The NorCAS is a prospective, population-based registry covering the former northern health region, which includes the northern counties of eastern England and northern Cumbria [see Additional file 1], and for this paper is referred to as the north of England. This region comprises a population of about three million, with approximately 35 000 births each year during the study period, of which approximately 780 births each year (2.2%) included a congenital anomaly and were therefore recorded in NorCAS [18, 19]. Data are collected from this population on congenital anomalies occurring in late miscarriages at 20 weeks or more gestation, in live births and stillbirths, and in terminations of pregnancy for fetal anomaly after prenatal diagnosis, whatever the gestation at diagnosis [19, 20]. Cases are reported to the register from multiple sources including antenatal ultrasound, fetal medicine, cytogenetic laboratories, the regional cardiology centre, pathology departments and paediatric surgery to ensure a high case ascertainment. Details concerning the method for data collection have been described previously [19, 20]. For this study, data on all pregnancies with a congenital anomaly delivered between 1st Jan. 1985 and 31st Dec. 2003 were extracted from NorCAS.
The NorCAS is one of only a handful of routinely collected health datasets to hold information on residential location of each individual at more than one point in time. Details of maternal address at booking appointment (the first official check-up in pregnancy, typically at a gestational age of 13 weeks) and delivery are recorded, providing an opportunity to explore residential mobility. We acknowledge that the NorCAS data represent a specific subset of pregnancies, that this dataset may not therefore be wholly representative of all pregnancies occurring in the region, and that migration may differ between pregnancies resulting in an infant with a congenital anomaly versus those resulting in a healthy infant (see discussion).
Eligible cases were those with a gestational age at delivery of ≥ 24 weeks (a viable delivery) [21]. This cut-off was chosen as 1) it excludes most terminations and more severe (spontaneously aborted) anomalies, and 2) a bimodal plot of gestational age at delivery is more typical of congenital anomaly pregnancies. The aim was to make the findings more easily generalised to the majority of pregnancies in the region that result in a healthy baby. Where a pregnancy resulted in more than one case being entered into NorCAS (i.e. multiple pregnancies with more than one baby with anomalies), the pregnancy was counted only once (46 pregnancies affected).
The addresses at booking and delivery were geocoded based on the address postcode, a unit of postal geography comprising approximately 15 households [22]. Grid references were assigned to the postcode centroid, which represents the geographic centre of the collection of adjacent addresses making up the postcode. Grid references were obtained from the ONS All Fields Postcode Directory [23], with postcode grid references at a 100 metre resolution for the years 1985-1997, and a 1 meter resolution for the years 1998-2003. The first line of each woman's address (house number/name and street) was also compared to enable the detection of a very local move (i.e. within the same postcode), although in these cases it was not possible to calculate the distance moved.
Data on maternal age (available for 97.5% of eligible cases) was also used in this analysis. As individual level data on socio-economic status were not collected, area level census-derived socio-economic indicators (the Index of Multiple Deprivation score 2004 (IMD) [24] and Townsend deprivation score (TDS) [25]) were assigned to each mother at booking and delivery. A TDS (based on unemployment, car ownership, owner occupation, and overcrowding) calculated at the electoral ward level (a unit of administrative geography containing on average 5 500 people [22]) was assigned to each postcode based on the electoral ward the postcode fell within, and using data from the relevant Census (i.e. 1981 Census data for 1985; 1991 Census data for the years 1986-1995; 2001 Census for the years 1996-2003). An IMD score (based on crime, education, skills and training, employment, health and disability, housing and services, income, and living environment) calculated at the lower layer super output area (SOA) level (a census based geography containing on average 1 500 people [22]) was assigned to each postcode based on the SOA the postcode fell within using data from the 2001 Census.
All statistical analyses were performed using the statistical software package SPSS 15.0 (SPSS Inc., Chicago). The Chi-square test was used to compare differences in proportions, independent sample t-tests to compare means, and Pearson correlation coefficients to explore correlations between variables. Multivariable logistic regression was used to explore determinants of mobility using 'residential move' as the binary outcome; variables were entered into the model simultaneously. For the categorical variables maternal age and IMD score, where a linear dose-response relationship seemed apparent by quintiles, these variables were re-entered into the model as continuous variables to maximize power in assessing the relationship.
In almost a third of cases the address at delivery was missing, mostly over the period 1987 to 1998 (47% missing). This missing data could introduce bias into our estimates of mobility, and if the address at delivery is more likely to be missing in those cases making a residential move (i.e. is differential with respect to mobility) this could result in a substantial underestimation of mobility in this population. In addition, over the period 1999-2003, we were not able to differentiate between a woman having the same address at booking and delivery and a woman having an unknown address at delivery because of a change in data entry method (the address at booking was automatically copied to the address at delivery field, and updated only if a move was recorded to have taken place). Again, this may result in an over-estimate of the number of non movers for these years. A sensitivity analysis was undertaken to assess the impact on mobility of applying different assumptions to these apparent non movers over the period 1999-2003, and to the missing data over the period 1987-1998. In addition, a small validation exercise was undertaken by cross-referencing a sample of randomly selected cases with the UK National Health Service National Strategic Tracing Service (NSTS) records. From this validation it was possible to assess whether moves had taken place during pregnancy when the NorCAS addresses at booking and delivery were the same, either because no move had taken place, or because they were automatically copied across and would previously have been unknown.
Ethical approval
The NorCAS has Patient Information Advisory Group exemption from a requirement for consent for inclusion on the register under section 60 of the Health and Social Care Act (2001) and has ethics approval (04/MRE04/25) to undertake studies involving the use of its data.