Identification of response bias on apparent pregnancy outcome after second trimester ultrasound

Objectives: To investigate whether there is a response bias in outcome studies after prenatal ultrasound and to quantify this potential effect by information source.

Materials and methods: All normal ultrasound scans between week 17 and week 24 and 6 days performed in the years 2004 and 2005 were investigated. A multinomial logistic regression model was applied to investigate the association between responders' outcome (questionnaire, phone interview and inquiry to birth clinic) and the following explanatory variables: maternal age, smoking status, body mass index, congenital anomaly status, low birthweight and preterm deliveries.

Results: From the 12 439 women, 7747 (62.3%) sent back the questionnaire, 3032 (24.4%) were interviewed by telephone and in 1660 cases (13.3%) the outcome was obtained from the birth clinic. Maternal age > 34 years [odds ratio (OR) 0.72, confidence interval (CI) 0.61-0.85/0.35, CI 0.29-0.42, telephone/birth clinic] and minor anomalies (OR 0.52, CI 0.28-0.98, birth clinic) were significantly underrepresented in nonresponders. Preterm delivery (OR 1.29, CI 1.11-1.50/1.30, CI 1.08-1.57), maternal smoking (OR 1.14, CI 1.07-1.25/1.31, CI 1.22-1.40) and stillbirths (OR 2.30, CI 1.09-4.87, birth clinic) were significantly, major anomalies (OR 1.83, CI 0.94-3.55/1.80, CI 0.79-4.10) were considerably overrepresented in these groups.

Conclusion: Spontaneous responding to prenatal follow-up questionnaires is significantly biased towards older and nonsmoking mothers with normal pregnancy outcome.