Neighborhoods and Health: Building Evidence for Local Policy. Annex B - Disparity Indices

05/01/2003

Health inequalities are of growing concern because they suggest that the advantages of good health are not equally available to everyone. Despite policies that promote public health and access to health care, health status measures show profound differences across groups. Health disparities have been clearly evident when members of racial and ethnic minorities are compared with the white population (Keppel, Pearcy, and Wagener 2002). The reduction in health disparities was one of the overarching goals of Healthy People 2000.

Another manifestation of health disparities that has not received as much public attention is place-based inequality. Studies have shown that the health of residents of certain disadvantaged neighborhoods is generally worse than the population as a whole (Geronimus 1996; Robert 1998; Roberts 1997). There are many possible explanations for these geographic patterns. Many cities are racially segregated, and segregation has been linked to poor health (Jackson et al. 2000). Moreover, poor housing, unsafe streets, and environmental contaminants might lower health status in some neighborhoods. The socioeconomic status of one's neighbors also has been suggested to have an indirect effect on health through various social influence processes (Robert 1998). Another potential factor is that some neighborhoods may be geographically situated so as make access to health care difficult. Although the causes of neighborhood health disparities are complex, there is growing concern about spatial inequalities and interest in explicitly addressing them.

If communities are to strive to reduce place-based disparities, they will need methods to determine where and to what degree such neighborhood health inequalities exist. At this point, no agreed-upon definitions or techniques are in place. Ideal methods would allow metropolitan areas to be compared on the degree to which they have neighborhood disparities. They would also allow those neighborhoods with extremely poor health to be identified. In this section, we first describe three approaches to quantifying neighborhood disparities on selected health indicators and test them using neighborhood indicators data from the five cities. After the technical description of the measures, we review the strengths and limitations of each method in revealing place-based inequality.

Methodology

The data for this analysis are four health indicators measured at the census tract level in each city. The health indicators chosen were age-adjusted death rates, teen birth rates, percentage of newborns whose mothers received prenatal care in the first trimester of pregnancy, and percentage of low-birth weight births. For each census tract, four years of data are used to achieve greater reliability. The data sources and methods used to calculate each of these indicators are described in section 2. These indicator data were used to calculate three measures of disparities among census tracts. The first technique described, the composite disparity index, produces a single score for a city on a particular health indicator. The next two types of indices identify the tracts that exceed a defined threshold of the health indicator based on the overall city rate.

Composite Disparity Index

The composite disparity index is a summary measure of the differences among the census tracts on their rates on a health indicator. It is an index of the amount of inequality among census tracts in the city. The index will be high when the indicator rates for tracts vary greatly from the overall rate of the city. This index captures extremes of both good health and poor health. The numerator of the index is the mean deviation.

The method of calculating the composite index draws upon the index of disparity for race described by Keppel and others (2002). The mean deviation was calculated by getting the absolute value for the difference in the rate of each tract and the overall rate for the city. Then all the differences were summed and divided by the number of tracts. The mean deviation was then divided by the city rate to create the index. Specifically,

figb1.1.gif where Xij is the rate on an indicator for tract i in city j, X.j is the rate on an indicator for city j, Nij is the number of census tracts in the city.

Extreme Distribution Count

Extreme distribution count is a method that counts the number of census tracts at the upper end of the distribution on a selected indicator. The percentage of a city's tracts that are in the high end of the distribution will be high if the distribution is skewed at the end of the scale (or at the low end if the indicator is in the other direction). The mean of the tracts plus 1 (or 2) standard deviation is the definition of high in this case. The notion of using the mean and standard deviation to identify neighborhoods of concern is consistent with the approach taken by Kasarda (1993) to identify disadvantaged neighborhoods.

The calculation simply compares the rate on the indicator of each tract with the mean plus one standard deviation for the whole city. The mean in this instance can be calculated either as the weighted mean of all of the tracts or as the rate for the whole city. The number of tracts that exceed this threshold is counted. The percentage of tracts that have extremely poor health is calculated by dividing the count by the total number of tracts in the city. This exercise is repeated using a two standard deviation threshold as well. In the case of percentage of births with first trimester care, we subtracted the standard deviation in order to examine the low end of the distribution.

Relative Threshold Count

Relative threshold count counts the number of census tracts that exceed twice the median for tracts in the city. The percentage of a city's tracts that are above this threshold is calculated to standardize for city size. An argument can be made for using the median rather than the mean plus standard deviation as a threshold (Hughes 1989). By definition, if the indicator is normally distributed, some tracts will exceed the mean plus one or even two standard deviations. This is due to the fact that as the variation in an indicator decreases, the standard deviation decreases. However, if an indicator is normally distributed, about two-thirds of the tracts should fall between plus or minus one standard deviation and about 95 percent should fall between plus or minus two standard deviations. Thus, some will exceed the threshold even as the gap narrows. Yet it is possible that no tract will exceed twice the median if there are no terribly unhealthy neighborhoods in the city. In other words, the use of the median is "distribution free." Thus, in a relatively geographically egalitarian city, the count of extreme tracts based on the median would be zero, while the small standard deviation in such a city could result in some fairly healthy tracts exceeding the threshold of the mean plus one or two standard deviations.

The calculation is simply a count of the number of tracts in which the rate is more than twice the median. In the case of a positive indicator, such as percentage with prenatal care, half of the median is used as the threshold. The percentage of the tracts exceeding the median is the count divided by the number of tracts in the city.

Results

Table B.1 displays the composite disparity index. For most indicators, the cities had similar index values both early and late in the 1990s. However, there are some notable exceptions, such as the decline in death rate and teen childbearing disparity in Cuyahoga County. It can also be seen that several of the cities have greater disparity on some, but not all, indicators. Nevertheless, each city shows a fairly high degree of disparity on at least one indicator.

Table B.1
Composite Disparity Index for Health Indicators by City
  Age-adjusted death rate Teen birth rate Pct. prenatal care in first trimester Pct. low birth weight
1990-94 1995-99 1990-94 1995-99 1990-94 1995-99 1990-94 1995-99
Cleveland/Cuyahoga County 79.4 34.6 138.8 89.0 46.1 46.4 11.7 12.3
Denver 47.9 45.0 73.6 66.6 51.7 43.1 34.5 34.0
Indianapolis/Marion County 28.9 29.1 68.7 60.4 31.8 32.9 13.3 11.5
Oakland 73.5 45.9 157.2 112.2 49.6 38.8 12.1 7.3
Providence n/a n/a n/a 90.3 n/a 25.7 n/a 11.7
Note: Although data were only available for the cities of Oakland and Providence, this analysis used the rates of Alameda County and Providence County as the reference points.

Tables B.2a and B.2b display the number and percentage of disparate tracts using the mean plus one and two standard deviation criteria. There is considerable variability across the cities on this measure. It can be seen, though, that very few tracts exceed the 2 standard deviation threshold, while more tracts exceed the 1 standard deviation threshold. Using this method, the most severe disparities are on the prenatal care indicator. This is in contrast to the composite disparity index, on which teen birth rate shows the highest score. The count of tracts above the mean plus one standard deviation also ranks the cities differently on inequality than does the composite score.

Table B.2a:
Number and Percent of Tracts Exceeding the "Extreme Distribution Threshold"
for Selected Health Indicators by City
  Number of tracts Mean and one standard deviation
Age -adjusted deaths Teen birth rate Pct. prenatal care in first trimester Pct. low birth weight
1990-94 1995-99 1990-94 1995-99 1990-94 1995-99 1990-94 1995-99
Number of tracts                  
Cleveland/Cuyahoga County 499 2 8 6 77 92 110 85 51
Denver 181 17 17 18 25 60 72 6 17
Indianapolis/Marion County 204 41 39 34 33 37 39 36 27
Oakland 107 1 1 1 1 19 20 19 13
Providence 37 NA NA NA 6 NA 5 NA 4
As percent of all tracts
Cleveland/Cuyahoga County   0.4 1.6 1.2 15.4 18.4 22.0 17.0 10.2
Denver   9.4 9.4 9.9 13.8 33.1 39.8 3.3 9.4
Indianapolis/Marion County   20.1 19.1 16.7 16.2 18.1 19.1 17.6 13.2
Oakland   0.9 0.9 0.9 0.9 17.8 18.7 17.8 12.1
Providence   NA NA NA 16.2 NA 13.5 NA 10.8
Table B.2b:
Number and Percent of Tracts Exceeding the "Extreme Distribution Threshold"
for Selected Health Indicators by City
  Mean and Two Standard Deviations
Age -adjusted death rate Teen birth rate Pct. prenatal care in first trimester Pct. low birth weight
1990-94 1995-99 1990-94 1995-99 1990-94 1995-99 1990-94 1995-99
Number of tracts
Cleveland/Cuyahoga County 2 2 6 17 14 23 15 10
Denver 4 6 5 3 39 41 3 6
Indianapolis/Marion County 2 5 7 8 6 3 6 5
Oakland 1 1 1 1 3 3 3 4
Providence NA NA NA 1 NA 0 NA 1
As percent of all tracts
Cleveland/Cuyahoga County 0.4 0.4 1.2 3.4 2.8 4.6 3.0 2.0
Denver 2.2 3.3 2.8 1.7 21.5 22.7 1.7 3.3
Indianapolis/Marion County 1.0 2.5 3.4 3.9 2.9 1.5 2.9 2.5
Oakland 0.9 0.9 0.9 0.9 2.8 2.8 2.8 3.7
Providence NA NA NA 2.7 NA 0.0 NA 2.7

Table B.3 presents a similar count, but twice the median is used to establish the threshold. This method generally identifies a greater number of tracts as exceeding the threshold. Cleveland, in particular, stands out as having a greater percentage of tracts that are classified as extreme according to this method. In fact, the rank order of the cities changes depending upon whether a median or mean plus standard deviation criterion is used.

Table B.3
Number and Percent of Tracts Exceeding the " Relative" Threshold
for Selected Health Indicators by City
  Number of tracts Twice the median
Age -adjusted death rate Teen birth rate Pct. prenatal care in first trimester Pct. low birth weight
1990-94 1995-99 1990-94 1995-99 1990-94 1995-99 1990-94 1995-99
Number of tracts
Cleveland/Cuyahoga County 499 15 8 162 153 7 13 64 42
Denver 181 3 1 26 17 38 41 6 0
Indianapolis/Marion County 204 1 1 40 23 1 1 3 6
Oakland 107 4 4 7 4 2 2 1 2
Providence 37 NA NA NA 0 NA 0 NA 1
As percent of all tracts
Cleveland/Cuyahoga County   3.0 1.6 32.5 30.7 1.4 2.6 12.8 8.4
Denver   1.7 0.6 14.4 9.4 21.0 22.7 3.3 0.0
Indianapolis/Marion County   0.5 0.5 19.6 11.3 0.5 0.5 1.5 2.9
Oakland   3.7 3.7 6.5 3.7 1.9 1.9 0.9 1.9
Providence   NA NA NA 0.0 NA 0.0 NA 2.7

Table B.4 shows how the cities are ranked on each indicator according to the three indices. For each of the indicators, the composite index shows a rather different ranking than do the two counts of extreme tracts based on either medians or mean plus standard deviation criteria.

Table B.4
City Disparity Rankings on Selected Health Indicators by Ranking Method
Age-adjusted death rate Composite Disparity Index Mean and one standard deviation Twice the median
1990-94 1995-99 1990-94 1995-99 1990-94 1995-99
Cleveland/Cuyahoga Cnty 4 2 1 2 3 3
Denver 2 3 3 3 2 2
Indianapolis/Marion County 1 1 4 4 1 1
Oakland 3 4 2 1 4 4
Providence n/a n/a n/a n/a n/a n/a
Teen birth rate            
Cleveland/Cuyahoga Cnty 3 3 2 3 4 5
Denver 2 2 3 2 2 3
Indianapolis/Marion County 1 1 4 4 3 4
Oakland 4 5 1 1 1 2
Providence n/a 4 n/a 5 n/a 1
Pct. low birth weight            
Cleveland/Cuyahoga Cnty 2 5 2 2 4 5
Denver 4 4 1 1 3 1
Indianapolis/Marion County 1 2 3 5 2 4
Oakland 3 3 4 4 1 2
Providence n/a 1 n/a 3 n/a 3
Pct. prenatal care in first trimester            
Cleveland/Cuyahoga Cnty 1 4 3 4 2 4
Denver 4 5 4 5 4 5
Indianapolis/Marion County 3 2 2 3 1 2
Oakland 2 1 1 2 3 3
Providence n/a 3 n/a 1 n/a 1

Discussion

The measures of health disparity differ from one another in several ways. Because the composite indicator provides a single score for the city, it can be used to compare among cities or compare a city with itself over time to see whether it is becoming less unequal. The other two methods identify specific neighborhoods that exceed a threshold of poor health. The number of such neighborhoods can be useful information to the city for planning health programs and determining how to target health resources. The percentage of such neighborhoods is a useful measure to compare cities with one another or to compare a region of changing size over time. Because the three approaches to measurement differ, it is possible for a city to display inequality on the composite index but to have no tracts that exceed the threshold for disparity in the other two indices. It is also conceivable that a city could have a few tracts with extremely poor health, but if the vast majority of tracts were quite similar on an indicator, the composite index would have a relatively low value.

An important limitation of any of these indices, though, is that they are influenced by the geographic boundaries of the metropolitan area used in the analysis. In this study, the central city or central county is used. If, instead, the entire metropolitan region had been analyzed, the results would have been different. Moreover, some of our geographic areas contain affluent suburbs, while others do not. For example, for the Cleveland study, we had access to data for the entirety of Cuyahoga County, in which Cleveland is located. The county contains many suburban municipalities, including some of the most affluent neighborhoods in the state. The analysis for Providence, though, focuses only on the city of Providence and does not include affluent suburban neighborhoods. In general, if the designated region is more economically homogenous, the disparities will seem less pronounced. This is due to the fact that the mean, median, or rate for the region is used as the basis for calculating the indices. If most of the census tracts are close to this measure of central tendency, the disparities will be small. Even if the overall health of the region is poor, there will be little inequality because all tracts are similarly poor in their health. Thus, researchers and planners making comparisons must carefully consider the choice of the regional as well as neighborhood boundaries.

A feature of all of these indices that should be noted is that they derive their reference point from within the city/county itself. For example, the extreme distribution count uses the rate on the indicator for the whole city/county in the calculation. An alternative would be to use a national rate in this calculation. If a national rate were used, the disparity for neighborhoods in the city would be relative to the nation. Such a count would not only reflect differences across the city but differences between the city and the nation on health indicators. National rates are available for many health indicators, especially those based on birth and death certificates. However, if a national rate is used it is important to be sure that the local indicators are being calculated using a similar methodology.

Despite these ambiguities and limitations, there is value in attempting to measure inequality and identify neighborhoods in which health indicators are extremely poor. The existence of neighborhood health inequality points to the need to examine the factors that may be responsible for such patterns. It also raises important questions about whether health care resources are being distributed in a fair and effective manner. Information about health disparities can be used to mobilize the community to address these conditions and monitor their progress.

View full report

Preview
Download

"report.pdf" (pdf, 5.74Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc1.pdf" (pdf, 36.01Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc2.pdf" (pdf, 80.73Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc3.pdf" (pdf, 81.09Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc4.pdf" (pdf, 80.81Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc5.pdf" (pdf, 80.84Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc6.pdf" (pdf, 80.25Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc7.pdf" (pdf, 44.47Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc8.pdf" (pdf, 44.54Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc10.pdf" (pdf, 44.74Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc11.pdf" (pdf, 42.16Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc15.pdf" (pdf, 51.62Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc19.pdf" (pdf, 38.08Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc23.pdf" (pdf, 42.28Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc24.pdf" (pdf, 38.21Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc25.pdf" (pdf, 39.36Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc26.pdf" (pdf, 40.5Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc27.pdf" (pdf, 41.26Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc28.pdf" (pdf, 42.45Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc29.pdf" (pdf, 41.76Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®

View full report

Preview
Download

"Annexc30.pdf" (pdf, 41.98Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®