Distance Traveled and Cross-State Commuting to Opioid Treatment Programs in the United States
Table 2
Multilevel model predicting patient travel distance to OTP.
Predictor
95% Confidence interval
Zero-order correlation
Regression coefficient
SE
Lower
Est.
Upper
Block one
USA Region
Southeast versus Northeast
0.145
1.354
1.808
2.413
Midwest versus Northeast
0.170
1.015
1.424
2.000
West versus Northeast
−0.215
0.169
0.576
0.807
1.130
Urbanicity
250 K–1 M versus Metro >1 M
0.129
1.051
1.359
1.758
<250 K versus Metro >1 M
0.160
1.223
1.683
2.316
Block two
Patient ZIP code area
0.006
1.345
1.360
1.375
Program ZIP code area
0.036
0.050
0.938
1.037
1.146
Block three
Age
30–43 versus 18–29
−0.021
0.013
0.954
0.979
1.004
43–81 versus 18–29
0.016
0.871
0.899
0.927
Female
<.01
−0.019
0.012
0.958
0.981
1.005
Race/ethnicity
Hispanic versus non-Hispanic white
0.022
0.680
0.709
0.740
Black versus non-Hispanic white
0.025
0.605
0.635
0.666
Other versus non-Hispanic white
0.036
0.836
0.896
0.961
Employed
0.005
0.012
0.972
0.995
1.019
Block four
Pain a reason for treatment
.01
−0.011
0.012
0.965
0.989
1.013
First methadone treatment
0.004
0.012
0.980
1.004
1.028
Strong urge to use
<0.001
0.018
0.966
1.000
1.035
Severe withdrawal
0.003
0.014
0.976
1.003
1.030
Prescription opioid use in past 30 days
0.016
1.051
1.085
1.119
Heroin use in past 30 days
−0.031
0.017
0.939
0.970
1.002
Interval estimates for each predictor have been exponentiated and can be interpreted as how travel distance is multiplied given a one-unit change in the predictor (or a contrast between one level of a categorical predictor and the reference category for that predictor). For predictors with multiple categories (i.e., USA Region, Urbanicity, Age, and Race/Ethnicity), the zero-order correlation is the multiple correlation when log distance is regressed on dummy variables.
*
; **.