Background: Although recall-based data are collected by survey when the vaccination records are not available, the preferred estimates remain the record-based ones due to the limited validity of recall-based data. However, the evidence on validity of maternal recalls is limited and varied across vaccine types. To close the gaps, we validated the maternal recall on vaccination against record-based data in six districts in Zhejiang Province, China. Methods: We used a cross-sectional survey of about 648 households with mothers who delivered in the last 12 months prior to the survey in October 2017, from six districts in Zhejiang Province. Vaccination status on five vaccine types scheduled before 12 months of age were collected through maternal recall and vaccination records. The level of agreement and recall bias between the two resources, the sensitivity and specificity of maternal recall were evaluated. Risk factors for maternal recall bias were also identified through logistic regression model for each type of vaccine. Results: The level of agreement between recall and record was above 90% across vaccine types, with the recall bias ranged from 2.2% to 9.7%. Recall bias due to over-reporting was slightly higher than that due to under-reporting. Recall bias was positively associated with high parity, home delivery, younger mothers, mothers with low education, and migrant mothers. Conclusions: This study indicated most of the vaccination status across vaccine types was accurately identified through maternal recall and supported the use of maternal recall to estimate the vaccination coverage as an alternative in the absence of record-based data.
Keywords: maternal recall; recall bias; vaccination; vaccination record; validity.