Effectiveness and parental acceptability of social-networking interventions for promoting seasonal influenza vaccination among young children: a randomized control trial (Preprint)


 BACKGROUND
 Seasonal influenza vaccination (SIV) coverage among young children remains low worldwide. Mobile social-networking applications such as WhatsApp Messenger are promising tools for health interventions.
 
 
 OBJECTIVE
 This was a preliminary study to test the effectiveness and parental acceptability of social-networking intervention that sends weekly vaccination reminders and encourages exchange of SIV-related views and experience among mothers via WhatsApp discussion groups for promoting childhood SIV. The second objective was to examine the effect on mothers’ decision-making for childhood SIV of additionally introducing time pressure for vaccination decision-making using countdowns of the recommended vaccination timing.
 
 
 METHODS
 Mothers of child(ren) aged 6-72 months were randomly allocated to Control, or one of two social-networking Intervention groups receiving vaccination reminders with (SNI/+TP) or without (SNI/-TP) a time pressure component via WhatsApp discussion groups at a ratio of 5:2:2. All participants first completed a baseline assessment. Both the SNI/-TP and SNI+TP groups subsequently received respective weekly vaccination reminders in Oct-Dec 2017, and participated in WhatsApp discussions about SIV moderated by a health professional. All participants completed a follow-up assessment in April-May 2018.
 
 
 RESULTS
 A total of 174 (84.9%, 174/205), 57 (71.3%, 57/80) and 60 (75.0%, 60/80) who were allocated into the Control, SNI/-TP and SNI/+TP, respectively, completed the outcome assessment. The social-networking intervention significantly promoted mothers’ self-efficacy for taking children for SIV (SNI/-TP: OR=2.69 (1.07-6.79); SNI/+TP: OR=2.50 (1.13-5.55)), but did not result in significantly improved children’s SIV uptake. Moreover, after adjusting for mothers’ working status, introducing additional time pressure reduced the overall SIV uptake in children of working mothers (OR=0.27 (0.10-0.77)) but significantly increased the SIV uptake among children of mothers without a full-time job (OR=6.53 (1.87-22.82)). Most participants’ WhatsApp posts were about sharing experience/views (52.1%, 226/434) of which 44.7% (101/226) were categorized as negative such as their concerns over vaccine safety/side effects and effectiveness. Though participants shared predominantly negative experience/views about SIV at the beginning of the discussion, the moderator was able to encourage the discussion of more positive experience/views and more knowledge/information. Most Intervention group participants indicated willingness to receive the same interventions (94.0%, 110/117) and recommend the interventions to other mothers (87.2%, 102/117) in future.
 
 
 CONCLUSIONS
 Online information support can effectively promote mothers’ self-efficacy for taking children for SIV but may not alone be sufficient to address maternal concerns over SIV to achieve a positive vaccination decision. However, the active involvement of health professionals in online discussions can shape positive discussions about vaccination. Time pressure on decision-making interacts with maternal work status, facilitating vaccination uptake among mothers who may have more free time job but having the opposite effect among busier working mothers.
 
 
 CLINICALTRIAL
 HKUCTR-2250, www.HKUCTR.com



Table of Contents
Objectives: This was a preliminary study to test the effectiveness and parental acceptability of a social-networking intervention that sends weekly vaccination reminders and encourages exchange of SIV-related views and experience among mothers via WhatsApp discussion groups for promoting childhood SIV. The study also examined the effect on mothers' decision-making for childhood SIV of additionally introducing time pressure for vaccination decision-making using the declining number of days remaining until the recommended optimal vaccination window closed. Conclusion: Online information support can effectively promote mothers' self-efficacy in taking their child(ren) for SIV but may not alone be sufficient to overcome maternal concerns about SIV and produce a positive vaccination decision. However, the active involvement of health professionals in online discussions can shape positive discussions about vaccination. Time pressure on decisionmaking interacts with maternal work status, facilitating vaccination uptake among mothers who may have more free time but having the opposite effect among busier working mothers.

Introduction
Seasonal influenza creates a substantial annual global disease burden. Young children are the most vulnerable age group [1,2], having higher viral loads and shedding virus for a longer period than adults and thereby are also important influenza viruses vectors to other household members [3].
Seasonal influenza vaccination (SIV) for children is therefore regarded as the most important measure to protect both children and the wider population [4] but uptake rates remain low in many countries [5][6][7]. In Hong Kong, children aged six months-12 years receive a subsidy under the Childhood Influenza Vaccination Subsidy Scheme (CIVSS) for SIV from private-sector general practitioners. This policy removes financial barriers by making the vaccine completely free for the parents of target children, though some GPs demand an additional small administration fee. Despite CIVSS, SIV uptake among young children in Hong Kong languishes around 30% [8,9]. Finding ways to improve SIV uptake thus remains crucial to reducing community influenza spread.
Sending vaccination reminders through mobile phone-based short message services (SMS) has been shown to promote vaccination uptakes, including for routine immunization and SIV in children [10][11][12][13] but reported effect sizes were small. A systematic review found that participants generally complaint that mobile phone SMS reminders were limited by formats and character set [14]. The proliferation of mobile messaging apps and smartphone use have made mobile messaging functions more flexible compared with traditional SMS. In Hong Kong, WhatsApp messenger is used by over 80% of the population [15], through the high penetration of smartphone use [16]. In addition to providing flexible messaging functions like message structure, formats and length, WhatsApp also permits social networking functions through creating multi-member online discussion groups.
Existing vaccination reminders for promoting childhood SIV uptake have usually contained information on influenza infection risks and SIV benefits [13,17,18] , key variables in cognitive theories of behaviour change [19]. However, studies suggest that people inflate risk from vaccination relative to risk from natural infection possibly due to biased media coverage of vaccine risk [20] or omission bias, the tendency to believe that an error of omission is less serious than that from commission [21]. Therefore, merely providing information on influenza infection risks and influenza vaccination benefits may be insufficient to overcome concerns over vaccine-related risks, an important impediment to SIV uptake [8]. According to dual-processing models, information is not processed systematically and deliberatively but is widely influenced by heuristic cues that require less effort to reach a quick and efficient decision [22,23], particularly when participants feel uncertain and lack cognitive resources such as time and energy to make a decision. Previous studies suggest that parental decision-making for children's vaccination is extensively modified by knowing other parents' vaccination decisions, indicating a strong social normative influence [8,24]. Others' behaviour provides important behavioural cues for social learning or imitation by indicating social approval, relieving safety concerns and increasing confidence in specific choices [8,24]. Therefore, knowing that other parents take their child for SIV can encourage hesitant parents to do the same.
This knowledge and experience sharing becomes more practical with messaging apps that enable social networking functions. However, few studies have examined the potential for socialnetworking interventions to promote parental decisions about SIV for their children.
Studies in behavioural economics and neuroscience have suggested that introducing time pressure in decision making could increase decision-makers' reliance on heuristic cues for decision making, mainly through the mechanisms of acceleration (i.e., switching to simpler strategies to speed up decision making) and selectivity (i.e., automatically omitting certain information and favoring certain information) [25][26][27]. It is also suggested that while individuals can efficiently integrate different cues to reach an optimal decision under some time pressure, those under high time pressure can only use limited cues that are more salient for them (e.g., heavily relying on negative cues) when making decisions [25,28]. Furthermore, time pressure may induce different affective states depending on individual capability to cope with the time limit and their cognitive load [26,27]. For individuals who perceived being able to make a decision within time limit and have more cognitive resource to perform the decision task, time pressure could make them energetic and active in seeking risk reduction strategies. Otherwise, time pressure may induce stress that subsequently lead to more reliance on anecdotal cues rather than statistical information in decision making and thereby impair their final decision [26,27,29]. Therefore, whether introducing time pressure can promote vaccination uptake or not may depend on how parents perceived the time pressure introduced in vaccination decision. Hong Kong runs an annual influenza vaccination campaign (October-December) that recommends parents obtain SIV for their 6 months-12 years-old children at least two weeks before the winter influenza season (January-March) allowing sufficient time for the body to produce antibodies following vaccination. Therefore, the recommended optimal SIV window starts from October till two weeks before the end of December annually, and as the winter influenza season approaches the optimal window diminishes, making vaccination decision-making for parents naturally time-constrained. This provides an opportunity to test the effect of introducing time pressure to parental SIV decisions.
This preliminary study tested the effectiveness and parental acceptability of social-networking interventions through the use of WhatsApp discussion groups for promoting children's SIV uptake in the context of Hong Kong. The specific objectives of this study were to:

Participants, group allocation and baseline assessment
Since mothers in Hong Kong are the primary decision-makers or significantly contribute to decisionmaking with fathers for children's immunization [8], this study only targeted mothers with at least one child aged 6-72 months to avoid confounding by gender effects. Other inclusion criteria were (i) Chinese communication fluency, (ii) having a Hong Kong network-connected smartphone with internet access, and (iii) having installed or being willing to install WhatsApp on their mobile phone.
These inclusion criteria were intended to limit subjects to be primary of Chinese ethnicity who comprise ~93% of the Hong Kong population to further minimize confounding by culture and language effects. Subjects were excluded if their eligible children had medical contraindications for immunization. Subjects were recruited before the 2017/18 CIVSS campaign started and excluded if their target child(ren) had already received SIV for the 2017/18 season. Eligible subjects were identified and recruited from previous samples of population-based random-dialled household telephone surveys and community outreach conducted by a commercial polling company previously used for successful population-based surveys [8,31]. All potential subjects were screened in a short telephone interview to confirm eligibility and obtain oral consent for study participation. Each consenting subject was later called by a part-time telephone interviewer for a ~10-min telephone baseline assessment interview. The baseline assessment collected data on both participants' and their children's SIV history, and other socio-demographic characteristics, participants' intention to take children for SIV during the 2017/18 CIVSS campaign, and baseline risk perceptions regarding childhood influenza and influenza vaccination. Before each telephone interview, the interviewer opened a sealed envelope which contained a random allocation sequence generated by computer to determine the subject's group allocation. Subjects who were allocated to an intervention group were notified that they would be participating in a WhatsApp discussion group during the intervention period to receive weekly vaccination reminders and share their views and experiences about SIV with other mothers and a group moderator. This being a preliminary study to test the effectiveness of social-networking interventions for promoting childhood SIV uptake, we aimed to recruit 200 subjects for the Control and 80 subjects for each of the two intervention groups, allowing for a 30% dropout rate in each group, to detect ~ 20% increase in vaccination uptake among the socialnetworking Intervention groups relative to the Control with a power of 80% and 95% confidence interval. To balance confounding between study arms and control group size, blocked randomization [32] was used to allocate participants to one of the three arms, using a ratio of 5:2:2 for group allocation. Neither participants nor part-time interviewers performing subject recruitment and allocation could be blinded to subject allocation but the interviewers who conducted baseline assessment were blind to the interventions arm (with or without time pressure) participants occupied.
The assessor of the primary outcome was blinded to all participant group allocation.

Vaccination reminders
The vaccination reminder comprised three messages. Message 1 introduced the CIVSS and doctors' recommendations for children's SIV; Message 2 addressed children's risk of seasonal influenza, and benefits and safety of SIV for children; while Message 3 addressed the number of days remaining for the recommended vaccination timing (days remaining from the date when the vaccination reminder was sent out to the date two weeks before the winter influenza season). While the vaccination reminders for SNI-TP contained Message 1&2, Message 3 (the time pressure component) was additionally included in the vaccination reminders for SNI+TP participants. All messages were constructed using information from the official websites of Hong Kong Centre for Health Protection and World Health Organization, and local published studies [33][34][35], and delivered in graphical format through WhatsApp. The messages contained mainly textual information but graphical information was also incorporated to represent some key themes (e.g., doctor's recommendation, eligibility of CIVSS and days remaining for optimal SIV window) and efficacy of SIV, aiming to improve audience's comprehension, and their attention and interest to read [36,37]. All messages were pre-tested using think-aloud interviews covering 10 eligible mothers to ensure their readability via a mobile phone and comprehensibility without inducing negative feelings. Multimedia Appendix 1 gives the finalized messages in both the Chinese and English but only the Chinese version was used in the intervention. Weekly vaccination reminders were assumed to be effective without increasing respondents' information load with a preference for receiving vaccination reminders during afternoon [14]. Therefore, vaccination reminders were sent to the intervention groups midafternoon on different weekdays, weekly over the CIVSS campaign period from October-December 2017. The first vaccination reminder was delivered two weeks after the CIVSS started and the last one delivered on 18 th December 2017, two weeks before the winter influenza season began. Overall, a total of eight vaccination reminders were delivered to the intervention groups over the eight-week intervention period.

WhatsApp discussion groups
In addition to delivering weekly vaccination reminders, the WhatsApp discussion group was also set up to provide positive peers support for mothers to make better-informed SIV decisions regarding their children. To control group size and facilitate group discussion, participants who were allocated to the intervention groups were then randomly allocated to one of the two SNI-TP and two SNI+TP WhatsApp discussion groups, each comprising ~40 mothers. In each WhatsApp discussion group, mothers could post their opinions and concerns about influenza and SIV and freely communicate with other mothers and the group moderator about their experiences of personal and child's influenza vaccinations. The project moderator monitored and facilitated the group discussions on a daily basis following standardized guidelines (Multimedia Appendix 2). In addition to delivering weekly vaccination reminders via WhatsApp discussion groups, the moderator also sent one additional message on a weekly basis to enforce exchange of positive views and experience about SIV. The moderator also addressed any questions, concerns or misunderstandings raised about influenza and influenza vaccination, if these were not first addressed by other mothers within the groups. Posting content irrelevant to influenza and influenza vaccination was discouraged. Participation rules were set and delivered in the discussion group immediately after the group was created. Participants were informed that those violating the participation rules, such as using offensive statements and harassment would be expelled from the discussion group. All members participating in the WhatsApp discussion group were encouraged to use Chinese for communication. Voice messages were discouraged and members were advised not to disclose names and other personal information to protect privacy. The WhatsApp discussion groups were closed by the project moderator two weeks after the last vaccination reminder was sent out.

Outcome assessment
In April-May 2018 after the winter influenza season, all participants were again contacted to report information on their children's SIV uptake before and during the 2017/18 influenza season. For participants who had more than one child eligible to the CIVSS, the vaccination status of each eligible child was recorded. Mother's intention to take their children for SIV in the next 12 months was also recorded. Risk perceptions regarding seasonal influenza and SIV for children were assessed again to examine whether any changes in perceptions occurred after the interventions. Participants' opinions about the interventions and their willingness to receive vaccination reminders via WhatsApp in the future were asked to assess the acceptability of the interventions. In addition, a total of 20 participants from the intervention groups were contracted in May-July 2018 for in-depth interviews to explore their opinions about interventions and the acceptability of using WhatsApp for promoting children's health. Figure 1 illustrates the study procedure and timing.

Data analysis
Pearson chi-square test was first conducted to compare participants' demographics, baseline perceptions, history of influenza vaccination and their target child's characteristics by intervention arm to assess randomization, and by follow-up status to assess selection bias.

Assessment of primary outcomes
Next, children's SIV uptake rate in 2017/18 was calculated for each group and compared between groups using Pearson-chi square test. Both the SIV uptake of all target children aged between 6-72 months and that of the youngest target child's SIV were compared across groups, because among families with more than one target child, the youngest one tends to be not vaccinated [38]. The intervention effect on children's SIV uptake was also examined by stratifying the analyses by participants' educational attainment, work status and household income to identify potential sociodemographic effect modifiers previously reported to be associated with parental acceptance of influenza vaccination for their children [39][40][41]. To further assess the effects of the interventions on vaccination uptake, a generalized estimating equations (GEE) logistic regression model was GEE can accommodate cases with missing outcome measures at some time points (cases with outcome measure at one time point will be counted) and the correlation between the outcome measures at different time points (i.e. the baseline and follow-up SIV uptake) [42]. Potential effect modifiers (e.g., participants' demographics) identified in the univariate analysis would be additional included into the GEE to test its interaction effects with both the time and intervention on the outcome. In the GEE analysis, participants' youngest target child's SIV status during the follow-up period was used as the outcome. Since the final SIV uptake of the target child(ren) of participants who dropped out at follow-up was unavailable, intention-to-treat (ITT) analysis was used as a conservative and sensitivity analysis by treating the lost outcomes as "not vaccinated" over the specific CIVSS campaign to compare with the complete-case analysis.

Assessment of the secondary outcomes
Excepting for effects on children's SIV uptake, the intervention effects on parental perceptions regarding influenza and SIV by intervention arm were also assessed using chi-square and similar GEE logistic regression modelling. All WhatsApp groups' posts were archived by the project moderator immediately before the WhatsApp discussion groups were closed. The mean number of posts per participant was calculated while the distributions of participants' frequency of posting across discussion groups were compared using Kruskal-Wallis equality-of-populations rank test. All discussion posts were examined to further explore participants' responses to the vaccination reminders, their perceptions and attitudes regarding influenza and influenza vaccination, and how they interacted with peers and the group moderator during the communication process. All posts were analyzed and coded by two researchers independently using content analysis. Each post was coded for the following categories: role (moderator or participant), format (text, picture, emoji or hyperlink), cyber-support (e.g., sharing views or experience and emotional exchange) and discussion topics (e.g., vaccine effectiveness, vaccine safety and side effects). More than one code could be assigned to one post. A coding scheme for cyber-support and discussion topics was first drafted and developed by the first author based on literature on online psychosocial support [43,44] and parental decision making for childhood influenza vaccination and vaccination attitudes [8,24], and refined throughout data analysis and the discussion of the research team. Then the refined coding scheme was used in QRS NVIVO 12.0 by the first author and a trained research assistant to independently code all the posts again. The interrater agreement between the two coders was assessed using the Cohen's Kappa (k) with a k of less than 0.6 indicating low agreement, which was then resolved by joint discussion between the two coders. How the moderator's involvement in the WhatsApp discussion could change the discussion direction about SIV among participants was also analyzed by plotting the time sequence of cyber-support behaviours of participants and the moderator in each discussion group. Parental acceptability of the intervention was first assessed by describing participants' opinions about the interventions and their willingness to receive vaccination reminders via WhatsApp in the future. In addition, thematic coding was conducted to identify themes and categories relating to parental acceptability of the interventions and using WhatsApp Messenger for child health promotion emerging from the in-depth interviews.
All quantitative data were analyzed using STATA 15.1 (StataCorp LLC, 1985-2017) while the textual data were analyzed using NVIVO 12.0 (QSR International, 2018) were more likely to drop out from the outcome assessment than were the Control (χ 2 2 =8.0, P=.02) but those who completed the baseline assessment and the outcome assessment did not differ by intervention condition in terms of their demographics, their target child's characteristics, past SIV uptake, baseline SIV perceptions and intention to take child for SIV (Appendix Table 1

Intervention effects on the target child's SIV uptake
The youngest target child SIV uptake rates were 37.9% (66/174), 33.3% (19/57) and 38.3% (23/60) in the Control, SIN-TP and SNI+TP, respectively. Chi-square test indicated that the interventions did not have significant effects on either the youngest target child's SIV uptake or all target child(ren)'s SIV uptake (Table 2). It also shows that the youngest child's SIV uptake appeared to be greater in SNI+TP for participants who did not have a full-time job (χ 2 2 =5.31, P=.07), suggesting that participants' work status may be a potential effect modifier (Table 2 and Multimedia Appendix 4).

Intervention effects on participants' perceptions of influenza and SIV
GEE were also conducted to examine whether change in participants' SIV perceptions from the baseline to the follow-up differed by intervention condition. The results showed that there were significant intervention effects on the change of participants' perceived self-efficacy in taking children for SIV, with participants of the SNI-TP (OR 2.69, 95% CI 1.07-6.79) and SNI+TP (OR 2.50, 95%CI 1.13-5.55) reporting more increase in confidence in taking their children for SIV than did the Control participants ( Figure 3 and Appendix Table 2 of Multimedia Appendix 3).  Table 3  and Emoji were also used (Appendix Table 3 of Multimedia Appendix 3). All relevant participants' and moderator's posts excluding the weekly vaccination reminders were coded for themes and categories relevant to cyber-support and discussion topics.

Discussion topics
The main discussion topics among participants' posts are shown in Table 5. The most common participant discussion topics were vaccination decisions followed by vaccination clinic and cost, vaccine safety and side effects and vaccine effectiveness ( These concerns or views about SIV seem to mostly reflect beliefs that SIV could weaken immunity, distrust about how the vaccine strain was estimated every year and a perception that vaccination is not a 'natural' process. Vaccination experience is distinguished from vaccination decision or plan because it mainly refers to participants' feeling about the vaccination process (e.g., injection pain) or after vaccination (more or fewer illnesses). Most participants' posts on medical eligibility of SIV and first-time influenza vaccination belonged to seeking information/opinions.  "Although there is mismatch, the vaccine is still effective for preventing influenza H1N1 or influenza B viruses"  "It (flu vaccination) is an additional protection for our children." Being mixed or neutral/purely seeking information about vaccine effectiveness (n=15, 29.4% (15/51))  "Is it true that one can still get a cold even after taking the vaccination but can protect against influenza?"  "Can influenza vaccination protect one against serious complications due to influenza?" Medical eligibility of SIV (n=40, 9.2% (40/434))  "I thought to take my daughter for flu vaccination today but she has a running nose and some cough. Is it OK for her to take flu vaccination?" Vaccination experience (n=33, 7.6% (33/434)) Positive (n=16, 48.5% (16/33))  "My child has taken the flu vaccination and he still feels very good now." Negative (n=12, 36.4% (12/33))  "My elder daughter took the flu vaccination for once but got more and severe sicknesses that year. Since then, she has never taken flu vaccination…" Mixed or uncertain (n=5, 15.2% (5/33))  "My two sons have taken the flu vaccination. One is three years old. He was given injection at the hip and he said no pain. Another is seven years old. He was given injection at arm. He said it was very painful and the pain lasted for two days." Doubtful or negative vaccination attitudes (n=26, 6.0% (26/434))  "Vaccination is to inject germs into the body"  "Is it necessary to take flu vaccination if my child is always healthy?"  "Too many vaccinations are not good for children"

Interactions between participants and the moderator during online discussion
To illustrate the change of participants' cyber-support behaviours as the moderator involved in the online discussion, participants' cyber-support behaviours were categorized into three types based on their potential effects on SIV uptake: positive cyber-support behaviours comprising sharing positive experience/views, sharing knowledge/information and positive emotional exchange; negative cybersupport behaviours comprising sharing negative experience/views and negative emotional exchange; and mixed/neutral cyber-support behaviours comprising sharing mixed/neutral experience/views, seeking information/opinions, and other emotional exchange. Figure 3 shows that although participants mainly shared their negative experience/views/emotions (blue bars) regarding SIV at the beginning of the online discussion, with the moderator's involvement throughout the discussion, the numbers of posts sharing positive experience/views, sharing knowledge/information and positive emotional exchange (red bars) increased. However, the discussion dynamic also indicates a less active participation in the discussion among the participants as the discussion proceeded.

Parental acceptability of the intervention
Of the 117 participants of the intervention groups who completed the outcome assessment, 115 (98.3%) reported reading the discussion posts at least several times a week during the intervention period and 105 (89.7%) had read more than one half of all discussion posts. Over 80% (95/117) indicated no concern over participating in the WhatsApp discussion groups. Of those expressing concerns, the most common concern was receiving misinformation or irrelevant information. Most Post-hoc qualitative interviews with 20 participants of the intervention groups were analyzed to clarify participants' in-depth opinions about the interventions (Appendix Table 4

Principal findings
This social-networking intervention, involving sending weekly vaccination reminders and encouraging exchanges of positive experiences and information among participants via WhatsApp discussion groups during an influenza vaccination campaign, did not significantly enhance children's SIV uptake. Two main reasons may explain why a significant effect of sending regular vaccination reminders was not identified. First, compared with previous studies that used vaccination reminders to promote routine childhood immunization [11,12], our study focused on promoting an optional vaccine, childhood SIV; parents have more risk-related concerns about optional vaccines [24]. Our qualitative data indicated that although the positive attributes of information from the moderator were appreciated by most participants, the information provided mainly improved knowledge, motivated contemplation and increased vaccination motivation. For participants who had already made the decision to take their children for SIV before joining in the discussion group, the information may prompt vaccination planning or be used as cues for taking action. For participants who had anti-vaccination attitudes or were hesitant to take SIV, the information was insufficient to change the psychological roots of the anti-vaccination attitudes [45] or remove concerns over vaccine risk and thereby cannot support a final decision for, or action on children's SIV. Second, compared with studies that found a positive effect of sending regular vaccination reminders for promoting influenza vaccination [10,13,17,18], vaccination reminders were delivered by a health professional researcher (the moderator) rather than a general practitioner in the primary care team who had access to the target children's medical records. Therefore, although information from the moderator was perceived by participants to be trustworthy, it may have been perceived as less relevant to children's healthcare, compared with information received directly from a general practitioner and thereby had less impact on parental SIV decision-making. However, excepting for children with chronic conditions, most parents and their children may not frequently interact with a primary care team. Therefore, although this reflects one potential weakness of our study, this may be more representative to a real public health scenario for promoting childhood SIV. Other studies suggest that even the healthcare providers' position on vaccine safety is being increasingly questioned by parents [46,47] involvement in vaccination discussions can create a more positive online experience. The internet has become probably the main information sources shaping negative parental attitudes around childhood immunization [48][49][50]. Active communication from health professionals may be sufficiently effective to combat vaccine hesitancy compared to attempts to control online media misinformation [51,52].
Despite not increasing SIV uptake among the target children, the social-networking intervention was significantly effective for promoting mothers' self-efficacy in taking their children for SIV. This is possibly due to the frequent posts of information about the vaccination clinics and cost that were shared by both moderator and participants through the online discussion. Previous studies also have found that online information support significantly increased parents' perceived self-efficacy in other child healthcare practices [53][54][55][56] and that peer experience-based information may be more likely to meet their information needs [57,58]. As parents' perceived self-efficacy for taking children for SIV is a significant predictor for children's SIV uptake [8], this is likely to facilitate future childhood SIV uptake. However, the discrepancy between the enhanced parental self-efficacy in taking child for SIV and the unchanged SIV uptake indicates that the direct effect of perceived self-efficacy on vaccination uptake is weak [8]. Enhanced self-efficacy should combine with positive vaccination attitudes to promote positive vaccination decision. However, the moderator was found to be the main source of knowledge and information about vaccine safety/side effect and effectiveness, while participants generally felt a lack of confidence in sharing their personal knowledge, particularly when there was a health professional (the moderator) in the group. Because experience-based knowledge and information from peers may be more powerful and persuasive for changing parents' attitudes [57,58], future studies should focus on how to encourage peers to share positive experience-based knowledge and information about vaccine safety/side effects and effectiveness for promoting childhood vaccination.
Including an additional time pressure did not significantly enhance childhood SIV uptake. However, subgroup analysis showed that children's SIV uptake significantly increased among mothers without a full-time job while declining slightly among mothers with a full-time job when the time pressure intervention was included. The qualitative data indicated that time pressure pushed participants to make a rapid decision, but those decisions can be either positive or negative. Unemployed and parttime-employed mothers may have more cognitive resource to deliberate the pros and cons of influenza vaccination and perceive that they have the ability to make the decision within time limit.
Therefore, under some time pressure, they may become more active in searching information to reduce the risk of influenza and efficiently integrate different cues to reach a positive vaccination decision. In comparison, working mothers face more pressure from work for childcare [41] and thereby tend to have more concerns over disruptive vaccination side-effects (proximal cost) than the risk of influenza (distal cost). Working mothers may also place more weight on the value of time taken from work to seek vaccination for their children [41] and thereby the negative cues that favor inaction (not vaccinate the child) may become more salient for them. As working mothers may have less cognitive resource to decide whether to take their children for SIV, the time pressure is likely to induce stress in decision making. Therefore, time pressure may enforce the influence of negative cues (e.g., side-effects of influenza vaccination) on the vaccination decisions among working mothers.

Limitations
This study had several main limitations. First, we only recruited participants who were users of WhatsApp or those who were willing to install WhatsApp in their mobile phone and thereby the sample may not be representative for the target population though the penetration rate of WhatsApp usage was very high in the population. Since almost all participants reported using WhatsApp on a daily basis, the data did not have sufficient variance to allow for examining the intervention effects stratified by WhatsApp usage. Second, a discussion group specifically for "influenza vaccination" may dissuade those uninterested in the topic, causing in-group biases. However, our analysis did not find significant differences in participants' demographics, perceptions of SIV, and SIV history and intention across intervention arms. Third, this was a preliminary study to test social-networking interventions effects on SIV uptake and as such the sample size was insufficient for detecting a small effect size. Fourth, data on children's SIV uptake were reported by parents and could not be validated from children's medical records and may be subject to social desirability bias. To reduce social desirability bias, participants were reassured about the anonymity of the study and the confidentiality of their data. Fifth, the WhatsApp discussion groups, out-of-office-hour discussions were not promptly monitored and addressed. The time lag in addressing participants' questions or concerns may have affected participants' subsequent participation in discussions and thereby SIV decision making. However, it is difficult to determine optimal moderator input in the WhatsApp discussion given the discussion group tried to encourage mutual support between participants. Furthermore, the infrequent emotional exchange among participants also indicated insufficient development of attachment to and friendships between group members, which could be a reason for why around half of the participants were 'lurkers', silent and passive members in the WhatsApp discussion. This represents to be a big challenge for the sustainability of online discussion. Future studies need to examine how to encourage information support from peers, moderate their emotional interactions and the optimized moderator participation.

Conclusion
The social-networking intervention for mothers was ineffective for increasing SIV uptake among young children but did effectively increase mothers' perceived self-efficacy for taking their children   Table 1 Comparison of demographics, the target child's characteristics, and history of SIV, baseline perceptions regarding childhood SIV and intention to take the target child for SIV among participants who completed the baseline and outcome assessments Appendix Table 2 Assessment of the intervention effects on parental perceptions using generalized estimating equations logistic regression Appendix Table 3 Characteristics of the posts from participants and the moderator in the WhatsApp discussion groups Appendix Table 4 Major themes and quotes from the qualitative interview on parental acceptability for the intervention using the WhatsApp discussion groups Multimedia Appendix 4: Appendix Figure 1 Change of child's SIV uptake by intervention condition and participants' working status