The Estimation of Gestational Age at Birth in Database Studies

Area of Science:

• Perinatal epidemiology
• Biostatistics
• Reproductive health

Background:

• Accurate estimation of pregnancy duration is essential for studies on prenatal medication safety.
• Gestational age data is frequently incomplete in clinical and administrative databases.
• This study addresses the challenge of estimating missing gestational age.

Purpose of the Study:

• To compare various methods for estimating missing gestational age.
• To identify the most accurate approach for determining pregnancy duration from incomplete data.

Main Methods:

• Utilized Clinical Practice Research Datalink and Hospital Episode Statistics data.
• Compared four methods: generalized estimating equations, multiple imputation, birth weight/sex-based estimation, and fixed value assignment.
• Used Hospital Episode Statistics gestational age as the gold standard.
• Conducted a simulation study to assess bias and mean square error.

Main Results:

• Generalized estimating equations (GEE) including birth weight showed the smallest average absolute bias.
• Assigning a fixed 39-week gestation to all missing values resulted in the largest absolute bias.
• GEE demonstrated the smallest mean square errors, while multiple imputation had the highest.

Conclusions:

• Generalized estimating equations provide the most accurate gestational age estimation when birth weight data is available.
• In the absence of birth weight, using fixed gestational age based on term/preterm status is a viable optimal approach.