Differences in Hospital Mortality Among Critically Ill Patients of Asian, Native Indian, and European Descent

Original Research CRITICAL CARE MEDICINE

Differences in Hospital Mortality Among Critically III Patients of Asian, Native Indian, and European Descent* Nadia A. Khan, MD, MSc; Anita Palepu, MD, MPH; Monica Norena, MSc; Najib Ayas, MD, MPH; Hubert Wong, PhD; Dean Chittock, MD, MSc; Morad Hameed, MD, MPH; and Peter M. Dodek, MD, MHSc

Background: It is unclear whether racelethnicity influences survival for acute critical illnesses. We compared hospital mortality among patients of Asian (originating from Asia or Southeast Asia), Native Indian, and European descent admitted to the ICU. Methods: Prospective cohort study ofpatients admitted to three lCUs Oanuary 1999 to January 2006) in British Columbia, Canada. Multivariable analysis evaluated hospital mortality for each ethnic group, adjusting for age, sex, APACHE (acute physiology and chronic health evaluation) II score, hospital, median income, unemployment, and education. To account for differences in case mix, multivariable analysis was also restricted to those patients admitted for the five most common lCU admission diagnoses (sepsis, pneumonia, brain injury, COPD, and ARDS) and adjusted for these diagnoses. Results: Of 7,331 patients, 21% were Asian, 4% were Native Indian, and 75% were of European descent. Crode mortality was 33% for Asian, 30% for Native Indians, and 28% for patients of European descent. After adjusting for potential confounders, Native Indian descent was not associated with an increase in mortality compared to European descent. Asian descent was associated with a signi6cantly higher mortality (odds mtio [OR], 1.22; 95% confidence interval [CI], 1.06 to 1.41; P = 0.005). After adjusting for case mix, this difference was DO longer seen. For patients admitted for COPD exacerbation, Asian descent was associated with a substantial increase in mortality (OR, 4.5; 95% CI, 1.56 to 12.9; P = 0.005). There were no significant differences in mortality by racelethnicity for patients who had any of the other common admitting diagnoses. Conclusion: Patients of Asian and Native Indian descent with acute critical ilbless did not have an increased mortality after adjusting for differences in case mix. (CHEST 2008; 134:1217-1222) Key words: Asian; ethniclty, ICU; mortality; Native Indian; nice Abbreviations: APACHE = acute physiology and chronic health evaluation; CI = confidence interval, OR = odds ratio; SES = socioeconomic status

R ace/ethnicity can influence survival for several medical disorders. For example, age-adjusted death rates due to stroke in Canadians of Asian descent "Frorn the Centre for Health Evaluation and Outcome Sciences (Drs. Khan and Palepu, and Ms. Norena), Department of Medicine; Division of Pulmonary Medicine (Dr. Ayas), HIV Clinical Trials Network (Dr. Wong), Department of Medicine; Division of Critical Care Medicine (Drs. Chittock and Dodek): and Department of Surgery (Dr. Hameed), Universityof British Columbia, BC, Canada. Dr. Khan holds a St. Paul's Hospital Physician scholar award, a GENESIS Young investigator award, and a Canadian Institute for Health Research New Investigator award. Dr. Ayas is a Canadian Institute for Health Research New Investigator as well as a Michael Smith Foundation for Health Research scholar. Dr. Palepu is a Michael Smith Foundation for Health Research senior scholar. www.chestjournal.org

are higher than those in the general population.' In addition, age-standardized mortality rates per 10,000 for death in Native Indian persons are higher than those in the general population for ischemic heart This work was done at the Center for Health Evaluation and Outcomes Sciences, University of British Columbia, BC, Canada. The authors have no conflicts of interest to disclose. Manuscript received April 16, 2008; revision accepted July 10, 2008. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.diestjournal. orglmisclreprints.shtml). Correspondence to: Nadia A. Khan, MD, MSc, 620-B 1081 Burrard

St, Vancouver, BC, Canada, V6Z 1Y6; e-mail: [email protected] 001: 10.13781chest.08-10l6 CHEST /134/6/ DECEMBER, 2008

1217

disease (14.1 vs 10.9), chronic lung disease (2.8 vs 2.4), and congestive heart failure (2.8 vs 1.5).2 Although many of these disparities in health outcomes are found in chronic medical disorders, it is less clear whether race/ethnicity affects survival after acute critical illness, especially as severity of illness plays a dominant role in outcomes of critical illness.' Previous studies evaluating racial/ethnic differences in survival among ICU patients show conflicting results, are mostly based on older data sets, and did not evaluate patients of Asian or Native Indian descent'> despite the recognized health disparities that exist within these populations.P Differences in outcomes according to racial/ethnic differences may provide novel insights into etiologies and pathogenesis of critical illness. Thus, the aim of the current study was to evaluate differences in mortality among patients of Asian (originating from East Asia or South Asia), Native Indian, and European descent who were admitted to three ICUs in British Columbia, Canada.

MATERIALS AND METHODS

24 h of ICU stay, a widely used predictor of hospital mortality among patients admitted to the ICU,3 also validated in a Chinese population." Although self-report is the optimal method for determining race/ethnieity," this is often not feasible in the ICU. In our study cohort, patient race/ethnic category (Asian, Native Indian, European descent) was determined by prospective review of the medical record by trained ICU nurses. In a review" of 100 random ICU hospital charts from the ICU database, the reliability of race/ethnic categorization was excellent (K = 0.81). As some ethnic minorities are more likely to have low income,IO·11 and socioeconomic status (SES) itself has an inverse association with ICU mortality,12.13 SES may be an important potential confounder when evaluating race/ethnicity. We assessed SES using area-level median income, postsecondary education (defined as any formal education after high school), and unemployment derived from the Canadian Census 200 1 socioeconomic file after patients' postal codes were converted into census dissemination areas using the Statistics Canada Postal Code Conversion me (2005). Area-level SES indicators strongly correlate with individual attributes'< and are a valid and widely used estimation of individual SES.15.16 Trained ICU nurses enter concurrent data daily from the medical chart into this database, and the database automatically acquires demographic data from the hospital admission, discharge, and transfer database. Accuracy of data are improved by automatic alerts for implausible values, electronic calculation of severity scores after manual input of the elements of these scores, and periodic testing of interrater and intrarater reliability.

Subjects and End Points The province of British Columbia has 37 ICUs, including 4 adult tertiary care ICUs, 7 large community ICUs, and 26 smaller-center ICUs. The ICU database is a registry of all patients admitted to ICUs at two tertiary care hospitals and one large community hospital located in two large multiethnic urban centers in British Columbia. The tertiary ICUs include a 15-bed ICU within a 400-bed hospital and a 27-bed ICU within a 6OO-bed hospital, and the community ICU is an 8-bed ICU within a 175-bed hospital. These ICUs admit both surgical and medical patients, and trained intensive care physicians coordinate all care. The severity of illness of patients in these ICUs as measured by the APACHE (acute physiology and chronic health evaluation) II score is moderate to high (mean scores of 22.6, 20.2, and 17.0). These scores are also consistent with those in other similar ICUs in British Columbia that were not included in this study. We reviewed this registry and identified all consecutive hospital admissions from January 1999 to January 2006 with an assigned race/ethnicity of Asian (South Asian or East Asian), Native Indian, or European descent aged 2: 18 years. Black and Hispanic patients were excluded due to insufficient patient numbers. We excluded all ICU readmissions to avoid counting two outcomes for the same patient. The primary end point for this study was hospital mortality.

Study Variables The ICU database contains prospectively collected information regarding age, sex, race/ethnic category, primary ICU admitting diagnosis, other ICU admitting diagnoses, underlying comorbid diagnoses, conditions diagnosed in the ICU, hospital mortality, hospital length of stay, and ICU length of stay. The ICU diagnoses are determined from a prospective chart review and are classified according to the dictionary of diagnoses developed by the Intensive Care National Audit and Research Council." The database also contains the APACHE II score based on the first 1218

Statistical Analysis Descriptive statistics were used to characterize the baseline characteristics stratified by race/ethnicity category. We calculated crude mortality and 95% confidence intervals (Cis) for each race/ethnic group and compared these rates using X2 testing. We also calculated diagnosis-speciflc mortality for the five most common ICU admitting diagnoses: septic shock/septicemia, pneumonia (with or without sepsis), primary brain injury, COPD exacerbation, and ARDS. Given that ARDS may coexist with sepsis and pneumonia, we also created an expanded definition for ARDS including those who had a secondary diagnosis of ARDS but a primary diagnosis of pneumonia or sepsis. As the mortality findings were not Significantly different using this expanded definition, we report the mortality among patients with ARDS as the primary admitting diagnosis. These diagnoses accounted for 41% of all primary admitting diagnoses for Native Indians, 27% for Asians, and 30% for patients of European descent. To estimate the independent association between race/ethnic category and hospital mortality, risk adjusted, logistic regression models were fitted by using hospital mortality as the dependent variable, ethnicity/race categories (Native Indian vs European descent; Asian vs European descent) as exposure variables and a predetermined set of independent variables as potential confounders. The independent variables fit in the models were selected based on clinical judgment, published literature, and a guideline document on evaluating race/ethnicity for epidemiologic studies," These variables included age, sex, APACHE II score, hospital, median income, percentage of unemployment, and postsecondary education. First, we fit the model with all patients included to assess the overall independent effect of race/ethnicity on hospital mortality in the study cohort. To adjust for differences in case mix, we fit a second model restricted to only those patients admitted for the five most common diagnoses, adding each of these primary diagnoses as a possible confounder. Then to explore whether mortality is associated with different race/ethnic groups for certain diagnoses, we stratified this adOriginal Research

jus ted analysis by diagnosis. The Hosmer-Lemeshow goodnessof-fit test statistic and residual plots were used to check model assumptions. We assessed for the presence of an interaction between sex and race/ethnicity for the outcome of hospital mortality in the entire for the study cohort. We additionally assessed for the presence of an interaction between hospital and race/ethnicity for mortality given that regional variations may exist in admitted patient characteristics, care delivery, and mortality between hospitals. The presence of interactions was assessed using the Wald test. Statistical analyses were conducted using statistical software (Statistical Analysis System, version 9.1.3; SAS Institute; Cary, NC). This study was approved by the Research Ethics Board at Providence Health CarelUniversity of British Columbia, Canada.

RESULTS

Patient Characteristics

Of 7,331 patients admitted to these three ICUs during the study period, 4% were Native Indian, 21% were Asian, and 75% were of European de-

scent. Patients were most often referred to the ICU from the emergency department (20%), general medicine units (15%), general surgery service (8%), and neurosurgery service (5%). Native Indian patients were younger, more likely to be female, to have drug and alcohol dependence, HN, and lower SES indicators compared with patients of Asian or European descent (Table 1). The severity of illness on admission to ICU as indicated by APACHE II score was similar across all race/ethnic groups. Hospital Mortality

Hospital mortalitywas Significantly different among the racial/ethnic groups (p = 0.005) for the entire study cohort: 32% among those of Asian descent, 30% for Native Indian patients, and 28% for those of European descent. The multivariate analysis for all patients showed that after adjusting for age, sex, APACHE II score, hospital, and SES indicators,

Table l-Characteristics of Patients According to RacelEthnic Category* Characteristics

Native Indian (n = 288)

Asian Descent (n = 1,522)

European Descent (n = 5,521)

Age, yr Female sex, % SES indices Median income] Percentage unemployed Percentage with postsecondary education Comorbid conditions, No. (%) Ischemic heart disease Diabetes mellitus COPD Peptic ulcer disease Congestive heart failure Stroke/transient ischemic attack History of cancer

48.4 ± 15.4 52

60.6 ± 17.6 40

59.2 ± 17.11 36

16.1 ± 9.3 16.2 ± 11.9 56.0 ± 15.3

20.3 ± 6.6 8.7 ± 6.6 62.0 ± 14.0

22.2 ± 8.1 9.3 ± 7.9 64.4 ± 13.7

HIY Injection drug user History of alcohol abuse ICU admitting diagnosis, No. (%) Pneumonia, with/without sepsis ARDS Sepsis COPD exacerbation Primary brain injury Other admitting diagnosis APACHE II score first 24 h Hospital days, median (IQR) ICU days, median (IQR) ICU mortality, % (95% CI) Hospital mortality, % (95% CI) Diagnosis-specific mortality, % Pneumonia, with/without sepsis

32 (Il) 54 (19) 32 (II) 31 (Il) 31 (Il) 34 (12) 22(8) 44 (15) 85 (30) 103 (36) 33 (Il) 7 (13) 63 (22) 8 (3)

6(2) 171 (59) 21.6 ± 7.6 16 (7-34) 5.7 (2.O-Il.7) 22.2 (17.4-27.0) 30.2 (24.9-35.5)

122 (8) 37(2) 211 (14) 20 (1) 27(2) 1,105 (73) 21.0 ± 8.2 14 (6-29) 3.7 (1.6-8.3) 21.4 (19.4-23.5) 32.5 (30.2-34.9)

521 (9) 174 (3) 657 (12) 151 (3) 129 (2) 3,889 (70) 20.4 ± 7.4 16 (7-33) 8.3 (1.6-9.3) 19.0 (18.0-20.1) 28.3 (27.1-29.5)

36 35 45 60 30

30 42 44 23 20

24

~m

~

Sepsis COPD exacerbation Primary brain injury

54 13 17

1,203 (22) 945 (17) 790 (14) 204 (4) 853 (15) 648 (12) 699 (13) 213 (4) 439 (8) 719 (13)

345 (23)

388 (11) Il5 (8) 40 (3) 224 (15) 198 (13) 170 un Il (7) 23(2) 74(5)

*Data are presented as mean ± SD unless otherwise indicated. IQR = interquartile range. tIn thousands of Canadian dollars. www.chestjournal.org

CHEST / 134/6/ DECEMBER, 2008

1219

Table 2-Unadjusted and Adju8ted 0& for H08pital Mortality Factors Age (per 10 yr) Male sex APACHE II score (per point) Native Indian vs European descent Asian vs European descent Hospital A vs hospital C Hospital A vs hospital B Median income ($1,000 Canadian) % Unemployment % Post secondary education

Unadjusted OR (95% CI)

Adjusted OR (95% CI)*

1.28 (1.24-1.32) 0.92 (0.83-1.02) 1.15 (1.14-1.16) 1.10 (0.85-1.42)

1.240.21-1.29) 0.88 (0.78--0.99) 1.13 (1.12-1.14) 1.02 (0.76-1.37)

1.22 (1.08-1.38) 0.48 (0.41-0.56) 1.62 (1.45-1.81) 1.00 (0.99-1.01)

1.22 (1.06-1.41) 0.51 (0.43--0.61) 1.31 (1.17-1.49) 1.00 (0.99-1.01)

1.00 (1.00-1.00 1.00 (1.00-1.00

1.01 (1.00-1.02) 1.01 (1.01-1.00

*N = 7,331, adjusted for age, sex, APACHE II score, race/ethnicity, hospital, and SES (median income, percentage with postsecondary education, percentage of unemployment).

Asian descent was associated with increased hospital mortality (odds ratio [OR], 1.22; 95% CI, 1.06 to 1.41; P = 0.005) compared with European descent (Table 2). Native Indian descent and SES indicators were not associated with increased mortality (Table 3). Although the mortality rate varied significantly by hospital, there was no interaction between race/ethnicity and hospital for this outcome. Further, there was no evidence for an interaction between race/ethnicity and patient sex. Diagnosis-Specific Mortality

After adjusting for differences in case mix among patients who had one of the five a priori most common admitting diagnoses, there were no Significant differences in mortality across racial/ethnic groups: Asian vs European descent (OR, 1.08; 95% CI, 0.85 to 1.38) and Native Indian vs European descent (OR, 1.16; 95% CI, 0.76 to 1.77). When we analyzed patients who had each of these five admitting diagnoses separately, the mortality rate among those admitted for COPD exacerbation was 60% for Asian descent, 13% for Native Indian, and 23% for Euro-

pean descent patients. After adjustment for potential confounders among patients admitted because of COPO, Asian descent remained associated with a substantialincrease in mortalitycompared to European descent (OR, 4.5; 95% CI, 1.6 to 12.9; P = 0.005) in this subgroup. Hospital mortality did not differ across the racial/ethnic groups for the remaining admitting diagnoses of septic shock/septicemia, pneumonia (with or without sepsis), primary brain injury, or ARDS (Table 1). DISCUSSION

We found that patients of Asian descent who are admitted to ICUs in an equal-access health-care system have higher hospital mortality than patients of European descent. However, after adjusting for differences in case mix, these differences were no longer Significant. There are many studies2.17-19 documenting health disparities between Native Indian persons and the general population. However, in our study, we did not find any Significant difference in hospital survival between patients of Native Indian or European descent who were admitted to these ICUs, APACHE II scores, an important determinant of hospital mortality, were similar in all racial/ethnic groups. Earlier studies evaluating the association between race/ethnicity and ICU outcomes have produced conflicting results. In one prospective study' of 249 patients admitted to an ICU between 1985 and 1986, black patients had higher hospital mortality compared with white patients even after adjusting for APACHE II score and insurance status. In a large national study; of 17,440 patients admitted to 43 ICUs in the United States (from 1988 to 1990), hospital mortality rates in black and white patients were similar (10% vs 9.6%, respectively). In a stud y20 of patients hospitalized because of sepsis in 68 US hospitals, the ICU case-fatality rates were not different for black, Hispanic, or white patients after adjusting for clinical characteristics and hospital site. However, patients of Asian or Native Indian descent

Table 3-Ho8pital Mortality Stratified by Primary Admitting Diagno8i8* Factors

Pneumonia (WithlWithout Sepsis)

ARDS

Sepsis

CO PO Exacerbation

Age (per 10 yr) Male sex APACHE II score (per point) Native Indian vs European descent Asian vs European descent Hospital A vs hospital C Hospital A vs hospital B

1.18 (1.05-1.32) 0.76 (0.53-1.10 1.10 (1.07-1.13) 0.90 (0.38-2.13) 1.27 (0.81-1.98) 0.77 (0.45-1.34) 1.35 (0.92-1.99)

1.23 (1.01-1.49) 0.93 (0.50-1.71) 1.130.07-1.19) 0.54 (0.11-2.80) 0.93 (0.41-2.10) 0.30 (0.01-6.56) 1.64 (0.86-3.10)

1.21 (1.10-1.33) 0.88 (0.66-1.18) 1.12 (1.10-1.15) 1.61 (0.90-2.87) 1.27 (0.82-1.98) 0.84 (0.50-1.43) 1.03 (0.76-1.39)

1.21 (0.82-1.77) 0.95 (0.45--0.99) 1.05 (0.98-1.13) 0.61 (0.07-5.35) 4.48 (1.56-12.9) 0.83 (0.29-2.40) 0.67 (0.29-1.52)

Primary Brain Injury

1.11 (0.1l6-1.42) 0.58 (0.22-1.52) 1.13 (1.05-1.22) 0.50 (0.05-5.04) 1.39 (0..'51-3.75)

*Data are presented as adjusted OR (95% CI). N = 2,166, adjusted for age, sex, APACHE II score, race/ethnicity, and hospital.

1220

Original Research

were not included in these studies. Thus, our study extends the findings from these studies by evaluating different racialJethnic populations using contemporary data in a setting where insurance status does not influence access to hospital or critical care services. The Canadian health-care system has a single government payer. Consistent with other studies, we found that APACHE II score on hospital admission is a significant predictor of outcome'' across racial groupS.7 In our ICU cohort, SES indicators were not significantly associated with increased mortality. Hutchings et al13 reported that among 51,572 ICU admissions in the United Kingdom, there is an inverse relationship between SES and mortality in surgical admissions but not in nonsurgical admissions. This finding is generally consistent with our study because the ICU admissions were predominantly (70%) nonsurgical. Also consistent with other studies,20.21 we found differences in mortality by hospital that may reflect differences in characteristics of the patient admitted (measured and unmeasured) and caredelivery practices at these sites. However, hospital site did not significantly affect the relationship between race/ethnicity and mortality. In our exploratory analysis, patients of Asian descent who were admitted because of COPD exacerbation have a significantly higher mortality than those of European descent. Mortality for patients who had other major admitting diagnoses including septic shock/septicemia, pneumonia (with or without sepsis), primary brain injury, and ARDS did not differ among racial/ethnic groups. Further, when we entered these admitting diagnoses including COPD exacerbation into the first statistical model, the excess mortality observed among patients of Asian descent was no longer seen. These data suggest that the excess mortality initially observed among patients of Asian descent may be due to differences in case mix and specifically driven by the high mortality among patients of Asian descent who were admitted for COPD exacerbation. This hypothesis-generating finding is of interest and may signal a biological susceptibility to severe COPD among persons of Asian descent, differences in care delivery,5.22 or cultural differences in preferred level of care/do-not-resuscitate orders. Jayes and colleagues-" determined in a national ICU study that visible minorities, including patients of Asian descent, had Significantly fewer do-not-resuscitate orders compared with white patients; and so, potentially, sicker patients of European descent may decline ICU admission compared with patients of Asian descent. There may also be underlying differences in health behaviors, including the widely recognized epidemic of tobacco misuse and COPD in Asia,24-27 health literacy, social charwww.chestjoumal.org

acteristics not captured by our study, and/or genetic vartations.P Moreover, bias among care providers, differences in cultural preferences in treatment, and efficacy of therapeutic interventions may contribute to imbalances in hospital mortality. This study has several limitations. First, we included patients from a single geographic area; thus, these findings may not be generalizable to other centers or other patients of Asian or First Nations descent. However, the hospitals examined in this study are similar to other similar-sized hospitals in British Columbia. Second, there may be some misclassification of race assignment in our study. Misclassification would tend to underestimate differences among racialJethnic groups and bias the results toward the null. Third, given the limited number of Native Indian patients in our cohort, this study would be underpowered to detect potentially small differences in mortality within this subgroup. Fourth, inherent in observational studies, there may be residual confounding including differences in health behaviors, medical treatments, do-not-resuscitate orders, bias among health-care providers, or emerging risk factors. Nevertheless, we were able to adjust for key factors including sociodemographic factors and severity of illness at baseline using a contemporary data set. While this study sheds light on mortality among Asian and Native Indian patients already admitted to the ICU, we could not determine mortality among critically ill patients who were not referred to the ICU or who refused ICU care. Although APACHE II scores for the first 24 h of admission to ICU were similar across the groups, there may be important racialJethnic differences in access to ICU that need further investigation.

CONCLUSION

Patient racelethnicity is widely recognized to influence the outcomes of several major chronic medical disorders. In a universal health-care system, patients of Asian and Native Indian descent who have acute critical illness do not have a Significant increase in mortality after adjusting for differences in case mix, compared with patients of European descent. The observation of increased mortality among Asian patients admitted because of exacerbations of COPD needs to be confirmed and underlying mechanisms explored.

REFERENCES

1 Sheth T, Nair C, Nargundkar M, et al. Cardiovascular and cancer mortality among Canadians of European, south Asian and Chinese origin from 1979 to 1993: an analysis of: 1.2 CHEST /134/6/ DECEMBER, 2008

1\221

million deaths. Can Med Assoc J 1999; 161:132-138 2 Vital Statistics Agency, British Columbia Ministry of Health. Regional analysis of health statistics for status Indians in British Columbia 1992-2002. Available at: http://www.vs.gov. bc.ca. Accessed August 25, 2008 3 Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease classification system. Crit Care Med 198.5; 13:818-882 4 Horner RD, Lawler FH, Hainer BL. Relationship between patient race and survival followingadmission to intensive care among patients of primary care physicians. Health Serv Res 1991; 26:531--542 5 Williams JF, Zimmerman JE, Wagner DP, et al. AfricanAmerican and white patients admitted to the intensive care unit: is there a difference in therapy and outcome? Crit Care Med 1995; 23:626-636 6 YoungJD, Goldfrad C, Rowan K. Development and testing of a hierarchical method to code the reason for admission to intensive care units: the ICNARC Coding Method; Intensive Care National Audit & Research Centre. Br J Anaesth 2001; 87:543--548 7 Oh TE, Hutchinson R, Short S, et al. Verification of the acute physiology and chronic health evaluation scoring system in a Hong Kong intensive care unit. Crit Care Med 1993; 21:698-705 8 Kaplan JB, Bennett T. Use of race and ethnicity in biomedical publication. JAMA 2003; 289:2709-2716 9 Wenner J, Norena M, Khan N, et al. Reliability of ICU admitting and comorbid diagnoses, APACHE II score, and race in an electronic ICU datebase [abstract]. Am J Respir Crit Care Med 2007; 175:A219 10 Li P. Ethnic inequality in a class society. Toronto, ON, Canada: Thompson Educational Publishing, 1988 11 Dunn JR, Dyck I. Social determinants of health in Canada's immigrant population: results from the National Population Health Survey. Soc Sci Med 2000; 51:1573-1593 12 Latour J, Lopez V, Rodriguez M, et al. Inequalities in health in intensive care patients. J Clin Epidemiol1991; 44:889-894 13 Hutchings A, Raine R, Brady A, et al. Socioeconomic status and outcome from intensive care in England and Wales. Med Care 2004; 42:943-951 14 O'Campo P. Invited commentary: advancing theory and methods for multilevel models of residential neighborhoods and health. Am J Epidemiol 2003; 157:9-13 15 Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health 1992; 82:703-710

1222

16 Carr-Hill R, Rice N. Is enumeration district level an improvement on ward level analysis in studies of deprivation and health? J Epidemiol Community Health 1995; 49(suppl 2): S28-S29 17 Acton KJ, Burrows NR, Moore K, et al. Trends in diabetes prevalence among American Indian and Alaska native children, adolescents, and young adults. Am J Public Health 2002; 92:1485-1490 18 Sewell JL, Malasky BR, Gedney CL, et al. The increasing incidence of coronary artery disease and cardiovascular risk factors among a Southwest Native American tribe: the White Mountain Apache Heart Study. Arch Intern Med 2002; 162:1368-1372 19 Tonelli M, Hemmelgarn B, Manns B, et al. Death and renal transplantation among Aboriginal people undergoing dialysis. Can Med Assoc J 2004; 171:577--582 20 Barnato AE, Alexander SL, Linde-Zwirble WT, et al. Racial variation in the incidence, care, and outcomes of severe sepsis: analysis of population, patient, and hospital characteristics. Am J Respir Crit Care Med 2008; 177:279-284 21 Moran JL, Bristow P, Solomon PJ, et al. Australian and New Zealand Intensive Care Society Database Management Committee (ADMC). Mortality and length-of-stayoutcomes, 19932003, in the binational Australian and New Zealand intensive care adult patient database. Crit Care Med 2008; 36:46·-61 22 Rapoport J, Teres D, Steingrub J, et al. Patient characteristics and ICU organizational factors that influence frequency of pulmonary artery catheterization. JAMA 2000; 283:2.559-2567 23 Jayes RL, Zimmerman JE, Wagner DP, et al. Variations in the use of do-not-resuscitate orders in ICUs: findings from a national study. Chest 1996; 110:1332-1339 24 Chen ZM, Xu Z, Collins R, et al. Early health effects of the emerging tobacco epidemic in China: a 16-year prospective study. JAMA 1997; 78:1500-1504 25 Chan-Yeung M, Ait-Khaled N, White N, et al. The burden and impact of COPD in Asia and Africa. Int J Tuherc Lung Dis 2004; 8:2-14 26 Kim DS, Kim YS, Jung KS, et al. Korean Academy of Tuberculosis and Respiratory Diseases. Prevalence of chronic obstructive pulmonary disease in Korea: a population-based spirometry survey. Am J Respir Crit Care Med 200.5; 172: 842-847 27 Samarasinghe D, Goonaratna C. Tobacco related harm in South Asia. BMJ 2004; 328:780 28 Ho LI, Ham HJ, Chen CJ, et al. Polymorphism of the 13(2)-adrenoceptorin COPD in Chinese subjects. Chest 2001; 120:1493-1499

Original Research