Accuracy of Mothers' Perception of Birth Size to Predict Birth Weight Data in Bangladesh

The prevalence of low birth weight (LBW) is an important indicator of child health and wellbeing. However, in many countries, decisions regarding care and treatment are often based on mothers' perceptions of their children's birth size due to a lack of objective birth weight data. Additionally, birth weight data that is self-reported or recorded often encounters the issue of heaping. This study assesses the concordance between the perceived birth size and the reported or recorded birth weight. We also investigate how the presence of heaped birth weight data affects this concordance, as well as the relationship between concordance and various sociodemographic factors.

We examined 4,641 birth records reported in the 2019 Bangladesh Multiple Indicator Cluster Survey. The sensitivity-specificity analysis was performed to assess perceived birth size's ability to predict LBW, while Cohen's Kappa statistic assessed reliability. We used the kernel smoothing technique to correct heaping of birth weight data, as well as a multivariable multinomial logistic model to assess factors associated with concordance.

Maternally-perceived birth size exhibited a low sensitivity (63.5%) and positive predictive value (52.6%) for predicting LBW, but a high specificity (90.1%) and negative predictive value (93.4%). There was 86.1% agreement between birth size and birth weight-based classifications (Kappa = 0.49, indicating moderate agreement). Smoothed birth weight data did not improve agreement (83.4%, Kappa = 0.45). Of the sociodemographic factors, early marriage was positively associated with discordance (i.e., overestimation).

An important consideration when calculating the LBW prevalence is that maternally perceived birth size is not an optimal proxy for birth weight. Focus should be placed on encouraging institutional births and educating community health workers and young mothers about the significance of measuring and recording birth weight.