Skip to main content

Table 3 Multinomial logistic multilevel model predicting type of opioid use: Helmert contrasts of individual age

From: Age differences in heroin and prescription opioid abuse among enrolees into opioid treatment programs

  PO & Heroin vs. Heroin Only PO Only vs. Heroin Only
  Odds Ratio 95% Confidence Interval   Odds Ratio 95% Confidence Interval  
  Lower Estimate Upper p Lower Estimate Upper p
Individual Age Within Programs         
20-25 vs. Younger 0.859 0.962 1.077 0.497 1.017 1.156 1.314 0.027
26-29 vs. Younger 0.873 0.915 0.959 <.001 0.930 0.980 1.033 0.446
30-39 vs. Younger 0.903 0.927 0.951 <.001 0.948 0.976 1.004 0.093
40-49 vs. Younger 0.913 0.931 0.948 <.001 0.944 0.964 0.986 0.001
50-59 vs. Younger 0.923 0.940 0.958 <.001 0.940 0.960 0.981 <.001
60-70 vs. Younger 0.888 0.926 0.965 <.001 0.927 0.973 1.021 0.269
Interview Year         
2006 vs. 2005 0.869 0.959 1.058 0.403 1.042 1.167 1.307 0.007
2007 vs. 2005 0.978 1.091 1.217 0.119 1.164 1.317 1.489 <.001
2008 vs. 2005 1.864 2.071 2.300 <.001 1.962 2.215 2.500 <.001
2009 vs. 2005 1.933 2.146 2.382 <.001 2.134 2.404 2.708 <.001
Beale Urbanicity         
Medium vs. High Density 1.883 2.130 2.410 <.001 2.783 3.155 3.578 <.001
Low vs. High Density 2.837 3.792 5.067 <.001 4.706 6.220 8.221 <.001
U.S. Region         
Southeast vs. Northeast 2.093 2.468 2.911 <.001 15.955 18.867 22.310 <.001
Midwest vs. Northeast 3.092 3.750 4.548 <.001 9.259 11.281 13.745 <.001
West vs. Northeast 0.899 0.997 1.105 0.954 2.241 2.526 2.846 <.001
  1. Note: OTP averages of age contrasts were included in the model but are not shown here because it is not clear how to interpret them and the primary purpose of including them in the model is partialling out OTP-level age differences so that individual patient age contrasts can be unambiguously interpreted as pooled within-OTP associations. All associations were tested for significance using z-tests, in which the coefficient was divided by its standard error and compared with the normal distribution. Coefficients and confidence limits were exponentiated to odds ratios for presentation here. Sample sizes for this analysis were 29107 individual enrolees nested within one of 85 OTPs.