Clustered lot quality assurance sampling: a tool to monitor immunization coverage rapidly during a national yellow fever and polio vaccination campaign in Cameroon, May 2009

Abstract

SUMMARYWe used the clustered lot quality assurance sampling (clustered-LQAS) technique to identify districts with low immunization coverage and guide mop-up actions during the last 4 days of a combined oral polio vaccine (OPV) and yellow fever (YF) vaccination campaign conducted in Cameroon in May 2009. We monitored 17 pre-selected districts at risk for low coverage. We designed LQAS plans to reject districts with YF vaccination coverage <90 % and with OPV coverage <95 %. In each lot the sample size was 50 (five clusters of 10) with decision values of 3 for assessing OPV and 7 for YF coverage. We 'rejected ' 10 districts for low YF coverage and 14 for low OPV coverage. Hence we recommended a 2-day extension of the campaign. Clustered-LQAS proved to be useful in guiding the campaign vaccination strategy before the completion of the operations.

The effect of clustering on lot quality assurance sampling: a probabilistic model to calculate sample sizes for quality assessments

“…Combining sampling in clusters and LQAS increases the versatility of the tool. However, as others have noted, failing to account for the clustering in the design of the LQAS system typically leads to inflated misclassification errors [7,8]. This paper provides the general theory for sample size calculations for C-LQAS that accounts for the clustering in the design using a beta-binomial model.…”

“…Otherwise, the lot is classified to be of acceptable quality. The design of a C-LQAS system is defined by three numbers: the number of clusters sampled, m ; the number of items per cluster sampled, k ; and the decision rule for classification, d [7,8]. The number of defective 'units’ across the m*k elements are tallied and compared to the decision rule in the same way as the basic two-way LQAS.…”

“…However, if the sample size is calculated using the binomial model that assumes simple random sampling but there is clustering of the outcome, then the true misclassification risks are likely greater than the specified limits, α max and β max [7,8,11,12]. …”

“…However, previous work has shown that the implementation of an LQAS system in such a clustered manner when n and d are incorrectly calculated under the assumption of simple random sampling raises the misclassification risks above the specified bounds because of the correlation within clusters [7,8,11,12]. In 2010, Pezzoli et al describe an approach for developing a cluster-LQAS (C-LQAS) system where the total sample size, n , is based on the binomial distribution [12].…”

Predictors to parental knowledge about childhood immunisation/EPI vaccines in two health districts in Cameroon prior to the introduction of 13-valent Pneumococcal Conjugate Vaccines (PCV-13)

“…These two health districts were chosen because they were described as “Non-Performant” based on statistics from a rapid evaluation of vaccination coverage on the first phase of the National Immunisation days against Polio and Yellow Fever of 4-9 May 2009 [15, 16]. Yaoundé is located between Latitude 3° 52’ N and Longitude 11° 31'E.…”