NCI FACT SHEET: How Changes in U.S. Census Counts Affect NCI Cancer Rates
The National Cancer Institute (NCI) today released its annual update of
national cancer rates on its Web site,
http://seer.cancer.gov/csr/1975_2000/index.html. The Cancer Statistics Review includes incidence, mortality, and survival
rates for all cancers for the period 1975-2000 using new population estimates
from the U.S. Census Bureau. These data are also used to track cancer trends
over time.
1. How does NCI use census data to calculate cancer rates?
NCI's Surveillance, Epidemiology, and End Results (SEER) Program calculates
cancer rates using incidence, survival, and mortality data gathered through two
sources: incidence data from SEER, and death data from the National Center for
Health Statistics (NCHS).
To determine a cancer rate, NCI divides the number of cancer cases or deaths in
a given geographic area (numerator), by the total number of people in that area
as reported by the Census Bureau (denominator). The resulting cancer rate is
the proportion of people in that area affected by cancer. Because SEER rates
are calculated by dividing the numerator by the census-generated denominator,
the rates can be heavily influenced by changes or uncertainties in census
counts.
2. How often does the Census Bureau update and revise population estimates?
The Census Bureau routinely updates and revises population estimates every
year. The bureau calculates 'intercensal' estimates after a new census is
completed -- for example, using information from both the 1990 and 2000
Censuses, the bureau obtains better estimates for the 1990s. These revisions
are based on the most recent census information and on the best available
demographic data reflecting components of population change (namely, births,
deaths, net internal migration, and net international immigration).
Recalculating disease rates to reflect updated population estimates is standard
practice.
3. How did the Census Bureau's procedures change in 2000?
In the 1990 Census, respondents were asked to "select one" racial
classification (White, Black, Asian or Pacific Islander, American Indian or
Alaska Native.). The 2000 Census asked respondents to "select one or more" race
groups and separated the Native Hawaiian and other Pacific Islanders group from
the Asian group, resulting in a total of 31 different classifications. For
agencies such as the NCI and NCHS to continue reporting long-term trends in
disease rates for single-race groups, a method is needed to "bridge" these
multi-race classifications into a single race category. Such a method was
developed by the NCHS using information collected as part of their National
Health Interview Surveys. In collaboration with NCHS, the Census Bureau
produced a set of year 2000 population estimates that assigned everyone to one
race group only. The resulting 2000 estimates were then used to produce an
improved set of 1990-2000 population estimates. NCI and NCHS are making these
bridged population estimates available on their respective Web sites. These
revisions to the population estimates will affect the denominator in all SEER
cancer rate calculations for 1990-2000.
The Census Bureau has other studies and research in progress and expects to
make final intercensal bridged estimates available sometime in 2004.
4. Is there still uncertainty in the cancer rates and if so, where?
Although efforts were made to use the best available data and methods to
produce the bridged estimates, the calculations themselves could introduce a
small amount of uncertainty. However, because these calculations are based on
2000 Census counts, the majority of rates will have a greater degree of
certainty, not less. The potential for variability or uncertainty will be
greatest for the smallest population groups -- particularly specific race or
age groups and county-level residents.
In less populated areas such as rural counties, or in adjacent urban and
suburban areas where there was substantial migration of residents, a change in
the denominator can affect the county rate by as much as 20 percent -- unlike
large counties, where a small change in a large denominator will not affect
rates nearly as much.
If a new census population estimate is larger than an earlier one and the
number of cancer cases remains the same, the new rate will be smaller.
Similarly, the rate will increase if the census population decreases.
5. Can you give a specific example of an area where rates will be affected by
Census 2000?
Projecting population shifts in specific geographic areas is one of the
greatest sources of uncertainty in trying to produce accurate census counts.
For example, the population counts for Blacks in the Atlanta metropolitan area
are higher than previously estimated due to suburban migration that the Census
Bureau had not fully captured during the 1990s (before the 2000 Census
information became available). This means that cancer rates are actually lower
for Blacks in metropolitan Atlanta than originally calculated by SEER's use of
the Census Bureau's earlier population estimates. In general, differences
between projected populations and actual census counts are more likely to occur
in small areas such as counties and metropolitan areas than in entire states or
regions of the country.
It is important to note that these population changes will not affect actual
counts in the number of cancer cases or deaths, just the cancer rates
calculated for these populations. Furthermore, interpreting these new cancer
rates fully will take time since there are many ways in which population
changes have occurred.
6. Will these updated census counts significantly affect cancer rates?
No. Even with updated census figures, overall cancer rates for major cancers
will not change significantly. Some rates for less common cancers, cancers in
minority populations, smaller geographic areas (such as counties), or specific
age groups may be affected. For instance, updates to American Indian and
Hispanic populations will result in slight increases in their cancer rates.
Although it is too soon to tell the extent to which these rates may change
reported cancer rates, overall changes are relatively small -- for example, 1
percent or 2 percent.
7. Are there other factors that could cause cancer statistics to be adjusted?
Normal reporting delays (how long it takes to collect statistics on new
cases of cancer from SEER regional offices) and other adjustments, such as more
information on race, ethnicity, or tumor characteristics, could affect cancer
statistics. A study of these factors by NCI researchers was reported in the
Oct. 16, 2002 issue of the Journal of the National Cancer Institute
(volume 94, pages 1537-1545). This study focused on areas where quicker
accumulation of regional cancer statistics could give NCI more reliable numbers
for the most recent rates. The best estimate of rectification of reporting
delays and corrections to the most recent year of incidence data would be a
change in existing rates by about 3 percent to 6 percent.
8. Does the 1975-2000 SEER Cancer Statistics Review reflect corrections to
statistics that have been wrong for many years?
No. NCI is making small adjustments (about 1 percent to 3 percent) to some
cancer statistics, and larger adjustments to statistics in certain areas of the
country and to selected cancers. NCI statistical reports from previous years
have been the best rates that could be calculated given the data that were
available at those times. Adjustments currently taking place primarily relate
to rates in minorities based on the new categories, to improved completeness
and coverage of cancer statistics in certain geographic areas, or to population
updates based on the recent 2000 Census. Overall, however, the trends and rates
reported annually by NCI and its partners will not change substantially.
9. Where can I learn more about bridging and get other information about NCI
cancer statistics?
For more information on bridged population estimates, please go to
http://www.cdc.gov/nchs/about/major/dvs/popbridge/popbridge.htm.
For complete information on cancer statistics, please go to NCI's "Finding
Cancer Statistics" Web site at
http://surveillance.cancer.gov/statistics/.
For additional information on bridging and other modifications to NCI
population estimates, please go to
http://seer.cancer.gov/popdata/.
# # #
Back to Top |