Study setting

Newcastle upon Tyne, located within the Northern Region of England, has a current population of approximately 260,000 inhabitants. The population structure of the Northern Region is characterised by the low percentage of ethnic minorities, about 2% [18], and its relative stability with low levels of in and out migration. For example, among nearly 5,000 children aged between 1 and 11 years recruited into a study from 1996 to 1997, over 85% had lived at their address for most of their lives [19]. Residential mobility in pregnancy is also low: only 9% of cases notified to the population-based Northern Congenital Abnormality Survey (NorCAS) [20, 21] moved from the time of booking to delivery (Rankin J, personal communication).

During the 50 years following the end of the Second World War, the economy of Newcastle transformed from one dominated by heavy industry and coal production and trade to a service based economy by the early 1990s. This paralleled remarkable changes in societal factors; for example, the 1967 Abortion Act, the National Health Service (Family Planning) Act (1967), availability of free family planning services irrespective of age or marital status from April 1974, the Sex Discrimination Act (1975) and the Employment Protection Act (1975) were introduced during the study period.

PAMPER birth population

The PAMPER database contains birth details on all singletons born during 1961–92 to mothers resident within the city of Newcastle upon Tyne in Northern England. Information on multiple births was also collected, however it was excluded from these analyses as multiplicity is a known risk factor for the outcomes of interest of the PAMPER study, i.e. preterm birth and low birthweight. The boundaries of the PAMPER study area are shown in Figure 1 with the river Tyne forming the southern boundary of the study area. The PAMPER computer database of birth records was constructed using information from a number of sources (Figure 2). The primary source was paper-based neonatal records from the two major maternity hospitals at the time (Princess Mary Maternity, PMMH, and Newcastle General Hospitals, NGH). From the PMMH, delivery and neonatal records were available for the whole study period, 1961–92; the NGH records were available from May 1967 onwards. These neonatal records contained information on important maternal and fetal/infant characteristics and clinical information about the delivery (Table 1). Socioeconomic information included paternal and maternal occupation, marital status and housing tenure.

	NGH and PMMH neonatal records	Tyne & Wear Archives birth records	Birth ledgers
Mother's current surname	√	√	√
Residential address	√	√	√
Baby's sex	√	√	√
Date of birth	√	√	√
Vital status at birth	√	√	√
Place of birth	√	√	√
Plurality	√	√	√
Birthweight	√	√	-
Gestational age	√	√	-
Maternal age	√	√	-
Parity	√	√	-
Mode of delivery	√	√	-
Baby's surname	√	-	-
Paternal occupation	√	-	-
Maternal occupation	√ (for 1976–92)	-	-
Admission to Special Care Baby Unit	√	-	-
Resuscitation	√	-	-
Early mortality data with cause of death	√	-	-
Hospital morbidity data	√	-	-

Note: The following additional data on maternal and child characteristics are available in the PAMPER database either for the whole study period or for shorter periods: time of birth, date of discharge, discharge weight, date of death (in case of infant deaths), maternal blood group, marital status, housing (for the 1960s), details of previous births, placental weight, onset of labour (spontaneous vs induced), Apgar score, type of feeding on discharge, estimated date of delivery.

To capture home births, we additionally abstracted data from 'birth ledgers' (1961–1973), containing limited information on all births (Table 1). This data allowed us to obtain complete denominator information and to consider the changing proportion of home births (Figure 3A). We also used NGH birth records stored in the Tyne & Wear Archives (available from 24th April 1961), to complement information on key variables unavailable in birth ledgers (Table 1).

Each birth was georeferenced by postcode and/or grid reference. For births between 1961 and 1970 (prior to the introduction of postcodes), the address at birth was assigned a postcode from the 1991 postcode book or a grid reference. This allowed us to locate enumeration district (ED) of mothers' place of residence and hence to obtain the Townsend Deprivation Score (TDS), an area-based measure of material deprivation [22], at ED level (about 450 people in 200 households). TDS were calculated based on the 1971 (birth years 1961 to 1976), 1981 (1977 to 1986) and 1991 (1987 to 1992) Census data on unemployment, car ownership, owner occupation and overcrowding.

Stillbirths and infant deaths

We linked the dataset to information on stillbirths and infant deaths (including causes of death) from the Office for National Statistics (ONS) and to death data from the Northern Perinatal Mortality Survey (PMS) (available from 1981 onwards) [23]. Multiple births were retained in the PAMPER database for the linkage procedure, but subsequently removed from the singletons database. Among a total of 1,248 eligible stillbirths provided by the ONS, we were able to match 1,222 cases (98%) to the PAMPER database. Among the total of 1,532 eligible ONS infant deaths, 1,510 (99%) were matched to the PAMPER database.

As Gosforth in the north and some western residential parts of the PAMPER study areas were not part of the city of Newcastle upon Tyne prior to 1974, the ONS could not provide us with all stillbirths and infant deaths for these areas for this earlier period. However, we obtained death certificates and causes of stillbirth for cases known to us to be stillbirths and infant deaths. This may still have resulted in some missing infant deaths if a postneonatal death was not recorded in the hospital notes.

PAMPER database completeness and accuracy

Data entry staff (twelve individuals working 3-hour shifts) were trained in the medical terms/abbreviations used in the neonatal records and thus the percentage of errors was minimised. SVG was responsible for completing a descriptive 'summary' field, which contained the medical diagnosis and causes of death. In addition to the ONS and PMS data, stillbirth and infant death data were validated using birth record sources mentioned above.

At the initial stage of data entry, we double entered approximately 1% of the estimated total of 120,000 birth records for different decades of the study period (n = 1,474) to assess accuracy of the data entry results. At the final stage of data entry, the data were validated by checking for implausible values (e.g. implausible difference between date of discharge and date of birth, implausible birthweight by gestation combinations).

For data capture we used the 4D database software suitable for a simultaneous data entry by several people, for data manipulation we used Microsoft Office Access 2003.

Definitions

Stillbirths included were all babies born dead at 28 or more completed weeks of gestation. There were 12 cases (1%) recorded as stillbirths by the ONS with uncertain gestational age which were also included. Stillbirths with birthweight less than 500 g were excluded if gestational age was unknown. Infant death was defined as a death, following live birth, of an infant under one year of age. We defined preterm birth as birth at a gestational age less than 37 completed weeks and term birth as birth at a gestational age ≥ 37 weeks.

Data analysis

For descriptive statistical analysis we used the statistical software package SPSS for Windows, version 14.0. We used chi-square tests to test differences in proportions and independent-sample t-tests for comparison of means.

Ethical approval

The study received a favourable ethical opinion from the Sunderland Local Research Ethics Committee (SLREC 1071).

Comparison with other studies

National trends on the total fertility rates for 1960–1990 mirror temporal trends shown in our study, where we used parity as a measure of fertility; during the 1960s 'baby boom', the national total fertility rates peaked in 1964 followed by a subsequent decline with a lowest level in the mid 1970s and a slight increase afterwards [24]. It has been suggested that the reduction in total fertility is attributable to improved means of fertility control (1967 Abortion Act and improved contraception efficacy) between 1967–68 and 1975. We also believe that the National Health Service (Family Planning) Act (1967), availability of free family planning services irrespective of age or marital status from April 1974, the Equal Pay Act 1970, the Sex Discrimination Act 1975 and the Employment Protection Act 1975, all contributed to women's reproductive decisions. This resulted in a decline in the proportion of teenage mothers and a parallel increase in the proportion of older mothers after the mid 1970s, as well as the overall increase in the mean maternal age in all and primiparous women. The increase in maternal age from the early 1980s was reported locally [6], nationally [25, 26], in Europe [27] and in the United States [28, 29]. Our data show that the mean maternal age in all and primiparous women was U-shaped with a declining trend from 1961 to the mid 1970s followed by a steady increase, repeating the national trend [24]. As advanced maternal age is associated with a higher risk of preterm birth and low birthweight [2, 8], its rise from the mid 1970s reported here may have contributed to the observed increase in the percentage of preterm birth in the last decade. Thus a study suggested that delayed childbearing may play an increasingly important role in low-birthweight trends in the United States [30].

We report a steady increase in the overall mean birthweight starting from the mid 1970s, which we observed for term births only and despite the increase in the proportion of preterm births in the second half of the study period. Hence, the observed rise in the total mean birthweight is likely to reflect the increase in birthweight for gestational age for term infants. This was also observed in Norway, where an increasing trend was reported for term births for 1967–1998 [31], but not preterm (22–32 weeks) which were heavier in the first decade compared to the last, in contrast to our findings. Similar trends were also observed in Canada from 1981 to 1997 where the increase in mean birthweight was restricted to term infants [32]. A study based on the Northern Region of England population, with Newcastle as part of this population, reported that the increasing trend in higher birthweights continued in the 1990s [6]. An increase in mean birthweight has been also observed in other parts of England [33], nationally [34] and in other Western countries [31, 32, 35].

The proportion of preterm births declined in the second decade compared to the first, but it was followed by a steeper increase in 1981–92. To our knowledge, there are no population-based studies from the UK for comparison. Studies from other countries also reported the increase in the percentage of preterm birth from the 1980s [31, 36, 37]. Several factors may have contributed to this increase. Thus there was a two-fold increase in the percentage of caesarean section among preterm births from the early 1970s to the 1990s, as with advances in neonatal technology, survival of extremely preterm infants dramatically increased, which justified interventions for fetal or maternal indications at earlier gestational ages. Similarly, in Norway the increase in the percentage of preterm births was attributable to a dramatic increase in the percentage of caesarean section among births delivered between 28 and 35 weeks in the late 1980s-1990s compared to the 1960s-1970s [31]. The increase in births to older mothers, which are associated with a higher risk of preterm birth and a higher percentage of caesarean section due to a higher rate of complications of pregnancy, may also have contributed to this increase. Another factor may be a wider use of assisted reproductive technology in the UK from the late 1980s [38], which is associated with a higher risk of preterm birth in singletons [39, 40] and is more widely used among older women.

Townsend deprivations scores, which we calculated for each birth in the database to measure neighbourhood socioeconomic status, also changed over time: the scores seemed to improve for the most affluent quintile and deteriorate for the most deprived, thus making the gap between the affluent and deprived groups wider. This is in line with the widening socio-economic and health inequalities which are now well documented in the UK.

Strengths and limitations of the PAMPER birth record database

The population-based PAMPER birth record database contains historical high-quality birth data in a defined compact geographical setting over a 32 year period during which there have been significant changes in obstetric and neonatal services. The completeness of the PAMPER database both for the number of births and information collected for each birth is a major strength. National and local trends in the number of births in the UK confirm the temporal fluctuations also observed in the PAMPER study: the highest number of births at the beginning of the 1960s (a so-called 'baby boom'), followed by a decline in the 1970s and a further increase in the number of births during the 1980s [41]. The completeness of the data for the key variables described here is expressed in the low percentages of missing data for these variables.

The availability of accurate population-based gestational age, a major determinant of birthweight, is one of the leading strengths of the PAMPER database, as gestational age was not available in national birth statistics during the study period. Further, birthweight for live births was not collected in the UK at national level until 1975 (as part of the Child Health Births Notifications System). Without gestational age, interpretation of trends in birthweight could be misleading, as it is not possible to disentangle whether changes in birthweight are attributable to changes in rates of preterm birth or to changes in actual fetal growth. However, in the UK and elsewhere in the world there is a lack of information on the incidence of premature birth using accurate data by gestation [15].

The accuracy of the data for the key variables was ensured by multiple checking, internal (within the database) and external (with national and regional death data, and other local sources of birth record data) validation of the data.

The PAMPER database also has several limitations. The lack of information on some important determinants of fetal weight at birth such as maternal height, maternal smoking and exposure to environmental tobacco smoke, which have changed over time thereby affecting changes in birthweight, is disappointing. For example, an increasing trend in maternal height was reported in Scotland for 1980–2000 [42]. In the UK, the prevalence of smoking in women increased sharply during and after the Second World War, reaching the level of about 42–44% in the 1960s – early 1970s [43, 44] followed by a gradual decrease thereafter [44]. However, adjustment for year of birth should be able to control for the effect of temporal changes in any factors influencing birth outcomes.

The accuracy of gestational age estimates is important for epidemiologic studies of pregnancy outcomes. Different methods for gestational age assessment (based on the last normal menstrual period (LMP) or early ultrasound measurements) throughout the study period may introduce bias in gestational age estimation over time. Thus it has been suggested that higher rates of preterm birth may be reported if determination of gestational age is based on ultrasonographic dating alone [45, 46]. In the 1960s and 1970s, when gestational age estimate was based on LMP and, if the dates were uncertain, on the paediatric examination of the baby, it may have more uncertainty. However, while creating our birth record database, we made the recording of gestational age as objective and accurate as possible by accepting gestational age calculated from the recorded estimated date of delivery (EDD) (i.e. LMP based) for the majority of births rather than by entering gestational age recorded in the neonatal notes or birth records. For example, the percentage of gestational age records based on the recorded EDD for 1961–70 was about 87% of records with known gestational age. In this study the ultrasound age estimate has been used since the early 1980s only for pregnancies with uncertain date of LMP or if there was a significant discrepancy between the two estimates, therefore it should not bias gestational age estimates over time. Moreover, gestational age seems to be accurate in our study as birthweight distribution at early gestational ages has a single mode in contrast to other studies reporting bimodal birthweight distributions at early gestations with implausible high birthweights for gestational age [16, 17].