Skip to main content
         

Vol. 08 Issue 1, Early Spring 2003

New York State's Cancer Mapping Project: A New Approach Using GIS and Spatial Analysis
The Ribbon 

Elizabeth L. Lewis-Michl, Center for Environmental Health
New York State Department of Health

New York State Department of Health's Cancer Mapping Project (officially the Cancer Surveillance Improvement Initiative) uses geographic information system tools and spatial statistical tools to provide information about the incidence of specific types of cancer for counties and ZIP codes. This project, which began in 1998, is an ongoing effort to give New Yorkers more information about cancer in their communities. It is designed to help guide cancer prevention programs and future research on the causes of cancer.

Information about cancer incidence in each of New York's 62 counties, including New York City's five boroughs, already was available in data tables provided annually. In 1999 and 2000, the Cancer Mapping Project presented the data for the first time in maps. County maps showed relative cancer incidence for 11 categories of cancer -lung and bronchus, prostate, breast, colon and rectum, kidney and renal pelvis, bladder, liver and bile duct, thyroid, leukemia, lymphoma, and brain and other nervous system.

In 2000 and 2001, the Cancer Mapping Project developed maps and data appendices of relative cancer incidence by ZIP Codes for the most frequently diagnosed types of cancer, breast, prostate, lung and colorectal. These maps and additional information about the New York State Cancer Registry and cancer risk factors are available at www.health.state.ny.us or by calling the New York State Department of Health at 1-800-458-1158.

What Information do the Maps Provide?

Both the county and the ZIP Code maps show geographic areas shaded with a range of colors to indicate relative cancer incidence in relation to statewide incidence. For the county maps, the relative shading is based on the incidence rate -the number of newly diagnosed cancer cases per population for five years. The county rates are adjusted to make them comparable even if there are differences in age of the populations.

For the ZIP Code maps, relative incidence is shown for each geographic area using a standardized incidence ratio (SIR). This ratio indicates whether the number of new cancer diagnoses is higher, lower, or about the same as "expected "for a five-year period in each geographic unit. The SIR is calculated by dividing the actual observed number of cases by the expected number of cases in each county or ZIP Code. Expected incidence is the number of new cancer diagnoses that would be expected in that geographic area if the rate per population of cancer in this area were the same as in New York State as a whole.

Protecting the confidentiality of each person diagnosed with cancer limits how information can be presented in maps available to the public. Data grouped by ZIP Code are shown only for the most frequently diagnosed types of cancer (breast, prostate, lung and colorectal). However, confidentiality issues still occur when providing information at the ZIP Code level for some ZIP Codes with very small populations. For this reason, some ZIP Codes were combined to provide information for a larger number of cases.

Maps Create Visual Images that can be Misleading

In a map that is shaded according to relative cancer incidence, large geographic areas with small populations can contribute to a misleading visual impression.While the shading for high or low incidence for the large, rural areas can dominate the map visually, the incidence estimates for these sparsely populated areas are based on small numbers of cases. On the other hand, elevated or lowered cancer rates in large populations in densely populated areas may be almost invisible on the map if the areas are small in physical size, such as the boroughs of New York City.

In addition, chance variations in numbers in small populations with very small numbers of cases expected can create the appearance of large differences in comparative cancer incidence rates. For example, consider a small population where the expected number of cases of cancer to be diagnosed over a five-year period of time, is five. If six cases were diagnosed, the SIR would be 6/5, a 20% excess of cancer incidence. For comparison, in a larger population where 100 cases of cancer were expected, an additional case would result in a SIR of 101/100, a 1% excess. One way to help prevent misinterpreting a map is to use additional information to estimate the accuracy, or margin of error, associated with each cancer incidence rate. In the county maps, the estimated cancer rate is accompanied by the actual number of cases upon which it is based. The county map pamphlets also show a bar graph of each county's relative cancer incidence rate.

On the State Health Department's web site there is a visual aid showing a margin of error, called a confidence interval around each of the rates to indicate that the rate likely falls somewhere within this range. If the range includes the statewide rate, this indicates that the difference between this County's cancer incidence and the statewide cancer incidence is more likely due to chance.

Spatial Statistics Help with Interpretation of ZIP Code Maps

New York State researchers used a statistical technique to evaluate elevations of cancer incidence that would not be expected to appear just by chance in individual ZIP Codes and groups of ZIP Codes. This was done to help offset potentially misleading visual impressions and to help with interpretation of the ZIP Code maps. This technique uses a program called SaTScan, developed by Martin Kulldorff, Ph.D. and distributed by the National Cancer Institute.

A computer program randomly distributes the state's total number of cases of cancer among each of the state's ZIP Codes and possible groups of adjoining ZIP Codes (using only the age distribution and population size for each ZIP Code.) This random assignment is repeated 9,999 times, calculating SIRs for each ZIP Code and possible group of ZIP Codes in each of these simulations. For each simulated SIR, a value called a "likelihood statistic "is calculated. The likelihood statistic takes into account the number of observed and expected cases and the SIR. The results of the computer simulations are evaluated by identifying the most extremely unlikely elevation (highest value of the likelihood statistic) in each simulation.

The actual SIRs are compared with these simulated results to identify cancer incidence in any ZIP Code(s) that are extremely unlikely. ZIP Code(s)are considered to have a statistically significant elevation in cancer incidence if the likelihood statistic for that area is higher than 95% of the maximum likelihood statistics from the simulations. In this way, the actual cancer incidence estimates for ZIP Codes and groups of adjoining ZIP Codes are compared to the simulated values to decide what level of excess of cancer incidence is truly unusual, or not expected simply due to random or chance variation.

Follow-up and Next Steps

The ZIP Code maps show hatching (diagonal lines) and cross hatching to identify areas of unusual cancer elevation not expected by chance. The State Health Department developed a protocol to investigate unusual disease patterns and is using the protocol to investigate a five ZIP Code area in Suffolk County where breast cancer incidence is estimated at about 50% above the average statewide incidence.

The first step in the Coram/Mt.Sinai/Port Jefferson Station investigation defined the geographic boundaries for follow-up. Next, researchers looked at some factors that might explain the higher rate of breast cancer, such as incorrect population estimates, or unusually high screening rates, detecting more cases of the disease in its early stages. These did not appear to be important factors for explaining the area's elevated incidence.

Researchers also began reviewing existing environmental data, examining information about the area's air, water and soil.

This past summer, area residents in the affected communities attended an information session to hear about the status of the investigation and other aspects of the State's Cancer Mapping Project. The State Health Department collected additional information from residents about historical sources of exposure that may have been a concern. Researchers are factoring this information into the ongoing investigation.

Currently State Health researchers are evaluating the literature on the biology of breast cancer and other factors that may contribute to disease incidence. They are reviewing the toxicology of environmental agents, especially those that are known or suspected to increase the risk of breast cancer. Researchers will recommend future activities after this review is completed and the environmental exposure information from the community is evaluated.

The follow-up work being done in the Coram/Mt. Sinai/Port Jefferson Station investigation has moved beyond the scope of tools such as geographical information system and spatial statistics. However, these kinds of tools can be valuable in helping to identify areas where further research into disease incidence may be needed.

Cancer Booklets, including color versions of county-level cancer maps and graphs for 1992-1996, are available in portable document format (PDF) at http://www.health.state.ny.us/nysdoh/cancer/csii/nyscsii2.htm

Back to the top

Last Update 04.04.03