Census Tract Re-Geocoding Project

Delaware Cancer Registry

In response to public concern about cancer incidence and death rates, cancer clusters, and restrictions related to cancer registry data release, Delaware passed a law in July 2008 requiring the Delaware Division of Public Health (DPH) to provide maps of cancer incidence rates by census tract to the public every year. The first maps had to be published in December 2008, include cancer diagnosis years 2001–2005. To meet the deadlines, the Delaware Cancer Registry (DCR) had to provide geocoded cancer data to the DPH in only a few months.

Historically, it took nearly a month to geocode about 5,000 cases every year using a variety of sources. Sample checking often found that the geocoding was wrong. To make sure the census tract coding was correct and complete, DPH and DCR staff decided to widen the diagnosis period to 2000–2006 and to work with an industry expert to re-geocode the about 32,000 cases involved.

DCR staff researched geocoding service providers and contacted other state central registries for their advice on how to reduce the number of cases that do not geocode to an exact street address. DCR utilized TeleAtlas batch coding system and to accomplish an intensive quality assurance process before geocoding.

• Staff reviewed addresses for data entry consistency and validated them using the usps.com, zip-codes.com, ffiec.gov and factfinder.census.gov Web sites.
• Using Accurint, a LexisNexis database DCR was able to replace Post Office boxes with physical addresses.
• LACSLink software successfully converted rural route addresses to street addresses.

DPH expected the project would cost tens of thousands of dollars while actual cost was $3,000 to use the TeleAtlas batch coding service.

As a result of this project—

• The percentage of DCR’s 2001–2005 data coded to street address level (Census Tract Certainty 2000=1) improved from 81% to 96.5%.
• Census Tract 2000 codes changed for about 18% of cases.
• The process developed in this project is now the DCR’s standard practice for address review and geocoding. Case address at diagnosis is geocoded to Census Tract Certainty 2000=1 for about 98% of Delaware’s annual cases.
• The DCR can now provide consistent, reliable, and complete census tract geocoded data for cancer cluster investigations. Internal questioning of DCR’s geocoded data quality has stopped.

Many factors such as population mobility and small area populations make evaluation and interpretation of cancer incidence rates at the census tract level very difficult. To provide a firm basis for these analyses, the cancer registry address at diagnosis and census tract data must be accurate and complete.