Parental reminder, recall and educational interventions to improve early childhood immunisation uptake: A systematic review and meta-analysis

Vaccination is one of the most effective ways of reducing childhood mortality. Despite global uptake of childhood vaccinations increasing, rates remain sub-optimal, meaning that vaccine-preventable diseases still pose a public health risk. A range of interventions to promote vaccine uptake have been developed, although this range has not speciﬁcally been reviewed in early childhood. We conducted a systematic review and meta-analysis of parental interventions to improve early childhood (0–5 years) vaccine uptake. Twenty-eight controlled studies contributed to six separate meta-analyses evaluating aspects of parental reminders and education. All interventions were to some extent effective, although ﬁndings were generally heterogeneous and random effects models were estimated. Receiving both postal and telephone reminders was the most effective reminder-based intervention (RD = 0.1132; 95% CI = 0.033–0.193). Sub-group analyses suggested that educational interventions were more effective in low- and middle-income countries (RD = 0.13; 95% CI = 0.05–0.22) and when conducted through discussion (RD = 0.12; 95% CI = 0.02–0.21). Current evidence most supports the use of postal reminders as part of the standard management of childhood immunisations. Parents at high risk of non- compliance may beneﬁt from recall strategies and/or discussion-based forums, however further research is needed to assess the appropriateness of these strategies.


Introduction
The reduction in global mortality associated with vaccinations is second only to the introduction of safe drinking water [1]. According to the World Health Organisation, childhood vaccinations prevent an estimated 2-3 million deaths per year. Yet despite global increases in childhood vaccine uptake, rates remain suboptimal (<95%), with vaccine-preventable diseases still posing a public health risk [2]. Neither is this risk limited to low-and middleincome countries (LMICs). Factors such as poor access to healthcare, indigenous or ethnic status, a large family size and low educational achievement are associated with pockets of low coverage in high-income countries (HICs) [3].
Maintaining reductions in mortality from vaccine-preventable disease relies upon continued immunisation uptake that, during childhood, is reliant on parental decision-making and subsequent attendance at vaccine clinics [4]. However, several factors may act as barriers to childhood immunisation. Factors include parental concerns about vaccine safety, a lack of knowledge about the recommended schedule, pain caused by the injections, distrust of the medical community and difficulty accessing clinics [5]. Therefore, it is important to understand the effectiveness of interventions implemented by primary care settings that are designed to improve childhood immunisation. Interventions to increase childhood immunisation have been targeted at a variety of groups, including healthcare providers, healthcare practices and parents [6]. This review will focus on the effectiveness of interventions targeted at parents. Many strategies have been trialled, including financial incentives [7] and home vaccination [8]. However, as the majority of trials have addressed (a) the lack of schedule awareness using parental reminder systems and/or (b) knowledge about the safety and efficacy of vaccines through educational leaflets or discussion-groups, these interventions will be the primary focus of this review. Systems designed to remind parents that their child was due (reminder) or overdue (recall) their immunisations have been linked to a 1.5 times increase in uptake [9]. The effects of parental education are less clear, http with evidence presented both for [10,11] and against [3] their utility.
Previous reviews have focussed on the efficacy of intervention strategies in isolation and not all have made specific recommendations regarding childhood immunisations. Today, primary health care services are under increasing pressure to meet immunisation expectations at both an organisational and patient level [12]. In order to facilitate physician judgements about interventions to increase childhood immunisation, and to increase the efficacy of intervention implementation and policy updates, a review comparing the effectiveness of multiple interventions to be compared is timely. Therefore, a systematic review and meta-analysis was conducted to evaluate available evidence on parental interventions to improve childhood (birth to 5 years) vaccine uptake.

Literature search
A systematic literature search of five databases (MEDLINE, EMBASE, EMBAR, CINAHL and PsychINFO) was conducted in February 2014 using the OVID and EBSCOhost search platforms (with adaptation of terms for EBSCOhost). Search terms were predefined to allow a comprehensive search strategy that included text fields within records and Medical Subject Headings (MeSH terms). Terms related to immunisation, immunisation uptake, infants and young children and intervention study design. The OVID search strategy is reported in Table 1. This search was conducted as part of a wider review of barriers and facilitators of childhood immunisation and so included both qualitative and quantitative data. The present review refers only to quantitative intervention studies.

Study selection
Database search results were combined and duplicates were removed. Studies were screened for eligibility by the primary author, with uncertain citations discussed with J.M. Full-text reports were gained for all eligible studies. The reference lists of included studies were additionally searched for any relevant articles. A sample of studies was independently assessed for eligibility by J.M. to corroborate study selection. Any disagreements were resolved by discussion. Studies were eligible for inclusion in the systematic review if they reported interventions aimed at parents of children (≤5 years-old) due or overdue one or more routine immunisations, recommended to be administered by WHO, with outcomes that measured child immunisation uptake. Because of variations in the reporting of immunisation uptake [3] outcomes that addressed the uptake of individual or a combination of recommended vaccines were included. Studies without a control group and studies that did not provide outcome data in terms of the number of children completely immunised or up-to-date for their age were excluded from the meta-analysis. Interventions that met these criteria but for which only one study was found were also excluded from pooled analyses.

Data extraction and assessment of methodological quality
Study characteristics were recorded using a pre-defined data extraction sheet. Information was extracted on (a) study design, (b) country of study, (c) intervention (including type, population, setting, details and sample sizes), (d) outcomes (including the number of children completely immunised for their age, received at least one dose of the studies vaccine(s), or were vaccinated on-time), (e) study findings and (f) eligibility for inclusion within meta-analyses.

Risk of bias in individual studies
Risk of bias was performed by the primary author using the Cochrane Collaboration Risk of Bias Tool [13]. Studies were assessed as being at a high, low or unclear risk of six attributes: sequence allocation, allocation concealment, blinding, incomplete outcome data, selective reporting, and other sources of bias. Studies were assessed as 'unclear' when an attribute (e.g., blinding) was not or insufficient evidence to support a judgement was provided. Evidence of quality across studies was determined by the proportion of studies given each judgement for each methodological attribute assessed in the tool. Although assessment of study quality is reported here it was not used to weigh review findings.

Data analysis
Studies that were eligible for inclusion in the meta-analysis were grouped according to intervention type. Separate metaanalyses were conducted for each intervention type. Studies examining multiple interventions could contribute to several analyses. Where trials had a cluster randomised design, reported intra-cluster correlation coefficient (ICC) were sought. If ICCs were not reported, unadjusted values were included in the metaanalyses, accepting that this might overestimate the weight of these studies in the analysis. Risk difference values and 95% confidence intervals were used to calculate both individual and pooled effect sizes for the effect of each intervention on complete childhood immunisation uptake. Potential differences between studies were explored by sub-group analyses including where possible, the effect of the country of study income, time, frequency and method of intervention delivery and focus of intervention content.
Heterogeneity was assessed using Cochrane's Q statistic, with p < .10 denoting heterogeneity. Inconsistency across studies was measured using the I 2 statistic, with a value greater than 40% presenting evidence of moderate heterogeneity and signalling the need to use a random effects model [13]. Where heterogeneity was not reduced by sub-group analyses, variability in study method was discussed. Evidence of publication bias was investigated by examining the symmetry of the funnel plot and quantified using the Egger statistic, with p < .05 denoting evidence of publication bias. All analyses were performed using StatsDirect [14].

Selection of studies
The literature search generated 1577 articles. Following the removal of duplicates, 1040 of the remaining 1078 articles did not meet the inclusion criteria based on an appraisal of the abstract. This resulted in 86 full-text articles, which were examined in depth. Forty additional articles were identified from the reference lists of eligible papers and eight systematic reviews identified in the database search [3,6,[9][10][11][15][16][17]. One hundred and twenty-six full text reports were examined, and 48 qualitative studies removed for later qualitative analysis. Based on the criteria cited above, 32 intervention studies were ineligible, leaving 46 articles suitable for inclusion in the systematic review. Of these, a further 13 articles were excluded because of inadequate study designs and outcomes measures, and 5 [8,[18][19][20][21] because of a lack of comparable trials, leaving 28 articles suitable for meta-analysis (Fig. 1). Table 2 summarises the characteristics of studies included in the systematic review and meta-analysis. Of the studies included in the meta-analysis (n = 28), 16 studies were conducted in the United States, 5 in the UK or Republic of Ireland, 2 in Pakistan and 1 each in Australia, Ghana, India and Japan. Twenty-four randomised controlled trials (RCTs), three cluster RCTs and one sequentially allocation control trial were included. One cluster RCT [22] reported an ICC of zero. Consequently, no adjustment was made for clustering and clustering had no impact upon any findings reported.

Characteristics of studies included in the meta-analysis
The studies included a total of 14,936 parent-child dyads whose immunisation uptake was assessed. Eight studies had data on the complete uptake of both DTP and Measles vaccines; 12 on DTP; 5 on MMR; and 1 each on DTP and OPV; Hib, HBV and PCV7; and DTP and HepB.

Risk of bias for individual studies
Using the risk of bias tool, 12 studies (43%) were judged to have an overall high risk of bias, 4 (14%) as low risk, and 12 studies (43%) as unclear risk. Risk of bias judgements for studies included in the meta-analysis are shown in Fig. 2. Nine studies were judged to be at a high risk of selection bias, describing a non-random component in the sequence generation process [38,41], inappropriate allocation concealment [37,42], or both [26,30,31,44]. The blinding of parents and/or health professionals was not possible where interventions were provided face-to-face. Only six studies included a blind outcome assessor. In the majority of studies, blinding was unclear or judged to be of high risk because those administering the intervention also assessed outcomes [22,30,31,[36][37][38]40,44,48]. In the majority of studies, insufficient information was reported to provide a judgement regarding blinding. Approximately 10% of studies [36,38,46] were judged to be at high risk of attrition bias owing to high rates of exclusion and loss to follow-up. Only three studies [22,40,42] referred to protocols, so for the majority of studies it was unclear whether selected reporting had been an issue. No evidence of publication bias was found for included studies: for each pooled analysis, funnel plots were symmetrical, with studies published in each quarter of the plot. The associated Egger statistics were non-significant in each case.

Effectiveness of reminder-based interventions
Thirteen studies evaluated the impact of one or more methods of parental reminder. Pooled risk differences were calculated for the effect of postal and telephone reminders, as well as studies that utilised both methods in one study arm.

Telephone
Five studies (1 sequentially allocated control trial, 4 RCTs) examined the efficacy of telephone reminders [25,26,29,32,34]. Telephone calls were made and/or messages were left with parents to remind them that their child's immunisations were due or overdue. As heterogeneity was minimal, a fixed effects model was used. Receiving a telephone reminder (Fig. 3b) was associated with a significant 4% increase in immunisation uptake (RD = 0.040; 95% CI = 0.006-0.073, p = .019). One large study [34] dominated the analysis. Heterogeneity of findings was low and thus a random effects model produced similar findings.

Combined postal and telephone (recall and reminder)
Four studies (1 sequentially allocated control trial, 3 RCTs) assessed the impact of receiving a postal reminder letter and telephone prompt on childhood immunisation uptake [25,26,32,35]. The fixed effect model found that the receipt of both postal and telephone reminders (Fig. 3c) was associated with a significant 10.6% improvement in immunisation uptake compared with controls (RD = 0.106; 95% CI = 0.070-0.143, p < .001). Substantial heterogeneity was found between the studies, with individual study findings ranging from a 3.2% to 18.9% increase in uptake, meaning the size of the effect cannot be accurately determined. Nevertheless, use of combined reminders remained significant using a random effects model (RD = 0.113; 95% CI = 0.033-0.193, p < .006).
One study [26] was methodologically different from others in the group because it used a combination of postal and telephone reminders to inform parents of their child's appointment before they were due, whereas the remaining studies used one method to inform parents of their child's appointment details (recall), and only used another method if children remained unimmunised after 1 week (reminder). Excluding this trial reduced heterogeneity (I 2 = 31.6%), while the overall effect remained similar (RD = 0.147; 95% CI = 0.10-0.195, p < .001).

Effectiveness of education-based interventions
Seventeen studies evaluated the impact of parental education. Pooled risk differences were calculated for the effect of educational interventions, education and reminder and the support of a Lay Health Worker (LHW).

Immunisation education
Ten studies (2 cluster RCTs; 8 RCTs) examined the effect of providing parents with immunisation-based education [22,30,31,[36][37][38][39][40][41][42]. Parents were advised about immunisation or general child health before their child's immunisation appointment. Education was facilitated by a discussion with a trained professional, or by written information in picture card or leaflet format. One study provided parents with the details of how to access several written educational sources. Two studies included intervention groups who received education at different time points; intervention groups are pooled for these trials [36,39]. The overall fixed effect suggests that parental education significantly improved immunisation uptake by 8.3% (Fig. 4a; RD = 0.083; 95% CI = 0.056-0.110, p < .001). However, findings from individual studies ranged from a 1.6% decrease to 26% increase in immunisation uptake and substantial heterogeneity was found in the data. A random effects model reported a similar average effect of intervention (RD = 0.078; 95% CI = 0.013-0.142, p = .018).

Immunisation education and postal reminders
Five RCTs examined the efficacy of interventions that provided parents with some form of immunisation education in addition to a postal reminder [30,31,[43][44][45]. The fixed effect model found parental education and postal reminders (Fig. 4b) led to a 16% increase in uptake (RD = 0.16; 95% CI = 0.12-0.19, p < .001), although with substantial heterogeneity. Individual study findings ranged from 1% to 26% improvements, with the two largest studies reporting the greatest effect, being conducted in Pakistan [30,31]. A positive effect of education and postal reminders remained, using a random effects model (RD = 0.13; 95% CI = 0.01-0.25, p = .04).

Support from Lay Health Workers (LHWs)
Four studies (1 cluster RCT, 3 RCTs) examined the impact of parental education about immunisation and advice from a LHW [46][47][48][49]. For the purposes of this review, LHWs were defined as a health worker providing education about immunisation, but who had not received any formal healthcare training. LHWs comprised of volunteer mothers, O-level graduates and community workers. A significant effect of LHWs (Fig. 4c) was found using the fixed effects model (RD = 0.10; 95% CI = 0.05-0.15, p < .001). However, individual study findings were mixed (ranging from a decrease of 3% to an increase of 20%) and substantial heterogeneity was found. A random effects model did not reach statistical significance (RD = 0.11; 95% CI = −0.02 to 0.25, p = .09). Sub-group analyses accounting for the specificity of LHW advice found that specific immunisation advice [46,47] was associated with a significant 17% increase in immunisation uptake (RD = 0.17; 95% CI = 0.10-0.24, p < .001).

Discussion
There is evidence to support the efficacy of postal and/or telephone reminders, parental education, and parental education with postal reminders for improving child immunisation uptake.
Reminder-based interventions were significantly more effective than routine care independent of their method of delivery. This finding is comparable to that of a previous Cochrane review that found that reminder systems were efficacious for immunisation uptake across the lifespan [9]. The present review however, conducted separate meta-analyses for individual reminder strategies specific to childhood immunisations and found that postal reminders were more effective than telephone reminders. Hence postal reminders are recommended for use in primary care to improve childhood vaccine uptake. Moreover, postal and telephone reminders had an additive impact on uptake; their combined use was associated with a greater increase in immunisation uptake than the use of each strategy alone. However, this effect could be an artefact of the more intensive recall-reminder strategies used in these trials and suggests that recall strategies may be particularly effective in parents whose children may be at risk of non-attendance. There is a need for future research to explore the efficacy of this intervention in trials comparing children at high and low risk of non-compliance.
The overall group analysis suggested that educational interventions significantly increased childhood immunisation uptake. However, sub-group analyses suggested that this effect was driven by two factors: (a) the study occurring in an LMIC and (b) parents having a discussion with a professional expert, rather than receiving information in written form. Analysis did not suggest that the timing or intensity of education impacted upon its effectiveness. The baseline education levels of the participants enrolled in included studies may explain the increased efficacy of interventions conducted in LMICs. Approximately 50% of mothers enrolled in studies within this comparison were illiterate or had no education. Secondary levels of maternal education have been associated with a two-fold increase in childhood immunisation compared to mothers with no education [63]. Interventions that raise the basic level of parental knowledge are therefore more effective in areas where understanding is low compared to countries where it is comparatively higher and educational barriers to immunisation may be more subtle and linked to vaccine belief [2,4]. Contrary to a previous meta-analysis [11,64], the overall effectiveness of LHWs could not be recommended following the application of a random effects model [20,34,[46][47][48][49]. Sub-group analyses did suggest that parents who received specific vaccine support from an LHW might be more effective than general support that did not extend beyond topics covered by health visitors in routine care, suggesting an avenue for future research.
Keys to the efficacy of discussion-based educational interventions may come from qualitative findings that suggest that discussion with a trusted medical practitioner may facilitate immunisation compliance owing to the depth and clarity of understanding gained compared to the reported overwhelming nature of written information leaflets [65][66][67]. These findings suggest that providing parents with the opportunity to discuss immunisations in detail with a healthcare professional may further facilitate immunisation rates. However, due to the additional human resources needed to incorporate practitioner-lead discussion within primary care settings, policy planners may be mindful to reserve these strategies for vaccine-hesitant parents.
The overall utility of educational strategies within standard practice may be further questioned when examined alongside the results of trials that provided parents with both immunisation education and appointment reminders. Using the same methods in both rural and urban settings, the two Usman et al. trials [30,31] examined education and reminder strategies in separate and combined study arms. In both communities, improvements of uptake in groups who received the combined intervention were minimal in comparison to postal reminders alone. This finding has implications for policy as it suggests that reminder systems may be sufficient facilitators of childhood immunisation in the majority of cases, and that discussion-based strategies may be most effective in families with children at high-risk of non-compliance. Such strategies may increase compliance because they acknowledge parental concerns about vaccination. Addressing these concerns in a discussion with a medical professional regarding the risks and benefits of vaccination may change the parental attitudes, knowledge and beliefs about vaccination. Changes in attitude may facilitate behaviour change; facilitating a pro-vaccination decision further facilitates subsequent vaccine uptake. However, the effectiveness of such strategies must be tested in future trials.

Strengths and limitations
Whilst the findings of this review help to summarise the large body of literature on parental interventions for childhood immunisation uptake, several limitations were apparent. First, substantial heterogeneity was evident in all of the comparisons except telephone reminders. Although random effects models were utilised to investigate the mean distribution of underlying intervention effects, such models do not identify reasons for variation. Unexplained heterogeneity suggests there may be some differences in study method and/or services provided that explain why interventions were effective in some cases but not in others. Although there were too few studies within the published research sampled to explore this formally, the effect of differences in study context can be illustrated by the two Usman et al. studies [30,31]. These studies utilised the same method of allocation, intervention strategy and outcome measure but were conducted in urban and rural areas of Pakistan respectively. However, they found a 12.8% difference in immunisation uptake of parents who received postal reminders in favour of rural communities, suggesting factors not reported such as access to, and interaction with, healthcare services may contribute to the success of interventions. Owing to the limits of journal space, the authors were unable to provide additional detail regarding risk of bias judgements for individual studies and excluded studies however; such details are available on request.
Variation between studies may also explain the discrepancy between the results of the present study and two previous reviews on LHWs [11,65]. The present review examined the effects of education provided by LHWs and therefore did not include two [20,34] out of four studies [20,34,46,48] previously analysed. Instead, two additional studies [47,49] were included. This difference and the high levels of heterogeneity found between studies in each review may explain this disagreement and further exaggerates the limited conclusions that can be made by presently available studies in the field.
Second, many of the strategies used in reminder-based interventions may not be relevant to the parents of today. Mobile phones are owned by large majorities of people living in major countries around the world, 75% of whom use their mobile phones for texting [68]. For example, 96% of mobile phone owners in Indonesia, a country where less than 50% of children do not receive the three doses of DTP [2], use their mobile phones to text. The increasing commonality of access to technology including mobile phones and text messaging may offer increased effectiveness in terms of outcome and cost. Several small-scale studies have linked text message reminders to improvements in both adolescent [69,70], and childhood [71] immunisations. However, larger RCTs are necessary to make firmer conclusions about the efficacy of such interventions.

Conclusions
The findings of this review suggest that several interventions, particularly postal reminders, combined recall and reminder strategies and discussion-based education, can increase childhood immunisation uptake. The precise effectiveness of these interventions is likely to be influenced by numerous factors such as country of intervention and levels parental vaccine hesitancy that need to be explored by future trials. This review highlights the potential benefits to childhood vaccine uptake of incorporating parental interventions, particularly postal reminders into the standard management of childhood immunisations, and the use of recall strategies and/or discussion-based forums with parents whose children are at high risk of non-compliance.

Role of the funding source
This work was supported by a studentship from the Economic and Social Research Council awarded to the primary author. The funding source had no involvement in the design, analysis or interpretation of the findings presented in this report.