Validating child vaccination status in a demographic surveillance system using data from a clinical cohort study: evidence from rural South Africa

BACKGROUND: Childhood vaccination coverage can be estimated from a range of sources. This study aims to validate vaccination data from a longitudinal population-based demographic surveillance system (DSS) against data from a clinical cohort study. METHODS: The sample includes 821 children in the Vertical Transmission cohort Study (VTS), who were born between December 2001 and April 2005, and were matched to the Africa Centre DSS, in northern KwaZulu-Natal. Vaccination information in the surveillance was collected retrospectively, using standardized questionnaires during bi-annual household visits, when the child was 12 to 23 months of age. DSS vaccination information was based on extraction from a vaccination card or, if the card was not available, on maternal recall. In the VTS, vaccination data was collected at scheduled maternal and child clinic visits when a study nurse administered child vaccinations. We estimated the sensitivity of the surveillance in detecting vaccinations conducted as part of the VTS during these clinic visits. RESULTS: Vaccination data in matched children in the DSS was based on the vaccination card in about two-thirds of the cases and on maternal recall in about one-third. The sensitivity of the vaccination variables in the surveillance was high for all vaccines based on either information from a South African Road-to-Health (RTH) card (0.94-0.97) or maternal recall (0.94-0.98). Addition of maternal recall to the RTH card information had little effect on the sensitivity of the surveillance variable (0.95-0.97). The estimates of sensitivity did not vary significantly, when we stratified the analyses by maternal antenatal HIV status. Addition of maternal recall of vaccination status of the child to the RTH card information significantly increased the proportion of children known to be vaccinated across all vaccines in the DSS. CONCLUSION: Maternal recall performs well in identifying vaccinated children aged 12-23 months (both in HIV-infected and HIV-uninfected mothers), with sensitivity similar to information extracted from vaccination cards. Information based on both maternal recall and vaccination cards should be used if the aim is to use surveillance data to identify children who received a vaccination.