Accounting for spatial autocorrelation in the association between preventable congestive heart failure hospitalizations and neighborhood measures in New York City

background: Congestive heart failure (CHF) is the most common preventable hospitalization in the Unites States (US) yet, there is a dearth of research aimed at understanding how neighborhood attributes and spatial relationships among places impact hospitalization rates. We assessed the degree of spatial autocorrelation in CHF hospitalization rates and accounted for clustering to accurately describe associations with neighborhood compositional measures in New York City.

methods: Using 2007 inpatient discharge data (n= 23,058) from New York Statewide Planning and Research Cooperative System, CHF unique and readmission hospitalization rates among adults were calculated at the US Census block group level and examined for spatial autocorrelation.  Both ordinary least squares (OLS) and spatial autoregressive (SAR) error models were fit to determine the effect of sociodemographic area measures on CHF rates with and without accounting for spatial clustering.

results: Older age composition and greater proportions of non-Hispanic black residents, Hispanic residents, households in poverty, and adults without a high school degree were significant predictors of higher CHF hospitalization rates in both OLS and SAR models. However, the latter showed a better model fit and reduced spatial autocorrelation in residuals.  

conclusions: Our findings indicate that CHF discharge rates are impacted by neighborhood compositional measures underscoring the importance of population-level approaches to prevention. Furthermore, when evaluating associations between area effects and health, spatial autocorrelation should be assessed and accounted for in regression models.