Page 16 - IJPS-8-2

International Journal of
Population Studies Accuracy of mother’s reporting on child immunization in Yemen

– 23 months with vaccination cards available have received them. Taking into account the cutoff of ≤80% as a low value,
BCG. Among the 578 children whose mothers are their sensitivity was low for both vaccines, with BCG showing
source of information, only 39% have received it. Whether the lowest value (70%). Considering the specificity, the
we analyze children who have never taken any vaccine, values were even worse for both types of vaccines, and
only 5% of the children with vaccination cards fall in BCG again presented the lowest value (56%). PPVs were
this category against 56% of children whose information high for polio 3 and BCG (88 and 84%, respectively). In
rd
comes from the mother’s report. This is an indication of contrast, NPVs were low for both (38 and 36%).
two possibilities that may be complementary to each other: These results indicate an acceptable effectiveness of the
(1) mother’s recall of child vaccination is very incomplete instrument (mother’s report) while estimating vaccination
in the sense that it probably underestimates children’s coverage. However, some features should be highlighted.
vaccination status, and (2) vaccination cards are a key
source of children’s vaccination history. According to the results presented in Table 1, mother’s
information is more reliable when predicting children
3.2. The accuracy of mother’s reporting on child who were actually vaccinated, but it is more problematic
vaccination status when predicting those who were not vaccinated. This
means that mothers usually underestimate the number of
We tested mother’s precision while reporting their
children’s vaccination status regarding two main issues: vaccines taken by their children in Yemen, which in turn
(1) The overall quality of mother’s reports based on may underestimate children’s vaccination coverage, as for
both mother’s reporting information and subsequent most children, the main source of information on their
vaccination schedule is their mother (59% of children aged
vaccination card information and (2) the direction of 12 – 23 months in Yemen does not have a vaccination card
mothers misreporting, which indicates whether mother’s available). In fact, as we assumed that the gold standard
report mostly leads to underestimation or overestimation
of children’s vaccination coverage. for the assessment of coverage was the vaccination card,
it is estimated that the vaccination coverage is 76.92% and
3.2.1. Overall quality of mothers’ reports 79.51%, respectively, for BCG and polio 3 in our sample
rd
of children. The estimated coverage for the same sample
As proposed by Miles et al. (2013), we specified ≤80% based on mother’s recall is more than 10 percentage points
as being a poor validity of the instrument/source of
information (in this case, mother’s report on vaccination lower than that based on the vaccination card (64.10% and
status of children aged 12 – 23 months). For both vaccines, 66.34%, respectively).
we found low agreement between the mother’s reports in 4. Discussion
R1 and vaccination cards in R3, with accuracies of 67 and
71% for BCG and polio 3 , respectively. The previous sources of vaccination data in Yemen are
rd
the 1991 – 1992 and 1997 DHSs and the 2006 MICS, both
3.2.2. The direction of mothers’ misreporting offering access to cross-sectional data. Nevertheless, while
Table 1 shows the sensitivity, specificity, PPV, and NPV adequately analyzing the accuracy of parents’ recall, it is
while comparing the mother’s report at R1 with the crucial to have access to longitudinal information. This is
vaccination card at R3 (the latter being our gold standard). because in a household survey interview, caretakers are
Before analyzing the results, we highlight that due to usually first asked to show the vaccination card of each
our sample restrictions, the sampled children are in child, and only in case they do not have it are they asked
better socioeconomic conditions compared to the others to recall whether the child had received each of the doses/
although, on average, both groups of children (those vaccines. In this way, in cross-sectional data, the researcher
included or those not in the sample) do not have adequate will only have access to one source of information, either
living conditions as access to health facilities, water, the vaccination card or parents’ report. Based on the
sanitation, education, and food security are still scarce in a NSPMS longitudinal survey, it was possible to test the
country where poverty rates were estimated at 45% in 2012 accuracy of mother’s reports in Yemen by comparing
(IPC-IG and UNICEF, 2014). Although these conditions mother’s information at R1 with the information collected
can affect both groups homogeneously, this possibility at R3 (6 months later).
cannot be tested, which introduces a limitation to the study. It is important to highlight that although we also

The sensitivity values were higher than the specificity, tested mothers’ accuracy in a sample including all
and the PPVs were higher than the NPVs for both vaccines. children aged 12 – 59 months, we found that memory
When comparing the type of vaccine, we easily see that errors increased significantly with age. In this sense, to
there was relatively little change in sensitivity between minimize miscalculating vaccination coverage indicators

Volume 8 Issue 2 (2022) 10 https://doi.org/10.36922/ijps.v8i2.1274