Table 2.

Spatially-Lagged Bivariate Models Predicting Presence and Counts of Affordable and High-Cost SLHs (N=2332)

	Bivariate Correlates of Affordable SLHs						Bivariate Correlates of High-cost SLHs
	Logistic Regression (N=2332)			Zero-Truncated Poisson Regression (N=147)			Logistic Regression (N=2332)			Zero-Truncated Poisson Regression (N=49)
	OR	[95% CI]	p	IRR	[95% CI]	p	OR	[95% CI]	p	IRR	[95% CI]	p
Neighborhood resident & housing characteristics
Resident characteristics
Population density	1.10	[0.93–1.29]		0.99	[0.65–1.53]		0.53	[0.33–0.84]	**	0.30	[0.10–0.87]	*
Racial/ethnic minorities
% Hispanic	1.07	[0.90–1.28]		0.86	[0.52–1.44]		0.20	[0.10–0.39]	***	1.18	[0.35–3.91]
% African American/Black	1.34	[1.19–1.51]	***	0.79	[0.54–1.17]		0.76	[0.46–1.24]		0.37	[0.05–2.67]
% Asian	0.61	[0.46–0.81]	**	1.17	[0.61–2.23]		0.73	[0.46–1.17]		0.46	[0.06–3.41]
% w/o high school diploma	1.06	[0.89–1.26]		0.77	[0.46–1.30]		0.21	[0.11–0.43]	***	0.68	[0.10–4.58]
% Unemployed	1.17	[1.01–1.36]	*	1.46	[1.02–2.07]	*	0.69	[0.47–1.00]		0.52	[0.21–1.29]
% Below poverty	1.22	[1.04–1.42]	*	0.89	[0.59–1.34]		0.56	[0.35–0.89]	*	0.81	[0.33–1.98]
Housing characteristics
Median home value	0.79	[0.65–0.97]	*	1.23	[0.84–1.82]		3.01	[2.20–4.10]	***	1.35	[0.67–2.74]
Vacancy rate	1.08	[0.94–1.23]		0.96	[0.61–1.51]		1.12	[0.97–1.30]		1.29	[0.94–1.78]
Availability of alcohol outlets
Number of bars	0.95	[0.80–1.14]		1.40	[1.04–1.90]	*	1.30	[1.10–1.53]	**	1.07	[0.79–1.45]
Number of off-premises outlets	1.16	[0.99–1.35]		1.10	[0.84–1.45]		1.38	[1.06–1.79]	*	0.84	[0.55–1.29]
Accessibility of recovery resources
Number of treatment facilities	1.32	[1.17–1.49]	***	1.16	[1.09–1.24]	***	1.25	[1.06–1.47]	**	1.05	[0.81–1.36]
Number of AA meetings	1.13	[1.01–1.28]	*	1.15	[1.01–1.31]	*	1.19	[1.00–1.41]	*	0.92	[0.70–1.22]
Number of other self-help meetings	1.22	[1.10–1.36]	***	1.12	[0.95–1.32]		1.13	[0.92–1.40]		0.70	[0.38–1.31]

Notes. All models included a spatial lag covariate (rho) representing the spatially-lagged dependent variable. Rho was significant on all bivariate logistic regression models. All zero-truncated Poisson regression models included an exposure variable representing the natural log of the total number of housing units in the neighborhood (constrained to 1) as well as a rho. Rho was not significant in models predicting number of affordable SLHs but was significant in model predicting numbers of high-cost SLHs.

^*p<0.05;

^**p<0.01;

^***p<0.001