We investigated the possibility of preterm birth misclassification as a determinant of variation in its reported rates. Using a database of 497,105 deliveries from 17 hospitals, the best estimate of gestational age made at delivery and entered into the database at that time was recalculated from the menstrual dates and mid-trimester ultrasound scan. The recalculated completed weeks of gestation at delivery was compared with that made at birth. Calculation of estimated gestational age varied between hospitals due to inconsistencies in 'rounding' and 'truncating' the weeks of gestation at delivery. This resulted in preterm birth misclassification rates of up to 10.1%.