Prediction of Non-Receipt of Needed Medical Care in New York City Adults: Evaluation of Individual and Neighborhood Characteristic Interactions through Boosted Regression Trees
David M. Wutchiett, Columbia University
Rates of non-access to needed medical services have been found to be elevated among uninsured adults and sociodemographic subpopulations. Neighborhood characteristics have been linked to differences in numerous health outcomes. Despite policy efforts to curtail rates of non-access among New York City adults, reported rates of missed care have been persistent within recent years. Through use of the boosted regression tree technique, missed care event outcomes were modeled with individual and neighborhood characteristics as predictors. Data consisted of cross-sectional representative samples of New York City adults surveyed through the 2009 (n=9900) and 2010 (n=8622) annual Community Health Survey. Neighborhood characteristics were obtained through the Census’ 2007-2012 American Community Survey and were linked to 34 neighborhoods. Through measurement of relative influence and comparison of out of sample error across models, results suggest a significant role for neighborhood characteristics in prediction of health services non-receipt.
Presented in Session P2. Data and Methods/Applied Demography/ Spatial Demography/ Demography of Crime