Denominator Revisions
(June 2007)

Overview

The denominators (population estimates) used to calculate rates in four WISH modules (Mortality, Injury Mortality, Injury Hospitalizations, and Injury Emergency Department Visits) are revised each year to incorporate new data from the National Center for Health Statistics (NCHS). An additional methodological revision was made to the WISH denominator series in May, 2007. This note briefly describes these revisions.

The denominator revisions may affect rates by race/ethnicity generated in WISH. Rates by age group and sex are not affected, except for minor differences due to rounding. Users are advised to document the WISH access date to assist in reconciling differences in rates prepared at different times.

Annual Revisions

To provide rates by race and Hispanic ethnicity, WISH uses bridged race population estimates (exit DHS) from NCHS. These are published in annual "vintages" that correspond to the year of release (e.g., vintage 2006 estimates are released in 2006). When NCHS releases a new vintage, estimates for previous years extending back to the 2000 Census are revised to incorporate the newly available data. These race/ethnicity revisions are incorporated into the WISH denominators, meaning that rates by race/ethnicity for specific years may change as new vintages are incorporated.

However, we also adjust the NCHS estimates so that when summed by county, age group and sex they equal the estimates produced by the Office of Health Informatics (OHI), which are not revised. This means that rates by county, age group and sex for specific years in WISH will remain constant, despite the addition of a new NCHS vintage each year.

WISH users are advised to investigate the potential impact of revised denominators on conclusions drawn from rates by race/ethnicity generated in the affected WISH modules. In practical terms, compare rates generated at different times and see how big a difference the revised denominators make.

Methodological Revision

Adjusting the NCHS estimates so they equal the OHI estimates means that the distributions of populations by race/ethnicity in the two data series differ from one another. However, a methodological change was made in May, 2007 that reduced the extent to which the OHI estimates differ from the NCHS estimates. This change was applied to all estimates extending back to 1990. The change may affect rates by race/ethnicity generated prior to May, 2007, so WISH users are again advised to investigate the potential impact on conclusions derived from these rates.

The mean absolute percent differences between the OHI and the NCHS estimates for both methods ("New" versus "Old") are shown in the table below. These percentages are for 1990-2005 and are shown by race and Hispanic ethnicity.

Mean absolute percent difference between
OHI and NCHS estimates: 1990-2005

The table shows that the largest improvements are among the non-White populations: percent differences between the OHI and NCHS estimates were cut at least in half among three of four non-White groups and the percent difference among Hispanics was reduced by 2.2 percentage points, or 36 percent.

Methodological Description

Briefly, the methodological change involved inserting an additional step into the adjustment process. Under the old method the estimates were produced and adjusted so they equaled the OHI estimates in a single step, while the new method first produces "raw" estimates, then adjusts them to equal the OHI estimates.

Under the new method, the first step in creating the "raw" estimates is to divide the NCHS estimates in individual cells by the NCHS total county population. This produces a set of proportions that correspond to the distributions found in the NCHS bridged race file. These proportions are then applied to the OHI total county population, which results in the set of raw estimates. The adjustment step then consists of dividing the OHI estimates by the sum of the raw estimates for each corresponding county, sex and age group, which produces a set of adjustment factors. The raw estimates are then multiplied by the corresponding adjustment factor. This produces the WISH denominators by county, age group, sex, race, and Hispanic ethnicity.

The two-step procedure in the new method results in bridged race denominators that a) are adjusted so they equal the OHI population estimates; and b) generally are closer to the NCHS bridged race estimates compared with those produced under the old method.

RETURN to previous page

Last Revised:  March 01, 2012