Preterm subtypes by immigrants’ length of residence in Norway: a population-based study

We linked birth and immigration data for all births occurring during a 20 year period among the six largest immigrant groups and used births among non-immigrants as the reference. Norway was a suitable setting for this study because nearly all births in this country occur within the public health sector. We matched birth records dated between January 1^st, 1990, and December 31^st, 2009, from the Medical Birth Registry of Norway (MBRN) with files from the National Population Register. The birth registry includes demographic data, information on maternal health before and during pregnancy, gestational length at birth and birth outcomes. The National Population Register contain information on the mother’s country of birth and country of origin, her first date of immigration to Norway and her education level. Unique personal identification numbers enabled linking of the two sets of registry data. The MBRN granted access to the birth data and the Regional Ethics Committee of Norway East approved the linkage to the Population Registry.

We extracted data from records of singleton births among non-immigrant women and among immigrant women from six countries outside Scandinavia, which enabled an analysis by country of birth [5]. We excluded stillbirths that occurred at <28 completed gestational weeks (0.4%) and births that occurred before registered immigration (1.5%). We further excluded cases that were missing information on gestational length (3.8%) and cases with improbable birthweights based on gestational length and sex (z-scores >4 standard deviations (SDs) away from the majority standard; 0.5%) [26]. The final sample included 40 709 births among immigrants and 868 832 births among non-immigrants. The six minority groups comprised mothers from Pakistan (n = 10 096), Somalia (n = 8 094), Vietnam (n = 6 336), Iraq (n = 5 879), Sri Lanka (n = 5235) and the Philippines (n = 5 069).

PTD was defined as delivery between completed gestational week 22 + 0 days and week 36 + 6 days. We classified PTD by mode of onset into ‘spontaneous PTD’ or ‘non-spontaneous PTD.’ Non-spontaneous PTD included births that required medically induced labor or a cesarean section before labor. Spontaneous PTD included births that began spontaneously, including preterm labor or prelabor rupture of membranes. Between 1990 and 1998, gestational length was determined based on the last menstrual period (LMP) reported at the first antenatal visit. From 1999 onward, findings from an early second-trimester routine ultrasound were primarily used (97.5%) to determine gestational length, and the LMP was used when ultrasound data were unavailable [27].

The main exposures were country of birth and length of residence. Country of birth was the country where the woman’s mother resided at the time of delivery. Length of residence was calculated as the interval between the year of first immigration to Norway and the year of delivery. In the final analyses the length of residence was categorized as 0–4, 5–9, 10–14 or ≥15 years. The first category (0–4 years) was further subdivided into <2 years and ≥2 years; however, estimates were the same for the two subcategories, so the subcategories were subsequently collapsed.

Based on the available literature, we selected co-factors associated with the length of residence and PTD [2, 28]. We extracted data on the following variables:

The means of the covariates were compared using the t-test/analysis of variance (ANOVA). Categorical covariates were compared using the χ² test and the χ² test for trends. To directly assess the relative risks (RRs) of the outcomes with 95% confidence intervals (CIs) we used univariable and multivariable Poisson regression in generalized linear models to estimate the associations between the length of residence and subtypes of PTD. Robust variance estimator (Huber-White) was applied to the models to adjust for data dependencies in mothers who gave birth more than once. See Tables 1 and 2 for categorizations of the covariates.

First, we compared non-immigrants with immigrants with various lengths of residence. In the first model, we adjusted for demographic covariates which included the year of delivery, maternal age, parity and the size of the labor unit. In the second model, we adjusted for maternal and fetal health-related covariates which included gestational and pregestational diabetes, hypertensive disorder, smoking (‘yes’, ‘no’ and ‘undisclosed’), placenta disorders, SGA, LGA and major birth defects. Second, we applied the same models to immigrants only, where we also included the country of birth in the model. We used recent immigrants (0–4 years) as the reference. For these analyses among immigrants only, we needed a country reference group and chose Iraqi women, as this group had similar PTD rates as Norwegian women. Due to the relationships between age at the time of immigration, the year of arrival and the covariates already in the model, these variables were added separately to the model as categorical covariates; however, the pattern of the results remained unchanged. Interaction terms between the length of residence and age at immigration and between the length of residence and the country of birth were not significant. Models were checked for goodness of fit. A two-sided p-value of 0.05 was considered statistically significant. The analyses were conducted using SPSS version 20 (IBM Corp., Armonk, NY, USA).

An application of the models to live births only did not change the pattern of results; this was also the case when we excluded major birth defects and when we restricted the sample to women living in the capital area. To reduce bias arising from the change in the method of determining gestational length, we limited the analyses to 1999–2009, which did not change the association pattern either. Moreover, in the analyses limited to immigrants, we stratified by early (gestational weeks 22–31) and late (gestational weeks 32–36) PTD to assess whether the relationship with the length of residence remained consistent.

Immigrants differed from non-immigrants with respect to demographic covariates and these characteristics were also associated with the length of residence (Table 1). The number of births in each minority group varied with the length of residence, reflecting changes in migration patterns to Norway over the 20-year period. The six minority groups analyzed comprised 2.6% of births in Norway in 1990 and 7.4% of births in 2009. Women from Pakistan and Vietnam had the highest proportion of the longest residencies (≥15 years) before delivery, while women from Somalia and Iraq had the highest proportion of the shortest residencies (0–4 years). Maternal and fetal health-related covariates associated with immigrants’ length of residence were gestational diabetes, pregestational diabetes, hypertensive disorder and an SGA or LGA infant (Table 2). There were significant differences in the gestational length at term between country groups (Table 3). Five of the six minority groups had a shorter mean gestational length in term spontaneous births compared with non-immigrants. Vietnamese women had the shortest gestational length, which was 6 days shorter than the longest gestational length, which was observed among Somali women. Gestational and pregestational diabetes and giving birth to an SGA infant were more common among most immigrant groups, whereas hypertensive disorders and giving birth to an LGA infant occurred less often in immigrants compared with non-immigrants (Tables 2 and 3). In univariate analyses, the risk factors for PTD were mainly similar for immigrants and non-immigrants, except for the level of education, where a high education level was protective among non-immigrants, whilst there was no association between education level and PTD among immigrants (data not shown).

Over the period under study, 48 191 cases of PTD were noted. Among both immigrants and non-immigrants, 70% of cases of PTD were classified as spontaneous, and 86% were classified as late PTD (weeks 32–36). Overall, the PTD rate was 29% higher among immigrants (6.8%) compared with non-immigrants (5.2%, Table 4). This difference was due to an increased rate of both subtypes of PTD. Filipino women were most likely to experience PTD, with a 52% increased rate compared with non-immigrants, whereas Somali women were the least likely to experience PTD.

The observed spontaneous PTD rate among immigrants was 21-57% higher across all categories of residence compared with the rate among non-immigrants (Table 4). Rates were stable across the length-of-residence categories among women from four of six countries; the exceptions were Sri Lankan women (among whom the rate increased) and Vietnamese women (among whom the rate decreased). Spontaneous PTD was most frequent among Vietnamese women (6.3) and the least frequent among Somali women (3.1), whilst the rate was 3.6 among non-immigrants. Although Somali women with the longest residencies (≥15 years) had a high spontaneous PTD rate of 7.7, this trend was non-significant across the length of residence. The spontaneous PTD rate was inversely related to the mean gestational length, in days, in term births in each minority group (R² = 0.905, p < 0.001; Figure 2).

In adjusted models, the likelihood of spontaneous PTD among immigrants remained 20-60% higher than among non-immigrants across all categories of residence (Figure 3). This pattern persisted in analyses limited to immigrants only (Table 5). Significant disparities by country of birth remained.

The observed non-spontaneous PTD rate among immigrants increased with the length of residence (p trend <0.001, Table 4). This pattern was found in women from four of the six countries; the exception was for Sri Lankan and Iraqi women; for the latter group a low number of cases destabilized the rate among the longest length-of-residence category. Overall, the observed non-spontaneous PTD rate was highest among Pakistani women (2.8) and lowest among Vietnamese women (1.2), whereas the rate was 1.6 among non-immigrants. The observed non-spontaneous rates by country group were not associated to the average gestational length (data not shown).Also in adjusted models the likelihood of experiencing non-spontaneous PTD increased linearly with longer length of residence (p trend <0.001, Figure 3). After adjusting for demographic factors, recent immigrants (residence of 0–4 years) had similar chances of experiencing non-spontaneous PTD as non-immigrants, whereas immigrants with lengths of residence of 5–9 years, 10–14 years and ≥15 years had 18% (aRR 1.18, 95% CI 1.03,1.35), 43% (aRR 1.43, 95% CI 1.20,1.71) and 66% (aRR 1.66, 95% CI 1.41,1.96) increased chances, respectively, of experiencing non-spontaneous PTD (Figure 3). When we additionally adjusted for health-related covariates, the effect measures were reduced, mainly due to the inclusion of pregestational diabetes and hypertensive disorders. These conditions were positively associated with increasing length of residence in the adjusted model.

In the models that included immigrants alone, we observed a similar pattern of results, even after adjusting for the country of birth (Table 5). Relative to immigrants with residence of 0–4 years there was a significant trend with increasing length of residence for non-spontaneous PTD, but not for spontaneous PTD. The association with residence was even stronger for early non-spontaneous PTD (weeks 22–31), which was 33-203% more likely among women with residence ≥5 years. In contrast, the results for late (weeks 32–36) non-spontaneous PTD did not differ from those of the overall analyses.

During the later period of the study, when we had information on gestational length that was determined using both dating methods, PTD rates among immigrants were 7.5% based on the LMP dating and 6.2% based on ultrasound dating (p < 0.001) (data not shown). The difference was due to a 21% reduction in spontaneous PTD rates, whereas non-spontaneous PTD rates were unaffected. Among non-immigrants, the PTD rate did not differ by dating method (both were 5.3%) (data not shown).

The influence of the length of residence on the likelihood of experiencing non-spontaneous PTD was robust across subanalyses and was independent of the age at immigration and the period of arrival. A trend in overall PTD was also present, consistent with findings from Canada [25, 29]. However, in our study, no subgroup of immigrants had lower risks of PTD than the host population, thus we did not find any evidence of a ‘healthy migrant’ effect. This could be due to the particular characteristics of immigrants that arrive to Norway. Unlike countries such as Canada, Norway did not allow unskilled labour immigration between 1975 and 1994, at which point Norway became associated to the European Union Single market [30]. Thus, there might have been less pressure towards selection to migration to Norway in comparison to other receiving countries. As to recent immigrants, we did not confirm an elevated PTD risk, as reported from Denmark [11].

We found that the association between the length of residence and non-spontaneous PTD was mainly explained by an increase in maternal and fetal risk factors. Of these, diabetes and hypertensive disorder had the strongest influence. In the general population, studies from the US have shown increasing occurrence of diabetes [22], obesity [21] and cardiovascular risk factors [23] across immigrants’ lengths of residence. A higher risk of illness during pregnancy with longer length of residence has been found in other migrant populations [31]. In pregnancy, diabetes and hypertensive disorders disturb the placental-fetal unit and both are common medical indications for the interruption of pregnancy before term [32, 33]. These disorders not only predispose the infant to preterm delivery but also directly contribute to mortality. However, if our findings were simply due to an increase in morbidity, we would expect to observe a similar association with spontaneous PTD, given the many shared determinants of the two subtypes [3, 34]. We did not observe this pattern. Our findings could therefore have been influenced also by a more frequent detection of risk conditions among women with longer residence. Improvements in the language skills of immigrants, as well as more knowledge on the part of providers, could improve communication and trust between the two. In turn, such processes are likely to affect the reporting and referral of complications [35].

After adjusting for a range of mediating factors, and relative to our reference of Iraqi women, Pakistani women retained an elevated risk of having a preterm obstetric intervention. This finding cannot be readily explained by our data and needs further investigation. Among other country groups, there were no significant differences relative to the reference.

Education level, as a proxy measure for SEP, is closely associated to adverse infant outcome [36]. In our study, information on maternal education level was missing among a high proportion of recent immigrants. Education level is not routinely registered among migrants upon arrival in Norway, but is gathered by questionnaires at a later stage, which explains this finding. The covariate was not included in the final models, as education level among immigrants was not a predictor for PTD in univariable analyses. This does not mean that there is an absence of a socioeconomic gradient in PTD among foreign-born women, but rather that the maternal educational level does not capture it in the same way as among native-born women [33, 37].

Spontaneous PTD receives a great deal of attention because it constitutes the majority of PTD cases [38]. Our results indicate that spontaneous PTD has limited value as an indicator of adverse neonatal outcome in ethnically mixed populations. After adjusting for a range of mediating factors and relative to our reference of Iraqi women, Vietnamese, Philippine, Sri Lankese and Pakistani women retained an elevated risk of having a spontaneous PTD. These minority groups in our sample were more likely to be registered as having experienced spontaneous PTD for two reasons. First, the likelihood of being classified as preterm for most minority groups corresponded with the mean duration of pregnancy in term births, suggesting a bias in the use of a uniform cut-off point for PTD. Second, a change in gestational age dating from ultrasound to LMP dating resulted in inflated PTD rates among immigrants, confirming previous findings among non-white women [39]. The change in dating method did not affect non-immigrants, which is consistent with previous findings that a change from LMP to ultrasound dating slightly increased rather than decreased the PTD rate [40, 41]. As gestational length has a strong genetic basis, this contributed to the persisting high spontaneous PTD rates we found in several minority groups. Somali women retained the lowest risk of all minority groups, which again is likely to be influenced by their longer duration of pregnancy compared to other groups.

The strengths of this study include the population-based design, with minimal selection bias; the high quality and completeness of the data on exposures (length of residence and country of birth); and the ability to distinguish between the modes of labor onset. The study included births over a 20 year period and although there was a secular increase in preterm caesarean section over this period due to improvements in neonatal care, these changes were adjusted for. Additionally, differential misclassification of the modes of labor onset and the occurrence of preterm prelabor rupture of membranes is believed to be minimal and unlikely to affect the observed associations. However, the present study also had limitations. First, we did not have information on covariates such as maternal height, BMI and previous PTD. In addition, the routine registration of maternal and fetal morbidity can be suboptimal; however, a validation of pregestational diabetes and hypertensive disorders in the MBRN was found to be satisfactory [42, 43]. Of SEP indicators, we only had information on maternal education level, and not on other indicators such as income or profession. Second, effects of the length of residence can be difficult to distinguish from cohort effects and the effect of the age at immigration; however, the associations found remained when controlling for these factors. Pakistani and Vietnamese women constituted a high proportion of women with the longest residences; this could have influenced the results for women in this category. Third, the routine determination of gestational length and fetal growth in Norway is interpreted according to majority standards. In addition to affecting preterm rates, it also affected classifications to SGA, and influenced the high prevalence of SGA infants born to foreign-born mothers [11, 25]. The alternative; country- or ethnic-specific standards [44], were not available to us, and their implications for clinical care is still controversial [15, 16]. To ultimately evaluate the clinical significance of the disparities in preterm subtypes, a determination of the optimal gestational ages of specific groups in terms of infant outcomes would be needed.