Use of Area-Based Poverty as a Demographic Variable for Routine Surveillance Data Analysis, Connecticut and New York City

BACKGROUND:  Socioeconomic status (SES) of individuals and their neighborhoods are often important determinants of the incidence of diseases and risk factors for them. However, no measures of SES have been adopted nationally for routine data collection and analysis of surveillance data collected by state and local health departments, with the exception of interview surveys such as the Behavioral Risk Factor Surveillance System.  While race/ethnic disparities often reflect differences in SES, collection of race/ethnic data is often incomplete, and interpretation of differences between groups and how to reduce disparities are fraught with complexity. Much data collected by state and local health departments includes the street address of residence of the affected person.  Current ease of geocoding enables identification of census tract of residence for 90-99% of street addresses. Census tract of residence can be linked to census and American Community Survey (ACS) data to determine neighborhood SES.

METHODS:  The Connecticut Emerging Infections Program and the New York City (NYC) DOHMH each have adopted using a standard neighborhood SES measure, percentage of residents below the federal poverty level, as a variable for routine data analysis, as recommended by the Harvard-based Public Health Disparities Geocoding Project. In Connecticut, recent surveillance data on influenza hospitalizations, invasive pneumococcal disease, campylobacteriosis and precancerous lesions were analyzed using either 2000 Census or 2010 Census combined with 2006-2010 ACS data. In New York City, data on all cause mortality 1990 and 2000 and tuberculosis incidence 2000 and 2008 were analyzed using 2000 Census data.

RESULTS:  Incidence of influenza hospitalizations and invasive pneumococcal disease in Connecticut, and all cause mortality and tuberculosis in NYC increased with increasing neighborhood poverty. Incidence of campylobacteriosis in adults and HPV-related precancerous cervical lesions in women 20-24 years old in Connecticut increased with decreasing census tract poverty. For four of the six analyses, data on race/ethnicity was complete enough to describe race/ethnic disparities and incidence by poverty level within race/ethnic groups. For all four conditions, incidence differences between high and low poverty groups were as great or greater than those described by race/ethnicity and within race/ethnic groups a relationship between relative poverty and disease incidence was observed.

CONCLUSIONS:  Analysis of surveillance data using census tract-level poverty describes disparities by socioeconomic status and provides a different perspective than race/ethnicity on disease epidemiology and potential prevention measures. Having a standard census tract-level measure would facilitate description of health disparities nationally and across states by SES.