• No results found

The human Papillomavirus vaccine in Kenya : predicting the uptake through the health belief model

N/A
N/A
Protected

Academic year: 2021

Share "The human Papillomavirus vaccine in Kenya : predicting the uptake through the health belief model"

Copied!
42
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Faculty of Social and Behavioural Sciences

Graduate School of Child Development and Education

The Human Papillomavirus Vaccine in Kenya:

Predicting the Uptake Through the Health Belief Model

Research Master Educational Sciences ‘12-‘13

Thesis 2

M-A van Stam

H. Vermandere, V. Naanyu, & F. J. Oort

28 August 2013

(2)

Full title:

The Human Papillomavirus vaccine in Kenya: predicting the uptake through the Health Belief Model

Short title:

Predicting the uptake of the HPV vaccine in Kenya

(3)

Authors:

M.A. van Stam

University of Amsterdam

Department of Child Development and Education Amsterdam, The Netherlands

H. Vermandere University of Gent

International Centre for Reproductive Health (ICRH) Gent, Belgium

V. Naanyu

Moi University College of Health Sciences School of Medicine Faculty

Eldoret, Kenya

F. J. Oort

University of Amsterdam

Department of Child Development and Education Amsterdam, The Netherlands

(4)

Abstract

Background: Providing Human Papillomavirus (HPV) vaccination to Kenyan women would

significantly decrease the high incidence rates of cervical cancer in the country. However, the impact of a vaccination program largely depends on a population’s compliance. Therefore in this study we examine the potential HPV vaccination coverage in Kenya.

Methodology: This longitudinal study collected data before and after an HPV vaccination pilot program. Through Structural Equation Modeling (SEM) we examined if the Health Belief Model (HBM) contains predictors for HPV vaccine uptake among 255 primary school girls.

Principal Findings: Willingness to vaccinate was not a significant predictor for vaccine uptake. Aiming to explain this gap, we found that uptake was only significantly associated with the HBM defined ‘cues to action’ variables that represented a lack of invitation (β = -0.60) and a lack of

promotion (β = -0.33). However, willingness to vaccinate was significantly related to susceptibility (β = 0.24), father as a barrier (β = -0.21), and self-efficacy (β = 0.40). Overall, the HBM constructs explained 64 % of the variance in uptake, whereas this was 43% for willingness to vaccinate. In addition, examining the modifying and direct effects of socio-demographic variables in the HBM, we found that knowledge, origin, religion, and assets were significantly influencing the (relations

between) the HBM constructs. The inclusion of these socio-demographic variables increased the explained variance of uptake from 64% to 86% and for willingness this became 47% instead of 43%.

Conclusions: The findings in our study partially support the application of the HBM. Willingness alone was an unreliable predictor for vaccine uptake. Therefore, we stress the importance of experimental designs to examine different promotional strategies. In addition, we encourage

incorporating socio-demographic characteristics of the target population in vaccination campaigns, for example by engaging community and religious leaders in the implementation of promotion programs.

(5)

Introduction

Cervical Cancer in Kenya

The female population of Kenya has to cope with high incidence rates of cervical cancer. The estimated age-standardized incidence rate per year is 23.4 per 100.000 [1,2]. In addition, the prognosis of women with cervical cancer in Kenya is poor. The age-standardised annual mortality rate is 5.4 times higher than in more developed regions (17.3/100.000 women vs. 3.2/100.000 women annually) [1,2]. These high incidence and mortality rates are related to the high prevalence of HIV [3], the low screening coverage in Kenya (only 3.2% of all women are screened every 3 years), and the absence of the Human Papillomavirus (HPV) vaccine in the national vaccination program [1]. If the 3-dose HPV vaccine becomes available in Kenya it would provide women on-going protection against several high-risk types of the HPV (i.e. type 16 and 18) [4–6].

-- Figure 1. An extended Health Belief Model. –

Health Belief Model

However, before Kenya adds the HPV vaccine to the national vaccination program, a situation analysis is necessary in order to prepare the introduction of the vaccine in terms of costs, infrastructure but also to assess readiness among the population [7,8]. To predict the potential coverage of the vaccine, many studies focus on the girls’ caregivers’ willingness to vaccinate [9–11]. In these studies the main interest is to find predictors and barriers for a high coverage of the vaccine [10,12,13]. Frequently these studies apply health related behaviour theories that include a variety of psychological factors (e.g., attitudes, beliefs, perceived barriers) to predict vaccine uptake [10].

An example of such theory is the ‘Health Belief Model’ (HBM), presented in Figure 1, an established model to predict vaccination behaviour [10]. Basically, the original HBM indicates that in order to take action for an individual (e.g. have one’s daughter vaccinated), this person would have to (1) perceive the disease at least as ‘moderately severe’; (2) perceive a susceptibility or vulnerability by the disease; (3) believe that there are benefits in taking the preventive action; and (4) not perceive major barriers obstructing the action [14]. Nowadays, the HBM is often extended with two additional 5

(6)

constructs: (5) ‘Self-Efficacy’, and (6) ‘Cues to Action’ [15–17]. Self-efficacy indicates the

’expectancies about one’s own competence to needed to influence outcomes’ [18], and Cues to Action

(CTA) are ‘the specific stimuli necessary to trigger appropriate health behaviour’ [17].

Furthermore, it is often suggested that socio-demographic variables influence the (relations between) HBM constructs [19]. Although these modifying socio-demographic factors are described as rather important, there is no clear description which variables are most important and there is no agreement about the way the socio-demographics influence the (relations) between HBM constructs (e.g. directly, mediated, or moderating) [20–22].

An additional point of discussion about the operationalization of the HBM is the outcome measure. The original HBM had actual behaviour as outcome (e.g. ‘vaccine uptake’). However, the aim of many studies is to predict the uptake before the vaccine is implemented. Therefore studies use attitudes (e.g. acceptance, intentions, and willingness) as outcome variable [10], assuming that these attitudes are direct predictors of behaviour [23,24]. However, in the context of health behaviour, attitudes do not always predict health behaviour [20,25].

In sum, there is no consensus about the operationalization of the HBM [16]. Consequently, the operationalization varies across studies, as well as methods applied to measure and validate the constructs in the HBM [9]. This lack of consensus might complicate the application of the HBM, however, it also encourages researchers to explore the additional value of innovative constructs and modifying variables in order to find a HBM that improves the prediction concerning the health behaviour of interest [19].

The HBM in Kenyan context

There still is a significant gap in HPV vaccine coverage research (applying the HBM) in Sub-Saharan Africa [10]. The few studies that examined health related behaviour in Sub-Sub-Saharan Africa through the HBM, found that only certain components of the HBM (perceived barriers, and self-efficacy) were related to the health behaviour [26–28]. However, these results should not be generalized to our study since no studies have been found that properly tested the HBM in order to predict the uptake of the HPV vaccine in Kenya.

(7)

The purpose of the present longitudinal study was to examine the applicability of the HBM to predict HPV vaccine willingness and uptake in Kenya. This general aim is specified into two underlying research questions. First, we examined the associations described in the HBM, and whether these HBM constructs predicted vaccine uptake [14,29]. Second, we examined the direct- and modifying effects of socio-demographic variables on the (associations between the) HBM constructs. In sum, the analyses of the data gathered in the pre- and post-vaccination survey, provided an answer to the question of applicability of the HBM in the context of HPV vaccination in Kenya. More concretely: is the HBM a useful tool to predict uptake or not?

(8)

-- Figure 2. Flow diagram of the recruitment and response of study participants.--

Methods

Ethics statement

The ethical committee of Moi Teaching and Referral Hospital, and Ghent University approved the objectives and design of this study. Written consent was obtained from all participants in the sample. Participants received no incentives for participation in the pre-vaccination survey. However, for the post-vaccination survey the participants received a small financial compensation of 200 Kenyan Shilling (US $2.34) for the time and effort they were willing to invest in a second time being interviewed.

Participants and Procedures

A flow diagram of the recruitment and response of participants within this longitudinal research design is presented in Figure 2. Ten public primary schools in Eldoret Municipality were randomly selected to participate in the pilot vaccination program, i.e. Gardasil Access Program (GAP). All girls in standard 4 to 8, approximately 9-13 years old, (n = 7000) of these school were invited to receive three free doses of Gardasil® (Human Papillomavirus Quadrivalent (Types 6, 11, 16 and 18) Vaccine, Recombinant). We randomly selected 472 mothers of eligible girls within these schools. The mothers, or guardians in case the biological mother was not the primary caregiver (in this paper all referred to as mothers), were invited to participate in the pre- and post-vaccination study in March 2012 and May 2013 respectively. The vaccination was provided in Moi Teaching and Referral Hospital, located in the center of Eldoret. At time of the post-vaccination study, the daughters of the participants could have received the three doses of the HPV vaccination.

The pre- and post- vaccination interviews were conducted by local medical students and nurses of Moi School of Medicine and Moi Teaching and Referral Hospital. The interviewers received training about cervical cancer and interview techniques. To achieve consistency in the interviews, standard guidelines for introductions, interviews, and informed consent requests were practiced. During the interview period the interviewers received ongoing supervision.

(9)

The pre- and post-vaccination interviews followed a similar interview procedure. First, to make an appointment, participants in the study were contacted by phone, or, in case the mother was not found, the interviewers went looking for them at the school or at the description of the living place. Next, the interview took place in the residence of the participant, the school of the girl, or any other place the participant felt at ease. The participants were interviewed in Swahili or English, depending on the preferences of the participant. Interviewers described the study as being about the attitude of mothers towards (the prevention of) cervical cancer. After the pre-vaccination interview (duration: 45 minutes) the participants were already informed about the post-vaccination interview (duration: 30 minutes). If participants were not able to participate in the complete post-vaccination interview, they were invited to answer three short questions about the uptake of their daughter. Written informed consent was requested before the pre-vaccination interview, and this was orally confirmed before the post-vaccination interview.

Measures

The measures compromising the study (outcome variables, HBM constructs, and socio-demographic variables) instruments are described below. The analyzed items in the questionnaires were based on the theoretical framework of the HBM [14,16,18].

Outcome variables: willingness and uptake

.

The main outcomes of the study were the

pre-vaccination measured attitude ‘Willingness to vaccinate’, and the actual behavior ‘Vaccine uptake’ obtained from the post-vaccination survey. Willingness to vaccinate comprised the sum score of ‘Would you vaccinate your daughter against cervical cancer?’ (1 = very unlikely – 5 = very likely), and ‘Will you let you daughter get vaccinated against cervical cancer through this program?’(1 = very unlikely –5 = very likely) (Cronbach’s α = .898). Vaccination uptake was assessed by a dichotomous variable (0 = received no HPV vaccine doses, 1 = received one or more doses of the HPV vaccine).

HBM constructs. To examine the associations between the HBM constructs and the outcome

measures of the HBM, the surveys included questions about the perceived severity, susceptibility, benefits, barriers, self-efficacy, and cues to action (Table 1).These items were operationalized in to nine latent HBM constructs: (1) severity; (2) susceptibility; (3) benefits; (4) safety concerns; (5) time

(10)

as a barrier; (6) father as barrier; (7) self-efficacy; (8) CTA lack of invitation; and (9) CTA lack of promotion (Cronbach’s alpha’s are presented in Table 1). ‘CTA: lack of promotion’ was the aggregated average of ‘lack of invitation’ per school. This variable is defined as ‘CTA: lack of promotion’ because the promotion of the vaccination was supposed to be carried out via the school. Most HBM constructs in the analyses were obtained from the pre-vaccination survey. Only the CTA variables were taken from the post-vaccination survey because a lack of invitation could only be measured after the vaccination program was carried out. Because the analyses of pre-vaccination constructs provide different implications than post-vaccination constructs, we analyzed and discuss the seven ‘pre-vaccination HBM constructs’ separately from the two ‘post-vaccination HBM constructs’. Means, standard deviations, and correlations across HBM constructs are provided in the Electronic Supplementary Material (ESM Table I).

--Table 1. Complete list of items used to assess Health Belief Model (HBM)

constructs.--Socio-demographic variables. According to the paper of Janz and Becker [30],

socio-demographic characteristics of the participants have direct and modifying effects on the (associations between) HBM constructs. Because we had little or no hypotheses about which socio-demographic indicators were most reliable in Kenyan context, we included 15 socio-demographic variables to explore their potential direct and modifying effects on the HBM constructs. The following 15 socio-demographic variables were included in the analyses: (1) age girl (range: 8-18); (2) age mother (range: 21-59); (3) marital status mother (0 = single, 1 = with partner); (4) number of children in household (range: 0-7); (5) awareness cervical cancer (range: 0-3); (6) knowledge HPV (range: 0-6); (7)

schooling mother (0 = no schooling - 5 = college/university); (8) schooling partner (0 = no schooling - 5 = college/university); (9) origin (0 = city, 1 = countryside); (10) religion (0 = Christian, no religion or other, 1 = Muslim); (11) working status mother (0 = no work, 1 = worked the last 12 months); (12) assets in household (range: 0-4); (13) house owner (0 = no, 1 = yes); (14) monthly rent (range: 200-15.000 KS); (15) SES school area of the girl (0 = slum, 1 = poor, 2 = rich).

(11)

Most of these modifying factors were obtained from the pre-vaccination survey, however three of them (knowledge, house owner, and rent) were only included in the post-vaccination survey. In addition, for some of the items (marital status, origin, and religion) answer options in the questionnaire were merged to facilitate interpretations (a full description of the answer options is available in the ESM Table III).

Statistical Analyses

To compare participants who completed versus did not complete the post-vaccination survey we performed an univariate analysis of variance (ANOVA). In addition, to ensure reliable

interpretation of the results, all variables were screened for multi-collinearity, linearity in the logit, missing data and outliers. HBM constructs were composed by taking together several items, i.e. converting them into latent factors (Table 1). Furthermore, to examine the clustering of the data per school due to the sample design, we performed a multilevel modeling procedure with binary outcome.

Structural equation modeling (SEM) was applied to answer the first research question concerning the examination of the associations described in the HBM. The following four models were fitted: (M1) HBM with willingness as outcome; (M2) HBM with uptake as outcome; (M3) HBM including CTA, with full mediation by willingness; (M4) HBM including CTA, with partial mediation by willingness. A detailed description of the specification of parameters in the models is described in the results section.

An explorative modeling procedure was applied to examine the second research question concerning the direct and modifying effect of the fifteen socio-demographic background

characteristics on (the associations between) the HBM constructs. We firstly examined per socio-demographic variable whether we could find a direct (e.g., Age girl→Uptake), mediated (e.g., Religion→Benefits→Uptake), or moderating effect (e.g., Knowledge*Severity→ Uptake) of the variable on uptake [22]. Second, we fitted the extensive model 5 (M5), including all significant (p<.05) direct, mediated and moderating effects of the socio-demographic variables in addition to the parameters specified in M3 (described in the previous section). Third, to facilitate interpretations, a final and more parsimonious model (M6) was created using an explorative stepwise model trimming

(12)

procedure: one by one removing non-significant parameters (p>.05) from the model (selection of variable was based on the highest p-value), while examining after every deleted parameter whether the model-fit did not decrease significantly.

Maximum Likelihood estimation (ML) was applied to evaluate the standardized parameters (beta, β) in M1 (HBM with willingness as outcome). While the remaining models, with a

dichotomous primary outcome variable (uptake), were fitted through the weighted least-squares estimator with mean and variance adjustment (WLSMV) [31]. The nine HBM constructs were allowed to correlate in all models.

In addition, all models were evaluated by assessing the efficacy of each model in predicting willingness and uptake (R2). Furthermore, for the path models (M3-M6) model fit was assessed with

the comparative fit index (CFI), the root-mean-square error of approximation (RMSEA), and the weighted root mean square residual (WRMR). Since the chi-square test is very sensitive to sample size and non-normally distributed data, this test was not included as indicator of the model fit. RMSEA values <.06, CFI >.97, and WMSR<1.0 indicate close fit [32,33].

(13)

Results

Preliminary Analysis

Of the 287 participants who completed the pre-vaccination survey, 256 (89%) participants agreed to participate in the post-vaccination survey. Of these 256 participants, 8% (n = 20) were not able to participate in the complete post-vaccination survey, they only provided the information about their daughter’s uptake of the vaccine through the short telephone survey. Non completers (n = 31) were similar to completers (n = 256) on all HBM constructs and socio-demographic variables with only one exception. Compared to completers the non-completers scored slightly lower on self-efficacy (F (1, 285) = 6.486, p = .011).

The dataset did not violate any of the underlying assumptions for structural equation modeling with dichotomous outcomes. In addition, the Cronbach’s alpha was found to be acceptable (>.75) for all latent HBM constructs. Furthermore, the HBM constructs did not contain any missing values [34]. However, one participant was deleted from analysis because she did not report whether her girl was vaccinated (Nanalyses = 255). Furthermore, socio-demographic variables that contained less than 5%

missing data were imputed using the expectation maximization method (EM) after concluding the data were missing completely at-random (Little’s MCAR χ2(259) = 257.583, p = .513).

The design of the vaccination program contained selection of girls on the school level, therefore multilevel analysis with a binary outcome was applied to examine the clustering of the data. The data revealed a significant clustering effect at the school level (ICC = 0.274), however this variance at the school level disappeared after including the aggregated CTA variable ‘lack of

promotion’ (Table II in the ESM). Therefore this aggregated variable is included as a predictor in the model building process.

Descriptives

Of the 255 participants who were included in the analyses, most participants (89%) were willing to have their daughter vaccinated (willingness>6) before the vaccination program was

introduced. However, only 31% effectively had their daughter vaccinated against cervical cancer with one dose or more. This positive attitude towards the vaccination was also reflected in the

(14)

vaccination measured HBM constructs: 98% saw cervical cancer as severe (severity>18), 74% thought it was (very) likely their daughter would develop cervical cancer (susceptibility>3), 98 % saw benefits for the health of the daughter (benefits>12), only low percentages saw barriers (27% safety>9, 5 % time>6, and 11% father>3), and 97% were confident they could let their daughter vaccinated if they wanted to (self-efficacy>4). In the post-vaccination survey, 27% of the participants mentioned they did not know where and when the vaccine was distributed. However, the average percentage of people mentioning this lack of invitation fluctuated per school (11% - 47%).

The socio-demographic characteristics of the participants can be summarized as follows: the average age of the daughter was 12; the average age of the mothers was 36; 76% of the participants had a partner; 3.5 was the average number of children in the household; 60 % had at least heard of cervical cancer (awareness); 77 % of participants answered more than three of the post survey knowledge questions correctly; the average level of schooling of the mothers was 1.5, and for fathers this was 1.8; 40% of the participants grew up in a city and 60% were from the countryside; 80% of the participants indicated to be Protestant, 15 % Catholic, 4% Muslim, 0.8% other or no religion (for analyses this was combined into Muslim vs. non-Muslim); 75 % of the participants had worked in the last 12 months; 58% possessed three or more of the requested assets; 23% owned a house; the average rent of participants that did not own a house was 2878 KS; and the socioeconomic status (SES) of the school area was divided in 40% slum, 24% poor, 37.5% rich.

--

Table 2. Structural Equation Modeling: subsequent Health Belief Models.--

Application of the HBM

To answer the first research question, examination of relations in the HBM, we fitted four subsequent versions of the HBM (M1-M4). These models are described in the following sections and an overview of the fit indices is presented in Table 2.

First, to examine how the seven pre-vaccination measured HBM-constructs predicted willingness and uptake, we fitted M1 and M2. In M1 it was examined how the seven pre-vaccination HBM constructs (severity, susceptibility, health benefits, safety concerns, time as a barrier, father as a

(15)

barrier, and self-efficacy) predicted willingness to vaccinate (Figure 3.a), whereas in M2 the same predictors were specified to predict the actual uptake of the vaccine (Figure 3.b). The seven predictors accounted for 44% of the variance in willingness (M1), compared to only 8% of the variance in uptake (M2). In addition, in M1 we found three significant predictors for willingness (standardized path coefficient susceptibility β = .25; father as barrier β = -.21; and self-efficacy β = .40), whereas none of the predictors were significantly associated with uptake in M2.

-- Figure 3. Health Belief Model to Predict the HPV vaccination willingness (a. M1) and uptake (b. M2).–

Next, to examine the assumption that willingness is a predictor for uptake, we fitted M3 (Figure 4). In this model the seven pre-vaccination constructs were specified to predict willingness, and willingness was specified to predict the uptake. In addition, to examine the effects of the post-vaccination measured CTA variables (lack of invitation and lack of promotion), these two variables where specified to directly predict uptake. This model provided good fit to the data, accounted for 43% of the variance in willingness, and improved the explained variance of uptake from 8% to 64% [CFI = 1; RMSEA = .00; WRMR = 0.60]. Similar to M1, the significant predictors of willingness included susceptibility (β = .24), father as barrier (β = -.22), and self-efficacy (β = .40). However, severity, benefits, safety concerns, and time as barrier were not associated with willingness. In addition, a lack of invitation (β = -.61), and a lack of promotion (β = -.33) were significantly negative associated with uptake. As expected based on M1 and M2, willingness was not significantly associated with uptake, and thus no significant indirect effects of the seven pre-vaccination HBM constructs on uptake were found.

M4 was specified to examine whether, after including the CTA variables, any of the seven pre-vaccination predictors had a direct effect on uptake in addition to the already in M3 specified indirect effect through willingness (M4: partially mediation by willingness vs M3: fully mediated by willingness). However, the model fit and explained variance of willingness and uptake did not improve significantly after including the direct effects [CFI = .99; RMSEA = .04; WRMR = 0.37;

R2willingness = .429; R 2

uptake = .650]. We therefore consider the more parsimonious M3 as the final version

of the HBM without modifying socio-demographics.

(16)

--Figure 4. Fully mediated Health Belief Model with Cues to Action to predict the HPV vaccination uptake (M3). –

Influence of socio-demographic variables in the HBM

To answer the second research question, whether there are socio-demographic variables that have an influence on (the found associations between) the HBM constructs, we fitted M5 and M6. However, before the 15 socio-demographics were included in the HBM, an exploratory analysis was performed to examine per socio-demographic characteristics whether there was a significant direct, moderating, or mediated effect on uptake.

All 26 associations between the socio-demographic variables and HBM constructs that were found to be significant were added in M5. This model included direct effects on uptake of awareness, knowledge, schooling, and origin; by HBM constructs mediated effects on uptake of knowledge, religion, and assets; and moderating effects of the age of the girls, knowledge, religion, and rent on the relations between HBM constructs and uptake (see Table IV in ESM for details). Associated with the high number of predictors, the model provided poor fit to the data. However, the by the model accounted variance of willingness increased from 43% to 76% and for uptake from 64% to 97% [CFI = .07; RMSEA = .29; WRMR = 4.75]. Significant predictors of willingness were susceptibility (β = .16), father as a barrier (β = -.16), self-efficacy (β = .27), rent (β = .43). Significant predictors of uptake were a lack of invitation (β = -.24), and a lack of promotion (β = -.23). In addition, the relation between a lack of invitation and uptake was influenced by religion (β = -0.98). Furthermore, severity was associated with knowledge (β = .76); benefits with religion (β = -.56); father as a barrier with religion (β = .58), and house owning (β = .17); self-efficacy with religion (β = -.38); and a lack of promotion with assets (β = -.21) and house owning (β = .32). All other parameters specified in M5 were not found to be significantly associated with willingness or uptake.

To facilitate interpretations of the effects of socio-demographic on the (relations between the) HBM constructs, M6 was specified as a parsimonious version of M5 (see Figure 5). After the

explorative stepwise model trimming procedure, the final model (M6) included the direct effect of knowledge and origin on uptake, and mediated effects of knowledge, religion, and assets on uptake. These parameters were specified in M6 as an addition to the parameters in M3.This final model

(17)

provided good fit to the data, and the model accounted for 47% explained variance of willingness, and 86% of explained variance in uptake [CFI = .98; RMSEA = .03; WRMR = 0.63]. All standardized path coefficients are displayed in Figure 5.

-- Figure 5. Health Belief Model to predict the HPV vaccination uptake with modifying socio-demographic variables

(18)

Discussion

Kenya has to cope with one of the highest cervical cancer incidence rated in the world [1]. If Kenya would introduce the 3-dose vaccine that protects against several high-risk types of HPV, this high incidence rate would decrease significantly [6]. A vaccination program’s efficiency depends on the populations attitudes towards the introduced vaccine [7], and on the organization of the program itself [35]. Therefore countries should assess the national readiness for HPV vaccine introduction. The Health Belief Model is an established health theory to examine a population’s compliance. In this model attitudes and expectancies concerning severity, susceptibility, benefits, barriers, self-efficacy, and cues to action are expected to be important predictors of the health related behavior [18]. This study examined whether the HBM can be applied to predict HPV vaccine uptake in Kenya, a country with little or no prior research on HPV vaccine acceptability and uptake.

Application of the HBM

The most striking conclusion about the application of the HBM to predict the HPV vaccination uptake was the substantial difference between willingness and uptake. In line with the study of

Becker-Dreps et al. [8], we found that the most participants (89%) were willing to let their daughter vaccinated. However, only 31% effectively vaccinated her daughter with at least one dose of the vaccine. In addition, there was no significant association between willingness and uptake, and we found no significant direct or indirect effects of the pre-vaccination HBM constructs on uptake. As several previous studies also described, this indicates that the pre-vaccination attitudes, including willingness to vaccinate, did not allow us to predict which participants would have their daughter vaccinated [36,37].

This seems to be contrasting the assumptions of the HBM is that people use their pre-behavior attitudes to decide on their behavior. However, in our study many people reported a lack of

information about the vaccination program: the inclusions of the CTA variables indicating a lack of information increased the explained variance of uptake from 8 to 64%. The significantly negative effects of the CTA variables show that if practical information concerning the vaccination program was not available for the participants (no invitation), they were not capable in translating their

(19)

intentions into real behavior. Therefore, the behavior of participants who reported a lack of information was not based on their own pre-vaccination attitudes but rather on the delivery of information by others. This finding is in line with the findings of Wigle et al. [35], in their literature review they describe reaching girls for three doses is a great challenge in low- and middle- income countries.

However, a high level of willingness might not be automatically related to a high level of uptake, the other way around (a low level of willingness) would be even more dramatic for the uptake rates. If even willingness is not accomplished, it is expected that the uptake will be extremely low (independent of the received information concerning the vaccination program). Consequently, the associations we found between pre-vaccination HBM constructs and willingness are potentially promising for future HPV vaccination programs that include a sound promotion and distributional system. Therefore we will briefly describe the effects of the seven pre-vaccination HBM constructs on willingness to vaccinate in the next paragraph.

Corresponding with the review by Brewer and Fazekas [10] susceptibility and self-efficacy were both significantly positive related to willingness. Indicating that when participants assumed it was more likely that her daughter would develop cervical cancer, and participants who were more confident to have their daughters vaccinated if they wanted to, were more willing to have their daughter vaccinated. Furthermore, the significantly negative effect of a father as barrier on willingness indicated that participants who assumed that the father of the girl would not approve to have his daughter vaccinated were less willing to have her daughter vaccinated. There is however no proof that the fathers were actually against HPV vaccination and withhold their daughters from being vaccinated. These results stress the importance for future vaccination program developers to focus on delivering reliable knowledge about cervical cancer and vaccination program to both fathers and mothers.

In contrast to the three significant indicators, severity, health benefits, safety concerns, and time as barrier did not prove to be significant predictors of willingness. This lack of predictive value of these variables is in line with earlier studies that found the HBM is more applicable for repeated health behaviors with relatively small health risks (e.g. wearing a seatbelt)[21]. In contrast, cervical cancer is, especially in Kenya, a relatively big health threat. This is supported by our findings; almost everyone

(20)

in our study indicated cervical cancer as a highly severe threat, a high amount of benefits of the vaccine, and almost no barriers to vaccinate (see paragraph Descriptives). This lack of variance in pre-vaccination attitudes due to overall agreement, might explain the lack of significant relations in the HBM. For future studies this result does not imply that these constructs should not be included in the study, however, researchers are encouraged to nuance the questions concerning these constructs to achieve more variance in the answers.

In sum, the high level of willingness was explained by several pre-vaccination HBM

constructs. However, in contrast to our expectations based on the HBM, uptake was only explained by the CTA-variables that represented a lack of information, and not by willingness or any of the pre-vaccination HBM constructs. These findings might raise the question why participants who were highly willing to have their daughters vaccinated, not actively sought for more information to have their daughters vaccinated? Probably some participants did because even in the schools where the promotion was poor, some of the participants had their daughters vaccinated. Therefore, in the next section it is described which socio-demographic characteristics influenced the relations in the HBM.

Influence of socio-demographic variables in the HBM

We examined significant direct, mediated, and moderating-effects of 15 socio-demographic variables on the (associations between) HBM constructs in an explorative analysis. Although a detailed description of all significant effects (Table IV in ESM) is not in scope of this paper, they can be further examined in future studies. We will discuss the most important direct-, and modifying effects (knowledge, origin, religion, and assets) specified in the final model M6 (Figure 5).

Knowledge indicated how much correct information people possessed concerning cervical cancer and the HPV vaccine. In the final model knowledge was significantly positive associated with uptake, severity and susceptibility. Indicating that people who possessed more correct information were more likely to have their daughter vaccinated, to see cervical cancer as highly severe, and to expect their daughter to develop cervical cancer. Although these relations were not causal (knowledge was a post-vaccination construct), we support the previous studies that reported that the delivery of knowledge to people is the key issue to increase the uptake in Africa [8,38–40].

(21)

Origin, the next socio-demographic variable that was found to have a significant effect in the HBM, was significantly positively associated with uptake. This indicates that participants raised in urban areas had vaccinated their daughters more often compared to participants raised in rural areas. This result is also found in the study of Reiter et al. [25], and we discuss several explanations for this effect. First, in Kenya, urban and rural residents differ in demographic composition and average SES [41]. In previous research, a lower average SES, as found in Kenya’s rural areas, is often associated with a lower uptake of HPV vaccinations [35]. Second, people who grew up in the countryside may have other impressions with regard to time and money that needs to be invested for vaccination: people who grew up in the countryside generally had to travel more to reach basic health care. Even though people with rural descents did not indicate time or distance more as a barrier, these experiences of the past might have influenced their current willingness to vaccinate negatively [42]. Third, it might be that people with a rural background more readily turn to traditional (or ‘biocultural’) medicine compared to urban residents, or are at least more suspicious towards new preventive techniques [42]. However, these hypotheses about the reason why people with a rural background reveal lower uptake rates need more research. For example by examining whether mobile vaccination teams that are able to vaccinate on multifarious location (e.g. schools, markets or a churches) increase the uptake rates of people with a non-urban background [43].

Next, religion was found to be an important influential factor in the HBM, indicating that participants with an Islamic background were less likely to rank cervical cancer as extremely severe, saw less benefits in the vaccine, more often ranked time and the father as a barrier, and had lower self-efficacy with regard to vaccinating their daughter if they wanted to. In addition, willingness was also found to be directly influenced by religion, with again a negative relation between Islamic religion and willingness to vaccinate. Comparable effects were found in the study of Marlow et al. and Rosenthal et al. [44,45]. An explanation that they provide was that Islamic people do not allow sexual intercourse before marriage, and thus the HPV vaccine would not be necessary. Wong [46] responded that in order to increase the immunization in Islamic context, vaccination promotion should engage religious and community leaders. This institution has a higher status and influence on Islamic people than for example the school institutions.

(22)

The number of assets was the last influential socio-demographic variable that was included in final model. ‘Assets’ was significantly positively related to father as barrier and a lack of promotion: participants who possessed a higher amount of the requested assets (clock, electricity, radio, and television) reported more often that the father was a barrier, and were more in schools where the promotion was low. These findings can be viewed as contrasting the hypothesis that a higher SES is related to a higher willingness. However, we have two explanations of these results. First, our data revealed that participants with more assets were significantly more likely to have a partner, evidently meaning that participants with a partner indicated more often that their partner’s opinion might be a factor they have to take into account. However, this does not explain the significant association with the lack of promotion at the schools. However, it might be suggested that people with a higher amount of assets were less in contact with the school, for example by the assistance of employees that bring and collect children from the school. Or the other way around: schools in areas with lower SES put more effort into promotion in order to assist the parents in achieving this free health benefit. Nevertheless, there is no empirical evidence to support these suggestions.

Overall it can be concluded that socio-demographic variables are considerably related to the HBM constructs. The implications of the findings for future research and practice are explained after mentioning the limitations that might have influenced the outcomes of our study.

Limitations

Mentioning limitations of the present study can provide useful insight and directions for future research. One limitation pertains to our measure of uptake; this was self-reported rather than verified by participants’ medical records. It might be that social desirability and the financial compensation, led to a higher level of reported uptake. However, interviewers indicated that participants were not ashamed to report refusal of vaccination. Second, although multiple item measures have better predicting power, susceptibility and father as a barrier were single item measures. In future we recommend increasing the number of questions concerning these variables. Third, we found major support for the as a lack of information operationalized ‘cues to action’ is in this study. However, the operationalization of cues to action is rather vague and different across studies (e.g. exposure to mass

(23)

media, recommendation of a physician, and reminder postcards). Therefore we cannot generally support that the addition of cues to actions in the HBM as valuable: it depends on the

operationalization. Fourth, the vaccines in our pilot were provided for free, therefore only travel costs were taken in consideration in our analyses. Although additional vaccine costs might form a major barrier, we were not able to include this as a barrier in our study.

Recommendations for future research

The general aim of our study was to examine the applicability of the HBM in Kenya. We found support for the HBM constructs in predicting willingness to vaccinate. However, only the CTA variables that represented the lack of proper information about the place and time of the vaccination were significantly associated with uptake. In addition, our study stresses the importance of including socio-demographic variables as modifying factors on the (relations between) HBM constructs. Especially religion, knowledge, assets and origin were found to be important influential factors.

Based on the findings in this study we have recommendations for research and practice. For future vaccination efficiency studies, we suggest careful use of attitudes (e.g. intention, acceptability, or willingness) as sole predictors for behavior (e.g. uptake). Based upon our study results, we suggest that countries need a certain level op ‘predictability of daily life’ before we can conclude that intention is translated into actual behavior. In developing countries people frequently have to face unexpected events (e.g., coping with extreme weather circumstances, road blocks, outbreaks of violence, and changing family situations), factors that might hinder the conversion of intentions into action. Of course these unexpected events happen in more developed regions too, however a significantly lesser degree. Consequently, people in for example Kenya, may be a hundred percent willing to vaccinate, however the daily problems or occupations might withdraw the person to actively perform this intended behavior. In our study the vaccination program might have failed to reach the participants due to an extreme rainy season, strikes of the hospital employees, and the threat of post-election violence in 2013. Therefore, we strongly recommend future studies aiming to predict the uptake of a certain vaccine to include a vaccination pilot in the design. This pilot can lead to a better prediction about the expected barriers and uptake of the vaccine. In addition, we encourage future studies to

(24)

further examine the modifying influences of socio-demographic variables since they might be applied to increase the efficacy of future promotion campaigns.

Subsequently, the results in this paper have great importance for future vaccination campaign developers. We support the empirical evidence that promotion and knowledge of the HPV-vaccination is the key to a higher level of vaccination uptake [12,46]. Although efficiently delivering knowledge might be a big challenge in developing countries [35], we encourage promotion campaigns to tailor their campaigns to the socio-demographic structure of the target population. For example, by creating mobile vaccination centers in rural areas or by engaging religious and community leaders in the promotion strategy.

(25)

Acknowledgments

We would like to acknowledge the excellent work of Beatrice, Jacky, Purity, and all

enumerators, in the (organization of the) collection of data that we analyzed for this study.

(26)

References

1.

WHO/ICO (2010) Human Papillomavirus and Related Cancers in Kenya. Summary

Report 2010. Available: www.who.int/hpvcentre.

2.

Ferlay J, Shin HR, Bray F, Forman D, Mathers C, et al. (2008) Cancer Incidence and

Mortality Worldwide: IARC CancerBase. Lyon. Available: http://globocan.iarc.fr.

3.

UNAIDS (2012) 2012 report on the global AIDS epidemic. Available:

http://www.unaids.org/en/media/unaids/contentassets/documents/epidemiology/2012/g

r2012/20121120_UNAIDS_Global_Report_2012_en.pdf.

4.

Markowitz LE, Dunne EF, Saraiya M, Lawson HW, Chesson H, et al. (2007)

Quadrivalent Human Papillomavirus Vaccine. Recommendations of the Advisory

Committee on Immunization Practices (ACIP). MMWR Recomm Rep 56: 1–24.

Available: http://www.cdc.gov/mmwr/preview/mmwrhtml/rr5602a1.htm.

5.

Schiffman M, Castle P, Jeronimo J (2007) Human papillomavirus and cervical cancer.

The Lancet 370: 890–907. Available:

http://www.sciencedirect.com/science/article/pii/S0140673607614160. Accessed 21

January 2013.

6.

Kahn JA (2005) Vaccination as a prevention strategy for human papillomavirus-related

diseases. The Journal of adolescent health : official publication of the Society for

Adolescent Medicine 37: S10–6. Available:

http://www.ncbi.nlm.nih.gov/pubmed/16310136. Accessed 15 November 2012.

7.

Green LW, Glasgow RE (2006) Evaluating the relevance, generalization, and

applicability of research: issues in external validation and translation methodology.

Evaluation & the health professions 29: 126–153. Available:

http://www.ncbi.nlm.nih.gov/pubmed/16510882. Accessed 2 February 2013.

8.

Becker-Dreps S, Otieno WA, Brewer NT, Agot K, Smith JS (2010) HPV vaccine

acceptability among Kenyan women. Vaccine 28: 4864–4867. Available:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2906622&tool=pmcentrez

&rendertype=abstract. Accessed 30 October 2012.

9.

Allen JD, Coronado GD, Williams RS, Glenn B, Escoffery C, et al. (2010) A

systematic review of measures used in studies of human papillomavirus (HPV) vaccine

acceptability. Vaccine 28: 4027–4037. Available:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3006186&tool=pmcentrez

&rendertype=abstract. Accessed 9 December 2012.

10.

Brewer NT, Fazekas KI (2007) Predictors of HPV vaccine acceptability: a

theory-informed, systematic review. Preventive medicine 45: 107–114. Available:

http://www.ncbi.nlm.nih.gov/pubmed/17628649. Accessed 30 October 2012.

11.

Dempsey AF, Zimet GD, Davis RL, Koutsky L (2006) Factors that are associated with

parental acceptance of human papillomavirus vaccines: a randomized intervention

(27)

study of written information about HPV. Pediatrics 117: 1486–1493. Available:

http://www.ncbi.nlm.nih.gov/pubmed/16651301. Accessed 14 February 2013.

12.

Agosti JM, Goldie SJ (2007) Introducing HPV vaccine in developing countries--key

challenges and issues. The New England journal of medicine 356: 1908–1910.

Available: http://www.ncbi.nlm.nih.gov/pubmed/17494923.

13.

Graham JE, Mishra A (2011) Global challenges of implementing human

papillomavirus vaccines. International Journal for Equity in Health 10: 27. Available:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3143925&tool=pmcentrez

&rendertype=abstract. Accessed 30 October 2012.

14.

Rosenstock IM (1974) The health belief model and preventive health behavior. Health

Education & Behavior 21: 351–363. Available:

http://heb.sagepub.com/content/2/4/354.full.pdf+html. Accessed 15 November 2012.

15.

Gerend MA, Shepherd JE (2012) Predicting human papillomavirus vaccine uptake in

young adult women: comparing the health belief model and theory of planned

behavior. Annals of behavioral medicine : a publication of the Society of Behavioral

Medicine 44: 171–180. Available: http://www.ncbi.nlm.nih.gov/pubmed/22547155.

Accessed 30 October 2012.

16.

Carpenter CJ (2010) A meta-analysis of the effectiveness of health belief model

variables in predicting behavior. Health communication 25: 661–669. Available:

http://www.ncbi.nlm.nih.gov/pubmed/21153982. Accessed 11 November 2012.

17.

Mattson M (1999) Towards a Reconceptualization of Communication Cues to Action

in the Health Belief Model: HIV Test Counsling. Communication Monographs 66.

18.

Rosenstock IM, Strecher VJ, Becker MH (1988) Social learning theory and the health

belief model. Health Educ Q 15: 175–183. Available:

http://www.ncbi.nlm.nih.gov/pubmed/3378902.

19.

Noar SM, Zimmerman RS (2005) Health Behavior Theory and cumulative knowledge

regarding health behaviors: are we moving in the right direction? Health education

research 20: 275–290. Available: http://www.ncbi.nlm.nih.gov/pubmed/15632099.

Accessed 4 November 2012.

20.

Noar SM, Mehrotra P (2011) Toward a new methodological paradigm for testing

theories of health behavior and health behavior change. Patient education and

counseling 82: 468–474. Available: http://www.ncbi.nlm.nih.gov/pubmed/21185144.

Accessed 27 November 2012.

21.

Weinstein ND (2007) Misleading tests of health behavior theories. Ann Behav Med 33:

1–10. Available: http://www.springerlink.com/index/617767105Q825656.pdf.

Accessed 27 November 2012.

22.

Baron RM, Kenny DA (1986) The moderator-mediator variable distinction in social

psychological research: conceptual, strategic, and statistical considerations. Journal of

(28)

personality and social psychology 51: 1173–1182. Available:

http://www.ncbi.nlm.nih.gov/pubmed/3806354.

23.

Beck K (1979) The effects of positive and negative arousal upon attitudes, belief

acceptance, behavioral intention, and behavior. The Journal of social psychology 107:

239–252. Available:

http://www.tandfonline.com/doi/abs/10.1080/00224545.1979.9922704. Accessed 24

January 2013.

24.

Ajzen I, Fishbein M (1977) Attitude-behavior relations: A theoretical analysis and

review of empirical research. Psychological Bulletin 84: 888–918. Available:

http://content.apa.org/journals/bul/84/5/888.

25.

Reiter PL, Brewer NT, Gottlieb SL, McRee A-L, Smith JS (2009) Parents’ health

beliefs and HPV vaccination of their adolescent daughters. Social science & medicine

(1982) 69: 475–480. Available: http://www.ncbi.nlm.nih.gov/pubmed/19540642.

Accessed 30 December 2012.

26.

Volk JE, Koopman C (2001) Factors associated with condom use in Kenya: a test of

the health belief model. AIDS education and

prevention : official publication of the

International Society for AIDS Education 13: 495–508. Available:

http://www.ncbi.nlm.nih.gov/pubmed/11791782.

27.

Hounton SH, Carabin H, Henderson NJ (2005) Towards an understanding of barriers to

condom use in rural Benin using the Health Belief Model: a cross sectional survey.

BMC public health 5: 8. Available:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=547902&tool=pmcentrez&

rendertype=abstract. Accessed 15 February 2013.

28.

Katikiro E, Njau B (2012) Motivating Factors and Psychosocial Barriers to Condom

Use among out-of-School Youths in Dar es Salaam, Tanzania: A Cross Sectional

Survey Using the Health Belief Model. Isrn Aids 2012: 1–8. Available:

http://www.hindawi.com/isrn/aids/2012/170739/. Accessed 15 February 2013.

29.

Gerend MA, Lee SC, Shepherd JE (2007) Predictors of human papillomavirus

vaccination acceptability among underserved women. Sexually transmitted diseases 34:

468–471. Available: http://www.ncbi.nlm.nih.gov/pubmed/17139233. Accessed 21

January 2013.

30.

Janz N, Becker M (1984) The health belief model: A decade later. Health Education &

Behavior. Available: http://heb.sagepub.com/content/11/1/1.short. Accessed 17 April

2013.

31.

Muthén B (1984) A general structural equation model with dichotomous, ordered

categorical, and continuous latent variable indicators. Psychometrika 49: 115–132.

Available: http://link.springer.com/article/10.1007/BF02294210. Accessed 15 July

2013.

32.

Kline RB (2011) Principles and Practice of Structural Equation Modeling. 3rd ed. New

York: The Guilford Press.

(29)

33.

Newsom J (2005) Practical approaches to dealing with nonnormal and categorical

variables: 1992–1994. Available:

http://www.upa.pdx.edu/IOA/newsom/semclass/ho_estimate2.pdf. Accessed 15 July

2013.

34.

Peterson RA (1994) Meta-analysis of Alpha Cronbach ’ s Coefficient. Journal of

Consumer Research 21: 381–391. Available: http://www.jstor.org/stable/2489828 .

35.

Wigle J, Coast E, Watson-Jones D (2013) Human papillomavirus (HPV) vaccine

implementation in low and middle-income countries (LMICs): Health system

experiences and prospects. Vaccine: 1–7. Available:

http://www.ncbi.nlm.nih.gov/pubmed/23777956. Accessed 17 July 2013.

36.

Sheeran P (2002) Intention—behavior relations: A conceptual and empirical review.

European review of social psychology: 37–41. Available:

http://www.tandfonline.com/doi/abs/10.1080/14792772143000003. Accessed 15

February 2013.

37.

Harris KM, Maurer J, Lurie N (2009) Do people who intend to get a flu shot actually

get one? Journal of general internal medicine 24: 1311–1313. Available:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2787941&tool=pmcentrez

&rendertype=abstract. Accessed 25 July 2013.

38.

Coleman MA, Levison J, Sangi-Haghpeykar H (2011) HPV vaccine acceptability in

Ghana, West Africa. Vaccine 29: 3945–3950. Available:

http://www.ncbi.nlm.nih.gov/pubmed/21481327. Accessed 17 April 2013.

39.

Francis SA, Nelson J, Liverpool J, Soogun S, Mofammere N, et al. (2010) Examining

attitudes and knowledge about HPV and cervical cancer risk among female clinic

attendees in Johannesburg, South Africa. Vaccine 28: 8026–8032. Available:

http://www.ncbi.nlm.nih.gov/pubmed/20887829. Accessed 6 March 2013.

40.

Harries J, Moodley J, Barone MA, Mall S, Sinanovic E (2009) Preparing for HPV

vaccination in South Africa: key challenges and opinions. Vaccine 27: 38–44.

Available: http://www.ncbi.nlm.nih.gov/pubmed/18977271. Accessed 5 March 2013.

41.

Newmann S, Garner E (2005) Social inequities along the cervical cancer continuum: a

structured review. Cancer Causes & Control 16: 63–70. Available:

http://link.springer.com/article/10.1007/s10552-004-1290-y. Accessed 25 July 2013.

42.

Molyneux CS, Mung’Ala-Odera V, Harpham T, Snow RW (1999) Maternal responses

to childhood fevers: a comparison of rural and urban residents in coastal Kenya.

Tropical medicine & international health : TM & IH 4: 836–845. Available:

http://www.ncbi.nlm.nih.gov/pubmed/10632992.

43.

Watson-Jones D, Baisley K, Ponsiano R, Lemme F, Remes P, et al. (2012) Human

papillomavirus vaccination in Tanzanian schoolgirls: cluster-randomized trial

comparing 2 vaccine-delivery strategies. The Journal of infectious diseases 206: 678–

686. Available:

(30)

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3414230&tool=pmcentrez

&rendertype=abstract. Accessed 27 May 2013.

44.

Marlow L, Waller J, Evans REC, Wardle J (2009) Predictors of interest in HPV

vaccination: A study of British adolescents. Vaccine 27: 2483–2488. Available:

http://www.ncbi.nlm.nih.gov/pubmed/19368790. Accessed 21 January 2013.

45.

Rosenthal DA, Dyson S, Pitts M, Garland S (2007) Challenges to Accepting a Human

Papilloma Virus (HPV) Vaccine: A Qualitative Study of Australian Women. Women &

Health 45 (2): 37–41. doi:10.1300/J013v45n02.

46.

Wong LP (2009) HPV information needs , educational messages and channel of

delivery preferences : views from developing country with multiethnic populations.

Vaccine 27: 1410–1415. doi:10.1016/j.vaccine.2008.12.045.

(31)

Tables

Table 1. Complete list of items used to assess Health Belief Model (HBM) constructs. HBM Predictors Pre/ posta Item wording (response options)

Minimum and maximum values

(used in analyses)

# items α

b

Severity pre How serious would it be if your daughter would have cervical cancer? (1=not serious at all–5=very serious)

6 = min. severity 30 = max. severity

6 .781

Your daughter having cervical cancer would be upsetting. (1=strongly disagree–5=strongly agree) Your daughter having cervical cancer would be disruptive for health. (1=strongly disagree– 5=strongly agree)

Your daughter having cervical cancer would be disruptive for romantic relations. (1=strongly disagree–5=strongly agree)

Your daughter having cervical cancer would be shameful for her. (1=strongly disagree–5=strongly agree)

Your daughter having cervical cancer would be shameful for family. (1=strongly disagree– 5=strongly agree)

Susceptibility pre How likely is it that your daughter would develop cervical cancer in the future? (1=very unlikely– 5=very likely)

1= min. susceptible 5 = max. susceptible

1 n/a

Benefits pre The vaccine will protect her health. (1=strongly disagree–5=strongly agree)

4 = min. benefits 20 = max. benefits

4 .917

The vaccine will protect her reproductive health. (1=strongly disagree–5=strongly agree) The vaccine will prevent her from having cervical cancer. (1=strongly disagree–5=strongly agree) The vaccine will prevent her from having genital warts. (1=strongly disagree–5=strongly agree)

Barriers pre

Safety concerns You think it might have unknown future side effects. (1=strongly disagree–5=strongly agree)

3 = min. safety concerns 15 = max. safety concerns

3 .882

You think it might interfere with her fertility. (1=strongly disagree–5=strongly agree)

You’re afraid the vaccine will not be administered safely (clean needles). (1=strongly disagree– 5=strongly agree)

Time You think vaccination always takes a lot of time. (1=strongly disagree–5=strongly agree)

2 = time is no barrier 10 = time is a strong barrier

2 .792

You think it’s inconvenient that she needs 3 doses. (1=strongly disagree–5=strongly agree)

Father You think your partner or her father won’t approve it. (1=strongly disagree–5=strongly agree)

1 = father is no barrier 5 = father is a strong barrier

1 n/a

Self-efficacy pre Are you confident that you could let your daughter get vaccinated if you wanted? (1=not confident at all–5=very confident)

2 = min. self-efficacy 10 = max. self-efficacy

2 .763

For you, if you want your daughter to be

vaccinated against cervical cancer, that would be. (1=very difficult–5=very easy)

Cues to action post

Lack of invitation Did you have trouble with getting the correct information regarding where and when? (0=no-1=yes)

0= received invitation 1= not received invitation

1 n/a

Lack of promotion School average of respondents indicating they had trouble with getting the correct information regarding where and when.

.11= min. lack of promotion .47 = max. lack of promotion

1 n/a

a

Measure obtained from pre- or post-vaccination interview.

bCronbach’s α indicates the reliability.

(32)

Table 2. Structural Equation Modeling: subsequent Health Belief Models. Model indices M1 HBM with willingness as outcome M2 HBM with uptake as outcome M3 HBM with CTA and uptake fully mediated by willingness M4 HBM with CTA and uptake partially mediated by willingness M5 HBM including all effects of the socio-demographics M6 HBM including significant effects of the socio-demographics CFI - - 1 0.993 0.071 0.979 RMSEA - - .000 [.000;.064] .043 [.000;.139] .285 [.277;.294] .030 [.000;.055] WRMR - - 0.597 0.373 4.746 0.631 R2willingness .439 - .430 .429 .759 .465 R2uptake - .076 .644 .650 .965 .860

90 % confidence intervals are in squared brackets. Nparticipants=255, NSchools = 10.

(33)

Figures

Figure 1. An extended Health Belief Model. 3. Benefits 2. Susceptibility 4. Barriers 1. Severity Health Related Behavior 6. Cues to action Socio-demographic variables 3. Benefits 5. Self-efficacy 33

(34)

Figure 2. Flow diagram of the recruitment and response of study participants.

Random sample of schools

Primary school girls (n = 472) of standard 4-8 were randomly selected using class lists from 10 randomly selected public schools within Eldoret Municipality. The number of girls per school were in proportion with the size of the school. The mothers received an invitation to the study via their daughters.

Completed pre-vaccination survey Filled in questionnaires (n = 289)

Excluded

Second interviews of mothers who were interviewed twice (n = 2)

Invited for post vaccination-survey

By phone or visiting school or description living place (n = 287)

Data available for analyses

Complete pre- and post-interviews (n = 236) Pre-interview and uptake information (n = 20) Lost cases

Participants who could not be found, passed away, moved, not able/willing to participate (n = 31)

Lost cases

Non-respondents to the invitation (n = 187)

(35)

Figure 3. Health Belief Model to predict the HPV vaccination willingness (a. M1) and uptake (b. M2). Numbers represent the standardized parameters (β). R2 represents the explained variance of the dependent variables by the predictors (N = 255). *p<.05 Severity Susceptibility Benefits Barrier: safety Barrier: time Barrier: father Self-efficacy .02 .25* .08 -.04 .40* -.21* .03 WILLINGNESS R2=44% a. M1 b. M2 Severity Susceptibility Benefits Barrier: safety Barrier: time Barrier: father Self-efficacy

-.04 -.02 .02 .02 .27 -.04 .04 UPTAKE R2=8% 35

(36)

Figure 4. Fully mediated Health Belief Model with Cues to Action to predict the HPV vaccination uptake (M3). Numbers represent the standardized parameters (β). R2 represents the explained variance of the dependent variables by the predictors (N = 255). *p<.05

CTA: lack of invitation

CTA: lack of promotion

.12 -.61* -.33* M3

Severity Susceptibility Benefits Barrier: safety Barrier: time Barrier: father Self-efficacy .01 .24* .08 -.04 .40* -.22* WILLINGNESS R2=43% UPTAKE R2=64% .03 36

(37)

Figure 5. Health Belief Model to predict the HPV vaccination uptake with modifying socio-demographic variables (M6). Numbers represent the standardized parameters (β). R2 represents the explained variance of the dependent variables by the predictors (N = 255).

*p<.05 M6 Severity Susceptibility Benefits Barrier: safety Barrier: time Barrier: father Self-efficacy Knowledge Religion WILLINGNESS R2 = 47% UPTAKE R2 = 86%

CTA: lack of invitation CTA: lack of promotion

Origin .12* -.19* -.12 -.12* .12 .13* -.14* .20* .17* .21* .01 .23* .07 -.03 .05 -.20* .42* .16 -.52* -.62* .20* .33* -.13* Assets .19*

(38)

Electronic Supplementary Material (ESM)

Table I. Correlations, means, standard deviations, and ranges, among model constructs.

Variablea 1 2 3 4 5 6 7 8 9 10 11 1. Uptakeb -- 2. Willingness .175 2.641 3. Severity .027 .150 14.291 4. Susceptibility -.040 .353 -.079 1.014 5. Benefits .146 .314 .458 .026 4.804 6. Barrier: safety .075 -.104 .132 -.145 .052 14.215 7. Barrier: time -.089 -.163 -.206 -.055 -.253 .079 1.180 8. Barrier: father -.067 -.404 .001 -.296 -.104 .241 .212 1.110 9. Self-efficacy .236 .524 .343 .108 .541 .038 -.324 -.249 2.418 10. CTA: lack of invitationc -.716 .124 .072 .076 .064 -.110 .039 -.024 .098 0.186 11. CTA: lack of promotion -.572 -.128 -.108 -.079 -.124 -.090 .098 .043 -.123 .372 0.028 Meansd 31% 8.85 26.82 2.80 18.41 7.63 2.77 1.81 8.67 30% .30 SD .46 1.71 3.88 1.03 2.24 3.90 1.12 1.19 1.62 .44 .18 Range 0/1 2-10 6-30 1-5 4-20 2-15 2-8 1-5 2-10 0/1 .11-.70 N = 255.

a A composite was formed for those variables with multiple indicators by taking the sum of the items. b

HPV vaccine uptake: 0 = received no injections; 1 = received ≥ 1 injection.

c Lack of invitation; 0 = no; 1=yes. d

(39)

Table II. Multilevel logistic regression model for Health Belief Model (HBM): odds ratio estimates. Parameter MI Empty model MII HBM with lower-level predictors MIII

HBM with lower level predictors and interactions

MIV

HBM with significant higher level predictor Fixed effects Intercept -1.300 (0.401)*** [-1.085;-0.515] -3.856 (1.385)** [-6.571;-1.142] -2.736 (0.845)** [-4.393;-1.080] -2.667 (0.776)*** [-4.188 -1.146] Individual level Severity -0.024 (0.052) -0.027 (0.053) -0.047 (0.053) Susceptibility 0.062 (0.164) 0.050 (0.167) 0.054 (0.167) Benefits 0.019 (0.103) -0.433 (0.249)+ [-0.921;0.056] -0.463 (0.241)+ [-0.936;0.009] Barrier: safety -0.015 (0.045) -0.016 (0.046) -0.013 (0.045) Barrier: time 0.194 (0.175) 0.203 (0.178) 0.152 (0.174) Barrier: father 0.021 (0.160) 0.010 (0.162) 0.042 (0.162) Self-efficacy 0.392 (0.154)* [0.090;0.694] 0.396 (0.157)* [0.088;0.704] 0.369 (0.150)* [0.075;0.663]

CTA: lack of Invitation -5.392 (1.544)***

[-8.418 -2.365]

-7.684 (2.768)** [-13.109;-2.259]

-7.362 (2.607)*** [-12.471;-2.252] CTA: Lack Of Invitation*

Benefits -1.626 (0.801)* [-3.197;-0.055] -1.723 (0.785)* [-3.262;-0.185] School level

School Average CTA Lack of Invitation

-5.447 (1.501)*** [-8.389;-2.506] Random effects

School level variance

Intercept 1.246 (1.116) 0.462 (0.680) 0.510 (0.714) 0.000 (0.000) Deviance, df 290.4, 2 229.2, 10 224.2, 11 213.0, 12 ΔDeviance, df ΔMI,MII = 61.271, 8 p>.001*** ΔMII,MIII = 4.940, 1 p=.026* ΔMIII,MIV = 11.248, 1 p>.001*** R2 .211 .228 .267

Standard errors are in parentheses, 95 % confidence intervals are in square bracket.

+

p <.10; * p <.05; ** p <.01; *** p <.001. Nparticipants=255, Nschools = 10.

Referenties

GERELATEERDE DOCUMENTEN

Churchill continued to seek ways of asking Stalin to allow British and American aircraft, flying from Britain, to drop supplies on Warsaw and then fly on to Soviet air bases to

The following attitudes were explicitly inquired about, after the loss and 6 months later, to guide the interviews: autopsy beyond consideration owing to religious reasons;

Daarom is in de periode September 2014 – Maart 2016 onderzoek verricht naar laag-dynamische systemen in het rivierengebied (project OBN 2014-63-RI Laag-dynamische systemen

Seeking to navigate and explore diasporic identity, as reflected in and by transatlantic narrative spaces, this thesis looks to three very different novels birthed out of the Atlantic

Nienaber en Die son struikel (1964) van Dolf van Niekerk, vergelyk hy die literêre meriete van Baker se teks met dié van tekste wat in daardie stadium deur

The cause of under-five malnutrition in Tanzania is multifaceted and according to UNICEF Tanzania, inadequate care and feeding practice particularly during the first

This study makes use of automated means to rank the topics discussed on the Dr Math service and to align them with the topics encountered in the South African National