1) | |
| Single parent | 19 (9.3) | 9 (8.0) | 2 (7.1) | 8 (12.7) | |
| Religion | Muslim | 198 (97.1) | 109 (96.5) | 28 (100.0) | 61 (96.8) |
| Christian | 6 (2.9) | 4 (3.5) | 0 (0.0) | 2 (3.2) | |
| Education | No formal education | 96 (47.1) | 50 (44.2) | 13 (46.4) | 33 (52.4) |
| Primary | 73 (35.8) | 40 (35.4) | 12 (42.9) | 21 (33.3) | |
| Secondary | 25 (12.3) | 16 (14.2) | 3 (10.7) | 6 (9.5) | |
| Tertiary | 9 (4.4) | 6 (5.3) | 0 (0.0) | 3 (4.8) | |
| University | 1 (0.5) | 1 (0.9) | 0 (0.0) | 0 (0.0) | |
| Occupation | Housewife | 68 (33.3) | 35 (31.0) | 13 (46.4) | 20 (31.7) |
| Livestock activity | 49 (24.0) | 27 (23.9) | 5 (17.9) | 17 (27.0) | |
| Business | 30 (14.7) | 21 (18.6) | 3 (10.7) | 6 (9.5) | |
| Casual labour | 26 (12.7) | 16 (14.2) | 3 (10.7) | 7 (11.1) | |
| Petty trade | 20 (9.8) | 8 (7.1) | 4 (14.3) | 8 (12.7) | |
| Formal employment | 11 (5.4) | 6 (5.3) | 0 (0.0) | 5 (7.9) | |
| Age, years | Median (IQR) | 38 (35–44) | 39 (35–44) | 40 (37–43) | 38 (34–43) |
| Number of children | Median (IQR) | 5 (4–6) | 5 (4–6) | 6 (5–6) | 4 (3–6) |
| IQR, interquartile range. Percentages are column percentages within ward. Age and number of children departed significantly from normality (Shapiro–Wilk P<0.001) and are therefore presented as medians with interquartile ranges; the corresponding means were 39.8 (SD 7.1) years and 4.96 (SD 1.83) children. Kruskal–Wallis tests across wards: age P=0.257; number of children P=0.120. | |||||
| Indicator | Overall (N=204) | Garbatulla (n=113) | Kinna (n=28) | Sericho (n=63) |
|---|---|---|---|---|
| At least one dose, n | 79 | 48 | 6 | 25 |
| 38.7 (32.3–45.6) | 42.5 (33.8–51.7) | 21.4 (10.2–39.5) | 39.7 (28.5–52.0) | |
| Two doses completed, n | 23 | 17 | 2 | 4 |
| 11.3 (7.6–16.3) | 15.0 (9.6–22.8) | 7.1 (2.0–22.6) | 6.3 (2.5–15.2) | |
| One dose only, n (%) | 56 (27.5) | 31 (27.4) | 4 (14.3) | 21 (33.3) |
| Completion among initiators, % (95% CI) | 29.1 (20.3–39.9) | 35.4 (23.4–49.6) | 33.3 (9.7–70.0) | 16.0 (6.4–34.7) |
| No dose received, n (%) | 125 (61.3) | 65 (57.5) | 22 (78.6) | 38 (60.3) |
| CI, confidence interval (Wilson method). Difference in uptake of at least one dose across wards: χ²=4.225, df=2, P=0.121; Monte-Carlo exact P=0.119. Ward-level estimates, particularly for Kinna, are imprecise because of small subgroup sizes and should be interpreted with caution. | ||||
| Reason | n (%) | 95% CI | Garbatulla n/N (%) | Sericho n/N (%) |
|---|---|---|---|---|
| Could cause infertility | 77 (62.6) | 53.8–70.6 | 40/66 (60.6) | 25/36 (69.4) |
| Vaccine may not be safe | 73 (59.3) | 50.5–67.6 | 36/66 (54.5) | 20/36 (55.6) |
| Do not consider it necessary | 61 (49.6) | 40.9–58.3 | 29/66 (43.9) | 17/36 (47.2) |
| Doubt vaccine effectiveness | 55 (44.7) | 36.2–53.5 | 35/66 (53.0) | 13/36 (36.1) |
| Religious or cultural beliefs | 49 (39.8) | 31.6–48.7 | 29/66 (43.9) | 16/36 (44.4) |
| Fear daughter would be stigmatised | 32 (26.0) | 19.1–34.4 | 15/66 (22.7) | 12/36 (33.3) |
| Discouraged by a friend | 28 (22.8) | 16.2–30.9 | 20/66 (30.3) | 6/36 (16.7) |
| Injection is painful | 22 (17.9) | 12.1–25.6 | 16/66 (24.2) | 4/36 (11.1) |
| Vaccine costs too much | 1 (0.8) | 0.1–4.5 | 0/66 (0.0) | 0/36 (0.0) |
| CI, confidence interval (Wilson method). Percentages exceed 100% in total because multiple responses were permitted. Kinna is omitted from the ward columns because only 21 caregivers there answered this item and the resulting estimates are too imprecise to be informative; full ward data are available on request. | ||||
| Knowledge item | Correct n | % (95% CI) | Domain |
|---|---|---|---|
| Has heard of cervical cancer | 170 | 83.3 (77.6–87.8) | Awareness |
| Has heard of HPV | 161 | 78.9 (72.8–84.0) | Awareness |
| Has heard of HPV vaccine | 151 | 74.0 (67.6–79.6) | Awareness |
| Aware that young girls are offered HPV vaccine | 143 | 70.1 (63.5–76.0) | Programme |
| Knows cervical cancer is preventable | 87 | 42.6 (36.1–49.5) | Cervical cancer |
| Correctly identifies what HPV infection causes | 77 | 37.7 (31.4–44.6) | HPV |
| Identifies vaccination as a preventive measure for HPV | 75 | 36.8 (30.4–43.6) | Prevention |
| Identifies sexual intercourse as mode of transmission | 72 | 35.3 (29.1–42.1) | HPV |
| Identifies vaccination as a way to prevent cervical cancer | 72 | 35.3 (29.1–42.1) | Prevention |
| Knows that not everyone infected develops symptoms | 62 | 30.4 (24.5–37.0) | HPV |
| Correctly identifies who may become infected with HPV | 45 | 22.1 (16.9–28.2) | Susceptibility |
| Has heard of Pap smear | 40 | 19.6 (14.7–25.6) | Screening |
| Correctly states ages eligible for HPV vaccination | 39 | 19.1 (14.3–25.1) | Programme |
| Correctly states the purpose of a Pap smear | 32 | 15.7 (11.3–21.3) | Screening |
| Knows screening is still needed after vaccination | 29 | 14.2 (10.1–19.7) | Screening |
| CI, confidence interval (Wilson method). Each item was scored 1 for a correct response and 0 otherwise; items not applicable because of a filter question were scored 0. The 15 items sum to the composite knowledge score reported in Table 5. | |||
| Group | n | Median (IQR) | Mean (SD) | Poor n (%) | Fair n (%) | Good n (%) |
|---|---|---|---|---|---|---|
| Overall | 204 | 6 (3–9) | 6.15 (3.75) | 133 (65.2) | 56 (27.5) | 15 (7.4) |
| Garbatulla | 113 | 6 (4–9) | 6.52 (3.82) | 70 (61.9) | 31 (27.4) | 12 (10.6) |
| Kinna | 28 | 6 (4–8) | 5.82 (3.56) | 19 (67.9) | 8 (28.6) | 1 (3.6) |
| Sericho | 63 | 5 (3–8) | 5.63 (3.70) | 44 (69.8) | 17 (27.0) | 2 (3.2) |
| IQR, interquartile range; SD, standard deviation. Knowledge categories: poor 0–7, fair 8–12, good 13–15. Kruskal–Wallis test of the composite score across wards: H=2.112, df=2, P=0.348. Overall 95% confidence intervals for the categories were poor 58.4–71.4, fair 21.8–33.9 and good 4.5–11.8. | ||||||
| Item (favourable response) | Favourable n | % (95% CI) | HBM construct |
|---|---|---|---|
| Rejects that vaccination encourages promiscuity | 92 | 45.1 (38.4–52.0) | Barrier |
| Agrees vaccination is cheaper than treatment | 90 | 44.1 (37.5–51.0) | Benefit |
| Agrees HPV could cause a serious health problem | 67 | 32.8 (26.8–39.6) | Severity |
| Would recommend vaccination for own child or relative | 64 | 31.4 (25.4–38.0) | Intention |
| Agrees prompt STI treatment reduces HPV infection | 64 | 31.4 (25.4–38.0) | Benefit |
| Would recommend that girls be vaccinated | 63 | 30.9 (24.9–37.5) | Intention |
| Rejects that the vaccine causes harmful side effects | 61 | 29.9 (24.0–36.5) | Barrier |
| Agrees vaccine reduces risk of acquiring HPV | 58 | 28.4 (22.7–35.0) | Benefit |
| Rejects that males cannot get cancer from HPV | 55 | 27.0 (21.3–33.4) | Susceptibility |
| Believes a child could be at risk of HPV infection | 53 | 26.0 (20.4–32.4) | Susceptibility |
| Agrees early sexual debut increases HPV risk | 45 | 22.1 (16.9–28.2) | Susceptibility |
| Agrees multiple partners increase HPV risk | 44 | 21.6 (16.5–27.7) | Susceptibility |
| Agrees vaccine reduces risk of some cancers | 40 | 19.6 (14.7–25.6) | Benefit |
| Considers self likely to contract HPV in lifetime | 36 | 17.6 (13.0–23.5) | Susceptibility |
| Agrees vaccine reduces risk of genital warts | 34 | 16.7 (12.2–22.4) | Benefit |
| Agrees HPV can be transmitted through oral sex | 25 | 12.3 (8.4–17.5) | Susceptibility |
| Rejects that Pap smear is unnecessary after vaccination | 20 | 9.8 (6.4–14.7) | Benefit |
| Agrees it is important for males to receive HPV vaccine | 9 | 4.4 (2.3–8.2) | Benefit |
| CI, confidence interval (Wilson method); HBM, Health Belief Model; STI, sexually transmitted infection. Negatively worded items were reverse-scored so that a favourable response always indicates an orientation supportive of vaccination. Items presented only to caregivers who had heard of the HPV vaccine were scored as unfavourable for those who had not, on the grounds that the absence of the belief is itself the state of interest; the number of caregivers to whom each filtered item was administered | |||
| Variable | Category | Vaccinated n/N | % (95% CI) | χ² (df) | P value |
|---|---|---|---|---|---|
| Age group, years | ≤34 | 12/40 | 30.0 (18.1–45.4) | 3.428 (2) | 0.180 |
| 35–44 | 52/118 | 44.1 (35.4–53.1) | |||
| ≥45 | 15/46 | 32.6 (20.9–47.0) | |||
| Marital status | Married | 55/155 | 35.5 (28.4–43.3) | 4.817 (3) | 0.186 |
| Widowed | 6/11 | 54.5 (28.0–78.7) | |||
| Divorced or separated | 11/19 | 57.9 (36.3–76.9) | |||
| Single parent | 7/19 | 36.8 (19.1–59.0) | |||
| Religion | Muslim | 76/198 | 38.4 (31.9–45.3) | 0.023 (1) | 0.881a |
| Christian | 3/6 | 50.0 (18.8–81.2) | |||
| Education | No formal education | 31/96 | 32.3 (23.8–42.2) | 10.562 (2) | 0.005 |
| Primary | 26/73 | 35.6 (25.6–47.1) | |||
| Secondary or higher | 22/35 | 62.9 (46.3–76.8) | |||
| Occupation | Housewife | 28/68 | 41.2 (30.3–53.0) | 41.501 (5) | <0.001a |
| Livestock activity | 7/49 | 14.3 (7.1–26.7) | |||
| Casual labour | 5/26 | 19.2 (8.5–37.9) | |||
| Petty trade | 13/20 | 65.0 (43.3–81.9) | |||
| Business | 15/30 | 50.0 (33.2–66.8) | |||
| Formal employment | 11/11 | 100.0 (74.1–100.0) | |||
| Number of children | 1–4 | 31/89 | 34.8 (25.7–45.2) | 0.739 (1) | 0.390 |
| ≥5 | 48/115 | 41.7 (33.1–50.9) | |||
| Ward | Garbatulla | 48/113 | 42.5 (33.8–51.7) | 4.225 (2) | 0.121 |
| Kinna | 6/28 | 21.4 (10.2–39.5) | |||
| Sericho | 25/63 | 39.7 (28.5–52.0) | |||
| Knowledge category | Poor | 31/133 | 23.3 (16.9–31.2) | 46.689 (2) | <0.001 |
| Fair | 33/56 | 58.9 (45.9–70.8) | |||
| Good | 15/15 | 100.0 (79.6–100.0) | |||
| Attitude category | Poor | 50/169 | 29.6 (23.2–36.9) | 36.177 (2) | <0.001a |
| Moderate | 20/26 | 76.9 (57.9–89.0) | |||
| Good | 9/9 | 100.0 (70.1–100.0) | |||
| CI, confidence interval (Wilson method). a One or more cells had an expected count below 5; the chi-square result was confirmed by a Monte-Carlo exact test based on 20,000 permutations, giving P=0.685 for religion, P<0.001 for occupation and P<0.001 for attitude category. Education was collapsed to three categories before testing because the tertiary and university categories contained 9 and 1 observations respectively. | |||||
| Variable | Category | cPR (95% CI) | aPR (95% CI) | P value | VIF |
|---|---|---|---|---|---|
| Age group, years | ≤34 (ref) | 1.00 | 1.00 | — | — |
| 35–44 | 1.47 (0.87–2.48) | 1.16 (0.72–1.88) | 0.546 | 1.31 | |
| ≥45 | 1.09 (0.57–2.07) | 0.97 (0.54–1.76) | 0.931 | 1.28 | |
| Marital status | Married (ref) | 1.00 | 1.00 | — | — |
| Widowed | 1.54 (0.86–2.75) | 1.89 (1.04–3.41) | 0.036 | 1.09 | |
| Divorced or separated | 1.63 (1.05–2.53) | 1.30 (0.83–2.04) | 0.258 | 1.14 | |
| Single parent | 1.04 (0.56–1.94) | 0.87 (0.55–1.38) | 0.544 | 1.19 | |
| Education | No formal education (ref) | 1.00 | 1.00 | — | — |
| Primary | 1.10 (0.72–1.68) | 0.67 (0.44–1.01) | 0.053 | 1.58 | |
| Secondary or higher | 1.95 (1.32–2.86) | 1.10 (0.58–2.09) | 0.765 | 2.08 | |
| Occupation | Housewife (ref) | 1.00 | 1.00 | — | — |
| Livestock activity | 0.35 (0.17–0.73) | 0.41 (0.20–0.81) | 0.011 | 1.38 | |
| Casual labour | 0.47 (0.20–1.08) | 0.49 (0.20–1.19) | 0.114 | 1.64 | |
| Petty trade | 1.58 (1.03–2.42) | 1.81 (1.12–2.92) | 0.015 | 1.39 | |
| Business | 1.21 (0.77–1.92) | 0.76 (0.45–1.29) | 0.307 | 1.66 | |
| Formal employment | 2.43 (1.83–3.23) | 1.22 (0.65–2.29) | 0.536 | 1.71 | |
| Ward | Garbatulla (ref) | 1.00 | 1.00 | — | — |
| Kinna | 0.50 (0.24–1.06) | 0.53 (0.26–1.10) | 0.086 | 1.15 | |
| Sericho | 0.93 (0.64–1.36) | 1.02 (0.72–1.45) | 0.897 | 1.16 | |
| Knowledge category | Poor (ref) | 1.00 | 1.00 | — | — |
| Fair | 2.53 (1.73–3.69) | 1.74 (1.12–2.70) | 0.013 | 1.70 | |
| Good | 4.29 (3.15–5.84) | 2.32 (1.40–3.84) | 0.001 | 1.92 | |
| Attitude category | Poor (ref) | 1.00 | 1.00 | — | — |
| Moderate | 2.60 (1.90–3.56) | 1.44 (0.99–2.09) | 0.059 | 1.60 | |
| Good | 3.38 (2.68–4.27) | 1.72 (1.16–2.56) | 0.007 | 1.44 | |
| cPR, crude prevalence ratio; aPR, adjusted prevalence ratio; CI, confidence interval; VIF, variance inflation factor; ref, reference category. Estimates are from a generalised linear model with Poisson family, log link and Huber–White robust variance. All variables associated with uptake at P<0.25 in bivariate analysis were entered simultaneously; all categorical predictors were entered as indicator variables. P values are for the adjusted estimates. A log-binomial model with the same specification failed to converge. A binary logistic model with the same specification exhibited complete separation and is not reported. | |||||