Reply to comments on Monitoring vaccination coverage: Defining the role of surveys.


Dear Editor, We thank Pond and Mounier-Jack for their comments on our paper, ‘‘Monitoring vaccination coverage: Defining the role of surveys” [1]. We agree that for many countries, administrative estimates of coverage are greatly inflated and misleading for programme planning purposes. The robustness of the WHO-UNICEF estimates of national immunization coverage (WUENIC) depends on the quality of the underlying data reviewed, which include administrative reports, as well as probability and non-probability sample surveys. In 2012, the Grade of Confidence (GoC) was introduced as a means of conveying uncertainty in WUENIC [2] and is low in the seven conflict-affected countries listed by Pond and Mounier-Jack. Table 1 shows that in five of these countries, vaccination cards were available for less than half the children surveyed; when card availability is low, it is particularly difficult to compare coverage trends. For example, in Nigeria, the proportion of children with DTP3 according to card was similar in surveys in 2010, 2011 and 2013, but in the EPI survey of 2010 a verbal history of vaccination was reported for 43% of children, more than double that of previous or subsequent surveys. Elsewhere, results from surveys did not always match expected trends (e.g. no apparent fall in coverage between surveys despite a 7 month stockout of DTP in one country), and some results were very unlikely (e.g. zero dropout between DTP1 and DTP3 in one Multiple Indicator Cluster Survey (MICS) (data from country reports at immunization_monitoring/globalsummary/wucoveragecountrylist. html)). The updated WHO guidelines on vaccination coverage surveys ( Vaccination_coverage_cluster_survey_with_annexes.pdf) discuss the challenges of using a new survey to compare with an older one, particularly an immunization coverage survey – these often lacked information on likely biases and confidence intervals were either not reported or not very meaningful from non-probability samples. The best way to compare results from different surveys is to plan a pair of surveys for such a purpose and work very hard to ensure standardised, well-documented and high quality data collection in both. Pond and Mounier-Jack suggest that two such surveys are feasible within each 5 years period. We would be reluctant to stipulate any particular interval as the usefulness of repeat surveys will depend in part on the likelihood of a change in coverage having occurred (which can be predicted from monitoring other indicators) [1] and the availability of accurate documentation of vaccination status on home-based or clinic records. Most of all, surveys should lead to action to strengthen programme performance and this is likely the weakest link in many countries, including those affected by conflict.