Area-level risk factors for adverse birth outcomes: trends in urban and rural settings

Significant racial and income disparities in birth outcomes exist in the US [1] and have persisted or even widened, despite a concerted research and practice effort aimed at reducing these disparities [2]. Previous research has shown that low-income individuals have higher preterm birth (PTB) and low birth weight (LBW) rates [3–7]. In addition, non-Hispanic African Americans compared to Non-Hispanic whites consistently have had at least twice the rate of LBW babies and 1.5 times the rate of PTBs [8], and previous studies controlling for socioeconomic status (SES) factors show racial disparities persist [9–11].

Health disparities across the urban–rural gradient also exist; both inner city and rural residents, compared to suburban, have poorer health outcomes [3]. The trends in urban–rural health disparities may be affected by demographic transitions, such that some rural areas have become more isolated as more people move into urban areas. Several explanations for urban–rural disparities in adverse birth outcomes have been suggested including higher prevalence of smoking, health care inequalities, and increased exposure to environmental hazards [12–16]. Taken together, these results suggest low socioeconomic status (SES) African American individuals in inner city and rural areas may be particularly vulnerable to adverse birth outcomes.

Previous research has not determined the extent that area-level socioeconomic factors contribute to birth outcome disparities or whether rurality modifies this effect. Several studies have found significant area-level socioeconomic adverse birth outcome disparities remained after individual variables were included in the model [14, 17–19]. However, some studies found that individual-level factors generally had stronger relationships with birth outcomes than area-level variables; and have also found that after accounting for individual risk factors that area-level poverty and other race factors were attenuated, and in some demographic strata did not have significant relationships with some adverse birth outcomes [17, 20, 21]. If there are specific area-level factors that play a role in determining individual-level risk, one might expect modification of the effect across urban–rural geographies, since several area-level environmental (e.g. air pollution) and social (e.g. isolation) risk factors are divergent across urban- rural settings.

Unlike the rest of the US, both urban and rural areas in the Deep South have significant non-Hispanic African American populations, allowing for characterization of racial health disparities across the urban–rural gradient. Alabama is consistently among the states with the highest PTB and LBW rates in the US. In 2010, the PTB rate in Alabama was 15.6% compared to the national average of 12.0%, and Alabama’s African American communities have particularly high PTB and LBW rates [22].

Here we examine whether contributions of individual or area-level factors to health disparities in the Deep South vary by urban–rural status over the last 20 years. We have two objectives: (1) determine whether 20-year temporal trends in PTB and LBW in Alabama vary by rurality and (2) determine if rurality modifies the effects of area-based racial composition or poverty with regards to individual-level risk for PTB and LBW.

The first part of our analysis found that RUCA codes defined isolated rural regions whereas population density metrics allowed for examination of highly urbanized areas. We found that long-term birth trends differed by rurality definition. Due to previously increasing national PTB rates, a concerted effort was initiated in 2006 to reduce PTB [1] and overall rates have been declining nationally and in Alabama since then, however our analysis shows adverse birth outcome rates are not decreasing in the most isolated rural regions. LBW rates in particular may still be increasing in these regions. In addition, we found that over the last 20 years, the most urban regions consistently had higher adverse birth outcome rates compared to other regions. In the case of LBW, the disparity between the most population dense and less population dense regions increased during the 1991–2005 time period, and the magnitude of the disparity was maintained through 2010.

Changing demographics and other individual-level characteristics over this time period partially explained the differing time trends, although isolated rural areas maintained elevated adverse birth outcome odds even after these adjustments. Rural regions captured by RUCAs have low commuting flows to metropolitan and micropolitan regions, indicating likely economic and social isolation. Previous literature has found that rural regions have increased prevalence of maternal smoking, and it has been found that lack of social capital is especially tied to maternal smoking in rural regions [12, 26]. Rural regions also have more unintended pregnancies and pregnant women in with poor cardiovascular risk factor status than their urban counterparts [13, 27]. Finally, maintaining full-time physicians locally, particularly family doctors and obstetricians is more difficult in rural regions [28], and there is increased time-to-care in rural regions [29], so medical advances that have allowed national decreases in adverse birth outcome rates might not be reaching isolated rural regions. The differing time trends by rurality seen in this analysis might be explained if any of these factors changed over time at different rates in isolated rural regions compared to non-rural regions. For example, while smoking prevalence is on a national decline, the prevalence of smoking among pregnant mothers might be decreasing at a faster rate in non-isolated rural regions.

Our analysis found that birth outcome disparities attributable to living in low-income African-American communities were heightened in population dense urban areas, compared to less-dense areas. This result is consistent with a previous study in Pennsylvania, where urban areas had higher preterm birth rates, even after controlling for race and income [20]. Much, but not all of the disparities by ZIP code-level demographic risk factors were explained by individual-level risk factors in our study. Our results may differ from previous studies that found stronger area-level relationships because we have included individual-level factors that account for significant area-level effects in previous studies. In particular, the method of payment may indicate the lack of adequate health insurance, and thus explain adverse birth outcomes associated with reduced access of care in inner city and rural regions [3]. One possible explanation for the remaining effect modification by rurality is that low SES residents in population dense areas have higher environmental burdens that may lead to adverse birth outcomes, such as increased exposure to heavy metals, air pollution, and higher summer temperatures [30–34]. For example, a study in North Carolina found seasonal preterm birth patterns to be more pronounced in urban residents, suggesting urban air pollution may explain the result [35]. Other routes for the effect modification by population density might involve social factors related to adverse birth outcomes, such as inner city crime, stress, racial discrimination, and built environment factors (such as liquor store density and housing damage). These factors may be more prevalent in low SES urban neighborhoods compared to non-urban low SES communities, and might explain the effect modification by rurality [36–39].

Alternatively, it has been previously posited that significant area-level effects may be found not due to causal relationships, but due to structural and residual confounding. Structural confounding may arise from socioeconomic and racial segregation, which in turn may lead to a lack of adequate distributions of individual-level factors within area-level strata [21, 40]. This may be addressed by choosing fewer strata (e.g. tertiles versus quartiles), but this in turn may lead to increased residual confounding [21].

There are limitations to the current study. Additional individual-level variables, such as smoking status, alcohol use, psychosocial factors, and disease states of the mothers, would be useful in further describing the relationships we found. Another limitation of this study was that we only had availability to warm-season months (May-September). It is possible that time trends and other relationships investigated differ by warm and cool seasons, limiting the generalizability of our results. Our current analysis does not account for area-level variable changes over the 20 year period. It also has been shown previously that ZIP code-level estimates provide less robust effect estimates than census tract level estimates [41] and that there at spatiotemporal mismatches when census-derived ZIP code tabulation areas (ZCTAs) are merged with postal ZIP codes, as was done in this study [42]. However, ZIP code-level analyses are preferable to larger and even more heterogenous areas, such as counties [13].

This study finds that adverse birth outcome rates remain higher in isolated rural and more population dense areas, and that these disparities are being maintained or increasing over time. This study also found that relationships between ZIP code-level poverty and adverse birth outcomes were significant in urban areas, even after accounting for individual-level risk factors. Interventions to reduce adverse birth outcomes among low SES populations have been found to be effective, particularly among rural populations, and can be implemented to reduce disparities [43].