County variations in seatbelt use: A multilevel modeling study

While seatbelt use varies from one state to another, less is known about variability in smaller geographic areas. Data from the 2002 Behavioral Risk Factor Surveillance System (BRFSS) and the US census were combined to identify factors responsible for county level differences in seatbelt use. Data were extracted from 73 counties with at least 100 BRFSS surveys (n = 51,139). Census data were used to form composite variables measuring population density, racial segregation, income disparities, and education. We also coded the state seatbelt law as primary (drivers can be stopped and ticketed for seatbelt violations) or secondary (drivers stopped for other violations can be ticketed). The seatbelt question was recoded into a binary variable (always, less than always). Multi-level modeling was performed using the HLM program (SSI, Lincolnwood, IL). The level 1 logistic regression model showed that age (p<0.0001), gender (p<0.0001), BMI (p<0.0001), Hispanic ethnicity (p<0.01), low income (p<0.01), high income (p<0.0001), high school dropout (p<0.001), and college education (p<0.0001) were all individual level predictors of seatbelt use. The level two analysis of intercepts (differences in county rates of seatbelt use) showed that state seatbelt law (p<0.0001), racial segregation (p<0.01), and population density (p<0.0001) were all significant predictors of county differences in seatbelt use after controlling for the level 1 model. These analyses support an ecological model of seatbelt use in which individual differences and socio-environmental characteristics interact to influence rates of seatbelt use in specific locations.