Our results show that DA average income is a useful indicator of area-level socioeconomic SGA inequality in Vancouver. Our finding of a statistically significant association of SGA by average income quintile is consistent with previous studies conducted using both area-level and individual-level income measures. In Canada, higher odds of preterm birth and SGA have been associated with lower average income even after adjustment for individual level characteristics [9, 10]. However, one study found individual-level maternal education to be independently and more strongly associated with both birth outcomes than average income . Unfortunately, our Registry data did not include individual-level maternal education, so we were unable to investigate this association. In the UK, it was found that there were higher odds of LBW in areas with greater area-level deprivation even after controlling for individual-level variables that were known risk factors for LBW [4, 7, 12]. Similarly, a study in the Netherlands found that neighbourhoods with lower median income had higher odds of SGA after adjustment for individual-level factors . In contrast to these studies, we included smoking during pregnancy, which has previously been associated with SGA [15, 25–27].
We found that mothers who lived in areas with lower average income were shorter and half of SGA cases were attributable to shorter height according to PAR%. It is important to consider some strong assumptions and unique characteristics of PAR% when interpreting this result. One assumption is that maternal height is a cause of SGA risk. Others are that additional risk factors are randomly distributed in the population and an intervention on maternal height would leave those distributions unchanged [38, 39]. These latter two assumptions are unlikely to be true and are sufficient reason to avoid a strong interpretation of the PAR% as a prediction of what would result after intervention on height. Finally, we note that PAR% is a function not only of the magnitude of a factor’s association with the outcome, but also the prevalence of the factor in the population . The large PAR% for height reflects, in part, our broad definition of shorter height, which resulted in high population prevalence. Because of these considerations, we interpret this result for height more qualitatively than quantitatively. Height is an important factor to consider in socioeconomic SGA inequality in greater Vancouver even if we do not know precisely the degree to which SGA risk (as defined in this analysis) would change in particular populations if average maternal height increased in future generations.
Despite these cautions of interpretation, our results may indicate a trans-generational aetiology for socioeconomic SGA inequalities in Vancouver. Based on evolutionary and human life-course theories, foetal growth is modulated in response to signals about the energy stores of the mother, such as maternal height (correlated with lean mass) and adipose tissue mass . Maternal height and adipose tissue differ, though, in their plasticity in adulthood. Whereas adipose tissue mass can increase or decrease throughout life in response to energy intake, human growth is thought to be canalized by age two (although some catch-up in height can occur during adolescence) [41, 42]. As such, adult height changes over generations rather than within individuals’ lives and is an indicator of multigenerational histories of conditions that facilitate growth . Adult height at a population level is known to increase or decrease over generations in response environmental conditions, but is constrained by the height of the parental generation . Our finding that mothers from areas with lower average income were shorter possibly indicates that historical variability between family lineages in access to favourable environmental conditions is mirrored within the social strata of Vancouver today. As such, the socioeconomic SGA inequality that we observed may be a consequence, in part, of socioeconomic differentials in historical access to environmental resources favouring growth. Another possible hypothesis is that taller height itself confers a competitive advantage in Vancouver’s society leading to greater upward social mobility.
However, our findings suggest that additional historical or contemporary causes of socioeconomic SGA inequality may also exist as evidenced by the strong negative association between current average income and SGA even after adjustment for maternal risk factors including height. Furthermore, average income had an adjusted PAR% of 14% for SGA, which was larger than that of any other risk factors except height and parity. Again, this result should be interpreted cautiously because the large PAR% for average income reflected, in part, our broad definition of lower average income. Future research should aim to elucidate the causal structures that bring about the association between SGA and contemporary area-level average income. Possibly, this association is the causal effect of fewer individual material resources among those living in areas of lower average income or might reflect poorer environmental quality in those areas. Although we found that area-level percentage employment and high school education were not associated with SGA when modelled together with area-level average income, it is important to note that these three area-level variables are geographically clustered in Vancouver and throughout Canada as evidenced by the principal components analysis upon which the INSPQ deprivation index is derived. Investigations of the causal structure of the association between SGA and average income should consider employment and education domains, as well. A possible reason that average area income, but not education, was associated with SGA might be that recent immigrants from other countries receive lower returns to years of schooling and experience than Canadians . On the other hand, lower average income might be associated with other contemporary or historical causal phenomena leading to higher SGA risk today. In this analysis, we tested all of the known maternal risk factors for SGA that were available to us and found that none of them could explain relative socioeconomic SGA inequality. However, shorter maternal height explained about half of absolute socioeconomic SGA inequality. More research is needed to determine if small babies born to shorter mothers are at increased risk of adverse health outcomes and, if so, how stature could successfully be targeted to reduce socioeconomic SGA inequality. Current maternal height cannot be changed through intervention, but the height of the next generation may be influenced by interventions that affect foetal and childhood growth.
Our results may indirectly point to the importance and potential utility of focusing on social determinants of health in research and policy related to socioeconomic SGA inequality. Although individual risk factors for SGA (other than maternal height) might not be viable targets of intervention to reduce inequality when targeted one by one, our analysis did not investigate the potential benefits of targeting multiple risk factors at once. But if this is what would be needed, a potentially more efficient approach could be to identify and focus on the fundamental causes of health inequality that Phelan et al.  posit must exist because socioeconomic mortality inequality has persisted for centuries even though the major causes of mortality have shifted from infectious to chronic diseases. In this view, the inequality in health behaviours and comorbid disease states that we see in Vancouver today are not the fundamental causes of socioeconomic SGA inequality, but are instead the contemporary manifestations of deeper processes leading to health inequality. Frohlich et al.  suggest that income is a proxy for opportunities, resources and constraints, which are the fundamental causes of health inequality in populations. According to this view, the multiple forms of socioeconomic health inequality seen in Vancouver today might be addressed, for example, by focusing on equalizing the distribution of economic, cultural and social resources. Whatever the conceptual framework, more research and policy directed at the fundamental causes of health inequality may be a way forward given the results of our analysis.
Our analysis has several limitations. Although DAs are the smallest area for which aggregated income data are available from the census and so may be more homogenous than larger areas, without data on individual income we could not determine whether the association that we observed between average income and SGA was an individual- or area-level effect or both. We were also unable to include important variables previously associated with SGA, such as second-hand smoke exposure  and maternal race/ethnicity [8, 15]. Another limitation was that our analysis used population data that were collected for clinical use rather than research. In particular, the sensitivity of classifying smoking during pregnancy might have been lowered due to its stigma and consequent underreporting by women to their healthcare providers. To the degree that smoking during pregnancy was under-reported, our calculated PAR% for SGA of smoking during pregnancy might also be underestimated.
An additional consideration is that although Perinatal Services BC has built-in validation rules and regular data quality checks , 61% of the sample was coded as non-smokers due to lack of data on smoking status during pregnancy. This recoding most likely resulted in some misclassification of smoking status. However, our estimate of 6% prevalence of maternal smoking is consistent with estimates found in an independent study . Furthermore, a previous sensitivity study using randomly selected clinical records from the Registry found that recoding missing values as non-smokers led to 75% specificity and 98% sensitivity in smoking status classification overall . While we cannot determine if misclassification was differential with regard to SGA status, the fact that smoking status was reported prior to birth (and observation of SGA status) makes differential misclassification of exposure less likely than if it had been reported after birth. For this reason, as well, we might have underestimated the importance of smoking in the pathogenesis of SGA overall as well as in its inequality between income groups.