Table 3. Patient Subgroup Analysisa.
| Item | <3-mo Subgroup | 3-mo Subgroup | ||||
|---|---|---|---|---|---|---|
| 18-mo Change | EMAF plus PF plus PPC vs EMAF | 18-mo Change | EMAF plus PF plus PPC vs EMAF | |||
| EMAF | EMAF plus PF plus PPC | EMAF | EMAF plus PF plus PPC | |||
| Mean MME | ||||||
| Estimate, mg/d (95% CI) | −1.1 (−2.8 to 0.5) | −3.4 (−5.0 to −1.8) | −2.2 (−4.5 to 0.01) | −1.5 (−4.0 to 1.0) | −2.4 (−5.2 to 0.3) | −0.9 (−4.6 to 2.7) |
| Cohen d | −0.08 | −0.24 | −0.16 | −0.05 | −0.09 | −0.03 |
| P value | .18 | <.001 | .05 | .24 | .08 | .62 |
| Urine screen | ||||||
| Estimate, % (95% CI) | 4.5 (2.3 to 6.7) | −0.6 (−2.7 to 1.5) | −5.1 (−8.0 to −2.1) | 13.3 (7.8 to 18.8) | −1.1 (−6.9 to 4.6) | −14.4 (−22.3 to −6.5) |
| Cohen d | 0.39 | −0.05 | −0.44 | 0.45 | −0.04 | −0.48 |
| P value | <.001 | .59 | <.001 | <.001 | .71 | <.001 |
| Treatment agreement | ||||||
| Estimate, % (95% CI) | 2.8 (0.7 to 4.9) | −0.3 (−2.4 to 1.7) | −3.1 (−6.0 to 0.3) | 16.2 (10.1 to 22.2) | 2.3 (−3.8 to 8.5) | −13.8 (−22.2 to −5.4) |
| Cohen d | 0.30 | −0.04 | −0.34 | 0.55 | 0.08 | −0.47 |
| P value | .01 | .75 | .03 | <.001 | .46 | .001 |
| PHQ-2 screen | ||||||
| Estimate, % (95% CI) | 14.0 (9.8 to 18.2) | 14.3 (10.2 to 18.4) | 0.3 (−5.5 to 6.1) | 16.5 (10.5 to 22.5) | 11.2 (5.1 to 17.3) | −5.3 (−13.6 to 3.0) |
| Cohen d | 0.61 | 0.62 | 0.01 | 0.54 | 0.36 | −0.17 |
| P value | <.001 | <.001 | .92 | <.001 | <.001 | .21 |
| PEG-3 screen | ||||||
| Estimate, % (95% CI) | 1.1 (−2.4 to 4.6) | 6.7 (3.3 to 10.1) | 5.6 (0.8 to 10.3) | 4.5 (0.1 to 9.1) | 8.6 (4.0 to 13.2) | 4.1 (−2.4 to 10.6) |
| Cohen d | 0.08 | 0.49 | 0.41 | 0.23 | 0.44 | 0.21 |
| P value | .53 | <.001 | .02 | .05 | .04 | .22 |
| Aggregated P valueb | <.001 | <.001 | <.001 | <.001 | <.001 | <.001 |
Abbreviations: AF, audit and feedback; EM, educational meetings; MME, morphine milligram equivalent daily dose; PEG-3, Pain, Enjoyment, General Activity; PF, practice facilitation; PHQ-2, Patient Health Questionnaire–2; PPC, prescriber peer consulting.
Subgroups are patients with opioid prescriptions in each of the most recent 3 months (3-month subgroup) vs patients with 0, 1, or 2 opioid prescriptions in the most recent 3 months (<3-month subgroup). Values were estimated using a linear mixed-effects model with the restricted maximum likelihood method. This approach was applied to assess the change in the fixed-effect curve from baseline to the end of the intervention period (18 months), within and among the 4 study arms.
The geometric mean of P values was used to calculate the aggregated P values, which can be used as a method for multiple-comparison correction to assess the null hypothesis that the true changes (or differences in changes) for all 5 of the outcomes are uniformly zero.