Comparison of school based and supplemental vaccination strategies in the delivery of vaccines to 5-19 year olds in Africa - a systematic review

Search strategy

A search was carried out to identify all relevant studies. Both published and unpublished literatures were searched up to March 30, 2017. No restriction was placed on the publication language. The following electronic databases were searched using both medical subject headings (MeSH) and free text terms relating to vaccination, children, adolescents and Africa (Supplementary Table 1): PubMed, Africa Wide, Cochrane Central Register of Controlled trials (CENTRAL), Cumulative Index to Nursing and Allied Health Literature (CINAHL), World Health Organization Library Information System (WHOLIS), Web of Science, PDQ (Pretty Darn Quick)-Evidence and Scopus. The following grey literature databases were searched for reports, non-peer reviewed and non-indexed papers: WHO, The Global Vaccine Alliance (GAVI) and UNICEF. The reference lists of the included publications were evaluated to identify other potential studies.

Study selection

The following criteria was used to select primary studies for inclusion:

(1) the study was either a randomised controlled trial (RCTs), non-RCT, cluster-RCT, interrupted time series, controlled before-and-after, cohort, cross-sectional or case-control;

(2) participants were school aged children or adolescents (5–19 years) living in Africa;

(3) SIAs or SBVs were the vaccination strategies under investigation;

(4) vaccination coverage, cost of vaccine delivery or effects of either strategy (SIAs or SBVs) on routine health services including EPI were reported as any of the outcomes. Retrieved articles were independently screened by two reviewers (HEC and LA). Where study eligibility was unclear, a review was carried out by a third independent study team member (BK).

Data extraction

HEC and LA independently reviewed each included study and extracted data using a piloted data extraction form. Where discrepancies arose HEC, LA and BK reached a consensus by discussion. Corresponding authors were contacted for any missing data needed during the extraction process.

Quality assessment of included studies

Experimental studies were assessed using the Cochrane Collaboration’s tool for assessing risk of bias²⁶, while the Hoy et al., modified tool was used for cross-sectional studies²⁷. Using the Cochrane checklist, studies were scored as ‘High risk’ or ‘Low risk’. Using the Hoy et al., checklist, cross-sectional studies were scored as High risk’ or ‘Moderate risk’ or ‘Low risk’ of bias. Studies with high risk of bias were rated as poor quality while studies with moderate and low risk of bias were rated as moderate and high quality respectively. Where discrepancies in quality assessment occurred, HEC and LA discussed to arrive at a consensus. The quality of studies reporting the cost effectiveness was not assessed since our main interest was only the delivery costs.

Data synthesis and analyses

Data was analysed using Stata v. 14.0. Results from the studies reporting vaccination coverage were expressed as percentages. Reported costs of the delivery strategies were standardised to United States Dollars (USD) if reported in a different currency. The costs of the strategies (SBVs or SIAs) and effects of the strategy on routine vaccination were presented in a narrative form.

A meta-analysis for vaccination coverage using a random effects model with inverse variance proportion was carried out. Pooled statistics for vaccination coverage were expressed as proportions with 95% confidence interval (95% CI). Subgroup analyses were carried to evaluate vaccination coverage per strategy, stratified by study settings or type of vaccine. Missing values were imputed in order to enable a complete case analysis. A sensitivity analysis was then carried out where imputations were done to assess if the results significantly differed due to the data imputations.

Visual inspection of the forest plots, the Chi-squared test for homogeneity with a significance level set at 10% and the I² statistic were used to assess and quantify any heterogeneity between study results. Heterogeneity was rated as ‘low’ for ≤ 49% and ‘moderate’ for 50–74% and ‘high’ for ≥ 75% using the I² statistic.

Literature search

Three thousand seven hundred and nineteen (3719) studies were identified through searching the electronic databases. A further 1461 were identified from grey literature. An additional five studies were identified from the reference lists of all the search outputs. After duplicates were removed, 4938 studies were left for screening. The titles and abstracts of the 4938 studies were screened and 4872 were not relevant and therefore excluded. The full text of the remaining 65 were retrieved and assessed for eligibility. From the 65 studies, 31 met our inclusion criteria (Supplementary File 2).

Characteristics of included studies

A total of 31 studies were included in this review. There were 20 cross-sectional studies^23,28–46, eight economic evaluation studies^47–54, one cluster-randomised trial⁵⁵, one epidemiological report⁵⁶ and one interrupted time series⁵⁷. The included studies were published between 1993 and 2016; only three studies were published before 2000^45,46,56. A total of 17 African countries (Figure 1) and five different vaccines were represented from the included studies. Except for four of the included studies that were written in French^44,46,50,53, the rest were in English. One of the study team members (HEC) is French literate and translated the four articles. In terms of vaccine delivery strategy, 20 and 11 studies assessed SIAs and SBVs respectively. Out of these 32 studies, 20 reported on vaccination coverage (Table 1 and Table 2), nine on the cost of the vaccination strategy (Table 3) and three on the effect on routine immunisation (Table 4).

Figure 1. Countries and vaccine delivery strategies represented by the 31 included studies.

Risk of bias and quality assessment

Using the Hoy et al., modified tool²⁷, a 10 item scale was used to assess the internal and external validity of the 20 cross-sectional studies. Ninety-five percent (19) of the studies were of high quality (low risk of bias) meaning further research is unlikely to change our confidence in the estimate of the study outcomes. Five percent (1) of the studies were of moderate quality (moderate risk of bias) meaning further research is likely to have an impact on our confidence in the estimate of the outcomes (Figure 2b). For the internal validity, the included studies defined which participants were considered to have been vaccinated (by self-report or vaccination card) and used the same data collection tool for all the participants. However, five studies did not mention if the tool used was standardised^{29,34,42,45,46}. Ten studies collected information from proxies (parents or guardians of vaccinated children)^{29–32,34,36,43–46}. All the studies calculated vaccination coverage as the ‘number vaccinated divided by the number of the targeted population’. For the external validity, all the studies had representative samples in terms of age and sex. Random sampling was used in all except four studies^33,35,37,39. Similarly, majority of the studies had a low non-response rate except three studies^33,40,46.

The Cochrane checklist was used to assess the clustered-randomised trial⁵⁵ and interrupted time series⁵⁷ study (Figure 2a).

Figure 2.

a. Risk of bias for experimental studies. b. Risk of bias for cross-sectional studies.

Vaccination coverage

SIAs: Twelve studies reported vaccination coverage for SIAs. Three studies reported coverage data on meningococcal serogroup A (PsA-TT) vaccine^29,30,44, five on measles vaccine^{31,32,34,56,57} and two on yellow fever vaccine^36,43. Each of the remaining two studies reported on one vaccine; cholera³⁸ and meningococcal bivalent polysaccharide A/C vaccine⁴⁶. Vaccination coverage for SIAs ranged from 48.5% to 98% (Table 1).

A meta-analysis of a pooled estimate showed high vaccination coverage for SIAs 86% (95% CI: 80%, 93%) (Figure 3a). Three studies were excluded from the pooled meta-analysis due to missing sample sizes^31,32,38. However, after imputation of the sample sizes based on vaccination coverage, no major difference was seen in the pooled coverage (Figure 3b).

Figure 3.

a. Forest plot showing vaccination coverage for SIAs. b. Forest plot showing vaccination coverage for SIAs with imputed data.

SBVs: Eight studies reported vaccination coverage for SBVs (Table 2). All the SBVs reported coverage of the HPV vaccine among girls aged 9–19 years. Four HPV vaccine studies reported a combination of a grade and age based approach for identifying the target girls for the vaccination^28,33,35,55 while three studies reported a grade based only approach^23,40,41. The remaining study identified girls for HPV vaccination by age⁴², (Table 2). Vaccination coverage of completed doses among the targeted population ranged from 42.4 – 97.4%. A meta-analysis of SBVs showed a pooled vaccination coverage of 75% (95% CI: 67, 83) (Figure 4).

Figure 4. Forest plot showing vaccination coverage for SBV.

Comparison of vaccination coverage: To compare the two vaccine delivery strategies, pooled estimates for SIAs and SBVs were evaluated. For this comparison and subsequent analyses, Huhn et al., study³⁶ was not included since its setting (displaced) is not similar to the other studies reporting SIAs. The SIAs showed a higher vaccination coverage than and SBVs (Figure 5).

Figure 5. Comparison of vaccination coverage per strategy.

Subgroup analyses for vaccination coverage

Subgroup analyses were carried out to evaluate if vaccination coverage varied by the study setting or by vaccine type (for SIAs).

Study setting: The settings evaluated were: urban only, rural and urban mixed, or rural only. For SIAs, vaccination coverage did not vary irrespective of the three study settings (Figure 6). However, vaccination coverage was highest in urban areas (97%, 95% CI: 96, 98) and lowest in a mixed setting (88%, 95% CI: 79, 98).

Figure 6. Subgroup analysis of SIA coverage per setting.

Similarly for SBVs, there was no variation in vaccination coverage across the three settings. Vaccination coverage in urban, rural and mixed settings were as follows: 79% (95% CI: 77, 81), 74% (95% CI: 63, 85) and 75% (95% CI: 53, 97) respectively.

Vaccine type: There were variations in vaccine coverage dependent on the type of vaccine delivered via SIAs (Figure 7). Of all the vaccinations, measles SIAs reported the lowest coverage 83% (95% CI: 72, 93). Highest vaccination coverage were reported for SIAs against meningitis and yellow fever at 96%, (95% CI: 94, 98) and 98% (95% CI: 95, 99) respectively.

Figure 7. Subgroup analysis of SIA coverage per vaccine.

Cost of vaccine delivery strategy

SIAs: Seven studies reported the costs of conducting SIAs (Table 3). The costs represent monies spent to deliver the vaccines to the total target population during the SIAs. The available data on costs of the SIAs was not age specific and therefore, the costs for 5–19 year olds only could not be calculated. The SIAs targeted children aged 6 months and older age groups with the number of vaccinated people across all studies ranging from 23921 to 12,649,448.

Majority of the total cost in SIAs was used to procure vaccines and consumables. The highest proportion (86%) for vaccines and consumables costs’ was reported by da Silva et al., for the combined yellow fever and meningitis A/C campaign in Senegal⁵³. Costs for salaries of health staff and international supervisors were another high expense during SIAs. Other reported minor costs were: training of personnel, social mobilisation, transport, maintenance of equipments, cold chain and management of adverse events following immunisation (AEFI).

SBVs: Two studies reported the cost of HPV vaccination in Tanzania and Uganda^47,48 (Table 3).In Tanzania, 4211 girls were vaccinated with three doses of HPV vaccine either based on age or class. The total economic costs for the SBVs were 349,400 USD. The total cost per fully immunised girl in urban areas were 66 USD and 100 USD for class-based and age-based approach respectively; and in rural areas, 78 USD and 107 USD for class-based and age-based approach respectively.. Administration and supervision (salaries) of the project and procurement of vaccines accounted for the major expenses. The reported minor expenses in the study included training, cold chain, waste management and social mobilisation. The observed difference between costs per fully immunised girl for class-based and aged-based selection may be due to the fact that more time and logistics were needed to select girls based on their age. In Uganda, 3038 girls in a rural setting were vaccinated with three doses of HPV vaccine using a class based approach. The total economic cost (cost of all resources used regardless of who paid) was 30,646 USD excluding the cost of the vaccine. The total economic cost per fully immunised girl was 9.5 USD. Salaries of staff accounted for the greatest part of the cost followed by micro-planning, staff training, community mobilisation (start-up costs). Other expenses included supplies, cold chain, vehicles and transportation.

Effect of vaccination strategy on routine immunisation

SIAs: Two studies reported the effect of SIAs on routine immunisation and health services. Mounier-Jack et al., reported on a 10 day meningitis A campaign in Mali, in 2010³⁷ while Verguet et al., reported on a 3 weeks measles campaign in South Africa in the same year⁵⁷. Both SIAs reported a negative effect on routine immunisation (Table 4).

Low attendance: Both studies reported a decrease in the number of children attending the child clinics during the vaccination period. Mounier-Jack et al., reported a 71–74% decrease in the number of children vaccinated during the campaign³⁷ while Verguet et al., reported an 8% decrease in the number of children who were weighed in the postnatal clinic⁵⁷.

Redeployment of health staff: During the SIAs, staff in charge of routine immunisations were either deployed as supervisors for the campaign. This led to the closure of routine immunisation services in some districts during the campaign period.

Cold chain management: In Mali, routine vaccines were relocated and stored at the regional and district level so fridges could be made available to store the campaign vaccine³⁷.

SBVs: One study in Rwanda reported the effect of HPV vaccination on routine immunisation³⁹ (Table 4). According to Torres-Rueda et al., the vaccination activity which lasted 2 days had no effect on routine immunisation services³⁹. There was no change in the demand for routine immunisation and health services.

Strengths and limitations

We comprehensively and systematically searched for the relevant literature: both peer-reviewed and non-reviewed records were obtained. This study adhered to the PRISMA guidelines of conducting systematic reviews. Nonetheless, our review had several limitations. First, age specific coverage was not reported in some of the retrieved studies. Second, we observed a high heterogeneity during the meta-analysis. The heterogeneity was likely due to differences across studies of factors such as the age groups included, study settings and period. Third, the only vaccine administered via SBVs was the HPV vaccine, a new and more expensive vaccine. Lastly, few studies reported the effect of SIAs or SBVs on routine immunisation and health services and therefore, our findings may not accurately reflect the effects.

Implication for policy and research

Local evidence is crucial in the review and development of immunisation policies. Both SIAs and SBVs are routinely used in Africa to vaccinate older children. In settings where school enrolment is low as is the case in many African countries, our results show SIAs may be more effective in reaching school age children and adolescents than SBVs. However, caution needs to be exercised to mitigate the negative effect of SIAs on the routine health services. In Africa, the SBV has mainly been tested in the delivery of two or three dose HPV vaccine to adolescent girls while SIAs have been used for diverse vaccines and on a larger scale. Further research is therefore needed to assess the sustainability of SBVs for nationwide delivery of vaccines to school age children and adolescent in resource constraint settings as is the case in Africa. In the event that SBVs are chosen as the main delivery strategy, complementary community activities can be set up to target out of school children. Our results re-iterate the importance of systematic evidence to best inform African authorities on the optimal delivery strategies of vaccines targeted at school age children and adolescents for immunisation.