Early ultrasound dating of pregnancy: selection and measurement biases

Abstract

Can using early ultrasound examinations to date pregnancy introduce information bias in perinatal research? Our purpose was to identify determinants of early ultrasound examinations and to compare early ultrasound to menstrual history dating. Between January 1987 and June 1989, 1159 white, largely middle class, prenatal patients were contacted for a prospective observational study. 876 (76%) agreed to participate. Of these 764 (87%) met the eligibility criteria for this analysis, namely singleton pregnancy, delivered after 20 weeks (spontaneous or induced, vaginal or c-section), with prenatal chart abstracted. Selection factors for early ultrasound identified in multivariate analysis were: bleeding in early pregnancy, OR = 1.9 (1.0, 3.5), attendance at health maintenance organization OR = 7.2 (3.4, 15), no insurance or Medicaid only OR = 0.3 (0.1, 0.6), and increasing time from last menstrual period to first prenatal visit in weeks OR = 0.89 (0.85, 0.93). In conformity with previous results, ultrasound dating of pregnancy led to a higher estimate of preterm delivery (10 vs 7.6%), a higher estimate of term delivery (87.2 vs 82.7%) and a lower estimate of postterm delivery (2.8 vs 9.7%) than dating by menstrual history, p < 0.001. Selection factors and measurement issues, such as those described here, could introduce bias and should be carefully considered in the design, analysis and interpretation of perinatal research.