Predictors of self and parental vaccination decisions in England during the 2009 H1N1 pandemic: Analysis of the Flu Watch pandemic cohort data

During the 2009 H1N1 pandemic, UK uptake of the pandemic inﬂuenza vaccine was very low. Furthermore, attitudes governing UK vaccination uptake during a pandemic are poorly characterised. To the best of our knowledge, there is no published research explicitly considering predictors of both adult self-vaccination and decisions regarding whether or not to vaccinate one’s children among the UK population during the H1N1 pandemic. We therefore aimed to identify predictors of both self- vaccination decisions and parental vaccination decisions using data collected during the H1N1 pandemic as part of the Flu Watch cohort study. Data were analysed separately for 798 adults and 85 children: exploratory factor analysis facilitated reduction of 16 items on attitudes to pandemic vaccine into a smaller number of factors. Single variable analyses with vaccine uptake as the outcome were used to identify variables that were predictive of vaccination in children and adults. Potential predictors were: attitudinal factors created by data reduction, age group, sex, region, deprivation, ethnicity, chronic condition, vocation, healthcare-related occupation and previous inﬂuenza vaccination. Consistent with previous literature concerning adult self-vaccination decisions, we found that vaccine efﬁcacy/safety and perceived risk of pandemic inﬂuenza were signiﬁcant predictors of both self- vaccination decisions and parental vaccination decisions. This study provides the ﬁrst systematic attempt to understand both the predictors of self and parental vaccination uptake among the UK general popula- tion during the H1N1 pandemic. Our ﬁndings indicate that concerns about vaccine safety, and vaccine effectiveness may be a barrier to increased uptake for both self and parental vaccination. (cid:1) 2017


Introduction
The H1N1 ('Swine Flu') pandemic represented a significant worldwide public health emergency. Although initial fears concerning its potential severity were unfounded, the public health impact was nonetheless significant. Across the UK, at least 457 people died [17]. In England alone, there were 1700 critical care admissions, 7879 hospital admissions, and 580,000 general practice (GP) influenza-related consultations during the pandemic year [24]. The highest rates of infection and hospitalisation were observed among children [8].
Despite the overall positive UK response to the H1N1 pandemic [17], the pandemic influenza vaccination programme was less successful. The programme aimed to achieve 75% vaccine uptake among priority groups [17] but, in England, only attained uptake of 34.5% [33]. Strikingly, vaccine uptake among children between six months and five years old (the ages offered the pandemic vaccine, [17]) was only 23.6% in England during the pandemic [33]. From a preparedness perspective-given both that pandemic influenza represents the most significant risk of civil emergency to the UK [6], and the apparent risk posed to children by pandemic influenza-there is a clear need to understand factors that influenced vaccine uptake during the UK pandemic.
Internationally, there is a range of literature examining factors associated with vaccine uptake during the H1N1 pandemic (see [2] for a systematic review). However, the literature concerning specific factors influencing vaccine uptake during the 2009 UK pandemic is relatively sparse. Of 37 articles included by Bish and colleagues in their systematic review [2], only four specifically concerned UK vaccination: Myers and Goodwin [23], Rubin et al. [30,31]; and Stokes and Ismail [37]. The two papers by Rubin and colleagues Rubin et al. [30,31] used data from the same 36 Department of Health telephone surveys; a further paper using this dataset has since been published [15]. An additional four quantitative studies examine factors associated with H1N1 influenza vaccination [7,27,35,39]. We are also aware of two qualitative studies [21,22].
Of these nine quantitative papers, six concerned the behaviour of the general public [7,23,15,27,35,39]. However, only four measured public vaccination uptake rather than intentions [15,27,35,39]. Furthermore, only one of these papers explicitly considered factors associated with parental vaccination intentions [31]. Indeed, a further examination of Bish and colleagues' systematic review reveals only four studies (including [31]) that specifically focused on parental vaccination behaviour (Setbon and Raude, 2010; Schwarzinger et al., 2010; Torun and Torun, 2010; cited in Bish et al. [2]). Although there is a broader literature concerning predictors of parental vaccination during H1N1 (e.g., [4,5,9,16,18,26,34]; see also Larson, Jarrett, Eckersberger, Smith, & Paterson for a broader systematic review of childhood vaccine hesitancy), we are unaware of any additional quantitative papers concerning predictors of UK parental vaccination during the H1N1 pandemic. Given the threat posed by a future influenza pandemic and the low uptake of vaccination in the prior pandemic (particularly among young children), further understanding of vaccine uptake among all ages of the UK general population during the H1N1 pandemic is critical.
Flu Watch was a prospective cohort study of households in England run between 2006 and 2011 to help improve understanding of influenza burden and the factors (demographic, social, and behavioural) associated with influenza transmission [11]. During the 2009 H1N1 outbreak (specifically during Spring 2010), a 'pandemic cohort' captured information on: (i) the clinical profile of the illness, and; (ii) the behavioural and attitudinal responses of members of the public towards vaccination and antiviral usage [11]. Where a household contained children (<16 years old), parents were asked to complete a questionnaire for themselves and questionnaires on behalf of any children. These data can therefore be analysed to determine predictors of both self and parental vaccination decisions (i.e., decisions to vaccinate oneself and/or one's child). Priority groups for pandemic influenza vaccination changed during the course of the pandemic [17,33]. At the time of surveying the Flu Watch 'pandemic cohort', those eligible for vaccination included social/healthcare workers, pregnant women, household contacts of immunocompromised people, seasonal influenza clinical at-risk groups, and healthy children aged 6 months to 5 years [17,33]. In this paper, we use the Flu Watch pandemic cohort data to provide an initial examination of the potential predictors of both adult self-and parental vaccination decisions during the H1N1 pandemic. In particular, we were interested in addressing: (1) what were the significant predictors of H1N1 vaccine uptake among the 2009 Flu Watch pandemic cohort? (2) do these predictors differ depending on whether they concern self or parental vaccination decisions?

Participants & design
Participants were a subset of the Flu Watch 'pandemic cohort' participants who: a) provided baseline demographic data for the Flu Watch cohort and completed a 16 item attitudinal survey concerning pandemic influenza vaccination during Spring 2010 (1953 of 3744 (52%) 'pandemic cohort' participants), and; b) reported being offered an influenza vaccine (specifically, all those responding ''Yes" to the question ''Have you been offered a flu vaccine since August 2009?"; 883 of 1953 (45%) of survey participants). There was a degree of uncertainty within the Flu Watch cohort around which vaccination was offered (e.g., 8% of individuals were unsure which vaccine they were offered), and all individuals who were offered the seasonal flu vaccine were eligible for the pandemic vaccine (see [17]). Participants in our sample therefore included a mix of those offered pandemic vaccine and/or seasonal vaccine.
Of these 883 participants, 798 (90.4%; sampled from 552 households) were adults and 85 (9.6%; sampled from 57 households) were children aged 0-15 years. Participants were not offered vaccination directly through Flu Watch (the H1N1 UK national pandemic vaccination programme was GP-based; [17]). Uptake of pandemic influenza vaccination (identified as those reporting ''Yes" to the question ''did you have a PANDEMIC flu vaccine in 2009 or 2010?") was reported for 340 (43%) adults and 58 (68%) children. Survey responses could only be submitted upon completion of all questions. Previous influenza vaccination status was derived predominantly from self-report responses to the pandemic cohort study, in combination with information from medical records (see [11] for the full Flu Watch study cohort profile). Information on the date of pandemic vaccination was requested, but inconsistently reported by participants. Further participant characteristics are outlined in Table 1. All participants gave written informed consent (proxy consent for children). The protocol was approved by the Oxford Multi Centre Research Ethics Committee (06/Q1604/103).

Materials & procedure
Participants were asked to report their agreement with each of 16 statements relating to pandemic influenza or pandemic influenza vaccination (e.g., ''I did not think I was at risk of pandemic flu"; ''I was too busy/had too little time to get vaccinated") using five-point Likert scales (Strongly Disagree -Strongly Agree). Separate surveys were completed online for each participant (adult aged ! 16 years or child aged < 16 years), with parents completing the survey on behalf of children. The parent was specifically asked about their own attitude towards their child being vaccinated rather than responding by proxy (e.g., ''I did not think that my child was at risk of pandemic flu"; ''I was too busy/had too little time to get my child vaccinated"). All attitudinal items are presented in the Supplementary information.
Data reduction. Exploratory factor analysis using the principalfactor method was conducted separately for adults and children; this method seeks to identify the lowest number of factors that can account for variance common to a set of items [12]. Items relating to attitudes likely to be negatively associated with pandemic influenza vaccination (e.g., ''I did not think that I was eligible for pandemic vaccine") were reverse coded prior to data reduction for ease of interpretation. Examination of the correlation matrices and Kaiser-Meyer-Olkin tests confirmed the suitability of our data for exploratory factor analysis (with thresholds of >0.4 and >0.5, respectively, considered suitable on the basis of previous literature; [19,14]). The change in gradient of scree plots (displaying eigenvalues for each factor in decreasing order) was used to guide factor extraction. Scree plots were favoured over other methods such as the Kaiser criterion, which apply rigid criteria, due to their greater suitability for exploratory analyses [10,40]. Where the change in slope for the scree plot was not clear-cut (e.g., interpretable as either 3 or 4 factors) we examined the rotated (orthogonal varimax) factor matrices (which are easier to interpret) for both possible factor groups and selected the final number of factors on the basis of their compatibility with expert knowledge and the literature.
Cronbach's alpha was computed to examine internal consistency for each identified factor. Provided acceptable levels of internal consistency were met (a ! 0.7); [25], the items loading on each individual factor were subsequently averaged to produce a single aggregate variable related to each factor for each participant. If the factor did not display sufficient internal consistency, then the items corresponding to that factor were analysed individually. Relative risks (RR) and associated 95% confidence intervals (CI) were approximated using Poisson regression models with robust variance estimates (as per [13,36]) in preference to odds ratios because uptake of pandemic influenza vaccine was not rare (of those offered vaccination, 43% of adults and 68% of parents chose to vaccinate themselves/their child respectively).

Factor analysis
Factor analysis produced three factors for adults' attitudes towards pandemic self-vaccination decisions (Table 2). First, an eight-item factor broadly related to vaccine safety, testing and side-effects, and concerns regarding the impact of influenza on time-off work (a = 0.82). The item relating to concerns about ''time off work/education because of pandemic flu" in adults was negatively correlated with all others in the same factor despite having been reverse coded prior to data reduction (see Data Reduction section above). The item was therefore returned to its original non-reversed state for incorporation into the aggregate variable for this factor. Second, a two-item factor related to vaccine safety and effectiveness (a = 0.82). The third factor fell below the threshold for acceptable internal consistency (a = 0.53) and so corresponding items were included individually.
Factor analysis yielded five factors relating to parental attitudes towards pandemic vaccination of their children (Table 3). Only the first factor (a six-item factor related to vaccine safety, side effects and effectiveness) met acceptable levels of reliability (a = 0.79).
All other factors fell below the threshold (all as < 0.57). Items corresponding to factors two-five were included individually for further analysis.
In order to assist with interpretation of the results, the individually included items corresponding to both self-vaccination and parental-vaccination are presented together in Table 4.

Predictors of pandemic influenza vaccination
Tables 5 (self) and 6 (parental) summarise the results of the univariate analysis. Uptake of pandemic vaccination was predicted by previous influenza vaccination for 88% of adult self-vaccination (RR 8.51, 95% CI 6.43-11.3) and 98% of parental vaccination. 2 Uptake of pandemic vaccination was also significantly associated with age group for both parental vaccination and self-vaccination decisions.
Uptake of pandemic vaccine among adults was almost 50% lower for those aged 65+ years (RR 0.54, 0.44-0.66), relative to 16-44 year olds. Among children, 5-15 year olds were more frequently vaccinated (RR 1.36, 1.04-1.79) relative to the under 5 s. Relative to professional vocations, self-vaccination behaviour was comparatively higher among students (RR 1.63 (1.29-2.07)) and lower among the retired (RR 0.79 (0.65-0.96)). Those working in a healthcare related occupation (RR 1.49 (1.19-1.85)) were more likely to be vaccinated than non-healthcare roles. Vaccination was also higher among those with one or more chronic conditions (RR 2.31 (1.99-2.69)) relative to those without a chronic condition. There were no relationships between parental vaccination decisions and any of these potential predictors; although having a chronic condition was suggestive of increased uptake (RR 1.30 (0.99-1.71)). Finally, for parental (but not self-) vaccination decisions, deprivation was associated with uptake of vaccination (p = 0.04); however, the direction of effect was not consistent across the quintiles of deprivation.
Having concerns about vaccine safety, testing and side-effects, and the impact of influenza on time-off work was associated with lower vaccination uptake (RR 1.60 (1.41-1.81) [some items scale reversed] -indicating a 1.6-fold increase in vaccination uptake per unit increase in the Likert scale), whereas belief in vaccine effectiveness and safety was associated with greater vaccination uptake (RR 1.92 (1.73-2.13)). All of the individual data items included for analysis were associated with uptake of pandemic

Table 4
Self-vaccination and parental vaccination items that did not reliably load onto factors and so were included individually for analysis. vaccine among adults. Items relating to concerns about influenza (severe nature, spreading the infection to others, believing oneself to be at risk of infection or complications) were significantly associated with greater pandemic vaccine uptake (RRs 1.1-1.3; all items p 0.01). A belief that natural infection provides stronger immunity than vaccination was associated with lower uptake of pandemic vaccine, as was a belief that one was ineligible for pandemic vaccine (RRs 1.2-1.3, both p 0.001 [scale reversed]). Finally, having the pandemic vaccine recommended by one's doctor was associated with greater uptake of pandemic vaccine (RR 1.64 (1.52-1.77)).
Concern over pandemic influenza vaccine safety, side effects and effectiveness was associated with less parental vaccination (RR 1.68 (1.38-2.04) [some items scale reversed]). Two of the individual parental belief items were significantly associated with having one's child vaccinated against pandemic influenza. First, a belief that one's child was at risk of influenza infection was associated with greater parental vaccination (RR 1.46 (1.22-1.76) [scale reversed]). Second, a belief that natural infection provides stronger immunity than vaccination was associated with less parental vaccination (RR 1.23 (1.09-1.40) [scale reversed]). Although not significant, there was an indication (p = 0.06) that a belief that one's child was eligible for pandemic vaccine was associated with less parental vaccination. No other attitudinal items were associated with uptake of pandemic vaccine.
Finally, there were no significant relationships between either self-or parental vaccination decisions and pregnancy, sex, region or ethnicity (p > 0.5). The small number of pregnant women that were vaccinated in our sample (n = 3) is insufficient to detect significant effects. There was potential geographical variation (p = 0.07) in adult self-vaccination with people in the South East having lower uptake than those in the South West (RR 0.65 (0.45-0.94)).

Discussion
Taken together, decisions to vaccinate oneself and/or one's child against H1N1 were broadly associated with: lower concerns about the safety and effectiveness of the pandemic influenza vaccine, greater perceived risk of influenza, and less belief that natural infection provides immunity. The concerns over vaccine safety/ efficacy and perceived risk of influenza, in particular, proliferate through the research concerning self-vaccination during the H1N1 pandemic both domestically (e.g., [15,23,30,37,27,39]), and internationally [2].
For childhood vaccination, the current observed relationships between perceived risk and child vaccination are consistent with Rubin and colleagues' Rubin et al. [31] finding that perceived risk mediated their observed relationship between National Health Service (NHS) work and likely vaccine uptake [31]. However, this Table 5 Results of the univariate analysis examining the relationship between potential predictors of vaccination uptake and self-vaccination. relationship between perceived risk and parental H1N1 vaccine uptake was not observed in a recent paper from the United States ( [16]; but cf. [26] 3 ; [34]). Furthermore, although we are aware of no published UK data that examines the relationship between concerns about vaccine safety/side effects and parental vaccine uptake, the relationship reported herein is consistent with international data. For instance, Bults and colleagues [4] reported fear of side effects as a primary reason for declining to vaccinate their child against H1N1 (see also [5,34]). Our findings in relation to demographic predictors of pandemic vaccine uptake demonstrated variable consistency with the UK H1N1 literature. For instance, our finding concerning the relationship between age and vaccination was consistent with findings that younger individuals would be more likely to accept the pandemic vaccine if offered it (16)(17)(18)(19)(20)(21)(22)(23)(24) year olds vs. 65+ year olds, [30]; [15][16][17][18][19][20][21][22][23][24] year olds vs. all other ages, [39]), but inconsistent with other research (e.g., [23] found no significant effect of age on intentions to vaccinate). Similarly, our finding that previous influenza vaccination was positively associated with vaccination against H1N1 was consistent with some existing UK literature [15,31], but not others (Myers and Goodwin [23] found no relationship between previous seasonal vaccination and pandemic vaccination intentions).
Although our findings regarding chronic illness were consistent with existing literature indicating that individuals with a chronic illness are more likely to get vaccinated or accept the vaccine if offered (e.g., Han et al. [15]), we did not replicate any significant effects of gender (i.e., that men were more likely to accept or intend to accept the vaccine, [15,39]) or ethnicity (although previous findings in the UK literature have been inconsistent, e.g., [7,30]) on adult self-vaccination decisions/intentions. These inconsistent effects are not without precedence: in their systematic review of factors associated with H1N1 vaccine uptake (including several of the papers cited above), Bish and colleagues report mixed evidence for several demographic factors (i.e., age, socioeconomic factors, and ethnicity) and vaccine uptake [2]. Further work is therefore needed to fully understand the relationship between demographic characteristics and uptake of pandemic influenza vaccination.
Given the consistency between the attitudinal factors associated with UK H1N1 self-/parental vaccination and the literature, there is clear scope to use the collective findings to strengthen the UK's plans concerning the communication and roll out of future a Some categories combined due to small numbers. b Unable to estimate RR due to no variation in the outcome for unvaccinated individuals (i.e. perfect prediction). 3 Although Painter and colleagues did observe a significant bivariate relationship between perceived susceptibility of one's child to H1N1 (i.e., fear of children getting swine flu) and willingness to allow the child to be vaccinated, this relationship was rendered nonsignificant in multivariate analyses [26].
pandemic influenza vaccination. Although there is limited evidence as to the efficacy of community education campaigns when used alone, combining these campaign with other interventions (e.g., client reminders, expanded clinic hours, home visits) does have a demonstrable effect on vaccine coverage [3,38]. Similarly, although there is little evidence regarding the effectiveness of parental vaccination campaigns, further research to develop interventions targeting parental perceptions regarding vaccination is recommended [32]. Given this, and on the basis of both our findings and the extant literature, we suggest that future development of pandemic influenza plans should consider the potential for multicomponent interventions that specifically target self-and parental perceptions of vaccination effectiveness/safety as well as perceived risk of oneself/one's child contracting influenza to increase uptake of a pandemic influenza vaccination.

Limitations
Although our analysis does, to the best of our knowledge, represent the first attempt to explore predictors of both self-and parental vaccination decisions among the general population during the UK H1N1 pandemic, there are some methodological limitations to consider. First, the Flu Watch cohort as a whole was not representative of the English population. Specifically, this cohort underrepresents young adults, non-white ethnic groups, individuals who are socially deprived, and those living in the North, the West Midlands, and London [11]. In addition, the pandemic cohort survey participants contained a relatively small parental vaccination sample. As a result, we were unable to conduct more detailed, multivariate analysis using this data. Caution is therefore advised when generalising the results reported herein to the UK population. We are also aware that worry about swine flu varied (although not greatly) over the course of the pandemic, with some decline by January 2010 (see Fig. 1 in [30]). As the pandemic cohort data was collected after this decline (in Spring 2010), we are unable to explore this effects of this potential attitudinal change on vaccine uptake within our dataset. We therefore recommend both: a) that all existing data concerning UK parental vaccination attitudes/decisions during the H1N1 pandemic be analysed and published, and; b) that attempts to collect data during any future influenza pandemic ensure greater coverage of parental attitudes towards vaccination and parental vaccination decisions across the duration of the outbreak.
In addition, although we did identify several significant predictors of vaccination, none of these individually accounted for more than 23% of variance in behaviour (prior adult vaccination). Despite this, the relationships observed between vaccination uptake and perceived risk, vaccine safety, and vaccine efficacy were broadly consistent not only with the existing UK H1N1 vaccination literature [15,23,30,37], but also with the role of perceived risk and efficacy in several health behaviour models (e.g., Health Belief Model, [28,29]; Theory of Planned Behaviour, [1]). We are therefore confident in the theoretical significance of our findings.
Finally, the international literature suggests several potential predictors of self-and parental vaccination that were not captured in the Flu Watch pandemic cohort survey. For instance, the source of information concerning pandemic influenza (e.g., official sources, [2]; national rather than local news, [18]) may be associated with self-and parental vaccination. Moreover, perceiving vaccination to be socially normative (i.e., recommended by individuals that are important to you) is associated with both self-and parental vaccine uptake (e.g., [2,20]). Further research should therefore focus on assessing a wide range of predictors of both self-and parental-vaccination during any future UK pandemic outbreak in order to explain as much variance in vaccine uptake as possible.

Conclusions
To the best of our knowledge, the current paper represents the first attempt to explicitly consider predictors of both adult self-vaccination and parental vaccination among the UK general population during the H1N1 pandemic. Broadly, our results were consistent with the extant literature concerning UK vaccination decisions/intentions during the H1N1 pandemic. Our central findings suggest that concerns over the efficacy and safety of the vaccine as well as concerns regarding the perceived risk of pandemic influenza are critical determinants in both self-vaccination and parental-vaccination. These findings could be incorporated into the development of future interventions designed to improve both self and parental vaccine uptake among the UK population in the event of a future pandemic influenza outbreak.

Conflicts of interest
None.