Department of Health Services Logo

 

Wisconsin Department of Health Services

_________

WISH Home

_________

Queries:

Behavioral Risk Factor Survey

Birth-Related
All Births
Low Birthweight
Teen Births 
Prenatal Care
Fertility
Infant Mortality

Cancer

Injury-Related Mortality

Injury-Related Hospitalizations

Injury-Related Emergency Dept. Visits

Mortality

Population

Violent Death

_________

_________

Questions or Problems?

WISH Cancer Data Limitations and Caveats
 

Cancer Data Limitations

When interpreting cancer data it is important to remember certain limitations. Identification of cancer cases in the Wisconsin Cancer Reporting System (WCRS) is dependent on the reporting by hospitals, freestanding clinics, treatment facilities, and physician offices as mandated by state law. Limitations include underreporting in areas close to neighboring states, especially Minnesota (see Wisconsin Cancer Data Bulletin (PDF, 108 KB)), potentially misleading trends due to reporting variation, small numbers of cases, and misclassification or underreporting of race/ethnicity.

Wisconsin has reporting exchange agreements with 19 states, but there still may be a loss of cases of Wisconsin residents who were diagnosed in other states. Apparent increases or decreases in cancer incidence over time may reflect changes in diagnostic methods or case reporting rather than true changes in cancer occurrence. Many calculations in Wisconsin Interactive Statistics on Health (WISH) involve small numbers of cases, and the resulting statistics should be interpreted with caution. Data on race and ethnicity are based on information from medical records and death certificates; incorrect or missing information for these variables can result in underestimated incidence and mortality in racial/ethnic groups.

The cancer data upon which WISH is based are dynamic. That is, data are always being updated and improved. For example, in the published report, Cancer Incidence and Mortality, 2002-2006 (released November 2009), 26,952 cancers were included in the analyses of 2006 data. As of December 2013, the database for 2006 contained information on 28,498 cancers. WCRS staff are constantly updating data for all years when new information becomes available. In this regard, all data are subject to change when appropriate. For purposes of analysis, the data are "frozen" (closed) annually in a database so that in all publications and WISH, numbers and rates are consistent. The date of closure for 1995-2011 incidence data included in the current cancer module of WISH was December 2013.

Caveats about Cancer Data Completeness

The incidence data in WISH include all cases entered and processed in the Wisconsin Cancer Reporting System (WCRS) database by December 2013. Statewide 2011 incidence data in the WISH Cancer Module are considered at least 95% complete by the North American Association of Central Cancer Registries (NAACCR) certification process. However, data for specific cancer sites or for individual counties are not necessarily as complete. Data for cancers diagnosed and treated in a non-hospital setting (such as melanoma and prostate cancers) are possibly less complete than data for cancers treated in hospitals (such as lung and colorectal cancers).

The primary reason for incomplete data is that some hospitals, outpatient diagnostic facilities and outpatient treatment facilities are not reporting their cancer cases to the WCRS as mandated, or are not reporting cases in a timely manner. Incomplete and late reporting of cancer data may result in underestimates of the true burden of cancer in Wisconsin and Wisconsin counties. The number of cases and rates may change in future reports due to late reported cases or corrections based on subsequent data from reporting facilities.

WCRS has data-exchange agreements with 19 central cancer registries in Alaska, Arkansas, Arizona, California, Colorado, Florida, Illinois, Iowa, Indiana, Kentucky, Massachusetts, Michigan, Mississippi, Missouri, Montana, Texas, Washington, West Virginia, and Wyoming. Completeness of out-of-state reporting from these registries depends on the extent of their identification of Wisconsin residents and their standards of quality.

Geographic Boundaries. Two major factors affect interpretation of Wisconsin cancer data tabulations by geographic location. First, many Wisconsin counties have a small number of cancer diagnoses or deaths in a year, resulting in case frequencies in the single digits. Such small numbers may easily double or triple (or decrease by equivalent amounts) from year to year. When years of data or primary sites are combined to produce larger numbers of events, the rates become more stable. County data in WISH are averaged over five-year intervals to provide more stable rates. Even with multiple years of data, however, the numbers may still be small and some random fluctuation is expected. To provide guidance in assessing the stability of rates, all county tables show the confidence interval for each rate; the confidence interval will include the true value 95 percent of the time. A large confidence interval indicates high variability of the "true" rate, while a small confidence interval indicates greater stability of the rate.

Second, data completeness and/or timeliness may vary by region. The largest reporting variations relate to out-of-state diagnosis and treatment of Wisconsin residents who live in the sparsely populated counties along the Wisconsin/Minnesota border and are seen in Minnesota hospitals. Reporting by Minnesota hospitals to WCRS is a voluntary process (established through memoranda of understanding that protect patient confidentiality) and is not covered under Section 255.04, Wisconsin Statutes. Therefore, the number of reports sent to WCRS from Minnesota can vary greatly from year to year.

Patient Race and Ethnicity. Historically, differences among facilities in recording patient race (information not required in medical chart, entered by proxy, or patient self-reporting) have led to a number of ambiguous or unknown race codes in the cancer data reported to WCRS.

American Indian Identification. The collection of accurate information on cancer incidence in American Indians is difficult because of underreporting and frequent misclassification of cancer cases in this population. To help address this problem, state cancer registries coordinate with the U.S. Bureau of Indian Affairs and link cancer case files to enrollment files of the Indian Health Service. This process identifies a certain number of American Indians otherwise misclassified in the cancer reports. Data in WISH reflect American Indian cases reported by facilities and also those identified through the linkage with the U.S. Indian Health Service.

Hispanic/Latino Identification. The NAACCR Hispanic/Latino Identification Algorithm (NHIA) assigned Hispanic/Latino ethnicity to cases, using variables of birthplace, gender, race and surname, to increase the number of cases identified as Hispanic in the registry during the years covered in WISH. For a complete description of NHIA, and other updates to reporting guidelines, visit the NAACCR website (exit DHS). The application of the NHIA increased the number of Hispanic cancer cases beyond the number identified by reporting facilities. The NHIA was developed and tested by NAACCR and endorsed by the Centers for Disease Control and Prevention to correct for documented underreporting of Hispanic/Latino cases by facilities. Caution should be used when comparing rates for Hispanics/Latinos with the rates for race groups because ethnicity and race are not mutually exclusive categories in the cancer modules of WISH. Hispanics/Latinos who identify themselves as white or any other racial group are included in the race category as well as in the Hispanic category.

Asian/Pacific Islander Identification. The information about cancers among Asian and Pacific Islander populations is also enhanced by state cancer registries using a NAACCR developed algorithm. The NAACCR Asian Pacific Islander Identification Algorithm (NAPIIA v 1.2.1) uses a combination of variables to classify cases as Asian/Pacific Islander for analytic purposes. It is focused on coding cases with a race code of Asian NOS (not otherwise specified) or Pacific Islander NOS to a more specific race category (Chinese, Japanese, Hmong, Korean, for example). The algorithm uses the following standard variables: race, last name, first name, maiden name, birthplace, and sex. The NAPIIA can be found at the NAACCR website (exit DHS).

RETURN to previous page

Last Revised:  February 12, 2014