Local discrepancies in measles vaccination opportunities: results of population-based surveys in Sub-Saharan Africa

Background The World Health Organization recommends African children receive two doses of measles containing vaccine (MCV) through routine programs or supplemental immunization activities (SIA). Moreover, children have an additional opportunity to receive MCV through outbreak response immunization (ORI) mass campaigns in certain contexts. Here, we present the results of MCV coverage by dose estimated through surveys conducted after outbreak response in diverse settings in Sub-Saharan Africa. Methods We included 24 household-based surveys conducted in six countries after a non-selective mass vaccination campaign. In the majority (22/24), the survey sample was selected using probability proportional to size cluster-based sampling. Others used Lot Quality Assurance Sampling. Results In total, data were collected on 60,895 children from 2005 to 2011. Routine coverage varied between countries (>95% in Malawi and Kirundo province (Burundi) while <35% in N’Djamena (Chad) in 2005), within a country and over time. SIA coverage was <75% in most settings. ORI coverage ranged from >95% in Malawi to 71.4% [95% CI: 68.9-73.8] in N’Djamena (Chad) in 2005. In five sites, >5% of children remained unvaccinated after several opportunities. Conversely, in Malawi and DRC, over half of the children eligible for the last SIA received a third dose of MCV. Conclusions Control pre-elimination targets were still not reached, contributing to the occurrence of repeated measles outbreak in the Sub-Saharan African countries reported here. Although children receiving a dose of MCV through outbreak response benefit from the intervention, ensuring that programs effectively target hard to reach children remains the cornerstone of measles control.


Background
The World Health Organization (WHO) comprehensive measles control strategy aims to reduce global measles mortality by at least 95% by the end of 2015 compared with 2000 estimates and achieve measles elimination in at least five WHO regions by the end of 2020. One of the components of the strategy is to achieve and maintain high levels of population immunity by providing high vaccination coverage with two doses of measles-containing vaccine (MCV) [1]. Children should receive the first MCV dose through the Expanded Program of Immunization (EPI). A second should be delivered either through periodic supplemental immunization activities (SIAs) or through EPI.
Moreover, since 2009, outbreak response immunization (ORI) is part of WHO strategy for response to measles outbreaks in measles mortality reduction settings [2]. If assessment indicates the risk of a large outbreak, nonselective reactive mass vaccination campaign should be implemented as soon as the outbreak is confirmed. The aim is to reduce morbidity and mortality in the immediate by delivering at least one dose of MCV to susceptible children.
Although measles mortality was reduced by 91% between 2001 and 2008 in the WHO African region countries [3], measles remains a leading cause of vaccinepreventable deaths in Africa [4]. Failure to deliver at least one dose of MCV remains the primary reason for high measles mortality.
In order to sustain gains in measles control and further reduce mortality, an understanding of whether each of these vaccination opportunities reaches children is essential. Vaccination coverage is the key indicator to measure the progress of measles control. Ministries of Health (MoH) report annual administrative coverage (i.e. number of doses delivered divided by estimated number of children in the targeted age group) based on population estimates. In many contexts, however, population estimates, are often not up-to-date (i.e. population censuses might not be frequently performed and vital event registration may be absent or partial), resulting in biased or inaccurate estimates. Consequently, population-based surveys are often the best available means to estimate vaccination coverage at both local and national levels.
Over the past eight years, the non-governmental organization Médecins Sans Frontières (MSF) supported MoH of several countries in sub-Saharan Africa in the implementation of ORI through non-selective mass vaccination campaigns. Following these campaigns and as part of the overall response to measles epidemics, punctual household-based measles coverage surveys were conducted [5][6][7][8][9]. The main objective of these surveys was to estimate vaccination coverage through ORI. These surveys also aimed to estimate measles vaccine coverage through previous vaccination opportunities: in EPI, SIA, or outbreak response campaigns. Here, we present a secondary data analysis of these different surveys to describe measles vaccine coverage through different vaccination opportunities in different settings in sub-Saharan African countries. The aim is to provide additional information on whether these different vaccination opportunities reach children in need.

Methods
We considered all household-based measles vaccination coverage surveys conducted by Epicentre after a nonselective mass vaccination campaign that evaluated different opportunities and for which the complete dataset was available. We defined a survey as a unique analysis of measles vaccination coverage. We considered a total of 24 surveys conducted between 2005 and 2011 in 23 sites in 6 countries (Chad, Central African Republic (CAR), Democratic Republic of Congo (DRC), Cameroon, Malawi and Burundi). Surveys were conducted in different locations within a country in Chad, DRC, Malawi and CAR. Two surveys were conducted at two different times in N'Djaména (Chad). A total of 13 of the study sites were considered as urban or a mix of urban-rural (N'djamena city, Matadi city, Mbuji-Mayi city, Blantyre, Lilongwe, Mzimba and Balaka/Machinga districts, Likasi and Lubumbashi cities and Kabo city), the others were considered as rural.

Design and settings of vaccine coverage studies
Surveys were conducted on average 30 days after the first day of the mass vaccination campaign (range: 1 day to 4 months). The age group included in the surveys varied according to the age group targeted by the ORI. All surveys employed probability proportional to size clusterbased sampling (Additional file 1: Table S1), except the surveys conducted in N'Djamena (Chad) in 2005 and 2010 which used Lot Quality Assurance Sampling. The first household to be surveyed in each cluster or lot was selected randomly by spatial-based sampling [8], GPSbased sampling, complete household enumeration or by the EPI method [10]. Method for selection of the first household was by preference by complete enumeration when logistical and security constraints allowed. Subsequent households were selected by proximity. Sample size varied according to the design and hypothesis about existing vaccination coverage (range: 942 to 6622 children).

Data collection
Trained surveyors conducted face-to-face interviews with child's main caregiver. Standardized, pre-piloted questionnaires were used and interviews were conducted in local language(s). Information was collected on demographic characteristics (age at the time of the survey or date of birth, sex) and vaccination status for different vaccine opportunities (Additional file 1: Table S1). When possible, information on vaccination history was verified with vaccination card. When card was not available, probing questions, such as place were the vaccination took place or injection site, were asked to minimize the risk of misclassification.
Vaccination status was collected for different opportunities. ORI implemented by MoH with the support of MSF (MSF/MoH ORI) and EPI were considered in all study sites. Although different SIAs had been implemented at different time points in these countries, only vaccination during the most recent was considered.  Maroua (Cameroon) and Kirundu (Burundi), SIAs were implemented before the ORI but were not included in the surveys. Lastly, in four countries (Cameroon, Malawi, Chad and CAR), we also evaluated MoH mass campaigns in response to outbreaks conducted before MSF/MoH ORI.
Administrative coverage was provided by local health authorities.

Data analysis
Children were considered as vaccinated through a specific opportunity if they were eligible for vaccination (i.e. in the target age group) and vaccination was reported by card examination or oral history.
Vaccine coverage and 95% confidence interval (95% CI) were estimated for each vaccine opportunity. EPI coverage was defined as the proportion of children vaccinated through the routine program among children aged 9 to 11 months (i.e. in the targeted window for this program) and 12 to 23 months (i. e. who should have recently received their MCV through EPI). SIA, ORI and other potential MoH reactive campaigns coverage were defined as the proportion of children vaccinated amongst the target age group at the time of the campaign. ORI coverage was calculated by age group.
If the delay between the last day of the MoH/MSF campaign and the first day of the vaccine coverage survey was <31 days, age at the time of the survey was considered to be similar to age at the time of the MoH/MSF campaign. If not, age at the first day of the MoH/MSF campaign was calculated.
To evaluate if children received the two recommended doses, we calculated the number of doses received by summing the number of occasions a child was considered vaccinated. Only children with all information on vaccine opportunities were included in this sub-analysis, otherwise number of doses was considered as missing. In Malawi and DRC, where EPI, SIA and ORI opportunities were evaluated, we estimated the proportion of children vaccinated through ORI according to the number of doses of MCV they received before, and the proportion of children for whom the ORI dose represent a first, a second or a third dose of MCV. For this sub-analysis, information on EPI vaccination was considered for all children aged 9 month or older included in the survey and not only for children 9 to 23 months of age.
To explore potential associations between vaccination coverage and explanatory variables, Poisson regression was conducted when several places were surveyed at the same time. Wald tests were used to test explanatory variables (p < 0.05). Estimations were weighed to account for survey designs. Data was entered in EpiData v3 (Odense, Denmark) and analysed with Stata v11 (College Station, Texas, USA).

Ethical considerations
Surveys were conducted as part of the monitoring and evaluation of the emergency response and were not submitted for formal ethical approval in the country of the study. The MSF Ethical Review Board exempts such surveys from review as they constitute part of the emergency response and use a standard and widely accepted survey protocol. All studies, however, were approved by local and national authorities through the convocation of exceptional meetings. In N'Djamena, surveys were conducted with the authorization of the national technical committee for the battle against epidemics; in Matadi and Mbuji Mayi (DRC), authorization was obtained from the Ministry of Health; in Maroua (Cameroon), a request was made to the Division of Operational Research within the Ministry of Health for specific ethical clearance for the survey and was granted; in Malawi, the study was implemented in collaboration with the MoH after obtaining permission to carry out the survey, in Katanga, authorization was delivered by national and provincial health authorities.
The main caregivers of all participants provided oral informed consent via a specific explanation in the local language as well as a written document providing information on the use of information and on confidentially.
No specific ethnic or identifying information was recorded and all participants were free to refuse participation in the surveys. Vaccination and care were provided free of charge irrespective of participation in surveys.

Results
Between 2005 and 2011 vaccination data was collected from 60,895 children living in six countries. Half of them were male (sex ratio = 1.0). Twelve of the surveys included children 6 months to 14 years old, 5 surveys included children 6 months to 4 years old and 1 considered children 6 months to 9 years old.

SIA coverage
The surveys conducted in Katanga district in DRC in 2011 showed the highest coverage estimates for SIA activities (Table 3). Among these, the greatest coverage was obtained in Lubumbashi city and the lowest in Kambove, both in DRC. In Malawi, SIA coverage also varied within the country. Reported SIA administrative coverage was typically very high and often above 95%.

Other mass vaccination campaigns
Surveys results showed that previous ORI conducted by MoH

Discussion
We present the results of 24 population-based measles vaccination coverage surveys conducted in between 2005 and 2011 in 28 locations in Sub-Saharan Africa. In 2008, the WHO African Region measles technical advisory group recommended establishing a measles pre-elimination goal, to be achieved by the end of 2012, with the following immunization targets: >90% national MCV routine coverage, with at least 80% coverage in all districts; and ≥95% SIA coverage in all districts [11]. Routine vaccination coverage reached high levels in several districts of Malawi and in Kirundu (Burundi), but most of the surveys showed insufficient EPI coverage.
Survey results showed that the highest vaccination coverage was obtained through ORI. For most infants 6 to 8 months of age, ORI provided the first MCV vaccination. These infants are at higher risk of complications and death in case of illness highlighting the effectiveness of vaccinating this population in ORI.
EPI coverage appeared to be lower among children aged 9 to 11 months than 12 to 23 months old. This might show a delay in the age of routine vaccination. The recommendation to enlarge the targeted age group for routine activities [12] should be encouraged and reinforced. This would allow the protection of children that missed the routine age window and slow the build-up of susceptible children contributing to the risk of an epidemic in areas with circulating virus. While EPI programs should be flexible to ensure vaccination, efforts are needed to ensure an early first dose administration as a first priority. In contexts where SIAs were implemented, coverage was low and far under the targeted 95%. Moreover, in many locations these activities were not implemented or often implemented late, without respecting the recommended interval between SIAs. For instance, in Katanga province, after the 2007 SIA few measles cases were reported. The next SIA was planned in 2010, but this was postponed [13] contributing to the causes of a large outbreak in 2010-2011 [14].
Often administrative coverage was higher than survey results. This was most significant for SIA, where all administrative estimates were higher than surveys results. Administrative estimates are often at national level and do not account for provincial or district differences. Population-based surveys can provide specific information on coverage and help target interventions. A model that considers both administrative and survey data has even been developed to characterize the performance of the activities leading to the estimated coverage and help to predict the effect of future vaccination activities [15]. This kind of exercise is very important to adequately assess the risk of outbreaks.
Despite several opportunities, a non-negligible proportion of children remained unvaccinated or had not received the recommended 2 doses. Conversely, for some children ORI provided the third or higher dose. Moreover, children not reached either by EPI program or by SIA were also less reached by ORI. These results highlight that children are not equally reached by vaccination activities and multiplying vaccine opportunities does not always imply that unvaccinated children are reached. Further work is needed to ensure that immunization activities reach unprotected children. And in case of limited resources, previous MCV vaccination activities and their coverage should be evaluated to better allocate resources and improve coverage.
These survey data are subject to limitations. First, surveys were conducted only in settings were measles control strategies were not efficient enough to avoid an outbreak, but where surveillance system detected an increase in measles cases. Vaccine coverage in these areas is likely to be different from places where no outbreak occurred or was detected during the last decade.
Second, although card confirmation is the preferred method for ascertaining coverage, this is not always possible and oral history is considered. For example, none of the children could show a vaccination card for SIA, and only an average of 16.9% of the children considered as vaccinated through EPI could show their vaccination card [range 1.8% -36.2%]. Vaccination status regarding ORI was better documented as half of the children considered as vaccinated could show their vaccination card [range: 0.0% -83.4%]. This proportion was generally high, except in Malawi and Maroua where quasi none of the vaccination status was confirmed by card. As a result, over-reporting or under-reporting of vaccination might have occurred, depending on the context and despite probing questions. However, previous studies in areas of high measles incidence have shown parental recall to be reliable [16].
Third, most of the surveys were conducted within a month after the end of ORIs. There is therefore low risk of recall bias for ORI coverage estimates, especially as there is less risk of card loss. Information on routine vaccination was collected for all age groups, i.e. for children up to 15 years of age. Recall bias, especially for older children for which vaccination in routine vaccination occurred long ago, might be expected. We therefore chose not to present EPI coverage for older age group but this likely resulted in an overestimation of the number of doses received in older age groups.
Finally, although standardized protocols and training of all surveyors was rigorously implemented across settings, there may be additional inaccuracies in the data related to individual interviewers and supervisors as well as the inaccuracies of population data used to select the samples.
It is also important to recognize that although population surveys are a rapid means to obtain a vaccine coverage estimate, without serological confirmation, estimates of coverage remain an inference.

Conclusion
Population-based surveys are important to have reliable estimates of MCV coverage and can serve as a tool to assess different vaccine opportunities and to have a better understanding of coverage variations among age group and settings. Control pre-elimination targets were still not reached in the studied sites and might explain the occurrence of repeated measles outbreak in Sub-Saharan African countries. Furthermore, despite different vaccine opportunities, the number of unvaccinated or not fully vaccinated children was high in some settings. Strategies to better target hard-to-reach children are needed.

Additional file
Additional file 1: Table S1.