VCSs are an important source of information on vaccination programs. They are all the more important if there are several sources of information on vaccination that can allow either triangulation of information in order to have a better estimate, or comparability of the information collected. This study consisted in evaluating the validity of the maternal recall compared to the vaccination documents in terms of the vaccination status of children aged 12 to 23 months in a health district of Senegal. Data based on vaccination documents were used as a reference because they are generally preferred over recall data in VCS [18, 19]. The results indicate that maternal recall is an insufficient source of information for estimating vaccination coverage. Indeed, vaccination coverages for all antigens based on maternal recall were lower than those derived from vaccination documents. These results were similar to those demonstrated in studies conducted in Nigeria [20] and the Northern Mariana Islands of the Western Pacific [21]. Nevertheless, these results are contrary to many others [18, 22–24]. This survey would have produced biased results that could lead to inappropriate public health measures if only the mother's recall had been relied upon. This is materialized through maternal recall bias where the percentage of under-reporting is greater for each antigen compared to over-reporting. Because routine immunization is a series of events spanning several months rather than a single event, parents' recall of the child's immunization status may be tainted by forgetting.
Concordance differed between antigen types ranging from 58.0–67.6%. The concordance was relatively good for the BCG, HepB-BD and MR1 vaccines. However, it was moderate for the OPV3 and Penta 3. This is very mixed and suggests that mothers may have difficulty remembering precisely all the vaccines their children have received, especially for certain types of antigens. The disagreement between the sources, however small, reveals a problem of quality in the provision of immunization services, especially in terms of notification of vaccination in the documents and communication with the mothers.
According to a systematic review, a sensitivity and specificity less than or equal to 80% is a reflection of low validity [25]. In this case, the sensitivity resulting from this study is high for all the antigens. This result is similar to those highlighted by studies conducted in China [22] and Tanzania [18]. On the other hand, the specificity is below the threshold of 80% for all the antigens. An identical situation was found in Nigeria [20]. In other words, parents of unvaccinated children reported that they are vaccinated. This meant that maternal recall was sometimes influenced by the concept that the practice of childhood vaccination was considered socially desirable [26], thereby leading to misclassifications. A study in this direction reveals that mothers whose children are insufficiently vaccinated tended to overestimate the vaccination coverage [27]. This low specificity indicates a high number of false positives and increases the risk of exposure to vaccine-preventable diseases.
Multivariate analysis shows that rural area was significantly associated with recall bias. This result is similar to that of the study conducted in Nigeria [20]. Also, the absence of advice on vaccination during ANC and PNC induces recall bias. Women living in rural areas face many challenges including limited access to education. Indeed, most women who live in rural areas have a lower level of education. This could create a barrier to communication with vaccinators and misunderstanding of vaccination information [22]. Prenatal and postnatal education should increase mothers' knowledge and confidence in vaccination.
This study could have important implications for the immunization program. Although a systematic review on the validity of maternal recall suggested that we do not yet have enough evidence to draw a firm conclusion on the subject [25], our results seem to argue in favor of using vaccination documents. compared to maternal recall. Indeed, mothers' reports of childhood vaccination may not always be accurate and may be subject to recall or social desirability bias [18, 22]. This raises the question of the reliability of declarative data for estimating vaccination coverage. However, in our context, we found a relatively high availability of vaccination cards, exceeding 80% [28]. Therefore, it would be crucial to set up an intervention targeting mothers, especially in rural areas, in order to improve the retention and proper storage of health records. At the same time, vaccinators should be made aware of the value of reporting complete and accurate information in order to allow more precise estimates of vaccination coverage.
This study has limitations. First, the study was conducted in a single health district in Senegal, so the results may not be generalizable to the entire country. However, it is important to note that the districts in Senegal are organized in the same way and the providers trained in the same schools [29]. Second, the study is cross-sectional; therefore, it is not possible to determine the temporal relationship between factors associated with recall bias and recall bias itself. Third, it should be noted that this study is based on declarative data provided by the participants themselves; which can lead to memory bias.