Quality of death rates by race and Hispanic origin: a summary of current research, 1999.

OBJECTIVES: This report provides a summary of current knowledge and research on the quality and reliability of death rates by race and Hispanic origin in official mortality statistics of the United States produced by the National Center for Health Statistics (NCHS). It also provides a quantitative assessment of bias in death rates by race and Hispanic origin. It identifies areas for targeted research. METHODS: Death rates are based on information on deaths (numerators of the rates) from death certificates filed in the states and compiled into a national database by NCHS, and on population data (denominators) from the Census Bureau. Selected studies of race/Hispanic-origin misclassification and under coverage are summarized on deaths and population. Estimates are made of the separate and the joint bias on death rates by race and Hispanic origin from the two sources. Simplifying assumptions are made about the stability of the biases over time and among age groups. Original results are presented using an expanded and updated database from the National Longitudinal Mortality Study. RESULTS: While biases in the numerator and denominator tend to offset each other somewhat, death rates for all groups show net effects of race misclassification and under coverage. For the white population and the black population, published death rates are overstated in official publications by an estimated 1.0 percent and 5.0 percent, respectively, resulting principally from undercounts of these population groups in the census. Death rates for the other minority groups are understated in official publications approximately as follows: American Indians, 21 percent; Asian or Pacific Islanders, 11 percent; and Hispanics, 2 percent. These estimates do not take into account differential misreporting of age among the race/ethnic groups.

Abstracting data from medical examiner/coroner reports: concordance among abstractors and implications for data reporting.

The purpose of this study was two-pronged: 1) to determine the level of concordance (agreement) between multiple records abstractors who extracted defined data elements from printed medical examiner/coroner (ME/C) death investigation records; and 2) to identify data items for which improved reporting could facilitate the effective use of ME/C reports and data. Four hundred ninety four printed death investigation records were obtained from 224 medical examiner/coroner offices throughout the United States. Trained abstractors were asked to extract information for 110 data elements from investigative reports. Additional data elements for each toxicology workup were abstracted from toxicology laboratory reports and six-digit AIS codes were also abstracted for each injury as described in autopsy reports. The ability of multiple abstractors to identify each data element and identically abstract the data was assessed using Kappa statistical methods. Level of agreement for many data elements was very good (>0.9), but for some data elements agreement was marginal to poor, especially for items related to toxicology, the nature of specific injuries, and dates, times of the occurrence of death and injury. Many data items can be easily abstracted from ME/C records. However, some data items seem difficult to abstract reliably in all cases. Standardizing the report formats used by ME/Cs and/or standardizing the electronic storage of ME/C data would make the abstraction of such data easier and improve the usefulness of ME/C data.