Comparing spatial distributions of infant mortality over time: Investigating the urban environment of Baltimore, Maryland in 1880 and 1920

Applied Geography 86 (2017) 1e7

Contents lists available at ScienceDirect

Applied Geography journal homepage: www.elsevier.com/locate/apgeog

Comparing spatial distributions of infant mortality over time: Investigating the urban environment of Baltimore, Maryland in 1880 and 1920 Sarah E. Hinman Leiden University College The Hague, P.O. Box 13228, 2501 EE, The Hague, The Netherlands

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 June 2015 Received in revised form 2 April 2017 Accepted 10 June 2017

Infant mortality is a sensitive indicator of urban environmental conditions, and investigating the geography of such an indicator provides insight into variables affecting public health in urban North America in 1880 and 1920. Geographic information systems (GIS) and spatial analysis now provide a means by which to view past infant mortality distributions from a new perspective, one not available at the time. This study makes use of data collected from the 1880 and 1920 Vital Statistics Death Records for Baltimore, Maryland - mapping each infant death to his or her place of residence. Previous work with the 1880 data indicates an uneven distribution of infant deaths with some degree of spatial clustering. The current study takes these findings a step further through the use of the local spatial autocorrelation statistic, Gi*, to identify the locations of clusters in one or both years. The aim of the comparison is to determine whether the location of infant mortality clusters remained the same over time indicating persistent environmental, and possibly demographic, challenges in certain neighborhoods. The data indicated hotspots of infant mortality in both years with persistence in the Fells Point area of Baltimore. The significant clusters appeared in neighborhoods with large African American and/or immigrant populations in both years. The hotspots in the primarily African American neighborhood were only significant in 1880 despite presenting some intriguing questions about what caused such a change, particular when the population in that part of the city did not change. This work offers insights into the spatial distribution of infant mortality in the past and clues regarding which parts of the city need additional investigation to better understand their social and environmental characteristics. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Infant mortality Baltimore Historical GIS Urban environment

1. Introduction Whether considering the past or the present, infant mortality provides a sensitive indicator of urban environmental conditions. While it is difficult to identify the precise variable(s) that sustained late nineteenth century rates around 25 percent of live births, examining the geography of infant deaths provides an opportunity to explore those urban neighborhoods that might have experienced particularly high infant mortality rates and poor environmental conditions. Between 1880 and 1920 North American cities went through the epidemiological transition resulting in the majority of deaths coming from chronic ailments rather than infectious diseases like typhoid fever (Elman & Myers, 1999). While the entire population benefitted from a reduction in mortality rates, infant

E-mail address: [email protected]. http://dx.doi.org/10.1016/j.apgeog.2017.06.015 0143-6228/© 2017 Elsevier Ltd. All rights reserved.

mortality rates declined dramatically during this time period. In Baltimore, Maryland, the infant mortality rate in 1880 was 250 infant deaths per 1000 live births. This rate decreased to 90 deaths per 1000 live births by 1920. While a single trigger of declining rates may never fully be teased apart from other related variables, or the causes may have varied from place to place, it is possible to compare the locations of infant deaths as a means of identifying places with potentially less salubrious conditions as a means of peeking into past environmental conditions. This is particularly true if there are statistically significant clusters of infant mortality providing a focal point or points rather than an entire city to investigate. Even more intriguing would be if clusters in 1880 persist in 1920. Knowledge gained about the correlations between infant mortality and local conditions has implications for how to investigate modern spaces for interventions relating to infant mortality and infant health. Frequently there is a lack of public health data,

2

S.E. Hinman / Applied Geography 86 (2017) 1e7

particularly local level geographic data, in the urban Global South. While current data and data specific to locations in the developing world would be best for understanding the ongoing high rates of infant mortality, historical datasets marking the beginning and end of the North American epidemiological transition provide an opportunity to explore a potentially comparable scenario with a complete dataset. Scholars of urban public health in the developing world highlight the need for consistent data collection in order to begin to resolve negative health outcomes (Konteh 2009). Additionally, Harpham (2009) emphasizes the need to better define “community” in low-income urban settings. One approach to identifying communities in need can be to work with individual level data, such as the infant mortality data in this paper, in combination with spatial statistics in order to locate areas facing a specific public health problem. The challenge of consistent and reliable collection of data relating to vital events such as births and deaths is raised by both Harpham (2009) and Konteh (2009). Therefore, the current study seeks not only to better understand the geography of infant mortality in a 19th century American city, but also to begin to explore what aspects of vital event data for infants is most useful for understanding drivers behind geographic patterns of infant deaths and finally to suggest an alternative means of defining a community or neighborhood that is neither dependent on official boundaries nor local qualitative information. The study of historical infant mortality patterns, decline, and the variables contributing to increased risk of early death present a fascinating puzzle. As with all areas of study the answers produced are dependent upon the particular the study framing (Gregory, 2008; Woods, Watterson, & Woodward, 1988). The current study draws on two related and connected lines of inquiry, both of which relate to our understanding of infant mortality in the late nineteenth and early twentieth centuries. The first are those scholars considering the macro scale trends of where within a country did infant mortality rates decline first and/or at what speed (Gregory, 2008; Preston & Haines, 1991; Woods et al., 1988). The second line of inquiry looks within a localized area at which variables played the most important role in reducing an infant's chances of survival (Williams, 1992; Haines, 1995; Thornton & Olson, 2001, 2011). Within the latter group the findings can be organized into socio-economic drivers, cultural group/ethnic identity, or environmental conditions. Where these three sets of variables become problematic is that they are intertwined. Socio-economic status will have an impact on the housing quality and thus access to clean drinking water, for example (Rochester, 1923; Williams, 1992). In the most conclusive studies of the role of culture in infant mortality, the socio-economic status of certain cultural groups was somewhat determined by other societal attitudes of the time in terms of types of employment (Thornton & Olson, 2001, 2011). Thornton and Olson (2001) discuss ideas along these lines in Montreal for the same time period. They found a distinct improvement or detriment, depending, in survival of infants regardless of ethnic group depending upon which ethnic group dominated a given street. For example, while French Canadian infants in Montreal regardless of the family's socio-economic status had reduced chances of survival. Yet, if the family lived in an area of mostly Irish Catholics the French Canadian infant had an increased chance of surviving the first year. Previous work with the Baltimore, Maryland data from 1880 presented here sought to further investigate the connection between infant mortality and the environment by exploring relationships with proximity to of industrial land use. While the results indicated no correlation between land use and infant mortality, there did appear to be some connections between residential location and proximity to the waterfront. Additionally, questions related to housing density (rather than population density) surfaced (Author).

It should be noted that as health overall improved during this period and both the improvement for infant health as well as the rest of the population was at least partly connected to infrastructure/environmental changes as well as changes in medical knowledge. As the body of literature specific to historical infant mortality is small and some reference to general mortality literature where appropriate broadens the foundation of this work. First and most importantly is the role of city-wide infrastructure improvements, particularly that of clean drinking water producing a reduction in infant and child mortality from water-borne diseases. What is interesting according to Ferrie and Troesken (2008) though is that the overall impact of a clean water supply had variable results depending upon city. The implication is that within a city a variable geography of improvement could also have existed. Environmental conditions and their relationship with historical infant mortality are difficult to illustrate, particularly in a detailed way that may shed light on nuances of causality. When working with aggregated data, correlations tend to be weak (Condran & Crimmins-Gardner, 1978). Alternatively, it is laborious to work with individual level data. The difficulties of using infant death data pale when faced with finding detailed and specific environmental data at the same scale. For example, a city might build and complete a sewerage system, but this does not mean all households acquire connections to the service and records of these connections tend to be limited (Colten, 2002). Therefore, the purpose of this study is first to determine if there is a spatial clustering pattern among infant deaths in Baltimore in 1880 and in 1920. Second, to identify if there are areas of the city in which infant mortality clustered in both 1880 and 1920 which could indicate persistent environmental disamenities and persistent poverty. By identifying such areas it is easier to then investigate the particular characteristics of those parts of the city rather than the entire city. Finally, as Baltimore was a border city between the north and south there are opportunities to examine racial differences in infant mortality patterns between the two years. The two years investigated represent, in general, the beginning and end of the epidemiological transition in North American cities and they are also census years so that future research could utilize data from the enumeration schedules. 2. Study area: Baltimore, Maryland Baltimore, Maryland (39.2904 N, 76.6122 W) is just one example of a North American city undergoing the typical processes of urbanization and industrialization in the late nineteenth and early twentieth centuries. Yet there are a number of features in the city's history that lend this location to being particularly good for a geographic study of infant mortality at the individual level. First, as a border city between the north and south, Baltimore has a long history of African Americans and whites living in close proximity to each other. Following the Civil War, the implications of this history and the subsequent northward migration of blacks from the rural south resulted in 16 percent of the city's residents bring black in 1880 (Groves & Muller, 1975). In the study of infant mortality the racial differences in the city along with the socio-economic implications could result in identification particular of geographies. Not only does Baltimore have a particular historical story of African American migration that predates the Great Migration of southern blacks to northern cities following World War I, but it is also a destination port city for migrants from Europe in a process that began to increase in numbers with the Irish Famine of the 1840s along with German immigration in the same decades. While not the same degree of destination for immigrants as New York or Boston, Baltimore played a notable role in this part of American history. Spatially the result of immigration to Baltimore was the

S.E. Hinman / Applied Geography 86 (2017) 1e7

development of strong cultural enclaves usually located near places of employment. Regardless of where immigrating families eventually ended up in the city, most began their time in Baltimore residing in Fells Point where the ships bringing them to America landed. The geographic patterns from these trends included many rounds of succession and invasion as newer arrivals sought the cheapest housing in Fells Point and the slightly more established immigrants moved away from that part of the waterfront and into neighborhoods dominated by their cultural group (Groves & Muller, 1975; Hayward, 2008; Rochester, 1923). So in this way the geographic patterns of culture could begin to appear amongst the data as well and could follow in the work of Thornton and Olson (1991, 1997, 2001, 2011). In the interim period between the two years tested here, 1880 and 1920, not only did medical knowledge evolve tremendously and causes of death predominantly shift from infectious to chronic, but public health infrastructure improved potentially changing the urban environment. Baltimore is less unique in this realm than the cities mentioned above, but due to the city's political tensions, the timing of infrastructure implementation is reasonably peculiar. Primarily this relates to the timing of when Baltimore built its comprehensive sewerage system. Most large American cities built sewerage systems prior to 1900. While public health officials in Baltimore consistently referenced the need for sewers to improve health in the city, it was not until the fire of 1904 that finally tipped the balance in favor of the infrastructure upgrade (Boone, 2003). The sewerage system building began in 1905 and was completed in 1911. In terms of timing this element should not have visible outcomes on the geography of infant mortality other than to see substantially reduced numbers of deaths given that the data analyzed predate and postdate construction and completion. This could imply then that non-environmental variables drove infant mortality patterns in 1920. 3. Materials and methods Historical GIS is a well-established field (Knowles, 2002; Donahue, 2004; Cunfer, 2005; Gregory & Ell, 2007; Knowles and Hillier 2008; Gregory, 2008). In connection to the study of turn of the twentieth century infant mortality, the use of GIS allows for the first time the mapping of data that were originally contained only in written form - death certificates - providing the opportunity to visualize the spatial patterns of historical infant mortality. Additionally, a number of spatial statistical techniques which have existed for over two decades are now well integrated into GIS software making it a fairly simple matter to test for spatial autocorrelation (Getis & Ord, 1992; Ord & Getis, 1995). Of interest is the opportunity of using modern spatial analysis techniques to explore historical infant mortality data in order to identify urban neighborhoods in need of more detailed investigation so that a clearer picture of urban environmental conditions can be drawn. 3.1. Data The data needed to develop a historical GIS that investigates the questions asked here include death certificates from Baltimore and base maps from which digital layers of the city's blocks in both years can be created. Infant death information was extracted from all death certificates for each year. For both years all death certificates had to be viewed to determine if record was for an infant death or not. Infants are defined as individuals between birth and one year of age and following Thornton and Olson (2001) those not surviving the first 24 h are excluded as these would be from congenital causes and not related to any of the possible variables discussed above. All death certificates had to be completed by the

3

physician attending the death which should ensure their accuracy, within reason, given medical knowledge at the time. While consistent recording keeping of births and deaths was a fairly new requirement in the late nineteenth century, such practices became mandatory for the Baltimore City Health Department in 1875 to comply with Ordinance 86 (Howard 1924). According to Howard (1924) despite the requirement for birth certificates beginning in 1875, this procedure was not reliable until 1915. Therefore for this study, it was not possible to map individual level birth data for 1880 resulting in the use of the count of deaths only for both years studied. Death records were kept more consistently due to the need to have the document for a cemetery burial. In 1880 each death certificate included space for the date of death, person's full name, age (including years, months, and days if known), race, place of death, both primary and secondary causes of death, duration of illness, and sex. The place of death listed above typically refers to the place of residence given the limited use of hospitals at that time and was confirmed via cross-referencing the addresses listed with contemporary maps of the city, in most instances the place of death in 1880 was residential in nature. In 1920 each certificate included much of the same information, but with a few additions. There was space both for place of death and place of residence along with the city ward number. Other additional information was father's name, father's birthplace, mother's maiden name, and mother's birthplace, information about the individual's length of stay in hospital if that was where the death occurred. Finally, there was a special section added for those individuals who lived for less than a day to identify the number of hours and/or minutes the newborn survived. Only information for those certificates of individuals 12 months of age or younger were recorded for the GIS, but all of the information contained on each death certificate regardless of its relevance to this project was included. For the purpose of the present study only the location of residence and race are of interest, but future work will investigate seasonality, cause of death, and ethnicity. 3.2. GIS development The data collected led to the creation of two databases. The first, for 1880, contains 2306 infant deaths that occurred in that year. Of these deaths, it was possible to map 1525 individuals to their location of residence. Due to data illegibility or missing addresses, naturally some data points could not be mapped. In order to create a map of points representing infant deaths it was not possible to use the usual geocoding methods available in GIS software as the addressing system used by Baltimore City changed in 1887. Instead, the infant deaths were mapped individually by using address information included on Sanborn Fire Insurance maps published in 1890, but only after locating a list of old and new addresses in the 1887 Baltimore City Directory. The addressing system for the city as a whole changed and was standardized in that year. The final hurdle to overcome in mapping the data from 1880 involved decisions related to accuracy. In a few instances the residential address of an infant was listed as “the corner of Pratt and Light Street,” for example, or even more vaguely “South Broadway.” As there appeared to be no bias in this recording method the events were mapped as close to their place of death as possible and given a ranking of the accuracy of the mapping from one to three. Those infant deaths with a level three degree of accuracy are those that are located on the correct block but not the correct house. Those infant deaths with a level two degree of accuracy are located on the correct corner, and those with level one accuracy are mapped to the actual residence. This ranking system will allow for a narrowing of the dataset should future analyses require a more precision and can absorb the more limited number of data points. All mapped points

4

S.E. Hinman / Applied Geography 86 (2017) 1e7

were included in the analysis as the data were ultimately aggregated to the city block level for statistical analysis therefore the exact household location was not needed. Once mapped those infants who did not survive the first day were removed leaving 1401 deaths for analysis. Fig. 1 illustrates the distribution of infant deaths in 1880. As race and racial segregation may play a role in clustering patterns the data are displayed here by race. The 1880 death certificates tended to record race as “white,” “colored,” “mulatto,” and a few “unknown,” all of these race classifications area included in Fig. 1. The second database contains the information for infant deaths in 1920. There were 1942 infant death certificates recorded for this year. Of these deaths, it was possible to map 1075 individuals to the place of residence. These events were mapped through heads up digitizing individual points using georeferenced Sanborn Fire Insurance maps published in 1915 to access the residential address information. While in 1880 there were four categories of race recorded in the database, for the 1920 data only “white” and “black” (written at the time as “colored”) were noted on the death certificates therefore Fig. 2 reflects this categorization change. To create a digital layer of city compatible with the 1880 deaths, city blocks were digitized from georeferenced images of the 1876 Hopkins Atlas of Baltimore City and County. The city annexed surrounding parts of Baltimore County in 1888 and 1918 thus increasing the physical size of the city during the intervening years of this study (Arnold, 1978). The city's last annexation occurred in 1918 and therefore the city's area reached its modern boundary at that time, although it would take longer for all of the open spaces to be filled in with buildings, parks, and other urban land uses. Much of the historical city center remained the same between 1880 and 1920. In both years the boundary used is the municipal boundary of Baltimore City.

3.3. Analysis methods To test whether or not infant deaths in 1880 and 1920 cluster locally this study used the Gi* statistic. This technique allows testing for local spatial autocorrelation. By looking within a dataset it can be determined if there are hot spots of events within the larger dataset. More importantly, the tools available not only

Fig. 2. Distribution of infant deaths in 1920.

confirm or deny that there is local spatial autocorrelation, but produce mappable information allowing for the visualization of the data hotspots. The Gi* statistic, developed by Getis and Ord (1992; Ord & Getis, 1995), allows for the testing of events within a series of increasing bandwidths with the premise of identifying cluster sizes that might indicate a source of environmental causality or to test a single bandwidth to simply confirm or deny local spatial autocorrelation following the equation [see: Hinman, Blackburn, & Curtis, 2006]. As infant mortality occurs as a result of a number of variables and it is unlikely to be related to a particular cause and there is no set distance to use, therefore by using the Optimized Hot Spot Analysis tool (OHSA) in ArcGIS 10.2.x was used. This allowed the tool to use the data themselves to determine the best possible bandwidth size to use with each dataset in order to identify local clustering patterns. There are a number of choices using the OHSA tool regarding how to input the original data. For the purposes of this project the data used were the count of events on each city block upon which they occurred. Only those blocks positive for infant deaths were used in the analysis. The information input into the Optimized Hot Spot Analysis tool for 1880 included 843 city blocks with a column of information summarizing the number of infant deaths that occurred on each of those blocks. The number of deaths on the blocks ranged from 23 to 1 with an average of 1.65. The OHSA tool returned a result using a fixed distance band around each event positive block of 652.33 m. For the 1920 data, 738 city blocks with summary information of the number of infant deaths on each block were analyzed using the OHSA tool. The number of deaths on the blocks ranged from 5 to 1 with an average of 1.3 events per block. In this instance the tool identified peak clustering occurring with a bandwidth of 374.15 m. 4. Results

Fig. 1. Distribution of infant deaths in 1880. The word “colored” from the original death certificates and being reported here as “black”.

Statistically significant clusters were identified in both of the years tested. In 1880 there were 62 hotspots located in two distinct neighborhoods (Fig. 3). The first of these areas comprising 26 clusters is in Fells Point a waterfront neighborhood to the east of the city center. The second area is slightly to the northwest of the city center and contains the remaining 36 significant blocks. In

S.E. Hinman / Applied Geography 86 (2017) 1e7

5

Fig. 4. Gi* results for the infant deaths in 1920.

Table 2 Demographic variables relating to percentage of race in Baltimore and within the data used.

Fig. 3. Gi* results for the infant deaths in 1880.

1880 the z-scores ranged from a minimum of 2.69 to a maximum of 4.23 (Table 1). In 1920 there were fewer significant hotspots of infant mortality. Of the 738 blocks with infant deaths, 35 of these were the center of a statistically significant cluster. These clusters, with one exception, were all located in Fells Point and extended eastward (Fig. 4). The one cluster located elsewhere in the city is in the west of the city about equidistant between the northern and southern boundaries. The z-scores for this year ranged from 2.85 to 5.82 (Table 1). Table 2 summarizes the racial characteristics of the city and the racial composition of the significant clusters while Figs. 5 and 6 visualize the geographic patterns. Figs. 5 and 6 display significant clusters with a buffer surrounding the hotspots illustrating the bandwidth used, so that the reader can see which infant deaths contributed to the clustering. In general the table and figures highlight the relative contribution of race to infant mortality and significant clusters to the city's racial profile.

Demographic Characteristics of Baltimore Total Population 1880 332,313 Percent of Total Pop. 1920 733,452 Percent of Total Pop. Overall Infant Mortality Total Infant Deaths Analyzed 1880 1401 Percent of Deaths 1920 957 Percent of Deaths Infant Deaths Within Hotspot Bandwidths Northwest Clusters Total Infant Deaths 1880 359 Percent of total Fells Point Clusters Total Infant Deaths 1880 301 Percent of total 1920 150 Percent of total

5. Discussion The results indicate a number of interesting points of discussion, including investigating the racial characteristics of the infants who comprise the clusters themselves and the neighborhoods in which they are located. At the same time, these results are limited by the lack of birth records, the incomplete nature of historical data, and for now not knowing more about the neighborhood populations as a whole. The locations of the groups of clusters are intriguing as these neighborhoods represent two in which one expects to see higher levels of infant mortality. To focus first on the neighborhood of clusters northwest of the city center in 1880 means looking into a part of the city that was home to an established African American enclave. One of the most interesting elements of this neighborhood is that it housed black residents from across the socio-economic

White

Black & Mulatto

278,584 83% 625,130 85%

53,729 16% Black 108,322 14%

White

Black & Mulatto

876 62.5%

473 33.76%

692 72.31%

264 27.59%

White

Black & Mulatto

153 42.62%

206 57.38%

White

Black & Mulatto

252 83.72%

49 16.28%

140 93.33%

10 6.67%

spectrum. This is particularly intriguing since more than 50 percent of the infants contributing to the clusters were black or mulatto while overall these groups only comprised approximately 34 percent of all infant deaths in the city. This will require further investigation than the scope of this paper allows, but of interest is the opportunity to explore in greater detail whether most infants who did not survive were living on alleys or if this was not a

Table 1 Z-scores for Gi* analysis in 1880 and 1920. Year

Number of Hotspots

Minimum significant z-score

Maximum significant z-score

1880 1920

62 35

2.69 2.85

4.23 5.82

6

S.E. Hinman / Applied Geography 86 (2017) 1e7

Fig. 6. Composition of statistically significant clusters by race in 1920. Buffer distances based upon bandwidth used in the Gi* statistic, 374.15 m, in order to illustrate all of the data points contributing to the statistically significant clusters.

Fig. 5. a & b Composition of statistically significant clusters by race in 1880. Buffer distance based upon bandwidth used in the Gi* statistic, 652.33 m, in order to illustrate all of the data points contributing to the statistically significant clusters.

relevant variable. Due to persistent environmental and socioeconomic disamenities alley life could decrease an infant's chance of death. Alleys were not well drained, their narrowness limited the amount of light reaching houses and frequently more alley houses packed would be into the same length of street than on the main street. Also, and rather obviously, households renting or purchasing a residence on an alley were doing so because these were the least expensive properties in the neighborhood. While the neighborhood in question had a relatively large African American population in 1880, white infant deaths contributed to the statistically significant clusters here, too. It is important to investigate the socio-economic status of these white families along with their location with respect to alley residences. This can help to determine overall neighborhood effects on infant health. What is additionally fascinating about the northwestern clusters of infant deaths in 1880 is that they disappear in the 1920 data. This is not to say that infant deaths did not occur in this neighborhood in 1920, but that the data available did not produce statistically significant local clusters.

The completion of the city's comprehensive sewerage system in 1911 may explain these changes. The benefits for all from comprehensive and sanitary removal of waste could have been a major contributing factor to improvements in infant mortality. Troesken (2004) supports such a claim through the evidence he presents for Philadelphia indicating that while the entire city's population experienced better health with regard to infectious diseases, the African American residents experienced proportionally greater benefits than the white population. It is feasible that similar trends occurred in Baltimore helping to explain some of the differences between the clustering patterns in 1880 and 1920 as improved sanitation and clean water leads to improvements for infant health and thus survival beyond the first year. In contrast to the northwest, Fells Point contained statistically significant clusters in both 1880 and 1920. This part of Baltimore traditionally was home to the most recently arrived immigrants in the city. Some African Americans resided in this part of the city but mostly on alleys during the study period. As with other east coast port cities, Baltimore acted as a first point of entry for thousands of European immigrants with the first large waves beginning to arrive in the 1840s and 1850s. In terms of originating locations of the immigrants who arrived in Baltimore, mostly followed the general trends of New York and Boston. Therefore, first Irish and German immigrants arrived. Still, the proportions were different in Baltimore than its northern neighbors. By 1880 other immigrant groups were also arriving from eastern Europe and the earlier immigrants from Ireland and Germany had, generally begun to move beyond their earliest homes near the waterfront. There are a few things to note relating to the infant deaths of 1880 that offer some insights into whose infants were not surviving in Fells Point. First, both within and outside of Fells Point if an infant's place of birth was noted as somewhere other than Baltimore that location was either a nearby state or Bremen, Germany. In 1880 information about the parents' place of birth was not incorporated into the death certificate but given that 1880 was a census year future research could potentially use record matching to learn more about whole households. Nonetheless, through a quick inspection of the last names of the deceased infants there appears to be a remarkable number that appear to be of both Irish and German origin, alongside of a number of more English names as well. By the

S.E. Hinman / Applied Geography 86 (2017) 1e7

7

1880s immigrants from what is now the Czech Republic were arriving in Baltimore and like their predecessors they first resided in Fells Point and then migrated to more homogeneous neighborhoods, though given the linguistic affinity with German tended to locate near or in predominantly German neighborhoods in northeastern Baltimore. Also in the 1880s Polish migrants began to arrive in Baltimore, and this is where the story of infant mortality in 1920 picks up. A look at the Fells Point infant deaths in 1920 indicate numerous parents whose place of birth was Poland. Additionally, quite a few parents born in Baltimore have last names that appear to be Polish or of other eastern European origin.

within the now identified hotspots. Additionally the particular contributions of alley dwelling to patterns of infant mortality within the now highlighted neighborhoods have the potential to highlight broader elements of the urban environment.

6. Conclusion

References

The results of this research identified local clustering of infant mortality in Baltimore for both 1880 and 1920. There was persistence of clustering in the Fells Point neighborhood between the study years. The neighborhoods in which clustering was found are predominantly those housing African Americans and recently arrived immigrants to the United States. Through the preliminary exploration of the racial composition of the infant deaths and neighborhoods within which clustering occurred a few statements can also be made. Given the overall composition of Baltimore as a city, as a place in terms of race relations, and a destination for immigrants there are opportunities for any number of combinations of socio-economic status, cultural practices, and changing environmental conditions to have played a role in the geography of infant mortality. None of the variables will be simple to tease apart as to do so will require record matching between the manuscript census and the infant death records along with additional work with census records to rebuild the neighborhoods around infant deaths. If these data are compiled in a historical GIS there are great possibilities to begin to identify which variables played which kind of role in the city's public health profile. It is quite possible that different variables had different degrees of significance in different parts of the city. To begin to pull apart the importance of socio-economic status, cultural practice, and environmental conditions has implications for our broader understanding of urban development and public health processes. In terms of urban development these data help to fill in parts of the urban history and planning story for Baltimore around the turn of the twentieth century. Decisions made in the past resulted in infrastructure that continues to play a role today. The infrastructure built in the early twentieth century in turn has implications for how the urban environment can be altered and manipulated in the future. Additionally knowledge gained from a better understanding of infant mortality drivers can help to inform public health decision-making and interventions in places with developmental profiles similar to Baltimore between 1880 and 1920. This study fills a gap in the existing literature by providing a unique insight into local spatial autocorrelation at the city block scale. Our understanding is now enhanced by the ability to visualize the location of statistically significant hotspots of historical intraurban infant mortality. This enables us to better understand the public health and historical environmental landscapes enabling us to present valid corroboration for example in the urban Global South where similar data are not available or collected. Scale is critical to our understanding of urban public health and through the use of local spatial statistics we can better identify neighborhoods without reliance on formal political boundaries. To take this forward one would next need to further investigate the individual and household characteristics of those deaths within the mortality hotspots so we can understand commonalities amongst impacted families. Future research directions could include further exploration of the ethnic and racial composition

Arnold, J. L. (1978). Suburban growth and municipal annexation in Baltimore, 17451918. Maryland Historical Magazine, 73(2), 109e128. Boone, C. G. (2003). Obstacles to infrastructure provision: The struggle to build comprehensive sewer works in Baltimore. Historical Geography, 31, 151e168. Colten, C. E. (2002). Basin street blues: Drainage and environmental equity in new orleans, 1890e1930. Journal of Historical Geography, 28(2), 237e257. Condran, G. A., & Crimmins-Gardner, E. (1978). Public health measures and mortality in U.S. cities in the late nineteenth century. Human Ecology, 6(10), 27e54. Cunfer, G. (2005). On the great Plains: Agriculture and environment. College Station: Texas A&M University Press. Donahue, B. (2004). The great Meadow: Farmers and the land in colonial concord. New Haven: Yale University Press. Elman, C., & Myers, G. C. (1999). Geographic morbidity differentials in the late nineteenth-century United States. Demography, 36(4), 429e443. Ferrie, J. P., & Troesken, W. (2008). Water and Chicago's mortality transition, 1850e1925. Explorations in Economic History, 45(1), 1e16. Getis, A., & Ord, J. K. (1992). The analysis of spatial association by use of distance statistics. Geographical Analysis, 24(3), 189e206. Gregory, I. N. (2008). Different places, different stories: Infant mortality decline in England and Wales, 1851e1911. Annals of the Association of American Geographers, 98(4), 773e794. Gregory, I. N., & Ell, P. S. (2007). Historical GIS: Technologies, methodologies, and scholarship. Cambridge: Cambridge University Press. Groves, P. A., & Muller, E. K. (1975). The evolution of black residential areas in late nineteenth-century cities. Journal of Historical Geography, 1(2), 169e191. Haines, M. R. (1995). Socio-economic differentials in infant and child mortality during mortality decline: England and Wales, 1890-1911. Population Studies, 49(2), 297e315. Harpham, Trudy (2009). Urban health in developing countries: What do we know and where do we go? Health and Place, 15(1), 107e116. Hayward, M. E. (2008). Baltimore's alley Houses: Homes for working people since the 1780s. Baltimore: Johns Hopkins University Press. Hinman, S. E., Blackburn, J. K., & Curtis, A. J. (2006). Spatial and temporal structure of typhoid outbreaks in Washington, D.C., 1906-1909: Evaluating local clustering with the G*i statistic. International Journal of Health Geographics, 5(13). Howard, William Travis, Jr. (1924). Public health administration and the natural history of disease in Baltimore, Maryland, 1797e1920. Washington, DC: Carnegie Institution of Washington. Knowles, A. K. (Ed.). (2002). Past time, past place: GIS for history. Redlands, California: ESRI Press. Knowles, & Hillier, A. (2008). Placing History: How maps, spatial data, and GIS are changing historical scholarship. Redlands, California: ESRI Press. Konteh, Frederick H. (2009). Urban sanitation and health in the developing world: Reminiscing the nineteenth century industrial nations. Health and Place, 15(1), 69e78. Ord, J. K., & Getis, A. (1995). Local spatial autocorrelation statistics: Distributional issues and an application. Geographical Analysis, 27(4), 286e306. Preston, S. H., & Haines, M. R. (1991). Fatal Years: Child mortality in late nineteenthcentury America. Princeton, New Jersey: Princeton University Press. Rochester, A. (1923). Infant Mortality: Results of a field study in Baltimore, MD, based on births in one year. Washington, DC: Government Printing Office. Thornton, P. A., & Olson, S. (1991). Family contexts of fertility and infant survival in nineteenth-century Montreal. Journal of Family History, 16(4), 401e417. Thornton, P. A., & Olson, S. (1997). Infant vulnerability in three cultural settings in montreal, 1880. In A. Bideau, B. Desjadins, & H. Perez Brignoli (Eds.), Infant and child mortality in the past. Oxford, England: Clarendon Press. Thornton, P. A., & Olson, S. (2001). A deadly discrimination among Montreal infants, 1860-1900. Continuity and Change, 16(1), 95e135. Thornton, P. A., & Olson, S. (2011). Mortality in late nineteenth-century Montreal: Geographic pathways of contagion. Population Studies, 65(2), 157e181. Troesken, W. (2004). Water, race, and disease. Boston: MIT Press. Williams, N. (1992). Death in its season: Class environment and the mortality of infants in Nineteenth-century Sheffield. Social History of Medicine, 5, 71e94. Woods, R. I., Watterson, P. A., & Woodward, J. H. (1988). The causes of rapid infant mortality decline in England and Wales, 1861-1921. Part I. Population Studies, 42(3), 343e366.

Acknowledgements Part of this research was funded through cooperative agreements (Award Numbers 01-CA-11242343-042 and 01-CA11242343-085) with the United States Forest Service, Northeastern Research Station.