180 likes | 303 Views
Quality of Race and Hispanic Origin Reporting on Death Certificates in the US. Elizabeth Arias, Ph.D. Mortality Statistics Branch Division of Vital Statistics National Center for Health Statistics. Outline. The Problem
E N D
Quality of Race and Hispanic Origin Reporting on Death Certificates in the US Elizabeth Arias, Ph.D. Mortality Statistics Branch Division of Vital Statistics National Center for Health Statistics
Outline • The Problem • Death rates for some race and Hispanic origin groups appear questionable • Data Sources • The Data Collection Process • Evaluation of the Problem • The National Longitudinal Mortality Study • Self Reported Race/Ethnicity vs. Proxy Reported Race/Ethnicity • Change in Data Quality over two decades: 1980s – 1990s • Applications • Use information to adjust observed mortality estimates • Does adjustment have an effect on race/ethnic mortality disparities?
The Problem • Numerator of rate • Death Certificate=race and Hispanic origin identification is recorded by funeral director as per family member report or, often, as result of his/her observation of decedent. • Denominator of rate • Decennial Census=race and Hispanic origin identification is self-reported or reported by family/household member while individual is still alive.
Evaluation of Problem • The National Longitudinal Mortality Study • Current Population Surveys (CPS) & 1980 Census Data linked, using the NDI, to NCHS Mortality Data • 26 Annual March CPS (1973, 1979-1998) and 1980 US Census sub-sample (~2.3 million records) • CPS: National Sample of Households (60,000-80,000), Non-Institutionalized Population, 96% Response Rate • Deaths Occurring between 1979-1998 (~253,000) • Provides Ability to Compare Race and Hispanic Origin Reported by an Individual While Alive to What Was Reported on their Death Certificate upon Death.
Evaluation of Problem - Method • Classification Ratios: Ratios of CPS race and Hispanic origin counts to death certificate counts for sample of identified decedents; basically, a ratio of row to column totals in a bivariate table of CPS (row) by Death Certificate (column) classification • Estimate Classification Ratios for two time periods: 1980s and 1990s to assess change over time in quality • Examples: • White Ratio = Total CPS White / Total DC White • Black Ratio = Total CPS Black / Total DC Black • AIAN Ratio = Total CPS AIAN / Total DC AIAN • API Ratio = Total CPS API / Total DC API
Classification Ratios Estimated by Selected Decedent Characteristics • Information Derived From Death Certificate • Sex • Age (10-year age categories, exception first and last [0-14, 85+] • Nativity (US Born, Foreign Born) • Rural/Urban Status (rural or urban residence at time of death) • Region of Residence (Northeast, Midwest, South, West residence at time of death) • Co-ethnic Concentration (dichotomous variable indicating whether decedent died in county with high concentrations of co-ethnics)
Effect of Death Certificate Race and Hispanic Origin Misclassification on Mortality Measures • Data • pooled number of deaths 1999-2001 from NVSS • bridged April 1, 2000 population census estimates for race groups • Census 2000 Summary File 2 based on April 1, 2000 Census for Hispanic sub-groups • Method • Observed Age Specific Death Rages (ASDRi) and Age Adjusted Death Rates(AADR) • ASDR and AADR adjusted for age-specific classification ratios
Classification Ratios for Race Categories, by Period1979-89 and 1990-98
Classification Ratios for Hispanic Categories, by Period 1979-89 and 1990-98
Effect of Correction on Age Adjusted Death Rates • Observed and Corrected Age Adjusted Death Rates and Percentage Relative to White • AIAN • Observed • 718.0 85% of White AADR • Adjusted • 943.2 110% of White AADR • API • Observed • 510.4 60% of White AADR • Adjusted • 540.1 64% of White AADR
Effect of Correction on Age Adjusted Death Rates • Observed and Corrected Age Adjusted Death Rates and Percentage Relative to Non Hispanic White • Hispanic • Observed • 672.1 79% • Adjusted • 707.1 83% • Mexican • Observed • 753.7 88% • Adjusted • 746.3 88% • Puerto Rican • Observed • 798.1 93% • Adjusted • 819.8 96%
Effect of Correction on Age Adjusted Death Rates • Cuban • Observed • 692.8 81% • Adjusted • 694.6 82% • Central/South American • Observed • 743.1 87% • Adjusted • 589.6 69% • Other Hispanic • Observed • 381.8 45% • Adjusted • 650.7 76%
Summary • 1) Agreement between self-report and DC report continues to be excellent for the white and black populations, but not for other groups • 2) However, there has been some improvement over the 2 decades for most groups • 3) This study revealed important relationships between decedents’ characteristics and the quality of race and ethnic reporting on DC • AIAN and Hispanic decedents were significantly more likely to be correctly classified in areas of high co-ethnic concentration • Foreign born Hispanic decedents were more likely to be correctly classified
Summary, Continued • Death Rates are profoundly affected by the quality of reporting on DC • Most glaring example is the case of AIAN mortality estimates • The Hispanic Mortality Advantage (Hispanic Mortality Paradox) is barely affected • Hispanic mortality remains almost 20% lower than that of Non-Hispanic whites • Answer to the Hispanic Mortality Paradox lies elsewhere
Contact Information:Elizabeth Arias EArias@cdc.gov301-458-4727