BACKGROUND: Geographic aggregation of census tracts provides a solution to the small numbers problem, when calculating health indicators at the sub-county level. However, choosing the same geographic scale for several health indicators is a challenge since the number health events for different health conditions can vary within a census tract. For example, myocardial infarctions and cancer are more common in older populations while adverse birth outcomes will be more numerous in younger populations. Census boundaries can also change over time.
METHODS: The Geographic Aggregation Tool (GAT, http://www.albany.edu/faculty/ttalbot/GAT/) was used to merge census tracts with few health events with neighboring areas to provide stable rates for small areas in New York State for a 5-year period which bracketed the 2010 census. The tool was set so at least 250 births and 60 deaths occurred in each area. This ensured an adequate number of health events for calculating stable indicators for low birth weight and life expectancy. The GAT was used in an iterative process, so merged areas did not cross large town boundaries (population>25,000) or county boundaries. To resolve problems with census tract boundary changes over time, a GIS application was used to identify these changes. The census tracts that changed between 2000 and 2010 were merged with neighboring tracts so health indicators can be compared over time. Other health indicators were obtained from the CDC Environmental Public Health Tracking Portal to evaluate whether these indicators could be displayed on the same geographic scale used for low birth with and life expectancy.
RESULTS: The GAT was successfully used to create sub-county areas for displaying health indicators based on health outcomes which are either common in younger or older populations. These indicators included low-birth weight and life expectancy. A review of other indicators for asthma, myocardial infarction, premature births and common types of cancer indicate they would also have stable estimates at this same scale. Indicators based on conditions such as birth defects, very low birth weight and less common cancers would not have adequate numbers to provide reliable estimates at this scale.
CONCLUSIONS: Various health indicators can be displayed using the same small areas if appropriate aggregation tools are used. With these tools, public health professionals will be better able to target areas for public health interventions. The remaining impediment in creating small-area indicators is the lack of health data which are accurately geocoded to the census areas.