Estimating onset time from longitudinal data in the presence of measurement error with application to estimating...

Area of Science:

• Maternal and Fetal Health
• Biostatistics
• Public Health

Background:

• Accurate gestational age (GA) estimation is crucial for neonatal care and outcomes.
• Current ultrasound-based methods for GA assessment are often inaccessible in low-resource settings.
• Maternal anthropometry offers a potential alternative for GA estimation.

Purpose of the Study:

• To develop and evaluate a statistical model for estimating gestational age at birth using longitudinal maternal anthropometry.
• To incorporate maternal information from the last menstrual period, accounting for its measurement error.
• To propose methods for validating predictions when gold-standard measurements have errors.

Main Methods:

• Development of a shared random parameter model for GA estimation.
• Utilizing longitudinal maternal anthropometry and last menstrual period data.
• Evaluation through simulation studies using a training-test set paradigm.
• Methodology for prediction validation with error-prone gold-standard measurements.

Main Results:

• The proposed shared random parameter model effectively estimates gestational age from maternal anthropometry.
• Simulation studies demonstrated the model's performance under various conditions.
• The validation methodology proved effective for assessing prediction accuracy with measurement error.

Conclusions:

• The developed model provides a viable approach for estimating gestational age in resource-limited environments.
• This methodology can improve pediatric care by enabling more accurate GA assessment.
• The study offers tools for robust prediction validation in the presence of data imperfections.