METHODS

Wirral, the area studied, has a stable population of about 340 000 with fewer than 1% from ethnic minorities. Socioeconomically the population structure is comparable to all England. Over the 12 years of the study there were no changes in the boundaries of the district.

Data on live births to Wirral residents between 1 January 1990 and 31 December 2001 were obtained from the Wirral Hospital database. This database includes births outside the district to Wirral residents. Data collected included the home postcode of the mother at the time of birth, birth weight, gestational age, sex, and whether the pregnancy was single or multiple.

Birth weight was measured electronically with the baby naked. For birth weight measurements, only singleton births were analysed. Ideally gestational age is measured using an ultrasound scan. Early in the study, scans were not routinely carried out on mothers, and the gestational age was calculated from the date of the last menstrual period. Routine scanning at 12 weeks gestation was introduced in the later years of the study and was carried out in all pregnancies from 1999 onwards. The birth weights and gestational age were converted into standard deviation scores (SDS) using software obtained from the Child Growth Foundation.⁷

The mothers’ postcodes were converted into Townsend material deprivation scores,⁴ using the Postcode to Enumerated District data set obtained from Manchester Information and Associated Services.⁸ These scores were then ranked in quartiles, the upper quartile indicating the most deprived quartile.

Relations between ordered birth counts (by Townsend quartile and year) were examined using the Goodman-Kruskal γ statistic as a measure of ordinal association with a 95% confidence interval.⁹ Relations between continuous data (weights and SDS) and ordered categorical data (Townsend quartile) were examined using Kendall’s τ-b measure of rank interdependence.¹⁰ The γ and τ-b statistics measure the strength of association between two variables without assuming any particular distributions of the data. Time trends in weights and SDS by deprivation group were examined using analysis of covariance.¹¹ Calculations were performed using StatsDirect software.¹² Calculated probabilities are quoted to a minimum of 0.001, values below which we refer to as highly significant.

Ethical approval to carry out this research was obtained from the Wirral ethics committee.

RESULTS

There were 48 452 live births to Wirral residents from 1990 to 2001. The annual number of live births fell from 5013 in 1990 to 3137 in 2001 (37%). Over the study period there was a 6% fall in the population of female residents aged 15–45. The 1193 multiple births during the study period were excluded from the main analysis. A further 3044 births had no birth weight recorded in the database, and 3245 records did not have a valid postcode. The overall number of exclusions was 7959, representing 16.4% of all births during the study period. Quality of data coding and processing improved progressively, with 21% of data excluded in 1990 falling to 9% in 2001.

The number of singleton live births fell consistently during the study period (fig 1). The overall percentage fall was 28%. The change in the lower Townsend (least deprived) quartile (45%) was more than triple that in the upper (most deprived) quartile (14%). There was a highly significant association between Townsend quartile and the reduction in birth numbers over time (γ = 0.045; 95% confidence interval (CI) = 0.036 to 0.054; p < 0.001). There was a significant fall in the proportion of male births, from 52.5% in 1990 to 50.9% in 2001 (table 1) (γ = 0.018; 95% CI = 0.006 to 0.030; p = 0.04).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Change in the number of singleton births in each Townsend quartile between 1990 and 2001.

There was a small and non-significant increase in the annual mean birth weight (table 2). Throughout the study period, mean birth weight was 141 g higher in the lower than the upper Townsend quartile (table 2). There was a highly significant association between birth weight and Townsend quartile (τ-b = −0.66; 95% CI = −0.72 to −0.60; p < 0.001).

Table 3 shows the mean SDS for birth weights in the four Townsend quartiles. The SDS corrects for the gestational age of the infant. A SDS of 0 represents the 50th centile, and negative scores indicate that the infant was below the national mean birth weight for gestational age. The only negative scores were seen in the upper quartile. There was a small, non-significant increase in the mean SDS for all babies born during the study period. There was a highly significant association between SDS for birth weight and Townsend quartile (τ-b = −0.062; 95% CI= −0.068 to −0.055; p < 0.001).

Table 4 shows the numbers and percentage of singleton low (< 2500 g) and very low (< 1500 g) birth weights. The absolute number of low birth weight babies fell significantly (τ-b = −0.56; 95% CI = −0.99 to −0.14; p = 0.009), although as a percentage of singleton live births there has been no significant change. No significant time trends were seen for very low birthweight babies.

There was a highly significant association between Townsend quartile and annual trend in numbers of low birth weight (γ = 0.08; 95% CI = 0.04 to 0.12; p < 0.001) but not of very low birth weight (γ = 0.07; 95% CI = −0.06 to 0.20; p < 0.3) babies. Numbers of low birth weight babies in the least deprived quartile fell disproportionately compared with those from the most deprived quartile (γ = 0.17; 95% CI = 0.09 to 0.25; p < 0.001). The proportion of low birth weight babies in the least deprived quartile fell significantly (γ = 0.08; 95% CI = 0.01 to 0.15; p = 0.03), whereas the same proportion from the most deprived quartile did not (γ = −0.01; 95% CI = −0.06 to 0.04; p = 0.66). There was no significant difference between these quartiles for changes over time in the numbers of very low birth weight babies. The results for the very low birth weight data, considering the small numbers involved, did not change by using exact (permutational) methods of analysis.

DISCUSSION

Studying trends in births between 1990 and 2001, we have shown that the fall in crude birth rate for singleton births on the Wirral (28%) was greater than that for England and Wales (12%). The significant change in the male to female ratio follows a pattern seen throughout Europe for which there is no apparent explanation.¹³ There was a considerable socioeconomic difference in that the crude birth rate in the least deprived areas fell by three times (45%) the rate in the most deprived areas (14%).

We have also shown a socioeconomic difference in birth weight trends. When these were corrected for gestational age using SDS, the significant upward trend in mean SDS for birth weight in the least deprived quartile was matched with virtually no change in the most deprived quartile. Elmén et al¹⁴ have suggested that birth weight standardised for gestational age is an important health indicator, as it correlates closely with socioeconomic area indicators and mortality. There was also a significant reduction in the percentage of babies in the least deprived quartile weighing less than 2500 g, whereas the percentage in the most deprived quartile did not change significantly. No significant time trends were seen for babies weighing less than 1500 g, but the numbers involved were not large enough to give strong conclusions.

Spencer et al,¹⁵ using the Townsend material deprivation score, found no social trends in numbers of live births in Sheffield between 1985 and 1994. The difference from our results may reflect an interdistrict difference or we may have identified a more recent trend. This could be addressed by monitoring national trends in birth numbers by postcode and area deprivation measurement.

Our findings, which showed a smaller reduction in the number of low birth weight babies in the most deprived quartile, are consistent with the ONS report.² Earlier research had suggested a national increase in mean birth weight in the preceding 30 years, particularly among the most disadvantaged.¹⁶

The smaller reduction in births in the most deprived quartile has great public health significance. Bambang et al⁵ reported a significant association between area based deprivation and perinatal deaths in the West Midlands. Guildea et al⁶ reported similar findings in Wales and also showed that deprivation was a significant risk factor for post-neonatal infant mortality. Also, Dowding and Barry¹⁷ have shown a clear social gradient in the overall prevalence of cerebral palsy in the Eastern Health Board in Ireland.

Low birth weight is associated with other adversity. Richards et al,¹⁸ in a longitudinal study of the 1946 British birth cohort, showed that birth weight was significantly positively correlated with cognitive function and educational attainment at age 8 years. Matte et al¹⁹ have recently shown that even in the normal ranges of birth weight there is a proportional increase in intelligence quotient (IQ) with increasing weight.

The trend that we have described in low birth weight has implications for later life, as low birth weight has been implicated as a risk factor for a number of chronic adult diseases.²⁰ While biological programming may mediate the link between low birth weight and adult disease, socioeconomic factors may also contribute. Bartley et al²¹ showed that low birth weight is associated with socioeconomic disadvantage in childhood and adolescence and with financial disadvantage at age 23 years.

There is evidence that targeting areas of deprivation may improve birth outcomes. Oakley et al,²² in a randomised controlled trial comparing antenatal social support with standard antenatal care in a socially disadvantaged population, showed that mothers receiving the intervention had babies with a higher mean birth weight, and fewer very low birth weight babies. The intervention mothers required fewer hospital admissions during pregnancy. They were more likely to experience spontaneous onset of labour and were less likely to require epidural anaesthesia. Intervention group mothers and babies were significantly healthier in the early postnatal weeks than controls. Teranishi et al²³ examined the influence of socioeconomic status on growth to 23 years in the 1958 British birth cohort. For low birth weight babies, catch up growth in height was more pronounced in social classes I and II than in other classes.

There are a number of potential weaknesses in our study. Did the exclusion of 6766 singleton births invalidate the deprivation score analysis? Analysis of the distribution of Townsend scores in the 2893 exclusions where there was a valid postcode showed that 949 (10.2%) were from the least deprived Townsend quartile and 522 (4.9%) were from the most deprived quartile. This indicates a potential bias, greatest in the least deprived group, that could have reduced the number of births in the early years of the study when the documentation of postcodes was least reliable. Reliance on the last menstrual period to calculate gestational age rather than ultrasound scanning could have influenced the calculation of the SDS for birth weight, as it tends to decrease the SDS.¹⁴ However, when the mean SDS for birth weight in the first six years of the study (when gestational age was calculated from the date of the last menstrual period) was compared with the last six years of the study no difference could be found.

It is not clear whether our findings are specific to the Wirral, or whether similar trends have occurred nationally. We suggest that routinely collected birth information and a suitable area based deprivation score should be more widely used to monitor trends. Currently, reliance on joint registration of births using the father’s occupational social class means that the most vulnerable group of mothers and babies may be missed.

Conclusion

We have shown that in the Wirral over the last 12 years, the reduction in live births has not been evenly distributed among social groups but has been significantly less pronounced in the most deprived areas. This increasing social inequality is further aggravated by the fact that the proportion of low birth weight babies has risen significantly in the most deprived areas of the Wirral compared with the least deprived areas. These public health findings are important to policy makers, as well as clinicians in maternity and primary and secondary care.