Acceptability of intranasal live attenuated influenza vaccine, influenza knowledge and vaccine intent in The Gambia

Highlights • Mothers of children who received LAIV preferred intranasal vaccines over injections.• The LAIV was viewed as safe, easy to give and less distressing than injections.• Influenza knowledge was higher in those with more education.• Influenza vaccine intent was very high, but not associated with influenza knowledge.• Based on acceptability, introducing influenza vaccination would be feasible here.


Introduction
Influenza is a major cause of acute respiratory infection globally, leading to a significant burden of morbidity and mortality [1]. Seasonal influenza has been well-studied in high-and middle-income countries, but neglected in Africa [2]. The rate of influenza-associated hospitalisations in children <5 years of age is approximately 3-fold higher in low-and middle-income countries (LMIC) compared to high-income countries [3]. One meta-analysis found that 99% of deaths attributable to influenzaassociated acute lower respiratory infections occurred in LMICs [4].
As a result of the increasing awareness of influenza-related disease in LMIC, in 2012 the WHO recommended that countries should consider influenza vaccination in high-risk groups including pregnant women and children <5 years [5]. The Gambia has no influenza vaccination policy and is yet to include influenza vaccination in the Expanded Programme on Immunisation (EPI) [6,7]. The NASIMMUNE study, a systems immunology research study in children aged 24 LAIV and offers a unique opportunity to study, for the first time, attitudes towards a new mode of vaccine delivery in this setting.
Factors that shape parental views on vaccination impact childhood vaccination rates and having insight into parental perceptions can be useful when introducing new vaccinations to a country [8]. High vaccine uptake in The Gambia through the EPI [9] has led to reductions in invasive disease secondary to Haemophilus influenzae type b [10] and Streptococcus pneumoniae [11]. Yet as more vaccines are added to the EPI, vaccine hesitancy might be a concern [12][13][14]. Knowledge and attitudes surrounding influenza and influenza vaccination in The Gambia are unknown. Gaining a better understanding of these factors can help guide future influenza vaccination strategies, including maternal influenza immunisation, shown to be beneficial to both mothers and infants in recent studies from South Africa and Mali [15,16].
The Health Belief Model (HBM) provides a framework of psychosocial constructs that may be determinants of health behaviour [17]. It posits that preventative health behaviours are more likely to be exhibited when an individual perceives they are highly susceptible to the disease, that the disease is serious, that the behaviour is beneficial, there are few barriers, and are cued into action [18]. These constructs have been shown to predict uptake of influenza vaccination and other health behaviours [19][20][21][22][23][24]. However, emotions experienced at the point of decision-making may be as important in health behaviour as a cognitive assessment of the risk, such as worry and anticipated regret, which have also been shown to be predictive of influenza vaccination [20,25,26]. Few studies have used the HBM to study influenza vaccine behaviour in LMICs [27][28][29][30][31].
We aimed to compare attitudes towards the safety, ease of use, and tolerability of LAIV between mothers with direct experience of their children receiving LAIV and mothers without such experience, and to establish whether their attitudes towards intranasal vaccines differ. Additionally, we aimed to quantify knowledge about influenza and to determine possible relationships between influenza knowledge, socio-demographic factors and willingness to accept influenza vaccination in pregnancy or for their children <5 (vaccine intent). Finally, we aimed to explore the relationship between health belief constructs and influenza vaccine intent in this cohort.

Study design
A cross-sectional survey was conducted in women !18 years at two sites (Sukuta and Faji Kunda health centres) five kilometres apart, in the coastal region of The Gambia in August and September 2017. In Sukuta, where NASIMMUNE was conducted, mothers of children who had been vaccinated with the intranasal LAIV (Nasovac-S Ò , Serum Institute of India Pvt. Ltd.) as part of the study were contacted sequentially and invited to an interview, up to a total of 150 (exposure group). A further 154 women from Sukuta who had not participated in the NASIMMUNE study were recruited at the same health centre. Due to wider community sensitization regarding the NASIMMUNE trial carried out in Sukuta, this group may have had some exposure to information about influenza and LAIV, therefore an additional control group of 150 women were recruited at Faji Kunda health centre (where there could have been no exposure to the study or community sensitization). These women were recruited through opportunistic sampling when attending for routine healthcare. The sample size was determined by the availability of mothers of children in the NASIMMUNE study (n = 168), with a 1:2 exposure to control ratio. Inclusion criteria were having at least one child <5, maternal age !18, resident in the area, and fluency in Mandinka (the most commonly spoken local language in the areas).

Data collection
Face-to-face interviews were conducted by trained fieldworkers who entered data in real-time into a questionnaire designed in REDCap TM [32]. The questionnaire was developed using simple terminology and was refined through cognitive pre-testing with field-workers and test participants. The questionnaire included seven sections: inclusion criteria, socio-demographics, vaccine intent, influenza knowledge, health-seeking behaviour, health beliefs, and LAIV acceptability. The influenza knowledge questions were designed de novo for this setting, selected and simplified from influenza knowledge questions used previously [31,[33][34][35] and refined following the cognitive pre-testing phase. Questions asked about symptoms, transmission, prevention, treatment, risk, and vaccination (see supplementary material: Appendix A). The responses were collated to form a score out of 15 points, which was converted to a percentage for analysis (score divided by 15, multiplied by 100).
Influenza vaccine intent was assessed for pregnancy and for children <5 by asking participants to respond to two statements: ''If I was pregnant, I would get a flu vaccine if it was free" and ''I would get a flu vaccine for my child under 5, every year, if it was free". Answers were recorded on a 4-point scale where 1 = agree strongly, 2 = agree, 3 = disagree and 4 = disagree strongly, with a ''don't know" option. In follow-up questions, unprompted reasons given for answers were coded into predefined categories by the interviewer.
Health belief constructs were assessed using statements answered on the same 4-point scale. HBM constructs assessed were: perceived susceptibility, severity, benefit, barriers and cues to action. Two additional concepts were also included: worry and anticipated regret. The answers given were converted into binary ''agree" or ''disagree" responses for analysis.

Ethical considerations
Ethical approval for the study was provided by The Gambia Government/MRC Joint Ethics Committee (SCC1555). Written informed consent was obtained from all participants. It was made clear that answers were confidential and anonymised, that they could withdraw at any time or decline to answer any questions.

Statistical analysis
Analysis was conducted using Stata Ò 12.0. Descriptive statistics were used to compare proportions between groups: Pearson's chisquared test or two-tailed Fisher's exact test (when one category had <5 participants) for categorical data; and Student's t-test or Wilcoxon rank-sum test for normally and non-normally distributed continuous data respectively. Univariate linear regression analysis was performed for predictors of influenza knowledge. Significance-testing was used for selection of variables to include in the multivariate model at a level of p < 0.2 to minimize type II error in selection [36]. The Cochran-Mantel-Haenszel test was used to analyse variations between groups for individual influenza knowledge question responses and to evaluate the difference in preference for intranasal or injection vaccinations between groups. Two-tailed Fisher's exact test was used to analyse perceptions of intranasal LAIV in the exposed group and associations between vaccine intent and health belief constructs. p < 0.05 was considered statistically significant.

Participants' characteristics
The 454 participants' characteristics who answered the survey are displayed in Table 1. There were significant differences between the exposure and control groups with respect to age, parity, education, husband's education and monthly household income.

LAIV acceptability
When asked ''If given a choice for your child between a flu vaccine injection and a nasal spray, which would you prefer?" a signif-icantly higher proportion of the exposure group stated they would prefer a nasal spray compared to the control group (93.3% vs. 34.9%, OR = 26.15, p < 0.0001, Table 2). The most common unprompted reasons for preferring the nasal spray were that it is easier to give, less painful and a perception of greater effectiveness (Fig. 1A). The most commonly stated reasons for preferring injections were a greater familiarity with injections and a belief that injections are more effective (Fig. 1B). In the exposure group, stating a preference for the nasal spray was associated with finding the LAIV less distressing (p < 0.001), safer or equally safe (p < 0.001) and easier or equally easy to give (p < 0.001) when compared to injections ( Table 2). Table 1 Socio-demographic and other characteristics of participants by group.

Influenza knowledge
The mean percentage influenza knowledge score was 68.0% and was not significantly different between exposure and control groups (69.2% vs. 67.4%, p = 0.0816), although differences were seen in individual questions (Fig. 2). Participants recruited at Faji Kunda had significantly higher scores than at Sukuta (70.4% vs. 66.8%, p = 0.0005). When comparing exposure and control participants from Sukuta, the knowledge was significantly higher in the exposure group (69.2% vs. 64.4%, p = 0.0001). Of the control participants from Sukuta, only two (1.3%) had been asked to participate in NASIMMUNE but did not.
The responses to individual knowledge questions are shown in Fig. 2. A significantly higher proportion of exposure participants could name all three core symptoms of influenza (11.3% vs. 3.0%, OR = 4.19, p = 0.001), while a significantly lower proportion of exposure group participants correctly identified that there is no cure for influenza (11.3% vs. 23.0%, OR = 0.43, p = 0.004) but thought ''medication" could cure influenza (86.7% vs. 71.1%, OR = 2.65, p < 0.0001). There were no significant differences between groups in any other questions.
Results of the univariate and multivariate linear regression analysis of factors that predict influenza knowledge are presented in Table 3. In the multivariate linear regression model, there was a significantly higher level of influenza knowledge in participants who had attended Arabic school, upper school or had higher education compared to no education, exhibiting an increased score with each stepwise increase in educational level. Additionally, participants whose husbands were students showed a significantly higher knowledge score than those whose husbands had no occupation (b = 13.31, p = 0.032). The multivariate model explained a significant proportion of the variance in influenza knowledge (adjusted R 2 = 0.1165, p < 0.0001).

Influenza vaccination intent
Almost all respondents stated that they agreed or agreed strongly with the vaccine intent statements (98.5% and 98.7% for pregnancy and children <5 respectively) and there was no difference in vaccine intent between groups (Table 1). Unprompted reasons given for intent are shown in Fig. 1C. There was no association between influenza knowledge and vaccine intent for pregnancy or children <5 as shown in Table 3 (p = 0.782 and p = 0.817 respectively).
The associations between health belief concepts and vaccine intent are presented in Table 4. For children <5, significant  associations were seen for perceived susceptibility, severity, benefit, and barriers, while for pregnancy, only perceived susceptibility, severity and worry were significantly associated with vaccine intent.

Discussion
There are few countries with seasonal influenza vaccination in sub-Saharan Africa, with none currently using intranasal LAIV [37]. The NASIMMUNE study, ongoing in The Gambia, provided an opportunity to study perceptions of intranasal LAIV between mothers whose children had recently received LAIV and women with no experience of LAIV. In the latter group, preference for an intranasal vaccine was moderate at 34.9%, with the most commonly stated reasons for preferring an injection relating to familiarity with injections and beliefs of greater effectiveness. Our results show that in those with direct experience of LAIV, the preference for nasal spray was significantly higher, in keeping with studies in high income countries [38,39]. Women in the exposure group stated, unprompted, that nasal sprays were easier to give than injections and that they were less painful, suggesting that the LAIV given in NASIMMUNE was viewed favourably. These results indicate that a future influenza vaccination programme in The Gambia using intranasal LAIV in children <5 would be received positively, particularly if the introduction were coupled with demonstrations or educational sessions to promote their safety, ease of use and effectiveness.
The influenza knowledge questions covered a range of influenza-related topics including symptoms, transmission, severity, high-risk groups, treatment and vaccination. The educational level in The Gambia is generally low, with many women illiterate, and never attending school (21.6%). The mean knowledge scores were no different between the exposure and control groups, but a good understanding of disease and health-risk concepts was seen overall despite the low formal education level. Due to community sensitization and possible knowledge transfer between NASIMMUNE study participants and non-study participants in Sukuta, a control group from Faji Kunda who had had no exposure to any trial information about influenza were included. Surprisingly, Faji Kunda had significantly higher knowledge than Sukuta overall. The effect was independent of education, which was broadly similar between sites, although knowledge did increase significantly with more education at both sites. The reason for the higher scores in Faji Kunda is unclear, but health beliefs within in the two communities may differ significantly and previous studies conducted at Faji Kunda may have influenced the results. Several large vaccine trials have recently recruited children through the Faji Kunda EPI clinic, which have included health, hygiene and nutritional education as part of community sensitization. One study from The Gambia has shown that such activities effectively disseminate information throughout the community [40], while others have shown that community social networks are important to knowledge, particularly when access to external information is limited [41]. Furthermore, educational women's groups can significantly improve health outcomes in low-income settings [42,43]. Community health beliefs may, therefore, be as important as formal education in regard to influenza-related knowledge.
When comparing groups from Sukuta alone, influenza knowledge was significantly higher in the exposure group, suggesting that participation in the NASIMMUNE study may have increased knowledge about influenza and vaccination. However, when asked what can cure influenza, a significantly higher proportion of the exposure group incorrectly answered ''medication", rather than ''nothing/no treatment". This could indicate that involvement in the study led people to erroneously believe that specific antiinfluenza medications are available. Children presenting with cough, fever and rhinorrhoea in the follow-up period of the NASIM-MUNE study were commonly given paracetamol, which may have been mistakenly regarded as a ''cure", possibly explaining the finding. This is a reminder that clinical trials can inadvertently spread misinformation in these settings and that steps should be taken to mitigate this risk through careful provision of information during informed consent processes and throughout studies [44][45][46][47].
Influenza vaccine intent was assessed by asking participants if they would accept an influenza vaccine if it was freely available for themselves during pregnancy or for their children <5. Over 98% of participants responded that they would accept the vaccination. The high intent could be explained by the clinical trials that have been conducted at both health centres for many years, which may have led to a high level of trust in healthcare in these communities. To avoid response bias, participants were reassured during the interview that they should not feel pressured to answer in the affirmative. Nonetheless, the high influenza vaccine intent observed may be, in part, due to acquiescence bias or social  desirability bias. However, the unprompted stated reasons for their answers confirmed that there is a strong belief that vaccines are safe and effective, and related to vaccines generally rather than influenza vaccines specifically. The high intent is also consistent with the high EPI coverage in The Gambia (95% for BCG and 81-99% for three Diphtheria-Pertussis-Tetanus doses) which is among the highest in Africa [9]. Neither higher influenza knowledge nor higher educational level were associated with vaccine intent. This may be due to the small numbers disagreeing with the intent statements, which also meant that the associations between health belief constructs and vaccine intent could not be quantified, but our results confirm previous findings elsewhere [19,20]. A larger sample size would be required to determine which health belief constructs are most predictive of vaccine intent in this setting.
A key limitation to this study is that opportunistic sampling was used. The sample may therefore not be representative of the wider community, which could explain the small observed differences in socio-demographic makeup between groups. The higher educational level and income observed in the exposure group may be due to more educated and affluent people being more willing to participate in clinical trials. Future studies should use probability sampling at more sites to better represent the knowledge, attitudes and perceptions towards influenza and vaccination in the country. Due to the high vaccine intent seen, the study was also underpowered to explore associations between intent and knowledge, so these should be interpreted cautiously. The intent questions specified that the vaccine was offered for free, so vaccine intent if the vaccine was only available for a fee remains unknown. Additionally, prior to being asked about LAIV preference over injections, they were not informed about the effectiveness of each vaccine, which may have altered their answers. Future studies could address these limitations in similar surveys, or use qualitative research methods to explore these attitudes in more depth, to gain an understanding of the underlying beliefs and motivations behind vaccine behaviour in The Gambia.

Conclusion
Willingness to undertake influenza vaccination during pregnancy, or to get annual seasonal vaccination for children <5 years if freely available was high, as was acceptability of LAIV in those with first-hand experience. Incorporation of the intranasal influenza vaccine into the childhood immunisation schedule in The Gambia in the future would likely be feasible from an acceptability perspective. Knowledge and understanding of health-related concepts surrounding influenza was generally good, though varied between communities, and was significantly related to higher educational levels. Despite reasonable health knowledge in this lowincome, low-education setting, more formal education would have a positive impact on influenza and health knowledge, and potentially have wider community benefits as well. Community-based educational interventions may also be beneficial in The Gambia.