United States Cancer Statistics (USCS)

Technical Notes: Statistical Methods: Suppression of Rates and Counts

Suppression for Reliability
When the numbers of cases or deaths used to compute rates are small, those rates tend to have poor reliability.¹ Therefore, to discourage misinterpretation or misuse of rates or counts that are unstable because case or death counts are small, incidence and death rates and counts are not shown in tables and figures if the case or death counts are below 16. A count of fewer than about 16 results in a standard error of the rate that is approximately 25% or more as large as the rate itself. Similarly, a case count below 16 results in the width of the 95% confidence interval around the rate being at least as large as the rate itself. These relationships were derived under the assumption of a Poisson process and with the standard population age distribution assumed to be similar to the observed population age distribution.

Suppression for Confidentiality
Another important reason for using a threshold value for suppressing cells is to protect the confidentiality of patients whose data are included in a report by reducing or eliminating the risk of disclosing their identity.² The cell suppression threshold value of 16, which was selected to reduce misuse and misinterpretation of unstable rates and counts in this report, is more than sufficient to protect patient confidentiality. (PDF-324KB)

Suppression for Bar Graphs
Because the incidence and death rates shown in the state-, sex-, and race-specific bar graphs are presented in rank order, we applied a criterion for suppressing data in addition to the threshold value of 16 cases. In these figures, incidence and death rates are not ranked or shown for any population groups of fewer than 50,000 people.

U.S. Census Regions and Divisions

Rates for U.S. Census regions and divisions were calculated by aggregating data reported from the states in each region and division. Only data from state registries that met USCS publication criteria were included in calculations of incidence rates for U.S. Census regions and divisions. Thus, where data for some states are excluded there is a potential for bias in the incidence rates for Census regions and divisions. We estimated cancer rates for Census regions or divisions with ineligible cancer registries by assuming that the incidence-to-mortality ratio in the portion of the region or division that was covered by eligible registries was the same as the incidence-to-mortality ratio in the portion that was not covered by eligible cancer registries. The age-adjusted incidence rates for U.S. Census regions and divisions are presented only if—

1. At least 80% of the population for the Census region or division was covered by cancer registries that met USCS publication criteria.



2. The 95% confidence intervals around the observed age-adjusted regional or division incidence rates based on data from eligible registries for each of six major cancer sites (prostate, female breast, male colorectal, female colorectal, male lung and bronchus, female lung and bronchus) included the estimate of the regional or division rate calculated using the specified criteria.

This Web site presents the observed age-adjusted incidence rates for all U.S. Census regions and divisions. Case counts for U.S. Census regions and divisions are presented if all state cancer registries in the region or division met the criteria for inclusion, unless the count for one state in the region or division is suppressed due to a count below 16.

Total United States

Cancer incidence rates for the United States are aggregate rates based on cancer cases reported from central cancer registries that met the USCS publication criteria and are the best estimates of the U.S. cancer burden available that are based on observed data. Case counts for the U.S. incidence rates for all ages combined are presented.

References

¹Brillinger DR. The natural variability of vital rates and associated statistics. Biometrics 1986;42(4):693–734.

²Doyle P, Lane JI, Theeuwes JM, Zayatz LM. Confidentiality, Disclosure, and Data Access: Theory and Practical Applications for Statistical Agencies. Amsterdam, The Netherlands: Elsevier Science; 2001.

Please note: Some of these publications are available for download only as *.pdf files. These files require Adobe Acrobat Reader in order to be viewed. Please review the information on downloading and using Acrobat Reader software.