Prematurity and low birth weight: geospatial analysis and recent trends

Abstract

Prematurity and low birth weight are of concern in neonatal health. In this work, geospatial analysis was performed to identify the existence of statistically significant clusters of prematurity and low birth weight using Moran’s I. Data was obtained from March of Dimes and the National Center for Health Statistics for the years 2015 to 2019. Analysis demonstrated the presence of hotspot (High-High) and coldspot (Low-Low) geographic clusters of these variables in regions across the United States. Additionally, factorial ANOVA was performed, and revealed the significance of demographic variables of interest. Given the strong relationship between these two variables, regions that are hotspots for one variable, but not the other, are of particular interest for further study.

Letter to the editor

Dear Editor,

It has been previously established that prematurity (PM) and low birth weight (LBW) are of concern when assessing neonatal health: Prior works have demonstrated the role of these variables in predicting neonatal morbidity and mortality [1]. Additionally, previous research has shown the role of both the health of the mother and her socioeconomic environment in the prevalence of these two conditions [2]. We aimed to use geospatial analysis techniques to identify whether statistically significant clusters of PM (< 37 weeks) and LBW (< 5.5 lbs) exist on a nationwide level and to further explore the socioeconomic determinants associated with those clusters.

We used birth and C-section data from the March of Dimes and the National Center for Health Statistics during the years 2015–2019 across 3105 US counties [3]. Moran’s I statistic was calculated to categorize individual counties as either Not Significant or as one of 4 statistically significant (p < 0.05) cluster classifications: High-High (H-H), High-Low (H-L), Low-High (L-H), Low-Low (L-L) [4]. In this attribution system, the first term designates the relative value of a given county compared to the national average; the second attribute reflects the relative value of neighboring counties compared to the national average. Demographic data was obtained from the American Community Survey (US Census Bureau). Factorial ANOVA was performed to evaluate the significance of contributory socioeconomic variables of interest at a significance level of 0.001.

Visualization of the cluster designations at the county level demonstrated clear geographic trends (Fig. 1). For both the PM and LBW analyses, there was an expansive H-H cluster that was persistent across the Southern states. There were 3 distinct expansive L-L clusters encompassing the New England states, the Midwest, and the Pacific Northwest. A LBW H-H cluster encompassed Colorado and northern New Mexico, yet this was not seen in the PM analysis. Similarly, multiple significant PM H-H clusters were identified in Texas, but not LBW clusters. Factorial ANOVA across clusters revealed significant contributions of various socioeconomic factors at a significance level of 0.001 for both the PM and LBW analyses (Tables 1 and 2).

Fig. 1

Geospatial mappings and analysis for (A) prematurity, (B) low birth weight, and (C) joint prematurity and low birth weight. Color designations reflect Moran’s I spatial categorizations. For joint mappings, High and Low designations represent agreement and Other represents areas of disagreement between the two variables

Table 1 Factorial ANOVA across preterm birth clusters. Asterisks reflect significance at a significance level of 0.001

Table 2 Factorial ANOVA across low birth weight clusters. Asterisks reflect significance at a significance level of 0.001

PM and LBW have previously been demonstrated to have a strong relationship, so it is unsurprising that the identified spatial clusters of these variables have substantial overlap, However, what is of particular interest are the regions that are clusters for one variable but not the other. For instance, there are regions of Texas where several counties are significantly higher in PM but not LBW. Conversely, there is a large region of Colorado where there is a substantial incidence of LBW despite that region not having high prematurity. Given that factors traditionally associated with prematurity would not explain this increase, it is important to look for other explanations. The Colorado Department of Public Health has previously proposed the contribution of high altitude to pregnancy-induced hypertension as a possible explanatory factor [5]. The inverse relationship in Texas is harder to attribute to an isolated cause; though, the prevalence of large, medically underserved immigrant communities in the identified regions is likely a contributing factor. The ANOVA findings in this study underscore the importance of many socioeconomic factors that differentiate the clusters, including race and various economic markers (e.g., SNAP, insurance type, educational status). Interestingly, the rural/urban character between clusters did not significantly differ in this analysis.

Since PM and LBW demonstrate similar geospatial patterns across the United States, and a strong relationship exists between these two factors, regions that are high in one variable and not the other are of particular interest for further study.