American Journal of Epidemiology
Copyright © 1998 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 148, No. 12
Printed in U.S.A.
Geocoding and Linking Data from Population-based Surveillance and the US
Census to Evaluate the Impact of Median Household Income on the
Epidemiology of Invasive Streptococcus pneumoniae Infections
Frederick M. Chen,1 Robert F. Breiman,1 Monica Farley,2 Brian Plikaytis,1 Katherine Deaver,1 and Martin S.
Cetron1
The emergence of drug-resistant Streptococcus pneumoniae poses new clinical challenges and may also
reflect a change in the epidemiology of S. pneumoniae infections. A variety of studies have shown that drugresistant S. pneumoniae infections are linked to antimicrobial use. It has been hypothesized that persons of high
socioeconomic status are at increased risk for a drug-resistant infection because of greater access to
antimicrobial drugs. To assess whether median household income is associated with increased risk of penicillinnonsusceptible S. pneumoniae infections, the authors geocoded and linked data from population-based
surveillance for invasive pneumococcal disease with data from the 1990 US Census. Among invasive
pneumococcal isolates from Atlanta, Georgia, in 1994, increasing proportions of penicillin-nonsusceptible
isolates were associated with higher median household incomes (x2 for trend, 15.17; p = 0.002). Despite higher
rates of invasive pneumococcal disease among blacks and persons who resided within lower median household
income areas, white patients in areas with higher median household income had a higher risk of being infected
with strains that were not susceptible to penicillin (Wilcoxon rank sum, Z = 2.66, p = 0.008). These findings
demonstrated the utility of geocoding and US Census data in describing the epidemiology of drug-resistant S.
pneumoniae and also provided more evidence that socioeconomic factors may influence the development of
drug resistance. Am J Epidemiol 1998; 148:1212-18.
censuses; drug resistance; socioeconomic status; Streptococcus pneumoniae
Streptococcus pneumoniae is the leading cause of
bacterial pneumonia, meningitis, bacteremia, and otitis
media. The emergence of drug-resistant S. pneumonae
poses new and difficult challenges for the treatment of
these infections (1,2). Recently, population-based surveillance in the metropolitan Atlanta, Georgia, area
revealed that more than 25 percent of invasive isolates
of S. pneumoniae were not susceptible to penicillin (3).
Antibiotic use, the key factor in the emergence of drug
resistance, appears to be higher among affluent
patients, who presumably have greater access to medical care (1, 3-7). Among persons with invasive pneu-
mococcal disease in the Atlanta study, whites had a
substantially greater risk of being infected with a drugresistant strain than did black persons (3). While racial
differences in antibiotic usage have been described,
race is also often linked to socioeconomic status in this
country (7). To assess whether socioeconomic status is
associated with increased risk of drug-resistant invasive pneumococcal infections, we linked surveillance
data from the Atlanta study with economic data
obtained from the 1990 US Census (8).
Received for publication September 2, 1997, and accepted for
publication May 5, 1998.
Abbreviations: CDC, Centers for Disease Control and Prevention;
PNSP, penicillin-nonsusceptible Streptococcus pneumoniae, STF,
summary tape file.
1
Childhood and Respiratory Disease Branch, Division of
Bacterial and Mycotic Diseases, National Center for Infectious
Diseases, Centers for Disease Control and Prevention, Atlanta, GA.
2
Emory University School of Medicine, VA Medical Center,
Atlanta, GA.
Reprint requests to Dr. Martin S. Cetron, Surveillance and
Epidemiology Branch, Division of Quarantine, National Center for
Infectious Diseases, Mailstop E-03, Centers for Disease Control and
Prevention, 1600 Clifton Road, Atlanta, GA 30333.
Surveillance
MATERIALS AND METHODS
From January through December 1994, populationbased surveillance was conducted for all isolates of S.
pneumoniae from normally sterile sites (invasive isolates) documented in the microbiology laboratories of
32 hospitals and one major reference laboratory serving the metropolitan Atlanta area (3). The eight-county region has a population of 2.34 million persons,
according to 1990 Census data (68 percent white, 29
percent black, and 3 percent other racial groups). On
1212
Geocoding Cases of Drug-resistant S. pneumoniae Infections
the basis of audits of microbiology laboratories, 74
percent of all invasive isolates from the study period
were sent to the Centers for Disease Control and
Prevention to determine their susceptibility to antimicrobial agents, as described (3).
Definition of antimicrobial resistance
The National Committee for Clinical Laboratory
Standards has defined three separate categories of
antimicrobial susceptibilities: susceptible (S), intermediate (I, formerly intermediate or low-level resistance),
and resistant (R, formerly high-level resistance) categories (9). This report refers to any isolate with intermediate or resistant antimicrobial susceptibility as nonsusceptible. Multidrug nonsusceptibility was defined as
nonsusceptibility to two or more of the following
antimicrobial drugs or drug classes: beta-lactam antibiotics and carbapenems, macrolides, trimethoprim-sulfamethoxazole, tetracycline, ofloxacin, and chloramphenicol (3).
Data management
Patient demographic data, including age, race, sex,
and street address, were obtained from standardized
active surveillance case report forms. These data were
stored at Emory University, geocoded (assigned census tract and block group numbers), and merged with
the 1990 US Census database to obtain socioeconomic information. Census data were obtained from the US
Bureau of the Census (US Department of Commerce,
Washington, DC), in the form of Summary Tape File
(STF) 3A data files on CD-ROM (8). Socioeconomic
data are available on the STF3A files at several different geographic levels, ranging in size from states to
block groups. The US Census geographic divisions
are, in decreasing order, state, county, census tract,
block group, and block. At the time of the 1990 US
Census, there were 61,231 census tracts in the United
States. Although the population of each census tract
varies widely, an average of 4,000 persons reside in
each. Census tracts are further divided into block
groups along street and neighborhood boundaries.
There are roughly four block groups within a census
tract. Block groups can be divided into census blocks,
the smallest level of geographic resolution in the census database. There are 7 million census blocks in the
United States (mean, 28 persons per block). US
Census socioeconomic data are available only to the
level of the census block group, not for individual census blocks.
Each census tract, block group, and block has a
unique numeric identifier within each county. The census tract is identified by a five-digit numeral with two
Am J Epidemiol
Vol. 148, No. 12, 1998
1213
decimal places (e.g., 123.45). For each census tract,
block groups and blocks are numbered with three-digit
integers (e.g., 123). The first digit of the block number
refers to the block group number. As a result, given a
street address, it is possible to geocode the address and
locate the census identifier number and its corresponding census tract, block group, and block.
By using the street address information for each case
of invasive pneumococcal disease, a census tract and
block group number was identified. This process was
accomplished by using LandView II, a geographic
information software system available on CD-ROM
from the US Bureau of the Census. In addition, a database maintained by the Atlanta Regional Commission
was utilized to locate new street addresses that were
missing from the 1990 Census database. Ten percent of
the isolates could not be geocoded. Some of these
addresses represented new neighborhoods or street
numbers absent in either database. Misspelled street
names and other data collection errors also contributed
to missing data.
Upon identification of the corresponding census
tract and block group number for each case, the median incomes of households within that block group
were accessible from the Census STF3A files. To
achieve a more-accurate point estimate, income distributions stratified by race were obtained for each block
group. By using the patient's reported race, a median
household income for the members of the patient's
race who resided within a particular block group was
calculated. Income estimates obtained from census
block group data have been shown to be statistically
valid measures of self-reported income (10, 11). In
addition to race and median household income,
patients were stratified by age with the information
provided by case report forms.
Data analysis
Overall incidence rates for invasive pneumococcal
disease were calculated by using 1994 population estimates provided by the Atlanta Regional Commission
(12). However, incidence rates incorporating socioeconomic or age data were calculated by using 1990 US
Census population data because 1994 data were
unavailable. Proportional changes in population over
the 4-year period were unlikely to be significant and
should not alter the interpretation of the calculated
rates (12). Incidence rates for black and for white
households were calculated for each of nine US
Census household income strata. Ninety-five percent
confidence intervals for risk ratios and p values were
calculated with the Wilcoxon rank sum test and
Fisher's two-tailed exact test. Calculations were performed with SAS (version 6.2) (SAS, Inc., Cary, North
1214
Chenetal.
Carolina) and the Epi Info statistical program (version
6.0a; Centers for Disease Control and Prevention,
Atlanta, Georgia).
RESULTS
A total of 716 cases of invasive pneumococcal disease were identified via active surveillance for 1994 in
metropolitan Atlanta. When 1994 Atlanta population
data were used, the overall incidence of infection with
invasive pneumococcal disease was 28 cases per
100,000 population; the incidence was 18 cases per
100,000 among whites and 48 cases per 100,000
among blacks. The estimated incidence for black children less than age 2 years (300 cases per 100,000) was
higher than that for white children in the same age
group (75 cases per 100,000) (figure 1).
Isolates from 528 cases were available for susceptibility testing (74 percent of all invasive pneumococcal
infections occurring in Atlanta during 1994). Periodic
laboratory audits demonstrated that patients for whom
isolates were tested did not differ significantly in age,
race, county of residence, and outcome from patients
for whom isolates were unavailable (3). A total of 137
isolates (26 percent) were nonsusceptible to penicillin.
A higher proportion of white children less than age 2
years were infected with a penicillin-nonsusceptible S.
pneumoniae (PNSP) isolate, compared with black children (45 vs. 23 percent; risk ratio = 2.02, 95 percent confidence interval 1.16—3.51) (figure 2). For other age groups, no
significant differences were observed in the proportion of
pneumococcal infections caused by PNSP strains.
Socioeconomic and isolate susceptibility data were
available from the combined census and surveillance
databases for 476 of the 528 cases (90 percent). After
stratification by median household income groups,
black households in low-income strata had much higher rates of invasive pneumococcal disease than did
white households in the same income strata (figure 3).
For example, in block groups with a median household
income of $7,500 per year, the incidence of pneumococcal disease was 127 cases per 100,000 black households and 12 cases per 100,000 white households (%2 =
24.5, p < 0.0001). In block groups with a median
household income of $20,000 per year, the incidence
of pneumococcal disease was 179 cases per 100,000
black households and 15 cases per 100,000 white
households (x2 = 106.2, p < 0.0001). The peak incidence of invasive pneumococcal disease in whites (57
cases per 100,000 households) occurred in middleincome strata ($30,000-$42,500). The incidence of
disease in middle-income black households was 169
cases per 100,000 households, with median income
$30,000, and 84 cases per 100,000 households, with
median income $42,500.
The incidence of invasive pneumococcal disease
was markedly lower in block groups with high median
household incomes (nine cases per 100,000 white
households at $87,500 and six cases per 100,000 white
households at $125,000). There were no cases of invasive pneumococcal disease reported in black households in block groups with median incomes of $87,500
and $125,000; however, there was a much smaller
number of black households in these income ranges
cases per 100,000
350
<2
>=65
FIGURE 1. Incidence of invasive pneumococcal disease by age and race. Using population-based surveillance, 716 cases of invasive S. pneumoniae were identified in Atlanta in 1994. Incidence rates were calculated by using 1994 Atlanta population data for blacks (solid line) and whites
(dotted line) in all age groups.
Am J Epidemiol
Vol. 148, No. 12, 1998
Geocoding Cases of Drug-resistant S. pneumoniae Infections
1215
% of PNSP isolates (n)
50
(15)
40
31
30
26
(4)
(2)
32
32
(22)
(24)
20
18
20
(5)
P
10-
1
<2
2-5
$
0
6-17
>=65
—-64
age (years)
FIGURE 2. Percentage of invasive pneumococcal isolates that were penicillin-nonsusceptible S. pneumoniae (PNSP) by age and race (shaded bars, blacks). By using population-based surveillance demographic data from Atlanta for 1994, the percentage of PNSP was analyzed by
age and race. White children less than age 2 years had an increased risk of infection with a penicillin-nonsusceptible isolate (risk ratio = 2.02,
95 percent confidence interval 1.16-3.51) when compared with black children of the same age. No statistically significant differences were
observed in other age groups.
cases per 10,000 households
20
2500
7500
12500
20000
30000
42500
62500
87500
125000
median household income
FIGURE 3. Incidence of invasive pneumococcal disease (per 10,000 households) in blacks (shaded bars) and whites (open bars) by median
household income. A total of 476 invasive pneumococcal isolates were geocoded and linked to 1990 US Census data at the block group level
to obtain median household income estimates. Incidence rates were calculated by using a denominator with units of households because of
limitations in the census database.
(9,232 black households vs. 108,279 white households) in the surveillance area.
Of the 528 isolates tested, 137 (26 percent) were
nonsusceptible to penicillin. The overall incidence of
invasive penicillin-nonsusceptible pneumococcal disAm J Epidemiol
Vol. 148, No. 12, 1998
ease was four per 100,000 persons among whites and
10 per 100,000 persons among blacks (11). However,
whites with invasive pneumococcal disease had a
higher proportion of infection with a strain that was
nonsusceptible to penicillin than did blacks (33 vs. 21
1216
Chenetal.
percent; risk ratio = 1.57, 95 percent confidence interval 1.17-2.11).
When the 476 isolates for which both isolate susceptibility and socioeconomic data were available
were used, increasing proportions of PNSP isolates
were associated with higher median household
incomes among cases (x2 for trend = 15.17, p = 0.002)
(figure 4). White patients with PNSP isolates were significantly more likely to live in block groups with
higher median household incomes (Wilcoxon rank
sum, Z = 2.66, p = 0.008); however, this association
was not detected for blacks (Wilcoxon rank sum, Z =
0.79, p = 0.428) (figure 5).
Patients with multidrug nonsusceptible isolates were
also more likely to live in higher median household
income block groups (Wilcoxon rank sum, Z = 2.89,
p = 0.004). This relation was not observed when data
were stratified by race (Wilcoxon rank sum for whites,
Z = 0.969, p = 0.33; Wilcoxon rank sum for blacks,
Z = 0.643, p = 0.52).
DISCUSSION
In Atlanta, patients living in census block groups
with higher median household incomes are at greater
risk for PNSP infections. The results suggest that the
higher proportion of nonsusceptible isolates among
suburban whites in Atlanta (previously reported)
reflects racial differences in socioeconomic status (3).
This observation raises the possibility that the association between median household income and antibiotic
resistance may be related to greater access to medical
care, with coincident excessive or inappropriate usage
of antibiotics.
Higher income was associated with PNSP among
whites; however, among black patients, no statistically
significant association between income and penicillin
nonsusceptibility was observed. A threshold income
level may exist, above which access to medical care
and use of antibiotic drugs increase, resulting in
increasingly common drug-resistance patterns. The
limited range of black household incomes may have
precluded the observation of such a resistance threshold level in this community. Unidentified potential differences in patterns of transmission of pneumococci
among both black and white communities may have
also played a role in the discrepancy. For example, we
did not assess the effect of overcrowding or use of
child care centers.
We did not find an association between socioeconomic status and infection with multiple drug-nonsusceptible strains after stratifying by race. In addition to
the explanations proposed above, the lack of statistical
significance may have been due to the small number of
multidrug nonsusceptible isolates per income strata,
resulting in limited power to observe a difference.
White children less than age 2 years had the highest
proportion of PNSP isolates. Antimicrobial access and
use are reported to be particularly high in this age group
(4, 13). Because information on attendance at child care
centers was not available, we could not evaluate an
association between PNSP disease and day-care centers.
% of PNSP isolates (n)
2500
7500
12500
20000
30000
42500
62500
87500
125000
median household income
FIGURE 4. Prevalence of penicillin nonsusceptible S. pneumoniae (PNSP) within each median household income group. Antimicrobial susceptibility data from 476 isolates of invasive S. pneumoniae were linked to 1990 US Census data. Within each median household income group,
the percentage of PNSP isolates is shown, x2 for trend = 15.17, p = 0.002.
Am J Epidemiol
Vol. 148, No. 12, 1998
Geocoding Cases of Drug-resistant S. pneumoniae Infections
1217
10
2500
7500
12500
20000
30000
42500
62500
87500
125000
median household income
FIGURE 5. Prevalence of penicillin nonsusceptible S. pneumoniae (PNSP) isolates within each median household income group, by race. The
percentages of PNSP isolates within each median household income group were further subdivided by race. Shaded bars indicate percentages
for blacks. For whites, there were significant differences in the proportion of PNSP isolates at different income levels (Wilcoxon rank sum = 2.66,
p = 0.008). The Wilcoxon rank sum for blacks = 0.79, p = 0.428.
The incidence of invasive pneumococcal disease
during the study period in Atlanta was higher than
rates previously described in other areas of the nation
(14-16). The extremely high incidence of invasive
pneumococcal disease among children, most notably
black children less than age 2 years (300 per 100,000
children), is of particular concern. Although the current 23-valent polysaccharide pneumococcal vaccine
is available for other high-risk groups, it is not efficacious in children less than age 2 years (17). These
results highlight the need for an effective pneumococcal vaccine for this age group; conjugate pneumococcal vaccines currently being evaluated contain
serotypes of S. pneumoniae associated with decreased
susceptibility to antimicrobial drugs and known to
cause invasive disease in children (3, 16).
The increased incidence of pneumococcal infections
among blacks has been described previously and has
often been noted to be independent of socioeconomic
differences (14-16, 18). Previous analyses utilized data
obtained at the level of the health district or census tract
and did not have the benefit of median household
income data at the block group level. This study demonstrated significant racial differences in the rates of pneumococcal disease at middle and lower levels of income.
These differences suggest the existence of risk factors other than socioeconomic status that may predispose certain blacks to pneumococcal infection. While
sickle-cell disease is one such example, its relatively
low incidence would not account for the racial disparAm J Epidemiol
Vol. 148, No. 12, 1998
ity in pneumococcal disease incidence. It has been
hypothesized that patients in groups with very low
median household incomes may have limited access to
medical and preventive care or live in overcrowded
conditions that predispose to pneumococcal infection;
however, we cannot explain the higher incidence rates
observed among persons from households with middle-income levels.
There is a strong influence of median household
income upon the proportion of PNSP isolates within a
population. Although lower median household
incomes increase the risk for invasive pneumococcal
disease, patients in higher income categories are at
higher risk for developing drug-resistant invasive
pneumococcal infections. The pattern of antimicrobial
drug use is probably a primary contributor to the emergence of drug resistance. Thus, at the high-income portion of the spectrum, excellent access to medical care
may translate into decreased risk of invasive disease,
but also a greater likelihood that infection will be due
to a penicillin-nonsusceptible isolate. Additional study
will be needed to evaluate whether the median household income is directly proportional to the amount of
antimicrobial use.
Limitations in the structure of the US Census database led to the use of a denominator with units of
households in the calculation of some incidence rates
in this report. The inability to derive an accurate estimate of the number of individuals in households within certain median household income strata precluded
1218
Chenetal.
the calculation of per capita incidence. We do not
believe that these calculations alter the interpretation
of the data. However, one should exercise caution
when comparing these incidence rates with rates calculated for individuals.
Pneumococcal disease is influenced not only by biologic factors, such as age and predisposing conditions,
but also by race and socioeconomic status. The racial
and socioeconomic factors that contribute to pneumococcal disease are complex and interwoven. This study
used novel methods to examine the influence of
socioeconomic and demographic factors on the risk of
invasive pneumococcal disease and PNSP infections.
Prospective, laboratory-based active surveillance systems, such as that in Atlanta, provide comprehensive
sources of data with potential to make compelling
observations. The ability to merge these data with US
Census data offers the opportunity to develop and
examine hypotheses regarding the impact of sociologic and economic factors on biologic events that shape
disease patterns within a population. Future studies
will be required to test these hypotheses with morespecific epidemiologic and laboratory tools.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
ACKNOWLEDGMENTS
14.
The authors are grateful to Wendy Baughman for her help
with the database, to the microbiology laboratories and
infection control practitioners in Georgia Health District III,
and to the epidemiology section of the Georgia Department
of Human Resources.
15.
16.
17.
REFERENCES
1. Klugman KP. Pneumococcal resistance to antibiotics. Clin
Microbiol Rev 1990;3:171-96.
2. Breiman RF, Butler JC, Tenover FC, et al. Emergence of drug-
18.
resistant pneumococcal infections in the United States. JAMA
1994;271:1831-5.
Hofmann J, Cetron MS, Farley MM, et al. The prevalence of
drug-resistant Streptococcus pneumoniae in Atlanta. N Engl J
Med 1995;333:481-6.
Reichler MR, Allphin AA, Breiman RF, et al. The spread of
multiply resistant Streptococcus pneumoniae at a day care center in Ohio. J Infect Dis 1992; 16:1346-53.
Ward J. Antibiotic-resistant Streptococcus pneumoniae: clinical and epidemiologic aspects. Rev Infect Dis 1981;3:254-66.
Centers for Disease Control and Prevention. Drug-resistant
Streptococcus pneumoniae—Kentucky and Tennessee, 1993.
MMWR Morbid Mortal Wkly Rep 1994;43:23-5, 31.
Melnick SL, Sprafka JM, Laitinen DL, et al. Antibiotic use in
urban whites and blacks: the Minnesota Heart Survey. Ann
Pharmacother 1992;26:1292-5.
Bureau of the Census. 1990 Census of population: social and
economic characteristics. Washington, DC: Department of
Commerce, Economics, and Statistics Administration, 1992.
National Committee for Clinical Laboratory Standards.
Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically. 2nd ed. Proposed standard M7-A3.
Villanova, PA: National Committee for Clinical Laboratory
Standards, 1993.
Cherkin DC, Grothaus L, Wagner EH. Is magnitude of co-payment effect related to income? Using census data for health
services research. Soc Sci Med 1992;34:33^U.
Krieger N. Overcoming the absence of socioeconomic data in
medical records: validation and application of a census-based
methodology. Am J Public Health 1992;92:703-10.
Atlanta Regional Commission. 1994 population and housing.
Atlanta GA: Atlanta Regional Commission, 1994.
Takala AK, Jero J, Kela E, et al. Risk factors for primary invasive pneumococcal disease among children in Finland. JAMA
1995;273:859-64.
Breiman RF, Spika JS, Navarro VJ, et al. Pneumococcal bacteremia in Charleston County, South Carolina: a decade later.
Arch Intern Med 1990;150:1401-5.
Bennett NM, Buffington J, LaForce FM. Pneumococcal bacteremia in Monroe County, New York. Am J Public Health
1992;82:1513-16.
Henneberger PK, Galaid El, Marr JS. The descriptive epidemiology of pneumococcal meningitis in New York City. Am
J Epidemiol 1983;117:484-91.
Butler JC, Breiman RF, Lipman HB, et al. Serotype distribution
of Streptococcus pneumoniae infections among preschool children in the United States, 1978-1994: implications for development of a conjugate vaccine. J Infect Dis 1995;171:885-9.
Filice GA, Darby CP, Fraser DW. Pneumococcal bacteremia in
Charleston County, South Carolina. Am J Epidemiol
1980;l 12:828-35.
Am J Epidemiol
Vol. 148, No. 12, 1998