Table 2. Logistic Regression Results for Buprenorphine Detectiona.
| Characteristic | Probabilityb | Adjusted odds ratioc | P valuec |
|---|---|---|---|
| Sex | |||
| Female | 0.859 (0.853-0.864) | 1 [Reference] | NA |
| Male | 0.844 (0.839-0.849) | 0.89 (0.86-0.93) | <.001 |
| Age, y | |||
| 18-24 | 0.786 (0.772-0.799) | 1 [Reference] | NA |
| 25-34 | 0.865 (0.859-0.871) | 1.75 (1.64-1.86) | <.001 |
| 35-44 | 0.877 (0.870-0.883) | 1.94 (1.81-2.07) | <.001 |
| 45-54 | 0.862 (0.854-0.871) | 1.71 (1.59-1.85) | <.001 |
| ≥55 | 0.856 (0.845-0.866) | 1.62 (1.49-1.76) | <.001 |
| US Census Division | |||
| East North Central | 0.819 (0.810-0.827) | 1 [Reference] | NA |
| East South Central | 0.882 (0.871-0.892) | 1.65 (1.54-1.79) | <.001 |
| Mid-Atlantic | 0.884 (0.875-0.893) | 1.69 (1.58-1.82) | <.001 |
| Mountain | 0.835 (0.821-0.849) | 1.12 (1.04-1.21) | .003 |
| New England | 0.881 (0.869-0.892) | 1.64 (1.51-1.78) | <.001 |
| Pacific | 0.804 (0.792-0.816) | 0.91 (0.86-0.97) | .002 |
| South Atlantic | 0.822 (0.813-0.832) | 1.03 (0.97-1.08) | .36 |
| West North Central | 0.881 (0.861-0.899) | 1.64 (1.44-1.88) | <.001 |
| West South Central | 0.830 (0.807-0.852) | 1.08 (0.97-1.22) | .17 |
| Health care practice specialty | |||
| Substance use disorder treatment | 0.836 (0.830-0.842) | 1 [Reference] | NA |
| Behavioral health | 0.870 (0.862-0.876) | 1.30 (1.24-1.38) | <.001 |
| Primary care physician | 0.847 (0.840-0.853) | 1.08 (1.04-1.13) | <.001 |
| Payer group | |||
| Private insurance | 0.847 (0.840-0.854) | 1 [Reference] | NA |
| Medicaid | 0.854 (0.848-0.860) | 1.06 (1.01-1.11) | .02 |
| Medicare | 0.852 (0.840-0.862) | 1.04 (0.96-1.12) | .35 |
| Uninsured | 0.853 (0.844-0.862) | 1.05 (0.99-1.12) | .11 |
| Collection year | |||
| 2015 | 0.850 (0.843-0.857) | 1 [Reference] | NA |
| 2016 | 0.843 (0.835-0.851) | 0.95 (0.90-1.00) | .06 |
| 2017 | 0.851 (0.842-0.860) | 1.01 (0.95-1.07) | .73 |
| 2018 | 0.861 (0.853-0.869) | 1.10 (1.04-1.16) | .001 |
| 2019 | 0.851 (0.842-0.859) | 1.01 (0.95-1.07) | .83 |
a
A logistic regression model of buprenorphine detection was performed containing collection year, US Census division, sex, age, health care specialty, and payer group. The total model fit was significant (χ2 = 1225.8; P < .001).
b
Marginal probability predictions (least square mean) and Sidak-corrected confidence intervals were estimated.
c
Adjusted odds ratios and P values were estimated for each covariate relative to the reference level.