© The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org.
Brief Report |
Race and Ethnicity: Comparing Medical Records to Self-Reports
Affiliations of authors: Research and Evaluation Department, Kaiser Permanente Southern California, Pasadena, CA (CNW, AMG, I-LAL); Center for Health Studies, Group Health Cooperative, Seattle, WA (SMG); Center for Health Research, Kaiser Permanente Northwest, Portland, OR (ELH); Department of Ambulatory Care and Prevention, Harvard Pilgrim Health Care, Boston, MA (MBB, LN, SWF); Department of Medicine, University of Washington, Seattle, WA (JGE); HealthPartners Research Foundation, Minneapolis, MN (SR); Division of Research, Kaiser Permanente Northern California, Oakland, CA (AA, LJH); Department of Medical Oncology, Dana Farber Cancer Institute, Boston, MA (KME)
Correspondence to: Carmen N. West, MS, MHA, USC/Comprehensive Cancer Center, 1441 Eastlake Ave., Room 3429, Los Angeles, CA 90033 (e-mail: carmenwe{at}usc.edu).
 |
ABSTRACT |
|---|
 TopNotes Abstract References |
|---|
Understanding and eliminating health disparities requires accurate data on race/ethnicity. To assess the quality of race/ethnicity data, we compared medical record classifications to self-report of a study of prophylactic mastectomy. A total of 788 women had race/ethnicity from both sources; 69.9% were 55 years of age or older, 38.3% were at least college graduates, and 67.8% were married or living with someone. There were 817 race/thnicity classifications for the 788 women, of which 758 (92.3%) were identical in the medical record and self-report. Sensitivity and positive predictive value were high (86.7%–97.2%) for whites, Asians, and blacks and moderate (64.0% and 68.1%) for Latinas. However, only one of 18 Native Americans was correctly identified in her medical record. Our results indicate that even if the overall accuracy of medical record classifications for race/ethnicity is high, such a finding may obscure substantial inaccuracies in the recording for racial/ethnic minorities, especially Latinas and Native Americans.
To understand and eliminate health disparities requires accurate data on race/ethnicity. In "Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care," the Institute of Medicine noted that, "From a public health standpoint, racial and ethnic disparities in healthcare threaten to hamper efforts to improve the nation's health" (1). In a separate report on cancer care, the institute noted that, "individuals who ... are members of racial or ethnic minority groups tend to have poorer cancer outcomes (2). Large clinical, health system, and government databases provide potentially useful sources for evaluating the effect of race/ethnicity on health and health care, yet the accuracy of race/ethnicity data from these sources is not clear (1,3). Several studies indicate important limitations in race/ethnicity drawn from automated databases (4–7). The purpose of this analysis was to compare patient self-report of race/ethnicity to medical record classification.
We used data from a prophylactic mastectomy study conducted under the auspices of the National Cancer Institute–funded Cancer Research Network, a consortium of research organizations affiliated with nonprofit integrated health care delivery systems. The goal of the Cancer Research Network is to increase the effectiveness of preventive, curative, and supportive interventions that span the natural history of major cancers through a collaborative research program within defined health system populations. The six systems participating in this study include the Group Health Cooperative, WA; Harvard Pilgrim Health Care, MA; HealthPartners, MN; and three Kaiser Permanente regions: Northwest (Oregon) and Northern and Southern California. Institutional review board approval was received at all sites, in accordance with assurances filed with and approved by the U.S. Department of Health and Human Services. The requirement for informed consent was waived for medical record abstraction, and consent was implied by return of the survey.
The study details have been published previously (8,9). All women members of the health systems aged 18–80 years with a contralateral or bilateral prophylactic mastectomy from 1979 to 1999 were included, as well as a random sample of women without prophylactic mastectomy. Women were identified and classified from automated enrollment, ambulatory care, hospitalization, and cancer registry data. Eligibility was confirmed through medical record abstraction.
We abstracted 1704 women's medical records; 1662 women had at least one noted race/ethnicity. The survey was mailed to 1202 living women with valid addresses; 812 (67.6%) responded, and 805 of these indicated at least one race/ethnicity. A total of 788 women had race/ethnicity information from both medical records and self-report. Of these 788 women, 113 (14.3%) had bilateral prophylactic mastectomy, 549 (69.7%) had contralateral prophylactic mastectomy, and 126 (16.0%) had neither.
We abstracted medical records using a standardized instrument, trained abstractors, and validated quality control program (10). Abstractors coded race/ethnicity from any source in the record into these categories: Asian/Pacific Islander/Filipina, African American/black, Latina/Hispanic, Native American, Caucasian/white, Other, and Unknown. If the medical record included multiple race/ethnicities in single entries, the abstractors coded all race/ethnicities. If, however, the racial/ethnic notations varied over multiple entries, the abstractors coded the most frequently recorded race/ethnicity.
Self-reported race/ethnicity was collected through a mailed survey. Nonrespondents to the initial mailing were sent a second mailing after 3 weeks and received a telephone reminder call 3 weeks following the second mailing. Women were asked to identify their race/ethnicity in response to this question: "To what race/ethnic group do you belong?" Women were instructed to select all that apply. The response categories were Asian or Pacific Islander, black or African American, Hispanic/Latina, Native American or Alaska Native, white or Caucasian, and Other.
We used self-reported race/ethnicity as the criterion standard because this is commonly accepted in the U.S. census and other settings as an accurate reflection of race/ethnicity, whereas medical record data are often based on the provider or staff observation and inference. When women reported or were noted in the medical record to have multiple race/ethnicity classifications, we used all classifications in the analysis. We calculated sensitivity and positive predictive value, with 95% exact confidence intervals, for each race/ethnicity classification. In addition, we also calculated kappa coefficients of agreement comparing the classification of whites to all nonwhites and of all federally designated minorities (blacks, Latinas, and Native Americans) to all others.
There were 817 race/ethnicity classifications for the 788 women, of which 758 (92.3%) classifications were identical in the medical record and self-report. Among the 788 women for whom we had race/ethnicity information from both medical records and self-report, 69.9% were 55 years of age or older, 38.3% were at least college graduates, and 67.8% were married or living with someone (Table 1). Sensitivity and positive predictive value were high (86.7%–97.2%) for whites, Asians, and blacks and were moderate (64.0% and 68.1%) for Latinas (Table 2). However, only 1 of 18 Native Americans was correctly identified in her medical record, and 15 were misclassified as being white.
|
|
When grouping women into white and nonwhite, sensitivity was 97.2% (95% confidence interval [CI] = 95.7% to 98.3%) and the kappa coefficient was 75.2%. However, when comparing federally designated minority groups to nonminority groups, sensitivity was poor (63.9%; 95% CI = 53.5% to 73.4%) and the kappa dropped to 66.5%. This contradiction was because most women in the study were white and the medical record correctly recorded them as such.
Our results generally are consistent with prior studies, although we found higher agreement in classifications of Asians. In a study of Veterans Administration dental outpatients, Boehmer et al. reported high correct classification of whites and blacks, modest correct classification of Latinos, and poor classification of Asians (6). In a broader and larger Veterans Administration study, Kressin et al. reported high agreement for whites and blacks and lower agreement for Asians and Hispanics (5). A number of other studies have reported challenges in race/ethnicity classification from various types of medical data, typically showing high accuracy for whites, intermediate accuracy for blacks, and lower accuracy for Asians and other groups (3–4,7,11).
Strengths of our study include a population of women in a community practice, whereas previous studies used select samples. The abstractors in our study underwent an extensive training and quality control process (8,9), and in a paper describing a sister study using the same staff, quality of abstraction was found to be high (10). We also had ongoing quality assurance, which captured and resolved inconsistencies on a regular basis. We allowed women to self-report multiple classifications on the survey. Women with a prophylactic mastectomy may not be representative of the general population, and recording of race/ethnicity for them may be different. Although our nonwhite groups accounted for only about 16% of the study population, this proportion is common in research and clinical practice.
Our results indicate that although the overall accuracy of medical record classifications for race/ethnicity is high, such a finding may obscure substantial inaccuracies in the recording for racial/ethnic minorities, especially Latinas and Native Americans. Inaccurate data on race/ethnicity can lead to erroneous conclusions in research and inadequate health care planning. Misclassifying individuals as white may mask health disparities, including differences in cancer care and outcomes, and prevent or delay efforts to understand and address the needs of specific subgroups. To support efforts to eliminate health disparities caused by race/ethnicity, self-reported classifications are preferable to medical record data.
|
NOTES |
|---|
This study was funded by the National Cancer Institute (5U19 CA079689, Increasing Effectiveness of Cancer Control Interventions, Edward H. Wagner, MD, MPH, Principal Investigator and 1R01 CA090323, Patient-Oriented Outcomes of Prophylactic Mastectomy, Ann M. Geiger, PhD, Principal Investigator).
|
REFERENCES |
|---|
|
TopNotesAbstractReferences |
|---|
(1) Smedley BD, Stith AY, Nelson AR, editors. Unequal treatment: confronting racial and ethnic disparities in health care. Washington (DC): The National Academies Press; 2003.
(2) Hewitt M, Simone JV, editors. Ensuring quality cancer care. Washington (DC): The National Academies Press; 1999.
(3) Pan CX, Glynn RJ, Mogun H, Choodnovskiy I, Avorn J. Definition of race and ethnicity in older people in Medicare and Medicaid. J Am Geriatr Soc 1999;47:730–3.[ISI][Medline]
(4) Gomez SL, Kelsey JL, Glaser SL, Lee MM, Sidney S. Inconsistencies between self-reported ethnicity and ethnicity recorded in a health maintenance organization. Ann Epidemiol 2005;15:71–9.[CrossRef][ISI][Medline]
(5) Kressin NR, Chang BH, Hendricks A, Kazis LE. Agreement between administrative data and patients' self-reports of race/ethnicity. Am J Public Health 2003;93:1734–9.
(6) Boehmer U, Kressin NR, Berlowitz DR, Christiansen CL, Kazis LE, Jones JA. Self-reported vs. administrative race/ethnicity data and study results. Am J Public Health 2002;92:1471–2.
(7) Blustein J. The reliability of racial classifications in hospital discharge abstract data. Am J Public Health 1994;84:1018–21.
(8) Herrinton LJ, Barlow WE, Yu O, Geiger AM, Elmore JG, Barton MB, et al. Efficacy of prophylactic mastectomy in women with unilateral breast cancer. J Clin Oncol 2005;23:4275–86.
(9) Geiger AM, Yu O, Herrinton LJ, Barlow WE, Harris EL, Rolnick S, et al. A population-based study of bilateral prophylactic mastectomy efficacy in women at elevated risk for breast cancer in community practices. Arch Intern Med 2005;165:516–20.
(10) Reisch LM, Fosse JS, Beverly K, Yu O, Barlow WE, Harris EL, et al. Training, quality assurance, and assessment of medical record abstraction in a multisite study. Am J Epidemiol 2003;157:546–51.
(11) Hahn RA, Mulinare J, Teutsch SM. Inconsistencies in coding race and ethnicity between birth and death in US infants: A new look at infant mortality, 1983 through 1985. JAMA 1992;267:259–63.[Abstract]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||