Using spatial regression methods to examine the association between county-level racial/ethnic composition and reported cases of Chlamydia and gonorrhea: an illustration with data from the state of Texas

Abstract

Background: Several studies have reported racial/ethnic disparities in the incidence of sexually transmitted diseases. However, very few studies have accounted for potential spatial dependence. Additionally, little is known about the relative magnitudes of the associations between county-level racial/ethnic composition and the 2 most commonly reported sexually transmitted diseases.

Methods: We used county-level data from the National Electronic Telecommunications System for Surveillance and the 2000 Census data to investigate the association between county-level racial/ethnic composition and reported cases of the 2 most commonly reported sexually transmitted diseases (chlamydia and gonorrhea) in Texas. We also estimated ordinary least square (OLS) models for comparison.

Results: Preliminary results from the spatial regression models indicated that the choice of spatial relationships criteria was important for model specification. The spatial error model (SEM) was superior to the spatial autoregressive model, spatial Durbin model, and OLS. The SEM for the 2 disease equations were further analyzed using a seemingly unrelated regression estimation (SURE) procedure. Although the SEM was superior to all models (using standard criteria), the coefficients were fairly stable across models. Our results showed that a unit change in percent black was associated with 1.6 (1.1 for Hispanic) and 3.3 (0.5 for Hispanic) percent change in chlamydia and gonorrhea rates (on average), respectively, compared with percent white.

Conclusion: Although there were no substantial differences in the magnitude of the estimated parameters, spatial regression models are potentially superior to OLS models and should be explored in future sexually transmitted disease studies.