Converging crises and maternal and child health: colonialism, extreme weather, and COVID-19 | Reproductive Health

This study examined the association between multiple disaster exposures and maternal and child health among 104,560 live births recorded in Puerto Rico’s vital records from 2017 to 2021. It also assessed effect modification by [1] geographic location within Puerto Rico and [2] colonialism, by comparing associations in Puerto Rico to those in Texas and Florida.

First, disaster periods were consistently associated with adverse maternal health outcomes, with particularly strong effects observed during the early and late pandemic periods. Second, disaster exposure did not significantly impact child health in Puerto Rico, and these associations did not vary by colonialism or geographic location. Third, although geographic-level effect modification results were mixed, gestational diabetes rates were higher in the early and late post-hurricane periods in Central and Eastern Puerto Rico compared to the Metropolitan region. Additionally, gestational hypertension rates were higher in the early and late pandemic periods in Southern Puerto Rico compared to the Metropolitan region. These findings are particularly notable given that Central and Eastern Puerto Rico were anecdotally among the most affected by Hurricane Maria, while the Southern region was heavily impacted by the earthquake events that coincided with the early pandemic. Lastly, colonialism modified the effects of disaster exposure on gestational diabetes in the late pandemic period and on excessive weight gain in the late post-hurricane, early pandemic, and late pandemic periods, with Puerto Rico experiencing a greater burden than Texas and Florida.

Overall, our findings share similarities with existing research while also highlighting key differences. While disaster exposure was associated with adverse maternal health outcomes, these associations remain underexplored in prior studies, and it is unclear whether our findings would be replicable in other populations. Nonetheless, our results suggest that pregnant individuals in Puerto Rico experienced heightened rates of post-disaster gestational hypertension, gestational diabetes, and excessive weight gain, with the strongest associations observed during the early pandemic period. This surge may reflect the effects of the pandemic alone or the cumulative impact of prior disasters compounded by the pandemic. However, we are unable to determine the relative contribution of these factors.

Prior studies have linked the COVID-19 pandemic to increased pregnancy-related stress and anxiety, driven by factors such as social isolation, unemployment, poverty, and intimate partner violence [47,48,49]. These psychosocial stressors, combined with heightened risks of food insecurity and disruptions to food systems during the pandemic, may have contributed to the elevated rates of adverse maternal health outcomes observed in our study [50, 51]. Additionally, these associations may reflect the cumulative effects of multiple disasters. Hurricane Maria, for example, was similarly associated with adverse mental health outcomes and disruptions in food systems in Puerto Rico [52,53,54,55,56,57].

Notably, our analysis of newborn health outcomes (Table 2) does not align with our initial hypothesis and shows some inconsistencies with prior studies. While our study did not find significant post-disaster differences in preterm birth or low birthweight, previous research has reported adverse impacts on children born in post-disaster settings. In particular, studies examining the effects of hurricanes, wildfires, and floods have found increased rates of preterm birth and low birthweight among disaster-exposed individuals, particularly among racial and ethnic minorities [7, 58,59,60,61]. However, prior research suggests that such findings may be influenced by live-birth bias. A key example is Hamilton and colleagues, who analyzed official birth records following Hurricane Katrina and initially found apparent reductions in preterm birth and low birthweight after the storm [62]. These findings were later challenged by researchers who estimated hypothetical birth rates under stable conditions and found that Hurricane Katrina was Likely associated with worse outcomes, with estimated risk ratios ranging from 1.30 to 1.35—contradicting the initial results [45, 62, 63]. Similarly, we found evidence of Live-birth bias in our analysis, given the significant decline in yearly Live births in Puerto Rico following Hurricane Maria. In 2017, there were 24,373 births, but this number steadily declined to 19,332 in 2021, representing a 21% reduction. More information can be seen in Table 7. It is likely that stable birth rates would have yielded different results in our analysis. However, this decline in total live births is plausibly linked, at least in part, to the impact of multiple disasters. Further research is needed to determine whether these reductions were driven by declining fertility, increased infant mortality, migration, or other factors affecting childbearing.

To the best of our knowledge, this study is among the few epidemiologic investigations to examine the impact of multiple disasters while considering colonialism as a determinant of health. While our results were mixed, the direction of association for gestational diabetes and excessive weight gain was consistently harmful. Stronger associations in Puerto Rico compared to Texas and Florida suggest that colonialism may have heightened the risk of disaster-related adverse maternal health outcomes. Additionally, live-birth bias, as discussed earlier, may have led to an underestimation of colonialism’s effects on maternal and child health. Puerto Rico experienced a substantial yearly decline in total Live births, ranging from 12 to 22%, whereas Texas and Florida saw only a 1–6% reduction (Table 6). This stark disparity suggests that colonialism may contribute to a differential reduction in live births under multiple disaster scenarios, further exacerbating health inequities.

Overall, geographic location within Puerto Rico did not significantly modify the relationship between disaster exposure and maternal and child health. However, some effect estimates in Table 3 suggest that proximity to disaster epicenters and more severe disaster experiences may be associated with adverse maternal health outcomes. Specifically, gestational diabetes was higher among individuals in the Central region during both the early (PR = 1.48, 95% CI: 0.86, 2.54) and late post-hurricane periods (PR = 1.57, 95% CI: 1.07, 2.30) compared to those in the Metropolitan region. Additionally, gestational hypertension was elevated in the Southern region during both the early (PR = 1.57, 95% CI: 1.16, 2.10) and late pandemic periods (PR = 1.72, 95% CI: 1.19, 2.49) compared to the Metropolitan region. These findings are particularly notable, as the Central region was reported to have experienced significant hurricane-related complications due to resource scarcity, while the Southern region was the epicenter of the earthquakes that occurred during the early pandemic period [64, 65]. These results align with prior research, including a study in China that found stronger adverse mental health effects among pregnant individuals living closer to earthquake epicenters [66].

Strengths & limitations

This study has several strengths. Importantly, this study uses a novel approach to examining colonialism as a determinant of health in the context of multiple disaster exposures—an area that remains largely understudied. Our observational and exploratory analysis provides an initial foundation for understanding these associations and guiding future research. Additionally, while cross-sectional designs traditionally limit causal inference, our approach mitigates some of these challenges by leveraging precise birth dates from official records, allowing us to compare outcomes against disaster timelines and reduce the risk of temporal ambiguity and reverse causation.

The study also benefits from the comprehensive nature of the NVSS, which serves as the official record of all live births in the U.S. This ensures that our dataset is fully representative of the study population. Furthermore, the use of geographic-level zip code data combined with FEMA disaster declarations allowed us to focus on areas most affected by disasters, reducing the risk of exposure misclassification. Lastly, the reliance on clinically diagnosed maternal health outcomes—certified and recorded by physicians—minimizes the potential for outcome misclassification, strengthening the validity of our findings.

Our study has several limitations to consider. First, the vital records data lacked unique identifiers, preventing us from accounting for multiple births from the same individuals. However, given that most births occur to first-time or non-consecutive repeat mothers, the likelihood of multiple births from the same individuals within the study period is relatively low. As a result, while this may introduce minor clustering effects, it is unlikely to significantly impact the overall observed associations. Second, selection bias may have influenced our findings, as the documented population exodus from Puerto Rico following disasters could include individuals at higher or lower risk for adverse outcomes, potentially underestimating or overestimating the true impact. Additionally, pregnancies most affected by disasters may have been more likely to result in miscarriage, stillbirth, or abortion, leading to an underestimation of observed associations [67].

While our study attempted to adjust for key confounders available in the data, residual confounding cannot be ruled out, as vital records do not capture key social and behavioral factors such as income, employment, or food security—factors that could influence pregnancy outcomes and be linked to disaster exposure. Moreover, our exposure classification does not account for individual-level variability in disaster experiences, which may bias our results toward the null.

The study lacks adjustments for multiple comparisons, increasing the possibility that some observed associations occurred by chance. We opted not to apply a Bonferroni correction, as it may be overly conservative and increase the risk of type II errors [68, 69]. Future studies may consider using false discovery rate (FDR) adjustments, which better balance the risk of type I and type II errors in analyses with multiple hypotheses [70].

Our live-birth bias analysis must be interpreted within the context of Puerto Rico’s long-term demographic trends. The island was already experiencing a significant decline in birth rates due to an ongoing economic crisis, which Likely contributed to emigration and slowed population growth. This trend may be attributed to the ongoing economic crisis that led to emigration and reduced population growth. Prior to Hurricane Maria in September 2017, Puerto Rico was already experiencing a significant population decline due to long-term economic challenges [71, 72]. However, Hurricane Maria in 2017 exacerbated this decline, leading to the steepest year-over-year drop in live births—from 28,000 in 2016 to 24,000 in 2017 and 21,000 in 2018. While birth rates began stabilizing between 2020 and 2023, this stabilization may have occurred sooner if not for the compounded effects of multiple disasters. These trends highlight how existing demographic and economic challenges intensified following the disasters, shaping Puerto Rico’s population trajectory. Nonetheless, we are unable to fully disentangle the effects of the preexisting economic crisis from the disaster aftermath when accounting for live-birth bias.

Further complicating our analysis is the issue of colonialism. Notably, Puerto Rico has long faced elevated rates of PTB and LBW, with studies showing that Puerto Rican women have a preterm birth rate of 13.2%, higher than other U.S. Hispanic subgroups [73]. Prior to Hurricane María, Puerto Ricans had a 23% higher rate of preterm birth, and a 35% higher rate of low birth weight compared to pregnancies in the continental United States [74]. These elevated rates are likely driven by systemic issues rooted in colonialism, including underfunded healthcare infrastructure, economic instability, and social inequities. Given this historically high baseline, the relative impact of these disasters on birth outcomes may appear less pronounced than expected. Moreover, factors such as declining birth rates during the preceding economic crisis and post-disaster population shifts may lead to misleading findings if the full context is not considered. Further, in our study, we operationalized colonialism as residing in a U.S. colony versus not residing in one. However, colonialism is a complex system shaped by numerous individual, community, political, and economic factors—including generational trauma, structural inequities, racism, and indigenous identity. These intersecting influences likely contribute to how colonialism affects maternal and child health outcomes. The lack of individual-level data in vital records limited our ability to explore these pathways. However, qualitative research based on in-depth interviews (currently in preparation) may provide additional insights that complement this analysis. Additionally, while birth vital records are a useful population-based data source, their design prioritizes de-identification in order to protect individual privacy, limiting the availability of detailed socioeconomic and geographic variables. However, our analysis adjusted for multiple SES proxies, including maternal and paternal education, maternal race, maternal and paternal age, marital status, birth payment method, and WIC participation. While direct income data was unavailable, these covariates capture key dimensions of SES, as WIC participation and Medicaid coverage during birth both require low-income eligibility. Therefore, while some residual confounding remains possible, it is likely minimal. We also considered alternative modeling approaches such as spatial, instrumental variable, and Bayesian hierarchical models to address unmeasured confounding; however, spatial models were not appropriate given our time-based exposure, instrumental variables were infeasible, and hierarchical models were limited by small sample sizes in some regions and concentration in the Metropolitan area. Future research linking vital records with richer data sources may further strengthen SES measurement and adjustment.