Incidence of Atherosclerosis by Race in the Dialysis Morbidity and Mortality Study: A Sample of the US ESRD Population : Journal of the American Society of Nephrology

Journal Logo

Clinical Nephrology

Incidence of Atherosclerosis by Race in the Dialysis Morbidity and Mortality Study

A Sample of the US ESRD Population

Parekh, Rulan S.*, †; Zhang, Lin; Fivush, Barbara A.*; Klag, Michael J.†, ‡, §

Author Information
Journal of the American Society of Nephrology 16(5):p 1420-1426, May 2005. | DOI: 10.1681/ASN.2004080661
  • Free


Black individuals have higher total mortality than white individuals (1). Reasons for this disparity are unclear but may relate to underlying disparities in access to health care, income, education, diet, lifestyle, and increased prevalence of comorbid conditions (24). In contrast to the pattern seen in the general population, however, blacks who are on dialysis have a much lower overall mortality (179 per 1000 person-years) than whites (279 per 1000 person-years) (5). Understanding why blacks who are on dialysis have a lower death rate than whites who are on dialysis may yield insights into how to reduce the excess risk in blacks seen in the general population (612).

Cardiovascular disease (CVD) mortality is the leading cause of death in people who are treated with dialysis, at least 10-fold higher than in the general population (13), and may explain some of the paradoxic association of race with mortality in ESRD patients. Although CVD mortality has been reported to be 20% lower in black than in white ESRD patients (12,14), death from nonatherosclerotic forms of heart disease, such as cardiomyopathy and arrhythmias, was included in the definition of CVD. Furthermore, 40% of ESRD patients have CVD at the initiation of dialysis (15), and the risk for recurrent atherosclerotic CVD (ASCVD) events is likely to be much higher in people with prevalent disease. Thus, small differences in the prevalence of ASCVD may explain the observed differences in whites compared with blacks. Previous studies (1517) that have adjusted for previous CVD have used data from the Center for Medicaid and Medicare Medical Evidence report, which have been shown to underreport preexisting conditions, especially CVD and smoking (18). Also, it is important to know whether the incidence of new ASCVD events differs by race to understand better ASCVD in the dialysis population. The aim of our study was to determine whether the incidence of new and recurrent ASCVD differs in whites and blacks in the US dialysis population using detailed patient information to characterize both prevalent ASCVD and CVD risk factors.

Materials and Methods

Study Design

We performed a longitudinal cohort study of US incident and prevalent ESRD patients and determined the incidence of ASCVD events in a 3-yr period. Patients were excluded from the study when they were younger than 18 yr, had a previous kidney transplant, or were on home hemodialysis. Patients were censored at death, transplantation, or 3 yr after enrollment in the study.

Data Sources

The study population was drawn from the Dialysis Morbidity and Mortality Special Study (DMMS) of the United States Renal Data System (USRDS). DMMS is an observational, prospective study of a random sample of the US dialysis population collected in four different waves. Patients in waves 1, 3, and 4 were drawn from a random sample of 25% of the hemodialysis facilities in the United States. The units were chosen randomly from the 550 dialysis units on the Medicare master list as of December 31, 1993. In wave 2, sampling included additional, new hemodialysis units that had been added to the Medicare master list after January 1, 1994, and peritoneal dialysis units. Patients in waves 1, 3, and 4 of the cohort entered the study on December 31, 1993, and were followed until December 31, 1996. In wave 2, patients entered the study on October 16, 1996, and were followed until September 23, 2000.

A total of 20,757 patients were enrolled in all four DMMS waves; 2696 patients were excluded because of missing USRDS identification or duplicative and/or incomplete information in the hospitalization file, leaving a total of 18,061 patients for analysis. Data are presented for the 16,103 patients of black or white race in the study.

Data were abstracted from the patients’ medical records by trained chart abstractors from the local ESRD network using standardized data forms developed by the USRDS Coordinating Center. Data abstracted included patients’ demographic and socioeconomic characteristics, comorbidities, laboratory values, dialysis modality, and dialysis adequacy. Supplemental information on race, gender, primary cause of ESRD, onset of dialysis, hospitalization, and mortality were obtained from USRDS. Data collected for DMMS and the USRDS have been previously validated (5,19).


Patients were stratified by prevalent ASCVD at enrollment into the study. Patients were defined as having prevalent ASCVD when the following conditions were listed as previous, present, or suspected on the DMMS initial abstraction form: Coronary heart disease, coronary artery disease, myocardial infarction, coronary artery bypass surgery, coronary angioplasty, abnormal coronary angiogram, cardiac arrest, cerebrovascular accident, peripheral vascular disease (PVD), amputation as a result of PVD, claudication, or transient ischemic attacks. Repeat analyses were conducted in people with suspected ASCVD defined as not having prevalent ASCVD, and this did not change the results (data not shown).

Other variables included in our analysis were age group (18 to 35, 36 to 50, 51 to 65, and >65 yr); race (white, black, and other); gender; ever smoking; diabetes; cause of ESRD (hypertension, diabetes, glomerulonephritis, cystic kidney disease, other urologic causes, other causes, and unknown); dialysis duration; current dialysis modality; body mass index; left ventricular hypertrophy (LVH); average monthly intradialytic weight loss; urea reduction ratio (URR); pulse pressure (average predialysis systolic BP − diastolic BP); and levels of total cholesterol, triglycerides, blood hemoglobin, and albumin. The product of serum calcium and phosphorus levels was also included. Calcium levels <4 mg/dl were corrected for low albumin levels (corrected calcium = serum calcium + 0.8*(4 − mean albumin) (20). LVH was determined by electrocardiogram or echocardiography. All variables were obtained at baseline entry into DMMS. Values for continuous variables that were out of measurable range in laboratory analyses were treated as missing. For continuous variables for which <10% were missing, missing values were set to the mean of the entire data set. For categorical variables, a missing category was created for LVH and smoking and included in the Cox regression analyses. No data were available on history of hypertension or family history of early coronary disease.


The primary outcome was new or recurrent ASCVD defined as the first ASCVD event after enrollment. Person-time was defined as the period from enrollment into DMMS to the date of first ASCVD event or date of censoring. ASCVD included coronary heart disease (CHD), PVD, and cerebrovascular disease (CVA). CHD consisted of acute myocardial disease; coronary artery bypass grafting; coronary angioplasty; or mortality from myocardial infarction, cerebrovascular disease, and sudden death. PVD was defined as amputation, abdominal aortic aneurysm, or iliac-femoral bypass. Last, CVA consisted of ischemic stroke or carotid endarterectomy. ASCVD events were assessed using Medicare data for hospitalizations and procedures and Health Care Financing Administration death notification forms. Only ASCVD events listed in the first two discharge diagnoses were used to define outcomes. Codes from the International Classification of Diseases, Ninth Revision, Clinical Modification were used to classify ASCVD outcomes.

Statistical Analyses

Analyses were stratified by prevalent ASCVD and race. For continuous variables, a t test or nonparametric tests were used for comparison of means and medians, respectively. χ2 or Fisher exact test was done to compare categorical variables. All continuous variables were categorized according to their distributions or kept as continuous variables.

Incidence of new or recurrent ASCVD, overall, and three subtypes of ASCVD (CHD, CVA, and PVD) was calculated by person-time and corresponding 95% confidence intervals (CI), assuming a Poisson distribution (21). The Kaplan-Meier method was used to generate survival plots, and the log rank test was used to test differences in cumulative incidence between whites and blacks. Cox proportional hazards regression analysis was used to assess whether the association of race with ASCVD incidence was independent of traditional ASCVD risk factors and dialysis-related factors that have been associated with ASCVD risk. Because prevalent patients were included in this analysis, we also performed Cox regression analyses stratified by duration of dialysis to determine whether any differences in risk of ASCVD between blacks and whites were due to survival after initiation of dialysis. Statistical significance was defined as P < 0.05 using two-tailed tests. Assumptions of the Cox model were tested and met. All analyses were done using SAS version 8.0 (SAS, Inc., Cary, NC).


Among a total of 16,103 patients, 8544 had prevalent ASCVD disease at the start of the study. Table 1 compares demographic and clinical variables by prevalent ASCVD and race groups. With or without prevalent ASCVD, whites were older, more likely to be male, less likely to have diabetes, and leaner. They also had a shorter duration on dialysis; higher pulse pressure, triglycerides, and hemoglobin levels; greater URR; and lower prevalence of LVH than blacks. Only 87 patients had HIV disease, 81 of whom were black and six of whom were white, in all four waves of the study. During the 3 yr of follow-up, 1012 whites and 420 blacks were censored for transplantation.

Table 2 shows the incidence rates of total ASCVD, CHD, CVA, and PVD by prevalent ASCVD, per 1000 person-years, with corresponding 95% CI. Mean follow-up time was 18.5 mo for whites and 21.9 mo for blacks. In people with and without prevalent ASCVD at baseline, the incidence of total ASCVD events was higher in whites. Whites also had a higher incidence than blacks of CHD, CVA, and PVD except in persons without a history of ASCVD at baseline, for whom the incidence of CVA was higher in blacks than in whites but not significantly so. In both whites and blacks, the incidence of recurrent events was approximately three times higher in people with ASCVD at baseline than the incidence of new ASCVD in patients with no history of ASCVD.

Figure 1 shows the incidence of ASCVD among blacks and whites during the 3 yr of follow-up, by prevalent ASCVD at baseline. Results of this analysis are similar to those seen using the person-years approach. Whites had a higher incidence of ASCVD than blacks throughout follow-up, independent of the presence of ASCVD at baseline (P < 0.0001 by log rank). The effect of prevalent ASCVD on subsequent ASCVD risk was of greater magnitude than race, with a much higher incidence of subsequent ASCVD events in those with prevalent disease (P < 0.0001 by log rank).

In Table 3, the crude and adjusted hazard ratios (HR) for an ASCVD event among whites compared with blacks, by prevalent ASCVD, are shown. In subjects with previous ASCVD, whites were 1.25 times more likely than blacks to develop a recurrent ASCVD event and 1.37 times more likely to have a coronary heart disease event, after adjustment for potential confounders. Whites also had a higher risk for CVA than blacks, but this HR was not significantly different from 1.0 in multivariate analyses. In patients without prevalent ASCVD, whites were 35% more likely to develop new ASCVD than blacks, with similar risk estimates for the development of CHD and CVA after multivariate adjustment. The CI around the HR for CVA, however, overlapped 1.0. In contrast, whites were not more likely to experience a PVD event compared with blacks, regardless of whether ASCVD was present at enrollment.

Similar analyses stratified by duration of dialysis are shown in Table 4. For patients who were on dialysis for <6 mo duration, whites were 1.42 times more likely to have recurrent ASCVD and 1.2 times more likely to have new ASCVD than blacks. Higher risk for recurrent ASCVD among whites compared with blacks was present for subjects who were on dialysis for up to 12 mo duration. Beyond 12 mo of dialysis duration, the HR were smaller in magnitude and the CI overlapped 1.0. Among patients with no previous ASCVD, there was also higher risk for new ASCVD events in whites than blacks that declined somewhat with longer duration on dialysis.


This national study of US dialysis patients demonstrates that whites on dialysis have at least a 25% greater risk for developing new or recurrent ASCVD events than blacks after adjustment for traditional cardiovascular and dialysis-related risk factors. This was true even when analyses were confined to subjects who were on dialysis for <6 mo. This excess risk was seen even though whites had a more favorable CVD risk factor profile than blacks, except for older age. Not unexpected, event rates were much higher for recurrent than for new ASCVD; the 3-yr incidence of recurrent ASCVD was >60%, and new ASCVD was >30%. Whites, however, had a higher ASCVD rate than blacks independent of the presence of ASCVD at baseline. Most of the difference between whites and blacks was due to a higher incidence of CHD in whites.

Our findings show that the incidence of new ASCVD in our study population is 30 times higher for blacks and 40 times higher for whites than in the general population (22). The tremendous burden of CHD in dialysis patients without ASCVD at enrollment, approximately 13% annually in whites and 10% in blacks, more than meets the National Cholesterol Expert Panel (Adult Treatment Panel III) definition of the highest risk group. Furthermore, the risk for recurrent ASCVD events in the dialysis population is five times greater than reported in the general Medicare population (23). These findings suggest a greatly accelerated rate of primary and secondary ASCVD in the dialysis population. All ESRD patients, regardless of race, should be considered as a high-risk group that requires special considerations for management of CVD similar to individuals with diabetes.

Previous studies have reported cardiovascular mortality with both new and recurrent events combined, which does not describe accurately the incidence of ASCVD in the dialysis population because of confounding by case mix. These studies also were not able to adjust for risk factors for mortality in the dialysis population and included nonatherosclerotic forms of CVD, such as congestive heart failure, cardiomyopathy, arrhythmias, valvular disease, and pericarditis in the outcome definition. With these caveats, findings of the present study mirror those of cardiovascular mortality outcomes (12,24). White ESRD patients in the United States experience up to 30% higher cardiovascular mortality than blacks (12,14,17). Our study extends these previous results to a more specific atherosclerotic disease outcome—incidence of nonfatal as well fatal events—and consideration of both traditional cardiovascular and dialysis-related risk factors.

Our findings are in direct contrast to the general population, in which whites have lower cardiovascular mortality than blacks. Consistent with the general population, however, CHD was the most common manifestation of atherosclerosis in both blacks and whites. CVA was the least common form of ASCVD in dialysis, and there were only marginal differences in the incidence rate of CVA by race. Because of the relatively small number of people who experienced stroke, however, power was limited to detect differences between whites and blacks.

Reasons for the shift to increased risk in whites compared with blacks after reaching ESRD are unclear, but improvement in access to health care may be one factor. In the general population, whites have better access to care than blacks (25). Lower access to health care before initiation of dialysis may have resulted in higher mortality, thus selecting a healthier population of blacks who do better than whites after treatment for ESRD commences. Once ESRD occurs, health care is provided by Medicare entitlement, so disparities in access to care are reduced. For example, use of cardiovascular procedures is similar in whites and black who are treated for ESRD (26). Access to care, however, is clearly not the only explanation. Mortality among white dialysis patients is also greater than among Asian or black dialysis patients in Canada, which has a national health care system with equal access to care (7).

Biologic differences between white and black dialysis patients may also account for some of the increased risk for ASCVD among whites. Potential factors include differences in lipid metabolism, inflammation, malnutrition, and dialysis dose. In the present analysis, the observed relationship was independent of serum triglycerides, body mass index, and serum albumin. Serum albumin is a measure of both nutrition and inflammation. We did not have data, however, on HDL cholesterol or Lp(a). HDL cholesterol is protective against atherosclerosis and tends to be higher in blacks than in whites, both in the general population and in people with ESRD (27). Thus, HDL cholesterol may have contributed to the lower risk seen in blacks. Lp(a), an atherogenic lipoprotein, is higher in blacks than in whites who are on dialysis, (28,29) and thus cannot explain the lower risk in blacks. More precise measures of inflammation, such as IL-6 and C-reactive protein, have not been studied systematically by race in the dialysis population and may be further contributing factors.

The strengths of this analysis include the large sample, which is representative of the US ESRD population. The large sample size permitted analyses stratified by duration of dialysis and demonstrated increased risk in both incident and prevalent dialysis patients. The comorbid conditions and risk factors were obtained from chart review, not from the Center for Medicaid and Medicare Medical Evidence Form, which underestimates prevalence of risk factors and comorbidity (18). In addition, we were able to adjust for risk factors that are unique to dialysis care, such as URR, that have been shown to be associated with increased mortality in the dialysis population. Our study also used “hard” outcomes such as surgical intervention; angioplasty; or a primary discharge diagnosis for acute myocardial infarction, stroke, and PVD to prevent overestimating the incidence of ASCVD. We did not use codes for angina, ischemia, congestive heart failure, intermittent claudication, or hemorrhagic stroke, which depend on either self-report of symptoms or may be difficult to diagnose accurately in the ESRD population. The codes used for CHD and CVA have been previously validated with an agreement of >80% (23,3032).

As detection of hospitalizations was based on Medicare billing, people with another primary insurance would not have been captured in our analyses, and our findings may be strictly generalizable only to the Medicare population. As noted previously, the cardiovascular risk factors included in our analyses were limited to traditional risk factors obtained through history or chart review at baseline entry into the study; thus, we were not able to study other potential risk factors, such as Lp(a) or homocysteine. Information on family history, history of hypertension, and medication use was also not available. Differences between blacks and whites in these unmeasured variables may have contributed to the observed differences in ASCVD risk and explain the similarity of risk estimated in both the unadjusted and adjusted analyses.

Whites on dialysis have a higher incidence of ASCVD than blacks on dialysis, and this difference is not explained by baseline ASCVD or dialysis-related risk factors. Greater risk for both new and recurrent ASCVD likely contributes to the higher risk for total mortality in whites than blacks in this population. Also, the incidence of new and recurrent ASCVD in both blacks and whites in the US dialysis population is significantly higher than the published rates of atherosclerotic disease in the general population. The dialysis population should be thought of as at very high risk for atherosclerotic disease and treated accordingly. Whites in particular are at greatest risk for ASCVD, and future studies to determine the underlying mechanisms for this differential risk are needed to diminish the rate of accelerated ASCVD in this vulnerable population.

Figure 1:
Incidence of new or recurrent atherosclerotic cardiovascular disease (ASCVD) over 3 yr of follow-up, by race and prevalent ASCVD in the US dialysis population (1993–1996).
Table 1:
Baseline characteristics of 16,103 dialysis patients by prevalent ASCVD and race (1993–1996)
Table 2:
Incidence rate of new or recurrent ASCVD, per 1,000 person-years, in 16,103 dialysis subjects after 3 years of follow-up stratified by prevalent ASCVD and race
Table 3:
Hazard ratio of new or recurrent ASCVD by prevalent ASCVD among whites compared to blacks in the US dialysis population (1993–1996)
Table 4:
Hazard ratio of new or recurrent ASCVD by prevalent ASCVD among whites compared to blacks stratified by duration of dialysis in the US dialysis population (1993–1996)


R.S.P. was supported by the Carl W. Gottschalk Research Scholar Grant/American Society of Nephrology and the National Institute of Diabetes and Digestive and Kidney Diseases 1K23DK02872-01A1. M.J.K. was supported by 5 K24 DK02856.

Published online ahead of print. Publication date available at

The data reported here have been supplied by the United States Renal Data System. The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy or interpretation of the US government.


1. Mortality patterns—United States, 1997. MMWR Morb Mortal Wkly Rep 48: 664 –668, 1999
2. Asher CR, Topol EJ, Moliterno DJ: Insights into the pathophysiology of atherosclerosis and prognosis of black Americans with acute coronary syndromes. Am Heart J 138: 1073 –1081, 1999
3. Davey Smith G, Neaton JD, Wentworth D, Stamler R, Stamler J; for the MRFIT Research Group: Mortality differences between black and white men in the USA: Contribution of income and other risk factors among men screened for the MRFIT. Multiple Risk Factor Intervention Trial. Lancet 351: 934 –939, 1998
    4. Cooper R, Cutler J, Desvigne-Nickens P, Fortmann SP, Friedman L, Havlik R, Hogelin G, Marler J, McGovern P, Morosco G, Mosca L, Pearson T, Stamler J, Stryer D, Thom T: Trends and disparities in coronary heart disease, stroke, and other cardiovascular diseases in the United States: Findings of the national conference on cardiovascular disease prevention. Circulation 102: 3137 –3147, 2000
    5. USRDS 1996 Annual Data Report. Bethesda, MD, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Division of Kidney, Urologic and Hematologic Diseases, 1996
    6. Owen WF Jr, Chertow GM, Lazarus JM, Lowrie EG: Dose of hemodialysis and survival: Differences by race and sex. JAMA 280: 1764 –1768, 1998
    7. Pei YP, Greenwood CM, Chery AL, Wu GG: Racial differences in survival of patients on dialysis. Kidney Int 58: 1293 –1299, 2000
    8. Reddan DN, Szczech LA, Klassen PS, Owen WF Jr: Racial inequity in America’s ESRD program. Semin Dial 13: 399 –403, 2000
      9. Salahudeen AK, Deogaygay B, Fleischmann E, Bower JD: Race-dependent survival disparity on hemodialysis: Higher serum aluminum as an independent risk factor for higher mortality in whites. Am J Kidney Dis 36: 1147 –1154, 2000
        10. Salem M: End-stage renal disease survival in blacks and whites. Am J Med Sci 323: 100 –101, 2002
          11. Wong JS, Port FK, Hulbert-Shearon TE, Carroll CE, Wolfe RA, Agodoa LY, Daugirdas JT: Survival advantage in Asian American end-stage renal disease patients. Kidney Int 55: 2515 –2523, 1999
            12. Bloembergen WE, Port FK, Mauger EA, Wolfe RA: Causes of death in dialysis patients: Racial and gender differences. J Am Soc Nephrol 5: 1231 –1242, 1994
            13. Foley RN, Parfrey PS, Sarnak MJ: Epidemiology of cardiovascular disease in chronic renal disease. J Am Soc Nephrol 9[Suppl]: S16 –S23, 1998
            14. Herzog CA, Ma JZ, Collins AJ: Poor long-term survival after acute myocardial infarction among patients on long-term dialysis. N Engl J Med 339: 799 –805, 1998
            15. Stack AG, Bloembergen WE: Prevalence and clinical correlates of coronary artery disease among new dialysis patients in the United States: A cross-sectional study. J Am Soc Nephrol 12: 1516 –1523, 2001
            16. Abbott KC, Agodoa LY: Hospitalizations for valvular heart disease in chronic dialysis patients in the United States. Nephron 92: 43 –50, 2002
              17. Ganesh SK, Hulbert-Shearon T, Port FK, Eagle K, Stack AG: Mortality differences by dialysis modality among incident ESRD patients with and without coronary artery disease. J Am Soc Nephrol 14: 415 –424, 2003
              18. Longenecker JC, Coresh J, Klag MJ, Levey AS, Martin AA, Fink NE, Powe NR: Validation of comorbid conditions on the end-stage renal disease medical evidence report: The CHOICE study. Choices for Healthy Outcomes in Caring for ESRD. J Am Soc Nephrol 11: 520 –529, 2000
              19. Completeness and reliability of USRDS data: Comparisons with the Michigan Kidney Registry. Am J Kidney Dis 20: 84 –88, 1992
              20. Block GA, Hulbert-Shearon TE, Levin NW, Port FK: Association of serum phosphorus and calcium x phosphate product with mortality risk in chronic hemodialysis patients: A national study. Am J Kidney Dis 31: 607 –617, 1998
              21. Hollander M, Koudstaal PJ, Bots ML, Grobbee DE, Hofman A, Breteler MM: Incidence, risk, and case fatality of first ever stroke in the elderly population. The Rotterdam Study. J Neurol Neurosurg Psychiatry 74: 317 –321, 2003
              22. Keil JE, Sutherland SE, Knapp RG, Lackland DT, Gazes PC, Tyroler HA: Mortality rates and risk factors for coronary disease in black as compared with white men and women. N Engl J Med 329: 73 –78, 1993
              23. Vickrey BG, Rector TS, Wickstrom SL, Guzy PM, Sloss EM, Gorelick PB, Garber S, McCaffrey DF, Dake MD, Levin RA: Occurrence of secondary ischemic events among persons with atherosclerotic vascular disease. Stroke 33: 901 –906, 2002
              24. Herzog CA: Dismal long-term survival of dialysis patients after acute myocardial infarction: Can we alter the outcome? Nephrol Dial Transplant 17: 7 –10, 2002
              25. Kressin NR, Petersen LA: Racial differences in the use of invasive cardiovascular procedures: Review of the literature and prescription for future research. Ann Intern Med 135: 352 –366, 2001
              26. Daumit GL, Hermann JA, Coresh J, Powe NR: Use of cardiovascular procedures among black persons and white persons: A 7-year nationwide study in patients with renal disease. Ann Intern Med 130: 173 –182, 1999
              27. Avram MM, Goldwasser P, Burrell DE, Antignani A, Fein PA, Mittman N: The uremic dyslipidemia: A cross-sectional and longitudinal study. Am J Kidney Dis 20: 324 –335, 1992
              28. Astor BC, Coresh J, Powe NR, Eustace JA, Klag MJ: Relation between gender and vascular access complications in hemodialysis patients. Am J Kidney Dis 36: 1126 –1134, 2000
              29. Longenecker JC, Klag MJ, Marcovina SM, Powe NR, Fink NE, Giaculli F, Coresh J: Small apolipoprotein(a) size predicts mortality in end-stage renal disease: The CHOICE study. Circulation 106: 2812 –2818, 2002
              30. Jollis JG, Ancukiewicz M, DeLong ER, Pryor DB, Muhlbaier LH, Mark DB: Discordance of databases designed for claims payment versus clinical information systems. Implications for outcomes research. Ann Intern Med 119: 844 –850, 1993
              31. Benesch C, Witter DM Jr, Wilder AL, Duncan PW, Samsa GP, Matchar DB: Inaccuracy of the International Classification of Diseases (ICD-9-CM) in identifying the diagnosis of ischemic cerebrovascular disease. Neurology 49: 660 –664, 1997
                32. Tirschwell DL, Longstreth WT Jr: Validating administrative data in stroke research. Stroke 33: 2465 –2470, 2002
                Copyright © 2005 The Authors. Published by Wolters Kluwer Health, Inc. All rights reserved.