Ascertaining variation in hospitalization risk among immigrants using small area analysis

5
Ascertaining variation in hospitalization risk among immigrants using small area analysis Peter Muennig , Haomiao Jia, Kamran Khan, Daniel J. Pallin Columbia University, Health Policy and Management, Mailman School of Public Health, 600W 168th Street, 6th Floor, New York, NY 10032, USA Available online 19 April 2006 Abstract Introduction. Data on hospitalization of immigrants to the United States are sparse, but small area analysis may improve the utility of existing data. Methods. Applying logistic regression analysis to 2000 census and hospitalization data for New York City, we examined the odds of hospital admission by major diagnostic category and global region of birth after controlling for covariates. We used individual-level covariates to control for age, race, and gender. By matching the patient's zip code of residence to census data, we then added median household income, the proportion of persons born in a particular global region, and the proportion of foreign-born persons living in the same zip code as independent variables. Results. The total proportion of foreign-born persons in a zip code predicts a lower hospitalization rate for most major diagnostic categories and most foreign-born groups. However, Africa-born persons have a higher odds of hospitalization for most major diagnostic categories up to 1.79 (95% confidence interval 1.73, 1.86) for blood and blood forming disorders relative to native-born persons. The odds of hospitalization among Africa-born persons for most conditions are over 3 times higher than other foreign-born groups. Hospitalization odds for Latin American-born persons were also higher than native-born persons across major diagnostic categories. Conclusion. Small area analysis generally predicts hospitalization rates that coincide with mortality studies and may serve as a useful tool for hypothesis testing in immigrant health. © 2006 Elsevier Inc. All rights reserved. Keywords: Emigration and immigration; Socioeconomic factors Introduction Almost 12% of the U.S. population is foreign-born (Larson, 2003). Relatively little is known about the health needs, risks, and desires of the newest Americans in part because national survey data are largely inadequate for these purposes. Collectively, immigrant populations have lower morbidity and mortality than native-born populations (Singh and Miller, 2004; Singh and Siahpush, 2002; Stephen et al., 1994). This may in part be due to selection of persons healthy enough to migrate, and in part due to significantly lower rates of risky health behaviors among recent immigrants (Shelley et al., 2004). This is thought to apply to most immigrant groups; however, samples of immigrant groups in national surveys are typically too small in size to produce reliable analyses (National Health and Nutrition Examination Survey, 2005). In larger data sets based on electronic billing data, such as the Healthcare Cost and Utilization Project, questions surrounding country of birth are not included (National Health and Nutrition Examination Survey, 2005; Muennig and Fahs, 2002). It is therefore of critical importance that alternative approaches to risk estimation be devised. One such approach is small area analysis, which has been demonstrated to be useful in analysis of county-level disability using statewide survey data, and at the neighborhood level using hospitalization data (Jia et al., 2004a,b; Muennig and Fahs, 2002). Elucidation of risk factors by immigrant group would not only help researchers begin to put the pieces of the immigrant puzzle together, it could potentially help answer questions such as why socioeconomic status plays less of a role in immigrant groups than among native-born persons, and why some groups seem to be at significantly greater risk of certain diseases than others (Shelley et al., 2004; Singh and Siahpush, 2002). Such data could also conceivably shed light on broader questions that Preventive Medicine 43 (2006) 145 149 www.elsevier.com/locate/ypmed Corresponding author. E-mail address: [email protected] (P. Muennig). 0091-7435/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.ypmed.2006.02.014

Transcript of Ascertaining variation in hospitalization risk among immigrants using small area analysis

(2006) 145–149www.elsevier.com/locate/ypmed

Preventive Medicine 43

Ascertaining variation in hospitalization risk among immigrantsusing small area analysis

Peter Muennig ⁎, Haomiao Jia, Kamran Khan, Daniel J. Pallin

Columbia University, Health Policy and Management, Mailman School of Public Health, 600W 168th Street, 6th Floor, New York, NY 10032, USA

Available online 19 April 2006

Abstract

Introduction. Data on hospitalization of immigrants to the United States are sparse, but small area analysis may improve the utility of existingdata.

Methods. Applying logistic regression analysis to 2000 census and hospitalization data for New York City, we examined the odds of hospitaladmission by major diagnostic category and global region of birth after controlling for covariates. We used individual-level covariates to controlfor age, race, and gender. By matching the patient's zip code of residence to census data, we then added median household income, the proportionof persons born in a particular global region, and the proportion of foreign-born persons living in the same zip code as independent variables.

Results. The total proportion of foreign-born persons in a zip code predicts a lower hospitalization rate for most major diagnostic categories andmost foreign-born groups. However, Africa-born persons have a higher odds of hospitalization for most major diagnostic categories – up to 1.79(95% confidence interval 1.73, 1.86) for blood and blood forming disorders – relative to native-born persons. The odds of hospitalization amongAfrica-born persons for most conditions are over 3 times higher than other foreign-born groups. Hospitalization odds for Latin American-bornpersons were also higher than native-born persons across major diagnostic categories.

Conclusion. Small area analysis generally predicts hospitalization rates that coincide with mortality studies and may serve as a useful tool forhypothesis testing in immigrant health.© 2006 Elsevier Inc. All rights reserved.

Keywords: Emigration and immigration; Socioeconomic factors

Introduction

Almost 12% of the U.S. population is foreign-born (Larson,2003). Relatively little is known about the health needs, risks,and desires of the newest Americans in part because nationalsurvey data are largely inadequate for these purposes.

Collectively, immigrant populations have lower morbidityand mortality than native-born populations (Singh and Miller,2004; Singh and Siahpush, 2002; Stephen et al., 1994). Thismay in part be due to selection of persons healthy enough tomigrate, and in part due to significantly lower rates of riskyhealth behaviors among recent immigrants (Shelley et al.,2004).

This is thought to apply to most immigrant groups; however,samples of immigrant groups in national surveys are typicallytoo small in size to produce reliable analyses (National Health

⁎ Corresponding author.E-mail address: [email protected] (P. Muennig).

0091-7435/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.ypmed.2006.02.014

and Nutrition Examination Survey, 2005). In larger data setsbased on electronic billing data, such as the Healthcare Cost andUtilization Project, questions surrounding country of birth arenot included (National Health and Nutrition ExaminationSurvey, 2005; Muennig and Fahs, 2002). It is therefore ofcritical importance that alternative approaches to risk estimationbe devised. One such approach is small area analysis, which hasbeen demonstrated to be useful in analysis of county-leveldisability using statewide survey data, and at the neighborhoodlevel using hospitalization data (Jia et al., 2004a,b; Muennigand Fahs, 2002).

Elucidation of risk factors by immigrant group would notonly help researchers begin to put the pieces of the immigrantpuzzle together, it could potentially help answer questions suchas why socioeconomic status plays less of a role in immigrantgroups than among native-born persons, and why some groupsseem to be at significantly greater risk of certain diseases thanothers (Shelley et al., 2004; Singh and Siahpush, 2002). Suchdata could also conceivably shed light on broader questions that

146 P. Muennig et al. / Preventive Medicine 43 (2006) 145–149

have perplexed international health researchers and domesticepidemiologists alike. For instance, why does Japan, which hasaround twice the prevalence of smokers as the U.S., have thehighest life expectancy in the world? (Marmot and Syme, 1976;Reed et al., 1982).

Proving causal associations on such topics will requirecontrolled studies, but more economical methods are needed toprioritize hypotheses for testing. Small area analysis is veryeconomical because it relies on combination of data from pre-existing databases. For instance, a previous small area analysisused sub-borough areas in New York City to estimate the risk ofhospitalization for immigrants, Puerto-Ricans, and native-bornpersons (Muennig and Fahs, 2002). The results of thehospitalization analysis proved to be highly consistent withmortality data, suggesting that this small area technique is validwhen applied to large groups over relatively large areas (Singhand Miller, 2004).

In this study, we examine variation in hospitalization rates bycountry of birth and major diagnostic category using a modifiedsmall area analytic technique. In this technique, a hospitaliza-tion data set is merged with census data. The analysis isconducted at the individual level, but variables missing from thehospitalization data set are linked to neighborhood character-istics at the zip code level.

Methods

We obtained all (de-identified) hospital admissions records for NewYork City residents from the year 2000 Statewide Planning and ResearchCooperative System (SPARCS) as well as population data from the 2000Decennial Census (Statewide Research Planning and Cooperative System(SPARCS), 2000; Census, 2000). The SPARCS data set containsinformation on 1,083,817 hospitalizations for the civilian population ofNew York City.

While the SPARCS data set includes information on patients' age, gender,and race, it does not contain information on region of birth or median familyincome. In order to include these two variables in our analysis, we matched eachsubject's zip code of residence from the SPARCS database to values obtained foreach zip code in New York City from the 2000 Decennial Census Long Form.Each subject is assigned a value equal to the proportion of a given immigrantgroup within the subject's zip code. We then treated all analyses as if they wereconducted at the individual-level (our unit of analysis) rather than the zip-codelevel.

After merging the two data sources, we used a logistic regression model toestimate the influence of country of origin upon hospitalization rates. Theregression approach to small area analysis has been found superior to directestimation, the synthetic method, and spatial smoothing (Jia et al., 2004a).

The following categorical variables were entered as independent predictorsof hospitalization: age (0–6, 7–17, 18–44, 45–64, 65+), gender, median familyincome (in increments of $1000), race (black, white, Asian), and the proportionof persons from a particular region of birth (continuous). We categorized regionof birth into 7 categories: Sub-Saharan Africa, the Middle East, East Asia,Other Asia, Latin America (including Mexico and the Caribbean), EasternEurope, other European countries, and also examined all immigrantscollectively. Finer groupings were not possible because they resulted ininadequate sample sizes within zip codes. The Middle Eastern countries includethose in North Africa, and other Asia includes South Asia, Southeast Asia, andthe Pacific Islands.

When analyzing groups by region of birth, the proportion of native-bornpersons was added as an independent predictor to control for the influence ofmultiple groups residing in a given zip code. Foreign-born persons werecategorized according to their self-reported region of birth. Persons whoidentified as native born were used as the reference group.

The regression was run once for each of the 7 regions of birth. In each run,the proportion of immigrants from the region of birth of interest in the zip codewas substituted as a continuous variable in the model.

We also performed separate logistic regression analyses using hospitaliza-tions for each Major Diagnostic Category (MDC) and “V-code” of theInternational Classification of Diseases, 9th revision (ICD-9) as the dependentvariable. Patients are assigned “V” codes when they make contact with thehealth system for reasons unrelated to an underlying condition. Becausepregnancy is in this category, it allows this common event to be compared withother medical events. We also examined specific diagnoses (ICD-9 3 digitcodes), which were felt to be of particular relevance to certain immigrant groups.

In our logistic regression models, hospitalization rates were generated bydividing age, race, and gender-specific hospitalizations obtained from individuallevel hospital data by corresponding population counts. When hospitalizationrates are very low, logistic models provide a good approximation for modelingcounts data. The median family income of the zip code, the proportion offoreign-born persons, and the proportion of foreign-born persons from a specificregion of birth were then added as covariates. Thus, we controlled for theproportion of persons born in a country other than the country of interest in ouranalysis. The model took the general form:

logitðpÞ ¼ b0 þ b1 � ageþ b2 � raceþ b3 � sexþ b4 � incomeþ b5� foreignþ b6 � nativeþ l

where “foreign” is the proportion of foreign-born from a given region and μ iszip code level random effect. Models for the overall foreign-born population didnot include the proportion of native-born persons.

Logistic models were constructed for each category of hospitalization andthe variables representing the proportion of persons from each region of theworld. Since hospitalizations for conditions by MDC occurred in a smallproportion of the population over the year, we did not convert odds ratios to riskratios (Zhang and Yu, 1998). We tested the models using the Hosmer andLemeshow goodness of fit test, and index plots of the residuals.

Results

The demographic profile of New Yorkers by country of birth(based on our analysis of 2000 census data) is presented in Table1. While 64.1% of the population is native to the U.S.,approximately 4.7% of the population was born in a U.S.territory, principally Puerto Rico, and 35.9% of the populationwas born outside of the U.S. or its territories. Approximately22% of New Yorkers report speaking English less than “verywell.”

All immigrant groups were represented in virtually all zipcodes within New York City (data not shown). However,African-born persons were disproportionately represented inlow- and high-income zip codes, particularly those surroundingthe United Nations building and those within Harlem. Thisgroup was represented in very small numbers in most otherneighborhoods (mean = 1.1%, range 0.09% to 7.49%).

Table 2 lists the estimated odds ratios for hospitalizationby MDC and region of birth relative to native-born persons.Most immigrant groups had a significantly lower odds ofhospitalization than native born groups for nearly alldiagnoses. One exception was pregnancy-related diagnoses,which were similar to native-born persons (OR = 1.03; 95%CI = 0.93, 1.13) and “V” codes (1.06; 95% CI = 1.05, 1.07),which include exposure communicable conditions and normalpregnancies.

The proportion of persons from Africa-born zip codes,however, was a strong predictor of hospitalization for infectious

andparasitic

diseases,endocrine/m

etabolicconditions,

hema-

tologicaldisorders,mentaldisorders,nervous

systemdisorders,

respiratorydiseases,

andskin

diseases(w

hichinclude

many

infectiousdiseases).

For

instance,relative

tonative-born

persons,foreign-born

personswere

atlow

erodds

ofhospital-

izationfor

infectiousand

parasiticdiseases

(OR

0.69;95%

CI=

0.67,0.70).

How

ever,the

proportionof

Africa-born

personsin

azip

codepredicted

anincreased

oddsof

1.62(1.58,

1.66)for

hospitalizationdue

tothese

diseases.Sim

ilarly,Latin

American

immigrants

were

atgreaterrisk

ofinfectious

andparasitic

diseases,endocrine/metabolic

disorders,respiratory

disorders,diseasesof

theskin,and

mentaldisorders

thannative-born

persons.How

ever,theodds

oftheseconditions

were

ofsm

allermagnitude

thanfor

Sub-S

aharanAfricans.

Wealso

examined

theodds

ofhospitalization

fortubercu-

losis,HIV,stom

achcancer,

breastcancer,

diabetesmellitus,

affectivedisorders,psychosis,alcoholuse

ordrug

dependence,and

chronicobstructive

pulmonary

diseaseby

theproportion

ofoverall

immigrants

inazip

code(data

notshow

n).Relative

topersons

frompredom

inantlynative-born

zipcodes,

theodds

ratiosfor

hospitalizationwere

lessthan

1.0for

allof

theseconditions

(range0.32

to0.97).O

fnote,for

allimmigrants,the

oddsratio

forhospitalization

forconditions

relatedto

alcoholuse

anddrug

dependencewas

0.32(95%

CI=0.32,

0.33),for

hospitalizationdue

toHIV

itwas

0.47(95%

CI=0.45,

0.49),and

fortuberculosis

itwas

0.88(95%

CI=0.76,

1.03).

Discussion

The

purposeof

thispaper

isto

testwhether

small

areaanalysis

might

serveas

auseful

toolfor

testinghypotheses

Table1

Dem

ographiccharacteristics

ofNew

York

City

basedon

2000Census

data(percent

oftotal

population)

Totalpopulation

8,008,278Median

agein

years34.2

Male

3,794,204(47.4)

Fem

ale4,214,074

(52.6)White

3,576,385(44.7)

Black

orAfrican

American

2,129,762(26.6)

American

Indianand

Alaska

Native

41,289(0.5)

Asian

787,047(9.8)

Native

5,137,246(64.1)

Born

inUnited

States

4,763,116(59.5)

Born

outside50

states(e.g.,in

territory)374,130

(4.7)Language

otherthan

English

3,554,805(44.4)

Foreign

born2,871,032

(35.9)Entered

1990to

March

20001,224,524

(15.3)Naturalized

citizen1,278,687

(16)Not

acitizen

1,592,345(19.9)

Speaks

English

lessthan

“verywell”

1,768,977(22.1)

Sub-S

aharanAfrica

68,599(2.4)

Middle

East

130,704(0.5)

Other

Asia

200,461(0.7)

East

Asia

352,757(13)

Latin

America

(IncludingMexico)

1,484,354(53.1)

Eastern

Europe

329,721(11.8)

Western

Europe

a227,055

(8.1)

aSom

egroups

ofpredom

inantlyEuropean

originwere

notincluded

inthe

analysis.These

includeCanadians,

Australians,

andNew

Zealanders.

Table 2Odds of hospitalization in the year 2000 for the foreign-born population by region of birth and major diagnostic category in New York City (relative to native-born persons with the same diagnostic category)

Major diagnostic category Sub-Saharan Africa Middle East Other Asia East Asia Latin America Eastern Europe Western Europe All immigrants

Infectious, parasitic (001–139) 1.62 (1.58, 1.66) 0.76 (0.74, 0.78) 0.76 (0.74, 0.78) 0.77 (0.75, 0.79) 1.23 (1.20, 1.26) 0.72 (0.70, 0.74) 0.69 (0.67, 0.71) 0.69 (0.67, 0.70), 0.96) 0.89 (0.88, 0.91) 0.91 (0.89, 0.93) 1.00 (0.98, 1.02) 0.94 (0.93, 0.96) 0.92 (0.90, 0.94) 0.93 (0.91, 0.95), 0.80) 0.77 (0.76, 0.79) 0.78 (0.76, 0.80) 1.20 (1.17, 1.22) 0.73 (0.71, 0.74) 0.68 (0.67, 0.70) 0.70 (0.69, 0.71), 0.69) 0.61 (0.59, 0.63) 0.63 (0.61, 0.66) 0.94 (0.91, 0.98) 0.68 (0.65, 0.70) 0.61 (0.59, 0.64) 0.70 (0.68, 0.72), 0.80) 0.82 (0.81, 0.83) 0.83 (0.82, 0.84) 1.23 (1.22, 1.25) 0.79 (0.78, 0.80) 0.61 (0.60, 0.62) 0.49 (0.48, 0.49), 0.86) 0.82 (0.79, 0.85) 0.84 (0.80, 0.87) 1.08 (1.04, 1.12) 0.82 (0.80, 0.85) 0.83 (0.79, 0.86) 0.78 (0.75, 0.80), 0.99) 0.94 (0.93, 0.96) 0.95 (0.94, 0.97) 1.03 (1.02, 1.04) 0.97 (0.95, 0.98) 0.92 (0.90, 0.93) 0.81 (0.80, 0.82),0.86) 0.84 (0.83,0.85) 0.87 (0.86,0.89) 1.22 (1.20,1.24) 0.79 (0.78,0.80) 0.77 (0.76,0.79) 0.68 (0.67,0.69), 1.00) 0.98 (0.96, 0.99) 1.01 (0.99, 1.03) 1.14 (1.12, 1.16) 0.96 (0.95, 0.97) 0.90 (0.89, 0.92) 0.85 (0.84, 0.86), 0.94) 0.92 (0.90, 0.94) 0.91 (0.89, 0.93) 1.06 (1.04, 1.08) 0.94 (0.92, 0.96) 0.88 (0.86, 0.91) 0.86 (0.84, 0.88), 0.98) 0.97 (0.96, 0.98) 0.95 (0.94, 0.96) 1.15 (1.14, 1.17) 0.95 (0.94, 0.96) 0.82 (0.81, 0.83) 1.05 (1.04, 1.06), 0.92) 0.94 (0.91, 0.97) 0.89 (0.86, 0.92) 1.23 (1.19, 1.27) 0.87 (0.85, 0.90) 0.79 (0.76, 0.81) 0.68 (0.66, 0.70), 0.99) 0.95 (0.92, 0.97) 0.93 (0.90, 0.96) 1.06 (1.03, 1.09) 0.99 (0.96, 1.01) 0.98 (0.95, 1.01) 0.80 (0.78, 0.82), 1.04) 1.12 (1.04, 1.21) 1.07 (1.00, 1.16) 1.03 (0.95, 1.11) 0.97 (0.90, 1.04) 1.00 (0.93, 1.09) 1.01 (0.94, 1.08), 1.09) 1.07 (0.97, 1.18) 1.16 (1.05, 1.28) 1.19 (1.07, 1.33) 0.89 (0.81, 0.97) 0.95 (0.85, 1.06) 1.03 (0.93, 1.13), 0.89) 0.93 (0.91, 0.95) 0.90 (0.88, 0.92) 1.12 (1.10, 1.14) 0.88 (0.87, 0.90) 0.78 (0.77, 0.80) 0.75 (0.74, 0.76), 0.95) 0.95 (0.94, 0.97) 0.97 (0.95, 0.98) 1.10 (1.08, 1.12) 0.94 (0.92, 0.95) 0.85 (0.83, 0.87) 0.82 (0.81, 0.83), 1.07) 1.07 (1.06, 1.08) 1.10 (1.09, 1.12) 1.05 (1.04, 1.07) 1.01 (1.00, 1.02) 0.96 (0.95, 0.97) 1.06 (1.05, 1.07)

147P.

Muennig

etal.

/Preventive

Medicine

43(2006)

145–149

Neoplasms (140–239) 1.16 (1.14, 1.18) 0.94 (0.92Endocrine, metabolic (240–279) 1.47 (1.44, 1.50) 0.78 (0.77Blood, blood forming (280–289) 1.79 (1.73, 1.86) 0.67 (0.64Mental disorders (290–319) 1.78 (1.76, 1.81) 0.79 (0.78Nervous system, sense (320–389) 1.30 (1.25, 1.34) 0.83 (0.80Circulatory system (390–459) 1.11 (1.09, 1.12) 0.98 (0.97Respiratory system (460–519) 1.37 (1.36,1.39) 0.84 (0.83Digestive system (520–579) 1.12 (1.10, 1.13) 0.98 (0.97Genitourinary (580–629) 1.15 (1.12, 1.17) 0.92 (0.90Pregnancy, birth (630–679) 1.05 (1.03, 1.06) 0.96 (0.95Skin, subcutaneous tissue (680–709) 1.27 (1.23, 1.30) 0.89 (0.86Musculoskeletal (710–739) 1.11 (1.08, 1.14) 0.97 (0.94Congenital (740–759) 0.98 (0.91, 1.05) 0.97 (0.90Perinatal conditions (760–779) 1.03 (0.94, 1.13) 0.99 (0.90Symptoms and signs (780–799) 1.25 (1.23, 1.27) 0.88 (0.86Injury and poisoning (800–999) 1.14 (1.12, 1.16) 0.93 (0.92Health status/health services (V01–V99) 0.96 (0.95, 0.97) 1.06 (1.05

All analyses control for age, race, income, and gender.

148 P. Muennig et al. / Preventive Medicine 43 (2006) 145–149

related to immigrant health. Our data reveal that hospitalizationis less common in zip codes with high proportions ofimmigrants, except for conditions related to reproduction.Given the ecological nature of the data, these observations couldbe explained by one of two hypotheses: (1) immigrants are lesslikely to be hospitalized, or (2) immigrants have similar orhigher odds of hospitalization than native-born people, but thirdvariable effects explain the observed odds of hospitalization.

Interpreting our data as a valid test of hypothesis (1) wouldexemplify the ecological fallacy, and thus hypothesis (2) mustbe entertained. However, our findings are generally consistentwith a large body of research suggesting that the arduousprocess of immigration selects for healthier individuals (Singhand Miller, 2004; Singh and Siahpush, 2002; Stephen et al.,1994). Moreover, our analysis of hospitalization by diagnosticcategories is highly consistent with mortality rates by cause ofdeath for immigrants overall (Singh and Siahpush, 2002). Thus,small area analysis seems to be a useful tool for testinghypotheses related to immigrant health.

The troubling finding in our study is that New Yorkers bornin Sub-Saharan Africa and Latin America had higher odds ofhospitalization for most categories. This potentially identifies avulnerable sub-population and, according to the aboverationale, mandates investment in studies that are moreanalytically rigorous than our small-area approach. Indeed,the incidence of parasitic disease is higher among persons bornin Africa and Latin America than in other groups, and higherrates of blood forming disorders might be expected amonggroups of African origin as well (Muennig et al., 1999).

On the other hand, African-born persons comprise thesmallest immigrant group in terms of absolute numbers (seeTable 1), are in small numbers in most zip codes, and aredisproportionately represented in high and low incomeneighborhoods (e.g., around the United Nations building andin Harlem). It is also an extremely diverse population, includinga wide variety of religious practices and geographic origins.

Afro-Caribbean populations in New York are known to beat lower risk of mortality than native-born whites, andAfrican-born populations nationwide are known to be atlower risk of overall mortality as well as mortality due to eachof the causes in Table 2 (Fang et al., 1996; Singh and Miller,2004). This suggests that small area analysis applied to thisgroup may be confounded by construct validity (i.e.,intragroup heterogeneity).

In national studies, such as the National Health Examinationand Nutrition Survey (NHANES), Mexicans and otherHispanics are generally found to be substantially healthierthan native-born persons regardless of country of birth (Singhand Miller, 2004; Singh and Siahpush, 2002; Sundquist andWinkleby, 1999). However, the high proportion of PuertoRicans in New York City – a population that mixes with otherHispanic groups and with higher hospitalization and mortalityrates than native-born persons – may contribute to skewing ofthese results (Muennig and Fahs, 2002). Even in thispopulation, poor health status for some diagnoses (Claudio etal., in press; Thorpe et al., 2004) may be counterbalanced bybetter health outcomes for others (Rosenberg et al., 2005).

New York City contains the most diverse immigrantpopulation in the United States, and one of the most diversein the word. For this reason, it may be a better laboratory forexamining specific groups than any other locality. Theupcoming New York City Health Examination and NutritionSurvey may prove to be a good data set for resolving this issue(NYC Health Examination and Nutrition Survey, 2005). Futurestudies might examine the interplay of ethnicity, place of birth,health, behavioral risk factors, socioeconomic risk factors, andgeographic risk factors.

This study was subject to a number of importantlimitations. Foremost, there were not enough available datato test for ecological bias in all of the outcomes and groupsunder study. While small area models explain up to 69% ofthe variation in hospitalization by country of origin, thereliability of such models is low when groups are notrepresented in many neighborhoods (Muennig and Fahs,2002). Second, we were unable to control for health insurancestatus, and thus underutilization. Finally, some of the variationwe noted between groups may be attributed to socioculturalfactors driving health-seeking behavior. This may work ineither direction; groups more likely to seek care may behealthier as a result and thus less likely to be hospitalized.Conversely, some groups may attempt to avoid electivehospitalizations.

Nevertheless, hospitalization is a proximal outcome that maybe partly dependent on cultural factors, and may partly relate toseeking medical care in the U.S. It is conceivable, for instance,that some people with relatives in New York City arrived fromtheir country of origin for the sole purpose of seeking hospitalcare and subsequently reported their relative's home address astheir permanent address in the U.S. Care seeking may beespecially relevant in considering the higher risk of hospital-ization among persons born in Latin America, since the flight isshorter, less expensive, and the social networks are moreextensive. To the extent that those from other countries seekcare in New York City and then return to their country of origin,our results are understated.

Our study points to a need for better information on the riskand risk factors for disease by immigrant group. As cliniciansincreasingly rely upon risk assessment in medical decision-making, country of origin may prove to be an importantdeterminant of whether to screen for disease, and whether tosuspect a given disease in patients with symptoms with whichclinicians are unfamiliar.

Recently, cultural competence has come to the fore as asignificant health issue (Carrillo et al., 1999). Although patientcommunication is paramount, the next logical question thatarises is whether clinicians have the knowledge base to prevent,diagnose, and treat disease once communication has beenestablished. Considerable attention has also recently been givento the inclusion of women and minority groups in clinical trials.Missing in this call for action are immigrants; the number ofimmigrants in the United States is on par with the number ofAfrican Americans and is growing at a faster rate (Gibson andLennon, 1999). Whether the United States will be able to profiton the human capital it has imported will be determined in part

149P. Muennig et al. / Preventive Medicine 43 (2006) 145–149

by the extent to which researchers understand immigrants'health needs.

References

Carrillo, J.E., Green, A.R., Betancourt, J.R., 1999. Cross-cultural primary care: apatient-based approach. Ann. Intern. Med. 130 (10), 829–834.

Census 2000, Census 2000 Release File 3, viewed 03/04 2004, http://www.census.gov/main/www/cen2000.html.

Claudio, L. J., Stingone,A.,Godbold, J., in press. Prevalence of childhood asthmain urban communities: the impact of ethnicity and income. Ann. Epidemiol.

Fang, J., Madhavan, S., Alderman, M.H., 1996. The association betweenbirthplace and mortality from cardiovascular causes among black and whiteresidents of New York City. N. Engl. J. Med. vol. 335 (no. 21), 1545–1551.

Gibson, C.J., Lennon, E., 1999. Historical Census Statistics on the Foreign-BornPopulation of the United States: 1850–1990. Bureau of the Census,Washington.

Jia, H., Muennig, P., Borawski, E., 2004a. Comparison of small-area analysistechniques for estimating county-level outcomes. Am. J. Prev. Med. 26 (5),453–460.

Jia, H., Muennig, P., Lubetkin, E.I., Gold, M.R., 2004b. Predicting geographicalvariations in behavioural risk factors: an analysis of physical and mentalhealthy days. J. Epidemiol. Community Health 58 (2), 150–155.

Larson, L., 2003. The Foreign-Born Population in the United States. Bureau ofthe Census, Washington.

Marmot, M.G., Syme, S.L., 1976. Acculturation and coronary heart disease inJapanese–Americans. Am. J. Epidemiol. 104 (3), 225–247.

Muennig, P., Fahs, M.C., 2002. Health status and hospital utilization of recentimmigrants to New York City. Prev. Med. 35 (3), 225–231.

Muennig, P., Pallin, D., Sell, R.L., Chan, M.S., 1999. The cost effectiveness ofstrategies for the treatment of intestinal parasites in immigrants. N. Engl. J.Med. 340 (10), 773–779.

National Health and Nutrition Examination Survey, 2005. 6/20/05, http://www.cdc.gov/nchs/data/nhanes/frequency/l13doc.pdf.

NYC Health Examination and Nutrition Survey, 2005. http://www.ci.nyc.ny.us/html/doh/html/hanes/hanes.shtml.

Reed, D., McGee, D., Cohen, J., Yano, K., Syme, S.L., Feinleib, M., 1982.Acculturation and coronary heart disease among Japanese men in Hawaii.Am. J. Epidemiol. 115 (6), 894–905.

Rosenberg, T.J., Raggio, T.P., Chiasson, M.A., 2005. A further examination ofthe “epidemiologic paradox”: birth outcomes among Latinas. J. Natl. Med.Assoc. 97 (4), 550–556.

Shelley, D., Fahs, M., Scheinmann, R., Swain, S., Qu, J., Burton, D., 2004.Acculturation and tobacco use among Chinese Americans. Am. J. PublicHealth 94 (2), 300–307.

Singh, G.K., Miller, B.A., 2004. Health, life expectancy, and mortality patternsamong immigrant populations in the United States. Can J. Public Health 95(3), I14–I21.

Singh, G.K., Siahpush, M., 2002. Ethnic-immigrant differentials in healthbehaviors, morbidity, and cause-specific mortality in the United States: ananalysis of two national data bases. Hum. Biol. 74 (1), 83–109.

Statewide Research Planning and Cooperative System (SPARCS). 2000. NewYork State Department of Health, 7/1/05, http://www.health.state.ny.us/nysdoh/sparcs/annual/t2000_01.htm.

Stephen, E.H., Foote, K., Hendershot, G.E., 1994. Health of the foreign-bornpopulation: United States, 1989–90. Adv. Data 241, 1–12.

Sundquist, J., Winkleby, M.A., 1999. Cardiovascular risk factors in MexicanAmerican adults: a transcultural analysis of NHANES III, 1988–1994. Am.J. Public Health 89 (5), 723–730.

Thorpe, L.E., List, D.G., Marx, T., May, L., Helgerson, S.D., Frieden, T.R.,2004. Childhood obesity in New York City elementary school students. Am.J. Public Health 94 (9), 1496–1500.

Zhang, J., Yu, K.F., 1998. What's the relative risk? A method of correcting theodds ratio in cohort studies of common outcomes. JAMA 280 (19),1690–1691.