Table 2. Changes in Extent and Intensity of Opioid and Benzodiazepine Coprescribing Associated With Centers for Disease Control and Prevention Guideline Release.
| Population or Opioid Episode Type | Commercial Insurance | Medicare Advantage | ||||||
|---|---|---|---|---|---|---|---|---|
| Long Term | Short Term | Long Term | Short Term | |||||
| Value (95% CI) | P Valuea | Value (95% CI) | P Valuea | Value (95% CI) | P Valuea | Value (95% CI) | P Valuea | |
| Extent of coprescription: adjusted rates of overlapping opioids and benzodiazepines, mob | ||||||||
| Change in level just before to just after guideline release, % | 0.87 (0.29 to 1.45) | .003c | −0.25 (−0.35 to −0.15) | <.001c | −0.14 (−0.72 to 0.45) | .64 | −1.02 (−1.21 to −0.83) | <.001c |
| Preguideline release slope, %/y | 0.52 (0.02 to 1.03) | 0.19 (0.13 to 0.24) | 1.09 (0.56 to 1.62) | 0.44 (0.32 to 0.56) | ||||
| Postguideline release slope, %/y | −0.95 (−1.44 to −0.46) | <.001c | −0.05 (−0.12 to 0.02) | <.001c | −1.06 (−1.49 to −0.63) | <.001c | 0.47 (0.35 to 0.59) | .70 |
| Intensity of coprescription: adjusted percentage of opioid prescription days with concurrent benzodiazepines among those with any overlap in the monthb | ||||||||
| Change in level just before to just after guideline release, % | −0.08 (−0.74 to 0.57) | .80 | −0.31 (−0.68 to 0.07) | .11 | −0.32 (−0.87 to 0.22) | .25 | −0.17 (−0.56 to 0.23) | .41 |
| Preguideline release slope, %/y | 0.57 (0.04 to 1.11) | 0.19 (−0.02 to 0.39) | 0.56 (0.09 to 1.02) | 0.28 (0.03 to 0.52) | ||||
| Postguideline release slope, %/y | 0.27 (−0.25 to 0.79) | .45 | 0.30 (0.03 to 0.56) | .53 | 0.84 (0.48 to 1.20) | .37 | 0.34 (0.11 to 0.58) | .70 |
P value for change in level tests whether change in level equals 0; P value for postguideline release slope tests whether postguideline release slope equals preguideline release slope.
Adjusted rates of overlapping opioid and benzodiazepine prescriptions filled represent predictive margins from a logit model that included the covariates patient age, sex, race/ethnicity, state of residence, and Elixhauser comorbidity flags calculated on a rolling 6-month basis. Adjusted rates of opioid prescription days with benzodiazepines represent predictive margins from negative binomial models specified with the same covariates as specified in the previous sentence, using the number of opioid prescription days in the month as an exposure variable (ie, the natural log of opioid prescription days was included in the model with the coefficient constrained to be 1). Separate models were specified for each patient population and opioid episode type (ie, commercial long term, commercial short term, Medicare Advantage long term, and Medicare Advantage short term). The SEs in the models were adjusted for clustering on individual patient.
Statistically significant after controlling for familywise error rate of 0.05 within table.