Skip to main content
NCBI home page
Public Health Reports logoLink to Public Health Reports
. 2020 Jun 18;135(4):472–482. doi: 10.1177/0033354920927854

Geovisualization and Spatial Analysis of Infant Mortality and Preterm Birth in Ohio, 2008-2015: Opportunities to Enhance Spatial Thinking

Elisabeth Dowling Root 1,2,✉, Emelie D Bailey 3, Tyler Gorham 4, Christopher Browning 5, Chi Song 6, Pamela Salsberry 7
PMCID: PMC7383761  PMID: 32552459

Abstract

Objectives

Geovisualization and spatial analysis are valuable tools for exploring and evaluating the complex social, economic, and environmental interactions that lead to spatial inequalities in health. The objective of this study was to describe spatial patterns of infant mortality and preterm birth in Ohio by using interactive mapping and spatial analysis.

Methods

We conducted a retrospective cohort study using Ohio vital statistics records from 2008-2015. We geocoded live births and infant deaths by using residential address at birth. We used multivariable logistic regression to adjust spatial and space–time cluster analyses that examined the geographic clustering of infant mortality and preterm birth and changes in spatial distribution over time.

Results

The overall infant mortality rate in Ohio during the study period was 6.55 per 1000 births; of 1 097 507 births, 10.3% (n = 112 552) were preterm. We found significant geographic clustering of both infant mortality and preterm birth centered on large urban areas. However, when known demographic risk factors were taken into account, urban clusters disappeared and, for preterm birth, new rural clusters appeared.

Conclusions

Although many public health agencies have the capacity to create maps of health outcomes, complex spatial analysis and geovisualization techniques are still challenging for public health practitioners to use and understand. We found that actively engaging policymakers in reviewing results of the cluster analysis improved understanding of the processes driving spatial patterns of birth outcomes in the state.

Keywords: GIS, geospatial analysis, spatial clusters, infant mortality, preterm birth, program planning, spatial thinking

Geographic information systems (GIS) have become ubiquitous in public health.1-4 Most state health departments routinely use GIS; the extent to which it is used at the local level varies, however, with many local health departments citing GIS as an area for workforce development.5,6 Training public health practitioners in geospatial technologies does not necessarily enable them to understand the patterns and processes that shape multiscalar patterns of health, unless they also learn to think spatially. Spatial thinking involves understanding relationships within and between spatial structures, and a wide variety of visualizations (from drawings to computer models) provides the means to communicate about them.7 In public health, spatial thinking permits researchers and practitioners to comprehend the relative locations of complex social, economic, environmental, and demographic interactions that produce patterns of health and disease. Spatial thinking and reasoning skills are not routinely integrated into public health curricula in institutions of higher education, however, despite being even more valuable today than previously because of the widespread use of geospatial technologies.

In 2016, the state of Ohio began using data analytics, including geospatial analysis and geovisualization, in new ways to better understand how to identify women at risk for poor birth outcomes and to inform initiatives addressing infant mortality. Geovisualization is a process by which tools such as GIS and mapping are used to support analysis of complex geospatial data in an interactive and comprehensible way. This effort was in response to the high infant mortality rate in Ohio, which consistently ranks 45th in the nation. Although rates in Ohio have decreased considerably since the early 1990s, from a high of 9.8 per 1000 live births in 1990 to 7.2 per 1000 live births in 2017, rates have stagnated since 2013. Furthermore, racial disparities (between African American and non-Hispanic white populations) in infant mortality in Ohio persist: African American infants die at nearly 3 times the rate as non-Hispanic white infants (15.6 vs 5.3 per 1000 births in 2017).8 Spatial variation in infant mortality also exists across Ohio, driven by racial segregation and social and structural inequalities in access to health care, quality housing, environmental conditions, employment and educational opportunities, and perceived racism and discrimination.9,10

Recognizing that a coordinated effort, grounded in Ohio-specific data, was needed to address sociospatial inequalities in birth outcomes, the Ohio Department of Health, the Ohio Department of Medicaid, and the Ohio Department of Higher Education jointly established the Infant Mortality Research Partnership (IMRP), a collaboration between state health agencies and researchers at institutions of higher education in Ohio. One objective of this government–academic partnership was to use GIS and geospatial analysis to map small-area estimates of infant mortality and preterm birth, find areas with higher-than-expected rates, and examine the placement and effect of various initiatives that address birth outcomes. One such initiative is the Ohio Equity Institute (OEI), a collaboration between the Ohio Department of Health and local partners that targets community-specific causes of infant mortality in 9 OEI-designated counties (Figure 1).11 The Ohio Department of Health selected these 9 counties because they collectively accounted for 59% of all infant deaths in the state and programs in these counties would be likely to have the greatest effect on preterm birth and infant mortality. Although the methods used by IMRP are not new14,15 (numerous studies examine spatial patterns of preterm birth16,17 and infant mortality),18-21 what is new is the partnership’s objective to increase state capacity for spatial thinking. IMRP challenged policymakers to examine spatial patterns in birth outcomes and reframe policy questions in terms of the interrelationship between health and complex social and economic issues.

Figure 1.

Figure 1

Open in a new tab

Major cities in Ohio and the 9 corresponding Ohio Equity Institute (OEI) counties. OEI comprises local programs that target community-specific causes of infant mortality.11 Map background shows population density (the darker the shade, the higher the density), and the thick black outline indicates OEI counties. Data source: Author analysis of US Census Bureau cartographic boundary files12 and 2014-2017 American Community Survey data.13

The objectives of this study were to (1) evaluate spatial patterns of infant mortality and preterm birth in Ohio and (2) examine changes in these outcomes over time, especially in OEI counties with programs that targeted reductions in these outcomes.

Methods

Data Sources

We conducted a retrospective cohort study of the spatial distribution of infant mortality and preterm birth in Ohio. We extracted data from the Ohio linked birth certificate and death certificate files for births that occurred from January 1, 2008, through December 31, 2015. We limited the sample to births with a gestation of >20 weeks. Although the medical literature generally agrees that 23 weeks is the lowest threshold for viability, we included earlier births because we were interested in the spatial patterns of mortality rather than individual determinants of early birth, and we did not want to risk biasing these spatial patterns with more stringent exclusion criteria. We defined preterm births as births occurring before 37 weeks of gestation (among births with a gestation of >20 weeks). We collected data on maternal age, race/ethnicity (non-Hispanic white, non-Hispanic black, non-Hispanic other, Hispanic, or unknown), education (<high school diploma, high school diploma or equivalent, some college, bachelor’s or advanced degree, or unknown), and address at birth from the birth records. The Ohio linked birth–death certificate files contained data on 1 132 756 births, 118 149 preterm births, and 8352 infant deaths during the study period.

We geocoded records by using the address at birth with a 95% successful match rate. Examination of ungeocoded records suggested there were proportionally more rural records and post office boxes than geocoded records. We assigned coordinates for latitude and longitude to a census tract, which we used as the unit of analysis for the study. We excluded data on 3470 births to mothers residing outside the state. We excluded an additional 33 779 records from the final analysis because they were not matched during the geocoding process. The final sample included 1 097 507 births, 112 552 preterm births, and 7193 infant deaths from all 2946 census tracts in Ohio.

We produced maps by using the US Census Bureau cartographic boundary files for state, county, and census tract boundaries.12 We obtained additional population data by census tract from the 2013-2017 American Community Survey.13

This project was reviewed and approved by The Ohio State University Institutional Review Board (approval no. 2016B0291).

Statistical Methods

We generated several sets of descriptive maps of infant mortality and preterm birth by using spatial empirical Bayes smoothing methods. We used smoothing methods because small-area estimates typically result in unstable rates. Rate smoothing is a technique used to (1) stabilize rates based on small numbers and (2) reduce extreme values in rates caused by various population sizes.22 We used a technique that computes a weighted average between the raw rate for each census tract and the local average (based on a first-order queens contiguity weights matrix), with weights proportional to the underlying population of births. We applied this method to all census tracts in Ohio. The method results in various levels of adjustment: rates in areas with a large population may have little adjustment, whereas rates in areas with small populations may have considerable adjustment. We implemented this technique in GeoDa version 1.12.123; details of this technique are described elsewhere.24 To display data on the descriptive maps, we applied a quantile classification system, which divides data into groups so that the total number of areas included in each class are approximately equal.

We used Kulldorff’s space and space–time scan statistics to identify groups of census tracts with higher-than-expected rates of infant mortality and preterm birth (high rates only, ie, hotspots).25,26 This method creates a large number of overlapping circular windows, centered on each census tract in the study area, which are allowed to vary in size to include a minimum of 1% to a maximum of 30% of the population at risk (ie, births). The number of observed and expected cases of infant mortality or preterm birth inside and outside each circle is tabulated and used to calculate a log likelihood ratio test statistic. The circle with the maximum likelihood is the most likely cluster, that is, the cluster with a significantly higher number of observed cases than the number of cases expected given the underlying population of births. We applied the discrete Poisson model using counts of infant mortality or preterm birth (cases) and total births (population denominator) in each census tract. We ran models with alternative maximums (up to 50% of the population at risk), but we found that models >30% identified similar clusters with the same set of core tracts. We evaluated significance by using standard Monte Carlo methods using 999 random replicas of the data set. We reported only clusters with a P value < .05. We also computed relative risks (RRs), calculated as the observed number of cases divided by the expected number of cases within the cluster, divided by the observed-over-expected cases outside the cluster. For space–time scan statistics, we used the same model parameters and set the minimum time aggregation to 1 month and the maximum to the entire study period. We used this maximum because we wanted to account for temporally persistent spatial clusters.

We conducted 2 cluster analyses. In the unadjusted analysis, we calculated the expected number of infant mortality or preterm birth events in a census tract by multiplying the total number of births in the census tract by the statewide rate. In the adjusted analyses, we calculated the expected number of events by using the predicted probabilities of infant mortality or preterm birth derived from a multivariable regression model containing maternal demographic characteristics that have an uneven spatial distribution: age, race/ethnicity, and education.27,28 This extra step allowed us to examine whether the observed clusters were due to the underlying characteristics of the birth population (eg, maternal age, race/ethnicity, or education) or some other factor not included in the multivariate model. We cleaned data and conducted the analysis in R version 3.4.2,29 SAS version 9.4,30 and SaTScan,31 and we created maps in ArcMap version 10.6.32

Results

Both infant mortality and preterm birth were centered on the urban areas of the state (Figure 2). We identified a region of high rates of infant mortality in the rural southern counties to the east of Cincinnati and Dayton. We also identified a band of high rates that ran along rural Appalachia from south to southeast.

Figure 2.

Figure 2

Open in a new tab

Empirical Bayes smoothed rate maps for (A) infant mortality and (B) preterm births as a percentage of all births in Ohio, 2008-2015. The Ohio Equity Institute (OEI) comprises local programs that target community-specific causes of infant mortality in 9 counties.11 Map classes were developed by using a quantile classification system. Data source: Author analysis of Ohio Department of Health linked birth–death certificate files and US Census Bureau cartographic boundary files.12

Infant Mortality

The unadjusted spatial cluster analysis of infant mortality (Figure 3A and Table 1) indicated 5 significant clusters centered on 6 major urban areas in Ohio (Cincinnati, Cleveland, Columbus, Dayton, Toledo, and Youngstown). Relative risks ranged from 1.52 to 1.94, indicating a 52% to 94% greater risk of death among infants born in these clusters than among infants born elsewhere in the state. When we adjusted these clusters for maternal age, race/ethnicity, and education (Figure 3B and Table 1), all the urban clusters disappeared, but we found a cluster (C1) spanning Cincinnati and the rural area to the east, indicating a higher-than-expected number of deaths given the age and race distribution of the population of births in that area (RR = 1.17; P = .01).

Figure 3.

Figure 3

Open in a new tab

Clusters (indicated by “C”) of infant mortality, birth cohort 2008-2015 in Ohio, resulting from the (A) spatial unadjusted model; (B) spatial model, adjusted for maternal age, race/ethnicity, and education; (C) space–time unadjusted model; and (D) space–time model, adjusted for maternal age, race/ethnicity, and education. The adjusted space–time model resulted in 1 cluster (C1), in which the relative risk of infant mortality was 75.3. The Ohio Equity Institute (OEI) comprises local programs that target community-specific causes of infant mortality in 9 counties.11 Data source: Author analysis of Ohio Department of Health linked birth–death certificate files and US Census Bureau cartographic boundary files.12

Table 1.

Unadjusted and adjusted spatial and spatiotemporal cluster results for infant mortality, birth cohort 2008-2015 (N = 1 097 507 births), Ohioa

Cluster Numberb Date Observed No. of Cases Expected No. of Cases Relative Risk Log Likelihood Ratio P Valuec
Spatial Analysis
Unadjusted (Figure 2A)
 Cluster 1 — 368 198 1.90 59.91 .001
 Cluster 2 — 359 198 1.86 54.82 .001
 Cluster 3 — 466 314 1.52 33.67 .001
 Cluster 4 — 137 71 1.94 24.25 .001
 Cluster 5 — 127 70 1.84 19.29 .001
 Cluster 6 — 87 48 1.81 12.51 .01
Adjusted (Figure 2B)
 Cluster 1 — 1432 1261 1.17 13.52 .01
Spatiotemporal Analysis
Unadjusted (Figure 2C)
 Cluster 1 2008-2015 368 198 1.90 60.12 .001
 Cluster 2 2008-2015 359 197 1.86 55.01 .001
 Cluster 3 2008-2015 466 314 1.52 33.54 .001
 Cluster 4 2008-2015 137 71 1.95 24.37 .001
 Cluster 5 September–November 2009 6 0 67.78 19.38 .01
 Cluster 6 2008-2015 127 69 1.84 19.30 .01
Adjusted (Figure 2D)
 Cluster 1 September–November 2009 6 0 75.34 20.01 .002
Open in a new tab

Abbreviation: —, not applicable.

aData sources: US Census Bureau cartographic boundary files12 and Ohio Department of Health linked birth–death certificate files.

bCluster numbers correspond to the numbers on maps in Figure 3.

c P values generated through Monte Carlo replications of the log likelihood ratio test statistic; P < .05 was considered significant.

The unadjusted space–time analysis of infant mortality (Figure 3C and Table 1) indicated that nearly all clusters were temporally persistent, meaning that the higher-than-expected number of infant deaths in these areas occurred across the entire study period. The clusters were again centered on 6 major urban areas of the state (Cincinnati, Cleveland, Columbus, Dayton, Toledo, and Youngstown), where the RR ranged from 1.52 to 67.78. In the space–time analysis adjusted for maternal age, race/ethnicity, and education (Figure 3D and Table 1), all clusters disappeared except for a cluster (Cluster 1) in south Columbus (RR = 75.3; P = .001).

Preterm Birth

The unadjusted spatial cluster analysis (Figure 4A and Table 2) indicated 8 clusters of preterm birth in 8 urban areas (Akron, Canton, Cincinnati, Cleveland, Columbus, Dayton, Toledo, and Youngstown). Three small rural areas east of Cincinnati and north and east of Columbus also had a higher-than-expected risk. Relative risks ranged from 1.14 to 1.43, indicating that infants born in these clusters had a 14% to 43% greater risk of preterm birth than infants born outside these clusters. When we adjusted for maternal age, race/ethnicity, and education (Figure 4B and Table 2), we found a reduction in RRs in all urban clusters. The 2 rural clusters to the east (C4) and south (C1) of Columbus became much larger, essentially covering the entire rural Appalachian region in the southeast portion of the state. These 2 areas had higher-than-expected rates given the age and racial/ethnic distribution of mothers who gave birth in these areas.

Figure 4.

Figure 4

Open in a new tab

Cluster analysis of preterm birth, birth cohort 2008-2015 in Ohio, resulting from the (A) spatial unadjusted model; (B) spatial model, adjusted for maternal age, race/ethnicity, and education; (C) space–time unadjusted model; and (D) space–time model, adjusted for maternal age, race/ethnicity, and education. The Ohio Equity Institute (OEI) comprises local programs that target community-specific causes of infant mortality in 9 counties.11 Data source: Author analysis of Ohio Department of Health linked birth–death certificate files and US Census Bureau cartographic boundary files.12

Table 2.

Unadjusted and adjusted spatial and spatiotemporal cluster results for preterm birth, birth cohort 2008-2015 (N = 1 097 507 births), Ohioa

Cluster Numberb Date Observed No. of Cases Expected No. of Cases Relative Risk Log Likelihood Ratio P Valuec
Spatial Analysis
Unadjusted (Figure 3A)
 Cluster 1 — 6259 4462 1.43 336.36 .001
 Cluster 2 — 7176 5763 1.26 169.94 .001
 Cluster 3 — 3012 2165 1.40 150.86 .001
 Cluster 4 — 1792 1292 1.39 87.46 .001
 Cluster 5 — 2123 1608 1.33 75.94 .001
 Cluster 6 — 1590 1265 1.26 38.97 .001
 Cluster 7 — 1543 1231 1.26 37.02 .001
 Cluster 8 — 440 312 1.41 23.35 .001
 Cluster 9 — 466 348 1.34 18.21 .001
 Cluster 10 — 707 574 1.23 14.33 .01
 Cluster 11 — 1381 1217 1.14 10.72 .048
Adjusted (Figure 3B)
 Cluster 1 — 11 574 10 406 1.13 69.98 .001
 Cluster 2 — 3813 3150 1.22 67.44 .001
 Cluster 3 — 4205 3716 1.14 32.01 .001
 Cluster 4 — 1243 1005 1.24 26.40 .001
 Cluster 5 — 2564 2220 1.16 25.86 .001
 Cluster 6 — 1033 823 1.26 24.93 .001
 Cluster 7 — 542 401 1.35 22.41 .001
 Cluster 8 — 536 406 1.32 19.13 .002
 Cluster 9 — 8759 8245 1.07 17.00 .003
 Cluster 10 — 641 523 1.23 12.54 .01
Spatiotemporal Analysis
Unadjusted (Figure 3C)
 Cluster 1 2008-2015 6259 4459 1.43 133.73 .001
 Cluster 2 2008-2015 7176 5766 1.26 169.22 .001
 Cluster 3 2008-2015 3012 2161 1.40 152.26 .001
 Cluster 4 2008-2015 1792 1290 1.40 88.15 .001
 Cluster 5 2008-2015 2123 1608 1.33 76.12 .001
 Cluster 6 March 2014–December 2015 566 367 1.55 46.49 .001
 Cluster 7 October 2011–December 2015 492 335 1.47 32.35 .001
 Cluster 8 2008-2015 2354 1995 1.18 31.14 .001
 Cluster 9 2008-2015 440 312 1.41 23.38 .001
 Cluster 10 June 2008–November 2013 374 264 1.42 20.13 .003
Adjusted (Figure 3D)
 Cluster 1 All years 11574 10407 1.12 69.77 .001
 Cluster 2 2008-2015 3813 3149 1.22 67.65 .001
 Cluster 3 July 2008–November 2012 2279 1873 1.22 41.97 .001
 Cluster 4 January 2008–February 2010 2098 1736 1.21 35.89 .001
 Cluster 5 March 2009–September 2014 618 447 1.38 29.33 .001
 Cluster 6 October 2011–December 2015 492 344 1.43 28.17 .001
 Cluster 7 January 2008–December 2015 1243 1005 1.24 26.42 .001
 Cluster 8 January 2010–July 2015 532 384 1.39 25.69 .001
 Cluster 9 March 2014–September 2015 498 363 1.37 22.65 .001
 Cluster 10 April 2009–August 2014 5679 5248 1.09 18.04 .047
Open in a new tab

Abbreviation: —, not applicable.

aData sources: US Census Bureau cartographic boundary files12 and Ohio Department of Health linked birth–death certificate files.

bCluster numbers correspond to the numbers on maps in Figure 4.

c P values generated through Monte Carlo replications of the log likelihood ratio test statistic; P < .05 was considered significant.

Similar to the space–time analysis of infant mortality, the space–time analysis of preterm birth indicated that most clusters of preterm birth were temporally persistent. The space–time clusters in Akron, Canton, Cincinnati, Cleveland, Columbus, Dayton, and Youngstown were nearly identical to the spatial clusters and spanned the entire study period (Figure 3C and Table 2). The cluster of preterm births in Toledo (C6) occurred in the last 2 years of the study period (March 2014 through December 2015), and a small cluster of preterm births to the east of Dayton (C7) occurred in the last 4 years of the study period (November 2011 through December 2015). The small cluster of preterm births to the north of Columbus (C10) occurred primarily during the early part of the study period (June 2008 through November 2013). When we adjusted for maternal characteristics, we observed changes in the timing of the clusters in the urban areas (Figure 3D). The 2 other spatial clusters of preterm births identified in the adjusted analysis were in rural Appalachia (C1 and C7), indicating a higher-than-expected number of preterm births given the age and racial/ethnic distribution of mothers who gave birth in these regions.

Discussion

Various statewide and community-based initiatives exist to reduce infant mortality in Ohio. By using geospatial analysis techniques, the IMRP explored the complex sociospatial relationships that are key to understanding spatial inequalities in infant mortality and preterm birth. Through a series of quarterly presentations and community-based workshops, the partnership presented results to a wide range of organizations engaged in health and human services. Partners were challenged to think critically about the spatial patterns observed in maps and question the social, economic, environmental, and structural factors that might influence those patterns. Here, we discuss our results in light of conversations with state and local partners.

In our study, infant mortality in Ohio showed a clear urban pattern. The Ohio Department of Health chose the OEI counties based on county-level aggregate data on infant deaths, and our spatial cluster analysis supports the selection of these 9 counties. County-level geographic analysis is good for large-scale policy decisions, but this type of analysis masks important within-county spatial variation. Many local health organizations engaged in the IMRP (eg, the Franklin County OEI) do not have the data or analytic capacity to develop census tract–level maps of birth outcomes. Our partners expressed surprise at the geographic location of some of the clusters, and they used the results of our analyses to consider various choices about resource distribution, where to target programs and interventions, and with which community organizations to partner. The maps also fostered conversations about the reasons for large-scale and small-scale inequalities, the availability of health services, neighborhood safety, housing opportunities, and homelessness.

Adjusting clusters for maternal age, race/ethnicity, and education resulted in the disappearance of many of the urban clusters of infant mortality. These results allowed us to discuss how urban residence does not increase the risk of infant mortality, per se; instead, the pattern reflects the concentration in urban centers of populations at highest risk of poor birth outcomes. Partners discussed needs specific to urban populations, programs that work well in urban communities, and the characteristics of successful programs in these communities.

That the clusters of preterm birth did not disappear entirely after adjustment for maternal characteristics was also discussed. Although some of the excess risk in these areas was explained by the concentration of the population at highest risk of poor birth outcomes, the persistence of many urban clusters implies that other factors contribute to clustering. We discussed social and environmental risk factors for preterm birth, many of which are described in the scientific literature (eg, air pollution,33,34 crime,35,36 neighborhood physical/social disorder).37 This discussion challenged the partnership to think about complex spatial interactions and social structures in urban areas that collectively affect maternal and child health.

Representatives from state health agencies were particularly interested in the rural clusters of infant mortality and preterm birth identified in the adjusted analyses. Although the overall population living in these areas is small compared with the populations in urban areas targeted by the OEI and other state-level programs, state partners saw these rural clusters as a way to expand the OEI program into rural communities.

The space–time analysis was challenging for the partnership to understand. The temporally persistent clusters of infant mortality indicated that the spatial extent of infant mortality did not decrease over time. We saw little evidence of improvement in Ohio’s urban areas. Partners were, understandably, disappointed with this finding. When we adjusted the space–time clusters for maternal age, race/ethnicity, and education, we saw a pattern similar to the pattern in the spatial analysis, suggesting that clusters were concentrated among young, poorly educated, African American mothers during the entire study period. Essentially, these maternal characteristics explain both the spatial pattern and the temporal persistence of infant mortality in urban areas. One partner suggested these findings implied that they should continue to focus on their target population (low-income, urban, African American mothers).

The unadjusted space–time analysis of preterm birth also suggested that few changes in preterm birth rates occurred over time. But the adjusted space–time analysis tells a different story. The temporal extent of most of the clusters changed quite a bit, which likely reflects changing demographic characteristics. These results stimulated discussions about whether the geographic location of communities at high risk of poor birth outcomes has gradually changed as a result of gentrification, suburbanization of poverty, and changes in the housing market. Cities in Ohio have changed demographically during the past 10 years; these changes have been driven by changes in urban economies and gentrification. In some areas, the low-income population is gradually being displaced from traditional urban communities.38 Columbus and Cincinnati have particularly low rates of unemployment and high rates of urban growth, whereas Cleveland has consistently lagged in economic growth.39 Partners discussed how the space–time clusters of preterm birth may reflect this dynamic. In addition, the 2 temporally persistent clusters in rural Appalachia reflect the ongoing economic crisis and entrenched population at high risk of poor birth outcomes in this region of the state. Local partners in particular were concerned about whether these processes are shifting the location of their target populations such that target populations may reside in new areas of a city or in rural counties bordering major metropolitan regions. Understanding changes in the geographic distribution of the population at risk of poor birth outcomes is critical for the success of programs tailored to this population.

Limitations

This study had several limitations. First, the geographic boundaries of the clusters detected in this study are approximations of the “true” clusters, especially because births were aggregated to census tracts. Although we know the general location of a cluster, we are uncertain about its exact boundaries. Furthermore, as with any ecological analysis, we cannot say that the entire population living in the cluster area has the same risk for giving birth to a preterm infant or having an infant die within the first year of life. Women have varying levels of risk, and these risks depend on their individual characteristics, behaviors, and family histories. Second, this geographic analysis uses residence at birth. Studies have shown that 10%-30% of women change residence between conception and birth.40,41 However, most of these moves appear to be local (eg, within the same city or county), and the characteristics of women who move are similar to those who do not move.42 Caution should be exercised when interpreting the results of geographic studies that use maternal residential address at delivery, especially if trying to ascribe the case of a cluster to a local program or event. Finally, it is important to keep in mind that because infant death is a rare event, the power of the statistical test may be too low to detect small clusters, especially in rural areas. This issue is likely compounded by the fact that records lost during the geocoding process (because the address could not be geocoded) were more likely to occur in rural areas.

Conclusions

Our study used data from a large, socioeconomically diverse population during a 7-year period to understand the dynamics of infant mortality and preterm birth in a state that has focused considerable public health resources toward reducing poor birth outcomes. One strength of our analysis was that it allowed for detection of changes over time as various maternal and child health initiatives were implemented in areas across the state. Furthermore, a strong government–academic partnership fostered a productive, iterative process by which statistical analyses and results were discussed and refined throughout the study to best meet the needs of state and local public health entities. Geovisualization and spatial analysis are excellent tools for engaging public health practitioners in spatial thinking. Results of spatial analyses can stimulate important discussions about the social and spatial determinants of health and the complex processes that lead to spatial inequalities. Although the original goal of the IMRP spatial analysis was to examine the spatial patterns of infant mortality and preterm birth and the effect of the OEI programs, we found that results led state and local organizations to think critically about the causes and consequences of spatial patterns and what they might mean for the future of outreach efforts, programs, and policies.

Footnotes

Declaration of Conflicting Interests: The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding: This research was funded by a grant from the Ohio Department of Medicaid, Ohio Department of Health, and the Ohio Department of Higher Education. Any opinions, findings, and conclusions expressed in this article are those of the authors and do not necessarily reflect the views of these agencies.

ORCID iD

Elisabeth Dowling Root https://orcid.org/0000-0002-9566-4031

References

  • 1. US Department of Health and Human Services GeoHEALTH platform. Accessed August 26, 2019 https://geohealth.hhs.gov/arcgis/home
  • 2. Centers for Disease Control and Prevention National Environmental Public Health Tracking Network. Accessed August 26, 2019 https://ephtracking.cdc.gov/showHome.action
  • 3. Centers for Disease Control and Prevention Interactive atlas of heart disease and stroke. Accessed August 26, 2019 https://www.cdc.gov/dhdsp/maps/atlas/index.htm
  • 4. Centers for Disease Control and Prevention Diabetes. Data & statistics. Accessed August 26, 2019 https://www.cdc.gov/diabetes/data/index.html
  • 5. Massoudi BL., Chester K., Shah GH. Public health staff development needs in informatics: findings from a national survey of local health departments. J Public Health Manag Pract. 2016;22(suppl 6):S58-S62. 10.1097/PHH.0000000000000450 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Drezner K., McKeown L., Shah GH. Assessing skills and capacity for informatics: activities most commonly performed by or for local health departments. J Public Health Manag Pract. 2016;22(suppl 6):S51-S57. 10.1097/PHH.0000000000000459 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. National Research Council Learning to Think Spatially. National Academies Press; 2006. [Google Scholar]
  • 8. Ohio Department of Health 2017 Ohio infant mortality data: general findings. 2018. Accessed August 28, 2019 https://odh.ohio.gov/wps/portal/gov/odh/know-our-programs/infant-and-fetal-mortality/reports/2017-ohio-infant-mortality-report-final
  • 9. Ohio Commission on Minority Health Achieving equity and eliminating infant mortality disparities within racial and ethnic populations: from data to action. 2015. Accessed March 5, 2020 https://www.mih.ohio.gov/Portals/0/Medical%20Expert%20Panel/IM%20White%20paper%20as%20of%20September%2021%208.26%20am%20sw.pdf
  • 10. Ohio Collaborative to Prevent Infant Mortality Ohio infant mortality reduction plan 2015-2020. 2015. Accessed March 5, 2020 https://odh.ohio.gov/wps/portal/gov/odh/know-our-programs/infant-and-fetal-mortality/reports/mortality-plan-15-20
  • 11. Ohio Department of Health Ohio Equity Institute. 2015. Accessed March 5, 2020 https://odh.ohio.gov/wps/portal/gov/odh/know-our-programs/maternal-child-health-program-mp-formerly- cfhs/oei/
  • 12. US Census Bureau Cartographic boundary files. 2018. Accessed March 21, 2020 https://www.census.gov/geographies/mapping-files/time-series/geo/carto-boundary-file.html
  • 13. US Census Bureau 2013-2017 American Community Survey 5-year estimates. 2018. Accessed March 21, 2020 https://www.census.gov/programs-surveys/acs/data.html
  • 14. Rushton G. Public health, GIS, and spatial analytic tools. Annu Rev Public Health. 2003;24:43-56. 10.1146/annurev.publhealth.24.012902.140843 [DOI] [PubMed] [Google Scholar]
  • 15. Cromley EK., McLafferty SL. GIS and Public Health . 2nd ed Guilford Press; 2012. [Google Scholar]
  • 16. South AP., Jones DE., Hall ES. et al. Spatial analysis of preterm birth demonstrates opportunities for targeted intervention. Matern Child Health J. 2012;16(2):470-478. 10.1007/s10995-011-0748-2 [DOI] [PubMed] [Google Scholar]
  • 17. Ponce NA., Hoggatt KJ., Wilhelm M., Ritz B. Preterm birth: the interaction of traffic-related air pollution with economic hardship in Los Angeles neighborhoods [published erratum appears in Am J Epidemiol. 2010;171(11):1248]. Am J Epidemiol. 2005;162(2):140-148. 10.1093/aje/kwi173 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Grady SC., Enander H. Geographic analysis of low birthweight and infant mortality in Michigan using automated zoning methodology. Int J Health Geogr. 2009;8:10. 10.1186/1476-072X-8-10 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Matteson DW., Burr JA., Marshall JR. Infant mortality: a multi-level analysis of individual and community risk factors. Soc Sci Med. 1998;47(11):1841-1854. 10.1016/S0277-9536(98)00229-9 [DOI] [PubMed] [Google Scholar]
  • 20. Hertel-Fernandez AW., Giusti AE., Sotelo JM. The Chilean infant mortality decline: improvement for whom? Socioeconomic and geographic inequalities in infant mortality, 1990-2005. Bull World Health Organ. 2007;85(10):798-804. 10.2471/blt.06.041848 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Szwarcwald CL., Andrade CL., Bastos FI. Income inequality, residential poverty clustering and infant mortality: a study in Rio de Janeiro, Brazil. Soc Sci Med. 2002;55(12):2083-2092. 10.1016/S0277-9536(01)00353-7 [DOI] [PubMed] [Google Scholar]
  • 22. Waller LA., Gotway CA. Applied Spatial Statistics for Public Health Data. John Wiley & Sons; 2004. [Google Scholar]
  • 23. Anselin L., Syabri I., Kho Y. GeoDa: An Introduction to Spatial Data Analysis . University of Chicago, Center for Spatial Data Science; 2018. [Google Scholar]
  • 24. Anselin L. Maps for rates or proportions. Published August 3, 2018. Accessed March 5, 2020 http://geodacenter.github.io/workbook/3b_rates/lab3b.html
  • 25. Kulldorff M., Nagarwalla N. Spatial disease clusters: detection and inference. Stat Med. 1995;14(8):799-810. 10.1002/sim.4780140809 [DOI] [PubMed] [Google Scholar]
  • 26. Kulldorff M., Athas WF., Feurer EJ., Miller BA., Key CR. Evaluating cluster alarms: a space–time scan statistic and brain cancer in Los Alamos, New Mexico. Am J Public Health. 1998;88(9):1377-1380. 10.2105/AJPH.88.9.1377 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Root ED., Gonzales L., Persse DE., Hinchey PR., McNally B., Sasson C. A tale of two cities: the role of neighborhood socioeconomic status in spatial clustering of bystander CPR in Austin and Houston. Resuscitation. 2013;84(6):752-759. 10.1016/j.resuscitation.2013.01.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Klassen AC., Kulldorff M., Curriero F. Geographical clustering of prostate cancer grade and stage at diagnosis, before and after adjustment for risk factors. Int J Health Geogr. 2005;4(1):1. 10.1186/1476-072X-4-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. R: A Language and Environment for Statistical Computing [computer program]. Version 3.4.2. R Foundation for Statistical Computing; 2017.
  • 30. SAS [computer program]. Version 9.4. SAS Institute Inc; 2014.
  • 31. SaTScan [software for the spatial and space–time scan statistics]. Version 7.0. Kulldorff M Information Management Services Inc; 2006.
  • 32. ArcGIS [mapping software]. Version 10.6. Esri; 2018.
  • 33. Ritz B., Wilhelm M., Hoggatt KJ., Ghosh JKC. Ambient air pollution and preterm birth in the Environment and Pregnancy Outcomes Study at the University of California, Los Angeles. Am J Epidemiol. 2007;166(9):1045-1052. 10.1093/aje/kwm181 [DOI] [PubMed] [Google Scholar]
  • 34. Brauer M., Lencar C., Tamburic L., Koehoorn M., Demers P., Karr C. A cohort study of traffic-related air pollution impacts on birth outcomes. Environ Health Perspect. 2008;116(5):680-686. 10.1289/ehp.10952 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Messer LC., Kaufman JS., Dole N., Savitz DA., Laraia BA. Neighborhood crime, deprivation, and preterm birth. Ann Epidemiol. 2006;16(6):455-462. 10.1016/j.annepidem.2005.08.006 [DOI] [PubMed] [Google Scholar]
  • 36. Masi CM., Hawkley LC., Piotrowski ZH., Pickett KE. Neighborhood economic disadvantage, violent crime, group density, and pregnancy outcomes in a diverse, urban population. Soc Sci Med. 2007;65(12):2440-2457. 10.1016/j.socscimed.2007.07.014 [DOI] [PubMed] [Google Scholar]
  • 37. Giurgescu C., Zenk SN., Dancy BL., Park CG., Dieber W., Block R. Relationships among neighborhood environment, racial discrimination, psychological distress, and preterm birth in African American women. J Obstet Gynecol Neonatal Nurs. 2012;41(6):E51-E61. 10.1111/j.1552-6909.2012.01409.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Zuk M., Bierbaum AH., Chapple K., Gorska K., Loukaitou-Sideris A. Gentrification, displacement, and the role of public investment. J Plan Lit. 2018;33(1):31-44. 10.1177/0885412217716439 [DOI] [Google Scholar]
  • 39. US Conference of Mayors and Ohio Mayors Alliance Ohio metro economies: GMP and job growth. 2018. Accessed September 28, 2018 http://www.usmayors.org/wp-content/uploads/2018/07/Ohio-Metro-Economies-Report-July-2018.pdf
  • 40. Bell ML., Belanger K. Review of research on residential mobility during pregnancy: consequences for assessment of prenatal environmental exposures. J Expo Sci Environ Epidemiol. 2012;22(5):429-438. 10.1038/jes.2012.42 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Fell DB., Dodds L., King WD. Residential mobility during pregnancy. Paediatr Perinat Epidemiol. 2004;18(6):408-414. 10.1111/j.1365-3016.2004.00580.x [DOI] [PubMed] [Google Scholar]
  • 42. Canfield MA., Ramadhani TA., Langlois PH., Waller DK. Residential mobility patterns and exposure misclassification in epidemiologic studies of birth defects. J Expo Sci Environ Epidemiol. 2006;16(6):538-543. 10.1038/sj.jes.7500501 [DOI] [PubMed] [Google Scholar]

Articles from Public Health Reports are provided here courtesy of SAGE Publications

RESOURCES