Change in social status and risk of low birth weight in Denmark: population based cohort study

Introduction

Low birth weight (birth weight <2500 g) is one of the main risk factors for infant mortality. Low birth weight occurs frequently in industrialised countries—for example in 5.3% of all births in Denmark.1 Fetal growth is determined by genetic as well as other factors2 3 4 5 6 and low birth weight and impaired fetal growth run in families.4 7 8 9 10 11 Maternal height and weight, socioeconomic group, smoking,2 5 12 13 and the intrauterine environment3 5 14 15 are important determinants of birth weight. Height of the father is correlated with the birth weight of the infant,5 16 but less so than height of the mother, and the correlation may depend on assortative mating—that is, women tend to choose men of similar height and weight as themselves.3 Studies of twins indicate that 40% of the variation in birth weight is attributable to genetic factors.15 However, the aetiology of birth weight in twins differs from that of birth weight in singletons.17

We studied the effect on birth weight in a subsequent pregnancy of changes in putative risk factors between pregnancies.18 Couples who have had a low birthweight infant are assumed to have been exposed to a sufficient set of causal factors to trigger this outcome (exposed cohort). Couples who had had a normal birthweight infant (unexposed cohort) are assumed not to have been exposed to a sufficient set of causes. The effect of eliminating or introducing possible exposures in the subsequent pregnancy was also studied. A similar design was used by Lie et al to study the recurrence of birth defects.19

Subjects and methods

Data were obtained from the Danish medical birth registry and the national bureau of statistics. The birth registry, established in 1973, contains data on all births in Denmark; 24 077 low birthweight infants born between 1980 and 1992 were identified among mothers who had at least two children.

All people born from 1945 onward and who are considered to be of reproductive age are included on the fertility database. This database links several databases to obtain the most complete possible data on family composition, cohabitation, education, and employment.20 Since 1968 all residents have been assigned a unique identification number at birth. We used this unique number to link children from the birth registry with their biological father, to obtain information on the employment status of both parents, and to determine the stability of the mother's relationship with her partner.

The exposed cohort consisted of all women who had given birth between 1980 and 1992 to a singleton weighing <2500 g (index child) and then to a subsequent liveborn infant (outcome child) (11 069). A 5% random sample of the general population of women who had had at least two singleton births during the same period served as the unexposed cohort (10 211). Since the time of death is often unknown for stillborn infants, they were excluded from analysis both in the index and outcome pregnancies.

The risk of having a low birthweight infant after the birth of the index infant was estimated as a function of the mother's status with respect to changes in partner, residence in one of Denmark's 275 municipalities, type of job, and social status between the births. The comparisons were performed within each cohort; couples without changes in any of the factors served as the reference group.

Social status was categorised as low, middle, or high, according to the job held at the time of pregnancy on the basis of the 10 point classification used by Statistics Denmark. Subjects were grouped by social status before analysis. The type of job was classified on the basis of the International Standard Classification of Industries.21

The partner with the highest social status determined the couple's social status at each birth. If data on both partners were missing in both pregnancies the couple was defined as not having experienced a change in social status (54 in the exposed cohort, 27 in the unexposed cohort). The mother's social status was used for women who were not cohabiting: 588 women in the exposed cohort and 351 in the unexposed cohort were not cohabiting within a year of the birth of the outcome child. The category of low social class included those who were unemployed or retired, those who performed unskilled manual work, and those with unspecified and unknown jobs. The middle category included office workers, students, skilled manual workers, and those working for or with their spouse or partner. The high social class category included all high ranking managers (responsible for at least 20 employees), high ranking office workers, (includes professors, doctors in a hospital, etc), self employed office workers (includes lawyers, accountants, business people, and medical specialists who own their own firm or practice), and owners of small businesses or shops.

Changes were examined using a logistic regression model in which low birth weight was the outcome variable and results were adjusted for potential confounders (parity, age of the mother, gestational age, gestational age squared (to obtain better control of confounding by a variable that is not linearly associated with the outcome on a logarithmic scale), interpregnancy interval (time from a birth to the next conception; ≤4 months v >4 months), and social status at the birth of the index child). Birth weight was also dichotomised at ≤2000 g. Additional analyses were made for the first two children and for outcome children born at term (39-41 weeks). A description of the two cohorts is given in Table 1.

To examine the effect of social mobility and change in partners on continuous birth weight we used a multiple regression model for each cohort after adjustment for the most important determinants.

Results

The overall risk of having an infant with low birth weight in the outcome pregnancy was 18.5% in the exposed cohort and 2.8% in the unexposed cohort. Women in the exposed cohort were younger, had lower social status, a lower degree of upward social mobility, and a higher rate of changing partner. In this cohort the biological father was less frequently recorded on the birth certificate. The average interpregnancy interval was 2.7 years in both cohorts (Table 1).

There was a higher risk of having a low birthweight infant when the social status of the couple declined between the two births and a lower risk when social status rose; this was significant in the unexposed cohort (Table 2). In the unexposed cohort the effect of changing social status was substantially the same regardless of the starting social status. In the exposed cohort the effect is mainly seen for those who changed from middle to low status. Adjustment for the cohabitation status of the mother did not change the impact of social mobility in either of the cohorts. No significant variation in risk was associated with change of male partner, municipality, or type of job.

Analyses of different subgroups of the two cohorts are summarised in Table 3. When analyses were restricted to first and second children and to children born at term, a decline in social status was a significant risk factor only in the unexposed cohort. A rise in social status was associated with a decreased risk of low birth weight in children born at term in both cohorts. Among the 4036 women in the exposed cohort whose index child weighed ≤2000 g, and the effect of a decline in social status was significantly less than for infants with a birth weight of <2500 g.

In the unexposed cohort the outcome infant was on average 80 g heavier than the index infant, but among infants born to women who had had a decline in social status the outcome infant was only about 40 g heavier (Table 4).

On multiple regression analysis gestational age and birth weight of the index infant were the most important predictors of birth weight for the outcome infant, but social mobility and social status were also associated with birth weight (Table 5). In the unexposed cohort the correlation coefficients between the two pregnancies calculated separately for women who changed partners (r=0.372; 0.314 to 0.329) and those who did not (r=0.471; 0.455 to 0.487) indicated more similarity in birth weight among those who did not change partners between the two pregnancies than among those who did.

Discussion

Our study corroborates the finding that women who have a low birthweight infant are more likely to give birth to another low birthweight infant than women who have had infants of normal birth weight.9 10 11 The most likely cause is related to maternal genes or the intrauterine environment. A decline in social status was found to be a comparatively strong predictor of low birth weight, which suggests that fetal growth is reduced under poor social circumstances. All results were adjusted for gestational age, and the effect was also seen in the subgroup of outcome children born at term in the unexposed cohort. In the exposed cohort the effect of a decline in social status was larger for newborn infants with a birth weight of ≤2000 g.

In the unexposed cohort a decline in social status was more closely associated with low birth weight than social status at the time of birth of the index child. The odds ratio of 1.24 in the exposed cohort represents an increase from 20% to almost 25% in the risk of having another low birthweight infant when compared with couples for whom none of the indicators of risk changed. The odds ratio of 2.32 in the unexposed cohort is a similar increase in excess risk (from 4.5% to 10%). The findings from the two cohorts are therefore similar on an additive scale, but the association was significant only in the unexposed cohort.

Many expect that lone mothers are at a higher risk of having a low birthweight infant, but including cohabitation status in the analysis did not change the risk associated with social mobility in either cohort.

In some cases poor health may have triggered a decline in social status,22 23 24 and health conditions rather than the change in social status may have caused low birth weights. These factors would, however, explain but a small part of the association since only severe diseases trigger a decline in social status in Denmark, which has an extensive social support system. Most women with health poor enough to initiate a decline in social status would probably also have low fertility and therefore not be considered in this study. Changes in social status have not previously been associated with low birth weight, but unemployment and low social status have been linked with poor reproductive outcome in several studies.12 13 25

The impact of paternal factors on birth weight remains unclear, but maternal half siblings have a much higher correlation for birth weight than paternal half siblings,3 14 and this is supported by our findings of little effect on birth weight of a change in partner. A change in male partner is not, however, expected to change the distribution of birth weight. If a paternal effect is randomly distributed, the average difference in birth weight between the two pregnancies would be expected to be unaffected by a change in male partner, but the correlation coefficient would be expected to be lower for those who change partner, which is what this study found.

Only limited data were available on environmental exposures, and these factors are probably more important in fetal growth than those that could be captured in this study. Area of residence is, for example, only a proxy measure for some environmental exposures such as clean drinking water or air pollution. A decline in social status may be related to lack of compliance in medical care26, and smoking or other factors that are linked with growth retardation may be related to downward social mobility, as they are for low social status.12 This study does not permit identification of the actual determinants responsible for low birth weight that may be related to a decline in social status.

Our study analysed a large cohort and was carried out without loss to follow up. The data are of good quality with regard to the timing of pregnancies, paternity, birth weight, residence, and job titles. Further research should identify changes in proximal fetal growth factors influenced by changes in social status. Women in poor social conditions should be given special attention during antenatal care.