Table 2.

Results of quasi-Poisson regression models analyzing linear associations of segregation and tree cover with firearm violence.

	Model 1 (main exposures only)			Model 2 (built environment added)			Model 3 (social environment added)
Exposure variable (scaled by std. dev.)	IRR	95% CI	p	IRR	95% CI	p	IRR	95% CI	p
ICE for race-income (higher = more privilege)	0.46	0.44, 0.49	< 0.001	0.47	0.44, 0.50	< 0.001	0.58	0.54, 0.62	< 0.001
Tree canopy cover	0.79	0.73, 0.84	< 0.001	0.84	0.79, 0.90	< 0.001	0.91	0.86, 0.97	0.005
Tract area				0.92	0.87, 0.97	0.001	0.93	0.89, 0.97	0.002
Barber shops				1.05	1.01, 1.09	0.02	1.05	1.02, 1.09	0.004
Beauty salons				1.00	0.95, 1.05	0.94	1.05	1.01, 1.09	0.03
Places of worship				1.12	1.08, 1.17	< 0.001	1.08	1.05, 1.12	< 0.001
Liquor stores				1.09	1.05, 1.13	< 0.001	1.07	1.03, 1.10	< 0.001
Unemployment rate							1.05	1.00, 1.09	0.03
Vacancy rate							1.23	1.18, 1.28	< 0.001
Low educational attainment							1.23	1.18, 1.29	< 0.001

Notes. Exposure variables were centered at the mean value and scaled by one standard deviation for comparability, then modeled as linear predictors. Models were quasi-Poisson regressions that included a lagged population offset, a fixed effect for each city, and controlled for street segment length and counts of barber shops, beauty salons, places of worship, liquor stores, convenience stores, and schools. Models also included thin plate spline terms (degrees of freedom = 29) to account for unmeasured spatial confounding. (Goldstick et al., 2015)