Improving the accuracy of birth notification data: lessons from the Birth to Ten study

Area of Science:

• Epidemiology
• Public Health
• Biostatistics

Background:

• Birth notification data is crucial for monitoring maternal and child health indicators.
• Longitudinal birth cohort studies provide valuable insights into population health trends.
• The Soweto-Johannesburg Metropole's Birth to Ten study offers a unique dataset for evaluating data quality.

Purpose of the Study:

• To assess the accuracy and consistency of birth notification data collected in the Soweto-Johannesburg Metropole.
• To identify potential sources of error and variability in birth notification records.
• To understand the impact of data inaccuracies on the reported prevalence of key infant health indicators.

Main Methods:

• Utilized birth notification forms from three health authorities for 5,448 singleton births (April-June 1990).
• Compared notification data with original hospital records for 539 births at Baragwanath Hospital.
• Analyzed data consistency across different notification forms and assessed accuracy for selected variables.

Main Results:

• Proportion of missing records ranged from 0.0% to 40.9%, with highest rates for non-mandatory variables.
• Agreement between notification and hospital records was high for infant sex (99.1%) but varied for continuous variables (e.g., gestational age 29.7%).
• Inaccuracies in birth weight and gestational age data led to misclassification of up to 25.7% for prematurity/post-term and 10.5% for birth weight categories.

Conclusions:

• Imprecise birth notification data can significantly distort the apparent prevalence of critical infant health indicators.
• Standardizing birth notification forms and improving data collection processes are essential for enhancing data consistency and accuracy.
• Selecting reliable and measurable health indicators is vital for improving the utility of birth notification data in public health surveillance.