Members of the geocoding project team at a recent meeting. From l to r, Jarvis Chen, Pamela Waterman, Nancy Krieger, and S.V. Subramanian

"Social inequalities in health, however real, can be ignored and made invisible if the data to document them are not collected," said Nancy Krieger, principal investigator of the project and an associate professor in the Department of Society, Human Development, and Health.

Five years in the making and available at no charge, the Public Health Disparities Geocoding Project Monograph is available at www.hsph.harvard.edu/thegeocodingproject/webpage/ monograph/. The tool offers for the first time a systematic way to monitor health disparities across a wide range of health outcomes by using area-based socioeconomic information from public data sources, such as the U.S. Census.

"This project provides state and local public health agencies with tools that can be used to conduct their own, much-needed analyses of the impact of economic position on neighborhood health status and outcomes," said Dan Friedman, an independent consultant who provides population and public health information services. He is a former assistant commissioner of the Massachusetts Department of Public Health.

U.S. health statistics rarely include individual-level socioeconomic information because their sources–such as cancer registries, notifiable disease surveillance systems, and hospital records–lack these data. As a result, influential national health reports have little to say about socioeconomic inequalities in health. Krieger has calculated that 70 percent of the objectives listed by the federal report Healthy People 2010, an official roadmap toward making Americans healthier, fail to include targets for reducing socioeconomic disparities in health due to a lack of related data.

Instead, the data are presented solely in relation to race/ethnicity–often measured crudely–and "we have no ability to assess socioeconomic influences on health within these racial/ethnic groups–let alone socioeconomic contributions to racial/ethnic disparities in health," said Krieger. "An obvious need is routine monitoring of health at local and national levels for all populations, stratified by socioeconomic position in addition to race/ethnicity and gender."

For that monitoring to happen, Krieger and a small team of researchers hammered at reams of public health and census records in Massachusetts and Rhode Island. The researchers set out to determine two things: what kind of information–when applied to what kind of group–would consistently tell officials something useful about health disparities. The answers could prompt public health officials to rethink their usual approaches to monitoring inequalities.

From the records, project director Pamela Waterman pulled out one million addresses in Massachusetts and Rhode Island to be geocoded, a process which involves assigning geographic codes (such as a census tract number) to a given location. Although computerized geocoding is a relatively quick process, addresses must first be reviewed–one by one–and extraneous information such as apartment numbers be deleted. This laborious task took several months of full-time work. Waterman sometimes spent hours at a single agency cleaning up addresses from databases that could not be moved offsite due to confidentiality concerns.

The project team, which included HSPH assistant professor S.V. Subramanian, chose a gamut of health outcomes to test, spanning from birth to death. The outcomes included low birthweight, childhood lead poisoning, tuberculosis, sexually transmitted infections, non-fatal weapons-related injury, cancer incidence, and mortality.

The researchers then created a variety of indicators of what they described as area-based socioeconomic measures–such as occupation, education, household incomes, and poverty levels–from U.S. census data. Typically characterized at a small area of geography, these measures can be thought of as representing the socioeconomic circumstances to which residents have been exposed. This information was linked to health records and tested at three levels of geography: ZIP codes (about 13,000 to 14,000 people), census tracts (about 4,000 people), and census block groups (about 1,000 people).

Powering these calculations were software programs designed by HSPH research scientist Jarvis Chen and used with the help of HSPH doctoral student David Rehkopf. In all, nearly 6,000 analyses were done (18 socioeconomic measures x 3 levels of geography x 18 specific outcomes, for the total population and stratified by race/ethnicity; by gender; and by race/ethnicity plus gender–for 2 states.)

Poverty, more so than education or occupation, at the census tract level emerged as the most consistent and powerful detector of health disparities–a compass to direct public health authorities to sources and locations of inequalities.

The finding held true again and again for a number of health outcomes. People living in impoverished parts of Massachusetts had mortality rates 1.3 to 1.4 times higher than residents in wealthy neighborhoods. Children in poor Rhode Island neighborhoods were more than nine times as likely to get lead poisoning than kids in richer communities. Residents in poor regions of Massachusetts faced more than eight times the risk of contracting tuberculosis than people living in well-off neighborhoods.

"Our goal was to come up with a valid, robust, easy to construct, and easy to interpret measure that could readily be used by any state health department for any health outcome," said Krieger.

"The emphasis in our country has been to look at health outcomes through the lenses of race and gender, which are important," said Waterman. "This project adds another lens to help us see a clearer picture of what is going on."

The team will soon start a training program that will bring health officials and researchers to HSPH to learn more about the project and its methodology, said Krieger.