The proposed research will be twofold in its purpose. First, it will utilize data from the 2000 census to continue my past research, based on several decenniel censuses, tracking patterns of racial housing segregation in the St. Louis MO-IL metropolitan

ejs

Volume 4

Spatial Mismatch and Access to Physicians Among African Americans:

Initial Findings and Directions for Future Research*

John E. Farley

Abstract: Three approaches are used to assess the extent of mismatch between the geographic distribution of the African American population and of Primary Care Physicians in the two major urban counties in the Illinois portion of the St. Louis Metropolitan area. The measures used are the index of dissimilarity, the number of physicians per census tract and per capita in predominantly white and African American areas, and the mean distance to the nearest physician for whites and African Americans. The index of dissimilarity and physicians per tract show clear evidence of mismatches, while the distance and per-capita measures do not. However, the latter two measures may be poor indicators of access due to disproportionate lack of automobiles among the area’s African American population. The continuing presence of two hospitals in the main African American area may also help explain the lack of mismatch on two measures. In other areas such as nearby St. Louis, MO, hospitals have left such areas. In the latter situation, it is hypothesized that physician-population mismatches may be more severe.

Introduction

It has as been well-documented that shortages of physicians and medical facilities exist in many African American neighborhoods. The purpose of this study is to update the evidence on this issue using data from the 2000 census along with data especially collected for this project, and to expand upon the present state of knowledge on this topic by doing the following three things: First, the study uses the index of dissimilarity to examine the degree of unevenness between the geographic distribution of the study area’s African American population and that of primary care physicians. Second, it compares the numbers of primary care physicians per tract in predominantly African American census tracts in the study area with the number of primary care physicians in predominantly white census tracts. And third, it examines the geographic distance between the study area’s African American population and the locations of primary care physician offices, using several different approaches to estimating this distance. The study was conducted in Madison and St. Clair Counties, Illinois, the two major urban counties in the Illinois portion of the St. Louis metropolitan area. The findings of this research are discussed in the context of research on locational effects on access to health care and data on auto availability and modes of transportation among African Americans.

Problem Statement and Literature Review

It is well-established in medical sociology that shortages of medical facilities exist in inner city neighborhoods with large minority populations and concentrated poverty. In 1998, 46 million people lived in areas designated by the Department of Health and Human Services as Health Professional Shortage Areas. Over half lived in central cities of metropolitan areas, with disproportionately large African American and Hispanic populations (Brink, 1998; National Health Service Corps, 2000). According to the Department, a minimum of 5,300 more physicians are needed in these areas, and ideally, 12,000 more are needed. Locally, a study by Confluence St. Louis in the early 1980s found just one physician per 32,000 residents in one neighborhood in the city of St. Louis, and fewer than 1 per 10,000 in several other low-income city neighborhoods (Confluence St. Louis, 1985). In the St. Louis Metropolitan area, Health Professional Shortage Areas (HPSAs) identified by the Department of Health and Human Services are heavily concentrated in East St. Louis and in the city of St. Louis. In the Illinois part of the St. Louis area (commonly referred to as the Metro-East area) as of 1997, the East St. Louis Service Area consisting of 22 census tracts in St. Clair Co. and one in Madison Co. is designated as an HPSA for Primary Medical Care. The medically indigent population of the Alton-Wood River area also qualifies that area as a HPSA for Primary Medical Care. In the Missouri part of the metropolitan area, several areas are designated as HPSAs based on their poverty population - Grace Hill-Cochran (13 census tracts in St. Louis city), North St. Louis (12 census tracts in St. Louis city and 2 in St. Louis Co.), Southeast St. Louis, consisting of 20 tracts in St. Louis city, and West St. Louis, consisting of 6 tracts in St. Louis city and 3 in St. Louis Co. Hence, as is the case nationally, local areas designated as Health Professional Shortage Areas are heavily concentrated in inner city areas, in this case mainly in East St. Louis and St. Louis City, along with some immediately adjoining areas. However, one important difference between St. Louis and East St. Louis is that, while all hospitals that were located in the predominantly African American and largely poor northern part of St. Louis city have closed, two hospitals remain open or immediately adjacent to East St. Louis in areas of predominantly African American and largely poor populations.

While we know there are shortages of providers in areas with poor and/or minority populations, we do not know precisely what this means with respect to geographic distance between these populations and health care providers. This is an important issue, because many poor and minority households in the area lack automobiles and - unless they are lucky enough to live near the area’s one rail transit line - they must therefore negotiate a sometimes difficult and slow bus transportation network, greatly increasing the travel time and difficulty to get health care - especially if its location is distant from where they live. As of the 1990 Census, 37.0 percent of African American households in East St. Louis and 41.6 percent of African American households in the city of St. Louis had no motor vehicle (U.S. Census Bureau, 2000). By 2000, these numbers had not improved much. The 2000 Census showed that overall in East St. Louis, 31.2 percent of all households had no motor vehicle and 31.5 percent of African American households had no motor vehicle (U.S. Census Bureau, 2001a, 2002b). Among the 52 percent of St. Louis residents who are African American, 36.2 percent had no motor vehicle - even more than in East St. Louis (U.S. Census Bureau, 2002b).

As suggested above, distance to health care facilities may be an important reason for the underutilization of health care facilities among minorities and the poor, particularly in areas such as East St. Louis where large proportions of the population lack automobiles and must rely on public transportation to get to health care providers.

There has been some, though not a great deal, of research on the effects of distance from health care facilities on utilization in the United States and elsewhere. Most of this research does show distance effects on health status, but the research findings are based largely on Veterans Administration hospitals (Piette and Moos, 1996; Burgess and DeFiore, 1994; Zwanziger, 1994; Holloway, Medendorp, and Bormberg, 1990), community hospitals in areas with largely rural populations effects (Gregory et al., 2000; Grummow, Gregory, and MacNamara, 1990; Shannon, Brashur, and Lovett, 1986; White, 1986), or was conducted in countries other than the United States (Bailey and Phillipe, 1990; Beland, et al., 1990; Burgy and Hafner-Ranabauer, 1998; Rossler et al., 1991; Jones and Bentham, 1997). Hence, the findings must be viewed as suggestive but not conclusive for hospitals in urban areas with high poverty rates, where lack of automobiles presents travel challenges in some regards similar to those faced by more widely-dispersed populations (VA patients, rural patients) who do have automobiles but face larger travel distances. Studies in the United States do indicate that ability to travel greater distances for health care is correlated to income (Bronsetin and Morrisey, 1990).

Another way in which this study extends previous research is through incorporation of the index of dissimilarity - a measure of the unevenness of two distributions across space (Taeuber and Taeuber, 1965). This index can be used to determine - on a scale of 1 to 100 - how geographically uneven are the distribution of population groups (e.g. African Americans) and of providers (e.g. physician’s offices). Statistically, the index tells us what proportion of the providers would have to be relocated in order for the distribution of the providers to match that of the population group. Within urban areas, this can provide a useful measure of how far we are from a situation of having the distribution of providers match that of a population group, and it allows us to determine whether the distribution of providers is more or less uneven than that of population groups (e.g. the segregation of blacks and whites). Obviously both racial segregation in the population and uneven distribution of physicians between black and white areas, for example, will influence the index of dissimilarity between providers and the African American population. It has been well-documented, in the St. Louis area, that the black and white populations remain highly segregated from one another (Farley, forthcoming, 2002, 1995, 1993, 1991, 1989, 1987, 1986, 1984a, 1984b, 1983). Although St. Louis is one of the most segregated metropolitan areas in the United States, segregation is the norm among metropolitan areas throughout the United States, especially in the Northeast and Midwest (Massey and Denton, 1993; Logan, 2001). Given this uneven population distribution, there is considerable potential for geographic separation between health providers and African American citizens. The side of the distribution that has not been measured in the way done here is physicians: if their distribution is not only also uneven, but uneven in a way that is opposite that of the African American population, the potential for separation between the two becomes extreme given the high unevenness in the distribution of the African American population. This study uses the index of dissimilarity to assess the extent to which this is the case.

Obviously, there are many causes of the considerable health differentials between white and black Americans. These differentials are large, both nationally and in the St. Louis/Metro East area. Recent studies have shown that the death rates of African Americans from cancer, stroke, diabetes, and AIDS all greatly exceed those of whites in St. Louis. In fact, the death rate from all causes combined is 56 percent higher for African Americans in St. Louis than it is for whites (computed from Missouri Department of Health, 2000). Among the more evident causes of these differentials are lower incomes, more dangerous jobs, greater exposure to criminal victimization, and higher proportions without insurance among the African American population.

At the same time, however, we know that in addition to all these factors, African Americans have fewer health providers in their neighborhoods on average. I propose that they may also be farther from these providers - a hypothesis which is directly tested by this research. There may also be a substantial overall segregation or separation of the African American population from the locations of health care providers - another hypothesis directly tested by this research. Moreover, such segregation or separation may contribute to the poorer health and higher mortality among African Americans. Recent research by Waitzman and Smith (1998) has indicated that economic segregation is a factor in mortality. For example, Chicago’s higher level of economic segregation (as measured by the degree of concentration of poverty and of wealth) may elevate its mortality by a factor of as much as 33% relative to cities with lower levels of economic segregation such as Seattle (Waitzman and Smith, 1998). Moreover, Massey and Denton (1993) have presented a strong case that racial housing segregation is a major reason for the type of concentrated poverty that Waitzman and Smith found in Chicago and elsewhere. Waitzman and Smith suggest that racial segregation may underlie their findings, and note that race was one of the few control variables that significantly mitigated the effect of spatial poverty concentration.

For all of the reasons outlined here and above, it is plausible that spatial segregation may be a contributing factor to the high levels of mortality among African Americans. It is, likewise, plausible that distance from providers may be an additional factor in the higher mortality of African Americans. However, in order to assess whether or not this is the case, it is necessary to first 1) document the degree of segregation between African Americans and health care providers, 2) determine the number of providers typically present in predominantly African American neighborhoods as compared to predominantly white neighborhoods, and 3) determine average distances from physicians for the African American and white populations.

Methods

Both the distance and separation/unevenness parts of this study utilize tract-level data on race from the 2000 Census of Population. These data were obtained from the Public Law 94-171 Redistricting Data, released over the Internet through the Census Bureau’s American Fact Finder Web site (U.S. Census Bureau, 2001a). The data were downloaded for all census tracts in Madison and St. Clair Counties, Illinois. For all measures, I used the following indicator of the number of African Americans in the census tract: the number of people in the geographic area who indicate black or African American only as their race.¹

In addition, I obtained lists of physicians by specialty from the medical staffs of every hospital located in Madison or St. Clair County. The list was cleaned to eliminate multiple listings of the same physician (because of physicians being on the staff of more than one hospital). For purposes of the present paper, the analysis was limited to primary care physicians, defined as including general or family practice, internal medicine, pediatrics, and obstetrics/gynecology (Ob/Gyn).

Out of the 349 primary care physicians with offices in Madison or St. Clair County, there were 21 physicians who had more than one office. When this occurred, both offices were included in the database, because proximity to either office should make it possible for a patient to have access to that physician. This yielded a total of 370 primary care physician offices in the database. In other words, a physician office is defined as a location at which one primary care physician provides services.

It was also very common for more than one physician to have an office at the same address. In such cases, the number of physician offices at this location was defined as the number of primary care physicians providing services at this address. This was done to reflect the reality that if there were, for example, four physicians at an address, this provided access to four physicians at this address. Using the database of addresses, geographic information system (GIS) technology were used to map the locations of the primary health care providers, and to identify the census tract of each provider's location.

For the first part of the study, I calculated the index of dissimilarity (Taeuber and Taeuber, 1965) in the distribution of African American population and of primary care physicians, using census tracts as the geographical unit to calculate the index. To contrast with this statistic, the index of dissimilarity for providers and the white population is also calculated (in this case, using only those who indicate white only as their racial identification). These statistics show the degree to which physicians are geographically distributed in a different manner from the geographic distribution of both the African American and white populations.

For the second part of the study, a comparison was made of the average number of primary care physicians located in predominantly white tracts (defined as less than 20 percent African American) and predominantly black tracts (defined as more than 80 percent African American.) This analysis was conducted for Madison County, St. Clair County, and the two counties combined. In addition, the number of physicians per thousand residents was also computed in each tract and averaged across tracts for predominantly white and predominantly African American census tracts.

For the third part of the study, mapping and geographic Information System (GIS) technology was used to overlay census tract racial data on computer maps showing the distribution of health care providers. From this overlay, it was possible to estimate and compare the mean distance to the nearest primary care physician for the area’s black and white populations. This statistic was computed in each census tract. For tracts with no physician, the distance was presumed to be the distance from the center of the tract to the location of the nearest provider in another tract. While this does not equal the distance for each individual, it is a good indicator of the average distance to a provider for people living in that tract. For instances in which one or more physicians were present in a tract, two measures were used: the distance from the nearest physician within the tract to the center of the tract, and the average of this distance and the distance from the physician's office to the farthest point in the tract. Neither is a perfect measure of the average distance to a physician within a tract, but they contain opposite biases, so taken together they provide a reasonable estimate of the distance. Since most people do not live at the center of a tract, the former measure understates the average distance to a physician, though it is probably more accurate as an average distance the farther the office is from the center. In contrast, the latter measure likely overstates the actual average distance, because it averages the distance to the center with the distance to the farthest point. For offices near the center, this measure is accurate, but the farther the office is from the center, the more it overstates the distance. In general, larger tracts will have greater average distances to physicians within the tract.

Finally, one other potential problem is generated by the use of the office closest to the tract center to compute the average distances within the tract - if there are multiple offices in different parts of the tract, this will not be a valid measure for most people in the tract. However, as a practical matter the impact of this may be small - the offices are highly clustered, so in most instances of multiple offices in the same tract, the offices are very close to one another - usually nearby on the same street and often in the same building. Thus, while the distance measure has its limitations, particularly within the same tract, it provides an approximation of how far people are from primary care physicians.

The averages were computed by race. To compute the overall distance for African Americans, the average distance to a physician for each tract was multiplied by the number of African Americans in the tract. These figures for each tract were summed and divided by the African American population of the county. A comparable procedure was used to compute average distances for whites. The result is an estimate of the average distance from a primary care physician for African Americans in each county, and for whites in each county.

Findings

Segregation/unevenness Table 1 presents indices of dissimilarity showing segregation or geographic unevenness between 1) white population and African American population, 2) white population and primary care physicians, and 3) African American population and primary care physicians for Madison and St. Clair Counties. Table 1 also shows the racial composition of each county as well as two measures of African American residential isloation/exposure, described below. In both counties, there is a significant geographic separation of physicians from both the white population and the African American population. For both whites and African Americans, the index of dissimilarity between the geographic distribution of the racial group's population and that of physicians is 58 or higher in all instances. This places it in the upper half of the possible range of the index of dissimilarity, 0 to 100. This is the case because, as shown in Figures 1, 2, and 3, physicians tend to cluster in a limited number of census tracts. In fact, in St. Clair County, exactly half the census tracts, 23 out of 56, have no physicians at all. On the other hand, 66 physicians - 31 percent of the county's primary care doctors - are in just two census tracts. Similarly in Madison County, the majority of census tracts - 36 out of 60 - have no physicians at all, while 70 primary care physicians - over 40 percent of the county's total - are located in just three census tracts.

This clustering makes it inevitable that the geographic distribution of physicians will be different from that of the population, and it is. The key question, however, is whether there is a significant racial difference in the extent to which this is the case. In other words, do doctors tend to cluster disproportionately in areas that are away from the African American population? The answer to this question appears to be "yes" in St. Clair County but "no" in Madison County. In St. Clair County, the index of dissimilarity between the white population distribution and the distribution of physicians is 58.7, while the index between the African American population distribution and the distribution of physicians is 70 - a difference of 11.3 points. This may not seem large, but in terms of variations in indices of dissimilarity in population distributions, it is quite substantial. For example, among all 331 metropolitan areas in 2000, an area with a black-white index of dissimilarity of 70 would have been the 30^th most segregated of all the areas, while an area with an index of 58.7 would have ranked 106^th. Put differently, an index of 70 would place a metropolitan area at the 91^st percentile of racial segregation, whereas an index of 58.7 would place it at the 68^th percentile - a substantial difference. Viewed in this context of an actual distribution of indices of dissimilarity, a difference of 11.3 points is substantial.

In Madison County, in contrast, the difference is much smaller. There, the index of dissimilarity between the distribution of physicians and that of the white population is 61.1, compared to an index of 62.9 between the distribution of physicians and that of the African American population. In this case, the difference is only 1.8 points, much less than is the case in St. Clair County. Thus, we conclude that in St. Clair County, but not in Madison County, there is a greater unevenness between the African American population distribution and the distribution of physicians than is the case for the distribution of the white population.

Why are the findings different for Madison and St. Clair Counties? The difference appears to be related to differences in the racial compositions and racial housing segregation patterns of the two counties. As shown in Table 1, Madison County has a much smaller African American population than St. Clair County, and that smaller population is also somewhat less segregated from the white population. In Madison County, just 7.3 percent of the population is African American - a total of 18,935 people. Of these, only 2,365 live in a census tract where the majority of the population is African American - in fact, there is just one tract in which more than half the population is African American. In addition, as shown by the median exposure index, half of all African Americans in Madison County live in census tracts that are at least 74 percent white. This reflects both the relatively small African American population of the county (as compared to St. Clair County) and its residential segregation index of 58.1, which is the lowest of any racially-diverse county in the St. Louis metropolitan area.

In St. Clair County, the potential for African Americans to be geographically separated from physicians is much greater, because far more of them live in overwhelmingly African American neighborhoods. In fact, as shown by the median exposure index for St. Clair County in Table 1, half of all the African Americans in St. Clair County live in census tracts where less than 2.5 percent of the population is white. This difference from Madison County reflects the much larger African American population of St. Clair County, 73,666 - nearly four times that of Madison County, and 28.8 percent of the total population of St. Clair County. It also reflects the fact that St. Clair County is more racially segregated than Madison County, with a black-white housing segregation index of 65.6 - 7.5 points higher than Madison County. With a larger African American population and a somewhat higher level of segregation, the potential for African Americans to be racially isolated is much higher. And in turn, this large, relatively more isolated population is more subject to avoidance in physician office-location decisions. This likely explains why in St. Clair County, but not in Madison County, the African American population is significantly more separated from physician office locations than is the white population.

__________________________________________________

Table 1. Separation Between Population and Physicians by Race, With Racial Composition and Residential Segregation Indicators, Madison and St. Clair Counties, 2000

St. Clair County Madison County

Segregation/Separation between:

Whites and African Americans 65.6 58.1

Whites and Primary Care Physicians 58.7 61.1

African Americans and Primary Care Physicians 70.0 62.9

Percent African American 28.8 7.3

Mean Exposure African Americans to Whites .308 .671

Median Exposure African Americans to Whites .025 .744

______________________________________________________

Physicians per Census Tract A similar pattern can also be seen in the analysis conducted in the second part of this study. In this analysis, I compare the average number of primary care physicians located in predominantly African American tracts with the average number located in predominantly white tracts. This analysis is shown in Table 2. It reveals that in both counties, the number of primary care physicians is greater in predominantly white tracts than it is in predominantly African American tracts. For purposes of this analysis, predominantly African American means 80 percent or more African American, and predominantly white means 80 percent or more white. In Madison County, there is just one tract that is at least 80 percent African American, and it has no primary care physicians. In contrast, 51 of the county's 60 census tracts are at least 80 percent white, and these tracts have an average of 1.90 primary care physicians per tract. In St. Clair County, there are 16 census tracts that are at least 80 percent African American, and these tracts have an average of 1.94 primary care physicians per tract. In contrast, there are 33 census tracts that are more than 80 percent white, and these census tracts have an average of 3.52 primary care physicians per tract - more than 1.5 more than in predominantly African American Census tracts. In both counties, then, the average number of physicians per census tract is at least 1.5 physicians greater in predominantly white tracts than it is in predominantly African American census tracts. This suggests that, for the many African Americans in St. Clair County who live in predominantly African American census tracts - and for the relatively few in Madison County who live in such a tract - the average number of primary physicians who have offices in the neighborhood is significantly less than is the case in predominantly white census tracts.

Table 2 also shows the number of primary care physicians per thousand population in predominantly white and predominantly African American census tracts. In St. Clair County, unlike the physicians-per-tract measure, this measure shows only a very small difference between predominantly African American and predominantly white tracts, and the difference actually slightly favors predominantly African American tracts. There are .673 primary care physicians per thousand population in predominantly African American tracts, compared to .638 per thousand in predominantly white tracts. This difference reflects the fact that predominantly African American census tracts in St. Clair County have smaller populations on the average than predominantly white tracts, mainly because the former have experienced substantial loss of population over recent decades. Over the past five decades, for example, the population of East St. Louis fell from about 85,000 in 1950 to around 31,000 in 2000. While this finding suggests that these tracts are not under-served relative to their populations, the fact remains that the relative lack of doctors in the geographic neighborhood may impede access to primary health care, especially given that many households in these census tracts lack a motor vehicle. In all four of the communities that have populations that are more than 90 percent African American, between about a quarter and a third of all households have no automobile available. In contrast, in heavily white communities, only between one in ten and one in twenty households typically lack an automobile.

_________________________________________________

Table 2.

Tracts with Less than 20 Percent African American Population

Physicians Tracts Physicians per Tract Phys. per 1000

Madison County 97 51 1.90 0.416

St. Clair County 116 33 3.52 0.636

Both Counties Combined 213 84 2.54 0.513

Tracts with More than 80 Percent African American Population

Physicians Tracts Physicians per Tract Phys. per 1000

Madison County 0 1 0.00 0.0

St. Clair County 31 16 1.94 0.673

Both Counties Combined 31 17 1.82 0.638

___________________________________________________

In Madison County, the relatively few African Americans who live in a predominantly African American area do appear to be under-served with respect to physician availability, since there are no primary care physicians in the one census tract that is more than 80 percent African American.

Distance to Physicians The findings concerning distance to physicians present a somewhat different picture from most of the analyses reported above. Among tracts that lack any physician, the distance from the center of the tract to the nearest primary care physician office was computed. This approximates the average distance of people in the tract from the nearest physician. The white and African American populations of these tracts were then used to compute the average distance to the nearest physician for whites and African Americans living in tracts with no physician. Contrary to expectations, in both counties, this distance proved to be slightly greater for whites than for African Americans. In Madison County, the average African American living in a tract with no primary care physician offices lived one mile from the nearest primary care physician office, while for whites, the average distance was 1.56 miles. In St. Clair County, the average African American living in a tract with no primary care physician offices lived .56 miles from the nearest primary care physician office, while for whites, the average distance was 1.11 miles. Hence, in both counties among people who lack a physician in their own census tract, African Americans were closer to the nearest physician than were whites.

_____________________________________

Table 3. Percent of Households with No Motor Vehicle and Percentage of Population that is African American, Selected Areas, 2000

Place Percent of Households with No Motor Vehicle Percent African American

Madison County 7.0 7.7

Alton 12.1 24.7

Collinsville 7.2 6.3

East Alton 11.5 0.9

Edwardsville 4.1 8.7

Glen Carbon 4.8 7.0

Godfrey 3.8 4.0

Granite City 10.6 2.0

Highland 6.5 0.1

Madison 17.4 42.1

Venice 29.4 93.6

Wood River 9.5 0.6

St. Clair County 10.4 28.8

Alorton 34.2 97.1

Belleville 11.1 15.5

Cahokia 9.9 38.7

Centreville 23.1 95.5

East St. Louis 31.2 97.7

Fairview Heights 5.0 17.1

O'Fallon 4.6 12.0

Swansea 4.6 8.6

Source: U.S. Census Bureau, 2002, Demographic Profiles, 100 Percent and Sample Data, Demographic Profile Data Search, World Wide Web, http://censtats.census.gov/pub/Profiles.shtml

__________________________________________________

Among tracts that have one or more physicians, we computed two estimates of the average distance of the population from a physician's office for whites and African Americans, as described above in the methods section. Using an average based on the distance from the physician's office to the center of the tract, we found that in Madison County, the distance to the nearest physician averaged .95 mile for whites and .77 mile for African Americans. In St. Clair County, this distance averaged 1.12 miles for whites and .59 mile for African Americans. As an alternative measure, using the average of the distance of the physician's office from the center of the tract and the distance between the physician's office and the farthest part of the tract, we found an average distance to the nearest physician in Madison County of 1.55 miles for whites, while for African Americans it was 1.31 miles. For whites in St. Clair County, the average distance according to this measure was 2.32 miles, while for African Americans it was 1.25. Despite their differences, both of these measures suggest that the average African American lives closer to a doctor than the average white resident. In Madison County, in tracts with physicians, the average black person lives about 0.2 miles closer to a physician than the average white person by either measure. In St. Clair County, where many whites live in larger tracts toward the fringes of the county, the difference is greater; by the first measure it is about .6 mile and by the second measure, just over a mile.

These findings paint a different picture than the mismatch indicators examined in the earlier part of the paper, particularly in the case of St. Clair County. In St. Clair County, we saw earlier that 1) there is a greater dissimilarity between the spatial distributions African Americans and physicians than between the distributions of whites and physicians, and 2) African Americans are less likely to live in census tracts that have physician offices. From this viewpoint, there is indeed a mismatch between where physicians are located and where African Americans live in St. Clair County.² However, this mismatch does not translate into a greater physical distance between where African Americans live and where physicians are located. In fact, in both counties, but more so in St. Clair County, the average African American is closer to a primary care doctor's office than the average white. This is because, on the whole, whites live in newer areas with lower population densities. Even though predominantly white census tracts tend to have more doctors' offices, these tracts are physically larger, so that the distances involved are greater. In terms of average distance to doctors, these larger tract sizes more than offset any mismatch between the distribution of physician office locations and the distribution of the African American population.

Nonetheless, caution must be exercised in drawing conclusions about access from this finding. Since they are more likely to lack automobiles, African Americans are more likely than whites to encounter transportation difficulties, and hence may have greater difficulty accessing physician offices even when the average distances to those offices are the same as or less than for whites. Since, as noted above, 2000 Census data indicate that between a quarter and a third of African American households in the study area lack automobiles, compared to only a tenth to a twentieth of white households, access may be severely impeded due to lack of an automobile. In terms of access, a physician a mile away may be "closer" to a sick person with an automobile than a physician .6 of a mile away is to one without an automobile. Because of such transportation limitations, the presence or absence of and number of physicians within a resident's immediate neighborhood may be as good an indicator, for those without motor transportation, as the distance to the nearest physician. And for this indicator, our data do suggest limitations in access for African Americans in both Madison and St. Clair Counties.

Discussion

These findings paint a rather complicated and mixed picture, but taken together they both suggest access problems to African Americans and give hints for future research as to where such access problems are likely to be maximized. First, there are some important differences between Madison and St. Clair County, and these differences reflect the different racial compositions and segregation patterns of the two counties. In St. Clair County, the index of dissimilarity between African American residential locations and primary care physician locations is quite high (70.0) and substantially exceeds the index for white residential locations and physician locations (58.7). In both counties, the number of physicians per census tract is significantly lower in predominantly African American tracts than in predominantly white tracts. However, Madison County has only one tract that can be described as predominantly African American, whereas St. Clair County has 16 such tracts, and a substantial proportion of the county's African American population lives in these tracts

Moreover, as can be seen in Figures 2 and 3, virtually all of the overwhelmingly African American tracts in St. Clair County are contiguous to one another and are located in and around East St. Louis. This creates a much larger area that can be identified as predominantly black than can be found in Madison County, with its smaller black population and its lower degree of racial housing segregation (Table 1). This relatively large predominantly black area has to some extent been avoided by primary care physicians, and this creates the mismatch shown in the dissimilarity data.

The extent to which this mismatch creates a problem of limited access to physicians among African Americans is unclear. On the one hand, in part because the tracts are smaller and the population density somewhat higher here than in predominantly white suburban areas, the black population of these areas is not farther from a primary care physician on the average than the white population in the suburban areas. But because 25 to 35 percent of the households in these areas lack automobiles, it is possible that, even though the mismatch in St. Clair County does not result in greater average distances from physicians for African Americans as opposed to whites, it may still result in a lack of access to physicians among African Americans. Those who lack automobiles must rely on limited public transportation to get to the doctor.

Are there situations in which a mismatch like that found in St. Clair County could also translate into greater average distances from physicians for African Americans as opposed to whites? The likely answer is "yes," because of two ways in which St. Clair County is unlike other areas with identifiable clusters of predominantly black census tracts. First, its population size and the physical size of its densely built-up area is smaller than other such areas, such as St. Louis City and County, Chicago, Cleveland, or Detroit. More important, unlike many of these areas, it has been able to keep hospitals open in predominantly African American areas and, by keeping these hospitals, it has also kept physicians who otherwise may have left. In St. Clair County, the two hospitals located in predominantly black areas have remained open while their counterparts in many other areas have closed. The hospitals are St. Mary's Hospital in East St. Louis and Touchette Hospital in Centreville. Both hospitals have physician office buildings adjacent, and the availability of these hospitals has undoubtedly been a factor in keeping physicians in these communities.

St. Clair County may in fact offer better access to physician services for its African American population than other locations with large African American populations that cover bigger areas and/or areas with larger populations, and that have had greater losses of hospitals. For example, in the northern half of St. Louis City, where most of the city's African American population of 178,266 lives, the number of hospitals is very small and the presence of community hospitals is nil. The only such hospitals there have closed: Homer G. Phillips Hospital in 1979 and the St. Louis Regional Medical Center (formerly St. Luke's Hospital) in 1997. When Regional was closed, it was replaced by ConnectCare which, according to its 2001 annual report, provided just 12 staffed inpatient beds out of a total of 24 Medicare/Medicaid beds (St. Louis ConnectCare, 2001). However, in November, 2002, it was announced by ConnectCare that these remaining beds would be closed as of December 15, 2002 (Vandewater, 2002). This leaves no general community hospital in the north half of the city of St. Louis, where most of the city's 178,266 African American residents live. There are two other hospitals – a VA medical center open only to veterans and a psychiatric facility - in the north half of St. Louis city, but neither provides general community hospital services comparable to those available at St. Mary's in East St. Louis or Touchette in Centreville (St. Louis City Department of Health, 2000). As community hospitals depart, so do many physicians, and the geographic area and the number of tracts in the resultant under-served, predominantly black area are much larger in places like St. Louis city than is the case in St. Clair County.

This can be seen in Figure 4, derived from the U.S. Census Bureau's American Fact Finder Thematic Mapper (U.S. Census Bureau, 2001b). This figure shows that the predominantly African American area of St. Louis city consists of 51 census tracts with African American populations of at least 66.4 percent. Indeed, 44 of these 51 census tracts have African American populations of at least 89 percent, and among all of these 44 tracts, the last 12 staffed hospital beds were eliminated when ConnectCare closed its inpatient beds in 2002. Together, the 51 tracts whose populations are at least two-thirds African American account for a large majority of the city's 178,266 African Americans, and they add up to about half of the city's 61 square miles. As a consequence, it is reasonable to hypothesize that in areas such as St. Louis, the impact of mismatches on the distance of African Americans from primary care physician office locations may be significantly greater than in St. Clair County, where the area of concentrated African American population is more limited (16 census tracts as compared to 51) and is served by two community inpatient hospitals.

For all these reasons, the logical next step in research on spatial factors in access to physicians among African Americans seems quite clear. The next step is to compare areas like St. Clair County (which have substantial concentrations of minority populations, but whose overall population and densely-populated area is limited) to other racially-segregated areas with larger populations like St. Louis City and County, Detroit, or Chicago - areas which in many cases have also lost doctors and hospitals to a greater extent than has taken place in East St. Louis and nearby communities in St. Clair County.

Figure 1. Madison County Primary Care Physician Locations and Tract Racial Composition

Figure 2. St. Clair County Primary Care Physician Locations and Tract Racial Composition

Figure 3. Madison and St. Clair County Primary Care Physician Locations and Tract Racial Distribution

Figure 4. Racial Composition of tracts in northern St. Louis City, 2000.

Notes

¹This indicator is necessitated by the new practice in the 2000 census of allowing respondents to give more than one racial identity. In the St. Louis metropolitan area, just 1.2 percent of the population indicated more than one race, so this measure includes the overwhelming majority of the area's African American population.

²However, there is less consistent evidence of such mismatches in Madison County, where the African American population is smaller and less spatially segregated.

References

Bailey, W., and D. R. Phillips. 1990. “Spatial Patterns of Use of Health Services in the Kingston Metropolitan Area, Jamaica.” Social Science and Medicine 30 (1): 1-12.

Beland, F., L.. Philibert, J. P. Thouez, and B. Maheaux. 1990. “Socio-spatial Perspectives on the Utilization of Emergency Hospital Services in Two Urban Territories in Quebec.” Social Science and Medicine 30 (1): 53-66.

Brink, Susan. 1998. “Caring for the Neglected Areas Untouched by a Physician Surplus.” In America's Best Graduate Schools 1998. ______: U.S. News and World Report.

Bronsetin, J. M., and M. A. Morrisey. 1990. “Determinants of Rural Travel Distance for Obstetrics Care.” Medical Care 28 (9): 853-65.

Burgess, J. F., and D. A. DeFiore. 1994. “The Effect of Distance to VA Facilities on the Choice and Utilization of VA Outpatient Services.” Social Science and Medicine 39 (1): 95-104.

Burgy, R., and W. Hafner-Ranabauer. 1998. “Utilization of the Psychiatric Emergency Service in Mannheim: Ecological and Distance-Related Aspects.” Social Psychiatry and Psychiatric Epidemiology 33 (11): 558-67.

Confluence St. Louis. 1985. Health Care for the Indigent Population in the St. Louis Region: A Report of the Health Care Task Force. St. Louis: Confluence St. Louis.

Farley, John E. (forthcoming). "Interracial Exposure Indices: When 'Average' is Not 'Typical.'" Forthcoming in Critical Demography 1 (2004).

Farley, John E. 2002. "Racial Segregation in the St. Louis Metropolitan Area, 2000." Edwardsville Journal of Sociology 2 (April). Available online at http://www.siue.edu/SOCIOLOGY/journal/FARLEYV2.HTM

Farley, John E. 1995. "Race Still Matters: The Minimal Role of Income and Housing Cost as Causes of Housing Segregation in St. Louis, 1990." Urban Affairs Review (formerly Urban Affairs Quarterly) 31 (November): 244-254.

Farley, John E. 1993. "Housing Segregation in the St. Louis Metropolitan Area, 1980-1990: Comparing Trends at the Block and Census Tract Level." Journal of Urban Affairs 15, 6 (November): 515-527.

Farley, John E. 1991. "Housing Segregation in the City of St. Louis: A 1988 Update." Urban Affairs Quarterly 26 (March): 442-450.

Farley, John E. 1989. Black-White Housing Segregation in the City of St. Louis: a 1988 Update. Regional Research and Development Services Report. 35 pp. Content of this report was also presented at a research colloquium jointly sponsored by Regional Research and Development Services and Department of Sociology and Social Work at SIUE, November 29.

Farley, John E. 1987. "Segregation in 1980: How Segregated are American Metropolitan Areas?" Pp. 95-114 in Gary Tobin (ed.), Divided Neighborhoods. Sage Urban Affairs Annual Reviews, Vol. 32. Newbury Park, CA: Sage Publications.

Farley, John E. 1986. "Segregated City, Segregated Suburbs: To What Extent Are They Products of Black-White Socioeconomic Differentials?" Urban Geography, 7,2 (March): pp. 180-87.

Farley, John E. 1984a. "P* Segregation Indices: What Can They Tell Us About Housing Segregation in 1980?" Urban Studies 21 (August), pp. 331-336.

Farley, John E. 1984b. "Housing Segregation in the School-Age Population and the Link Between Housing and School Segregation: A St. Louis Case Study," Journal of Urban Affairs 6, 4 (Fall), pp. 65-80.

Farley, John E. 1983. "Housing Segregation in U.S. Metropolitan Areas, 1980: The St.Louis Case," Urban Affairs Quarterly 18 (March), pp. 347-359.

Grummow, L. J., V. R. Gregory, and S. E. MacNamara. 1990. “Factors Influencing Utilization of Postdischarge Cognitive Rehabilitation Programs.” Health Services Research 25 (1, pt. 2): 197-211.

Gregory, P. M., E. S. Malka, J. B. Kostis, A. C. Wilson, J. K. Arora, and G.G. Rhoads. 2000. “Impact of Geographic Proximity to Cardiac Revascularization Services on Service Utilization.” Medical Care 38: 45-57.

Holloway, J. J., S. V. Medendorp, and J. Bromberg. 1990. “Risk Factors for Early Readmission Among Veterans.” Health Services Research 25 (1, pt. 2): 213-17.

Jones, A. P., and G. Bentham. 1997. “Health Service Accessibility and Deaths from Asthma in 401 Local Authority Districts in England and Wales, 1988-92.” Thorax 52 (3): 218-22.

Logan, John. 2001. "Ethnic Diversity Grows; Neighborhood Integration is at a Standstill." Report by the Lewis Mumford Center, State University of New York at Albany, April 3. Available on World Wide Web, http://mumford1.dyndns.org/cen2000/WholePop/WPreport/page1.html

Massey, Douglas S., and Nancy Denton. 1993. American Apartheid: Segregation

Luft, H. S., D. W. Garnick, D. H. Mark, D. M. Peltzman, C. S. Phibbs, E. Lichtenberg, and S. J. McPhee. 1990. “Does Quality Influence Choice of Hospital?” Journal of the American Medical Association 263 (21): 2899-906.

Massey, Douglas, and Nancy A. Denton. 1993. American Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.

McGuirk, M., and F. W. Porell. 1984. “Spatial Patterns of Hospital Utilization: The Impact of Distance and Time.” Inquiry 21 (1): 84-95.

Missouri Department of Health, Center for Health Information Management and Epidemiology. 2000. “Leading Causes of Death Profile by Race for St. Louis City Residents.” World Wide Web, http://www.health.state.mo.us/profiles/CauseOfDeath/CauseOfDeathRaceStLouisCity.html

National Health Service Corps. 2000. Who We Serve. World Wide Web, http://bphc.hrsa.gov/nhsc/Pages/about_nhsc/3D_who.htm

Piette, John D., and Rudolf H. Moos. 1994. “The Influence of Distance on Ambulatory Care Use, Death, and Readmission Following a Myocardial Infarction.” Health Services Research 31 (5): 573-91.

Rossler, W., A. Riecher, W. Loffler, and B. Fatkenheuer. 1991. “Community Care in Child Psychiatry: An Empirical Approach Using the Concept of Travel Time.” Social Psychiatry and Psychiatric Epidemiology 26 (1): 28-33.

St. Louis ConnectCare. 2001. 2001 Report to the St. Louis Community. World Wide Web, http://www.saintlouisconnectcare.com/about11.html

St. Louis City Department of Health. 2000. Public Health: Understanding Our Needs, Volume 1. St. Louis: St. Louis City Department of Health. Available online at http://stlouis.missouri.org/citygov/health/FinalDraft1.pdf

Vandewater, Judith. 2002. “ConnectCare Will Close Hospital on Delmar Boulevard Next Month.” St. Louis Post-Dispatch, November 14, p. c4.

Shannon, G. W., R. L. Brashur, and J. E. Lovett. 1986. “Distance and the Use of Mental Health Services.” Milbank Quarterly 64 (2): 302-30.

Taeuber, Karl E., and Alma F. Taeuber. 1965. Negroes in Cities. Chicago: Aldine.

U.S. Census Bureau. 2002a. Demographic Profiles, 100 Percent and Sample Data, Demographic Profile Data Search, World Wide Web, http://censtats.census.gov/pub/Profiles.shtml

U.S. Census Bureau, 2002b. American Fact Finder, 2000 Census of Population and Housing - Summary File 3. Detailed Tables. World Wide Web, http://factfinder.census.gov/servlet/DTGeoSearchByListServlet?ds_name=DEC_2000_SF3_U&_lang=en&_ts=54151256260

U.S. Census Bureau. 2001a. American Fact Finder, 2000 Census of Population and Housing - Redistricting Data (Public Law 94-171) Summary File. Detailed Tables. World Wide Web, http://factfinder.census.gov/servlet/DTGeoSearchByListServlet?ds_name=DEC_2000_PL_U&state=dt&_lang=en&_ts=25115666180

U.S. Census Bureau. 2001b. American Fact Finder, Thematic Map Server. World Wide Web http://factfinder.census.gov/servlet/ThematicMapFramesetServlet

U.S. Census Bureau. 2000. 1990 Census Lookup. World Wide Web, http://venus.census.gov/cdrom/lookup

Waitzman, Norman J., and Ken R. Smith. 1998. “Separate but Lethal: The Effects of Economic Segregation on Mortality in Metropolitan America.” Milbank Quarterly 76 (3): 341-73.

White, S. L. 1986. “Travel Time as Time Price and the Demand for Mental Health Services.” Community Mental Health Journal 22 (4): 303-13.

Zwanziger, J. 1994. “The Effect of Facility Characteristics, Eligibility Status, Age, and Travel Distance on the Demand for VA Inpatient Services.” Paper presented at the 12th Annual VA HSR&D Service Meeting, Washington DC, April 26 and 27.

*The author wishes to acknowledge support for this project from the SIUE Institute for Urban Research, and assistance with Geographic Information System (GIS) applications from Steve Galinski.