Life expectancy estimates are widely used to make comparisons between countries. Calculating life expectancy at the neighborhood level presents a challenge to health departments. Presenting health data at fine geographic scales can disclose confidential information. In addition, fine scale maps of life expectancy are often misleading due to small numbers of deaths in some areas.
METHODS:
We used death certificate data from 2008-2012 for New York State, excluding New York City. The data included street address and age at death. Deaths were geocoded and aggregated, by age group to the census tract level. These data were linked to population data provided by the census. We excluded tracts that had over 50% of the population living in group quarters. We also excluded tracts that were mostly covered by airports. If a census tract had less than 60 deaths, it was merged to one or more neighboring tracts using the Geographic Aggregation Tool (http://www.albany.edu/faculty/ttalbot/GAT) to provide more stable life expectancy estimates. Life expectancy at birth and the standard error of the estimate were calculated for each area using the SEPHO Life Expectancy Calculator (http://www.sepho.org.uk/viewResource.aspx?id=8943). If the standard error in these newly created areas was greater than or equal to 2 years, further aggregation was done. Next we linked socio-demographic data on poverty, income, health insurance and educational attainment (American Community Survey, 2008-2012) to the life expectancy estimates. The life expectancy estimates, along with socio-demographic factors were thematically mapped.
RESULTS:
We discovered inconsistencies between where populations and deaths were located in areas with large populations living in group quarters such as colleges, military bases, and nursing homes. Our final study area included 2,679 individual tracts. After geographic aggregation we had 2,415 areas with life expectancy estimates which ranged from 65 to 96 years. Short life expectancies were readily apparent in communities with high poverty, low income and low educational attainment levels. We found unusually long life expectancies in areas along the New York borders. We attributed long life expectancy estimates in border areas to residents dying in neighboring states thus artificial reducing death counts. After information on out-of-state deaths was obtained, life expectancy estimates in these areas decreased.
CONCLUSIONS:
Available tools exists for health departments to calculate life expectancy estimates at fine geographic scales. The estimates can be compared and mapped along with socio-demographic to provide useful information on where to focus public health initiatives.