Residential mobility during pregnancy in the north of England

BACKGROUND: Many epidemiological studies assign exposure to an individual's residence at a single time point, such as birth or death. This approach makes no allowance for migration and may result in exposure error, leading to reduced study power and biased risk estimates. Pregnancy outcomes are less susceptible to this bias, however data from North American populations indicate that pregnant women are a highly mobile group. We assessed mobility in pregnant women in the north of England using data from the Northern Congenital Abnormality Survey (NorCAS). METHODS: Data were extracted from NorCAS for 1985 to 2003. Eligible cases had a gestational age at delivery of ≥ 24 weeks (a viable delivery) (n = 11 559). We assessed mobility between booking appointment (average gestational age 13 weeks) and delivery for pregnancies where the address at booking appointment and delivery were known. The impacts on mobility of maternal age and area-level socio-economic indicators were explored using standard descriptive statistics. A sensitivity analysis and a small validation exercise were undertaken to assess the impact of missing data on the estimate of mobility. RESULTS: Out of 7 919 eligible cases for whom addresses at booking and delivery were known, 705 (8.9% (95% CI 8.3 - 9.5)) moved between booking and delivery; the mean and median moving distance was 9.7 and 1.4 km respectively. Movers were significantly younger (25.4 versus 27.3 years, p < 0.01) and lived in more deprived areas (index of multiple deprivation score 38.3 versus 33.7, p < 0.01) than non movers. CONCLUSION: Mobility in the north of England (9%) is considerably lower than that reported in North America and the only other study from the UK (23%). Consistent with other studies, mobility was related to maternal age and socio-economic status, and the majority of moves were over a relatively short distance. Although this population appears relatively stable, the mobility we have observed may still introduce misclassification or error into an exposure assessment relying solely on postcode at delivery, and migration should still therefore be considered a potential source of bias in future studies.