Table 2.
Outcomes and covariates and their measurement details.
| Outcomes and covariates | Measure | Collection point | Response type or analysis | Statistical test | |
| Primary patient-reported outcomes |
|
|
|
||
|
|
Risk awareness and recall | Opioid Risk Tool and risk assessment recall | Baseline, day 14, and at 3 months | Ordinal (three options) for baseline risk and risk recall | Cohen weighted kappa (within treatment arm) and χ2 tests (crude agreement) |
|
|
Self-reported opioid use | Yes or no if taking opioids | Days 1, 2, 4-6, and 14 and at 3 months | Dichotomous (yes or no) | Zero-inflated negative binomial or zero-inflated Poisson models |
|
|
Patient-reported preference for pain relief | Patient pain relief preference survey | Baseline | Five options for patient-reported pain relief preference | Cochran-Armitage χ2 test for trend |
|
|
Agreement on pain treatment between patient preference and provider decision | Patient preference vs electronic medical record documentation |
|
Five options for patient-reported pain preference and concordance with provider decision | Cohen weighted kappa, intraclass correlation coefficient, Bland-Altman plots (within treatment arm), and χ2 tests (crude agreement) |
| Secondary patient-reported outcomes |
|
|
|
||
|
|
Days to no opioid use | Number of pain medications taken daily | Days 1,2, 4-6, and 14 and at 3 months | Continuous (number of pills of each type) | Kaplan Meier and proportional hazards (time to full functionality) |
|
|
Functional status | Back Pain Functional Scale and the 20-Item Short Form Survey from the Medical Outcomes Study | Days 1, 7, and 14 and at 3 months | Likert scale (0-10) and composite score for five items (0-50) | Kaplan Meier, proportional hazards (time to full functionality), and random-effects mixed model to measure clinically important changes in functionality |
|
|
Satisfaction with pain treatment | American Pain Society Patient Outcome Questionnaire | Baseline; days 1, 7, and 14; and at 3 months | Likert scale (1-6) and dichotomous (satisfied vs not satisfied) | χ2 tests, Mantel-Haenszel summary statistics, and general linear model with log-linear link |
|
|
Trust in provider | Trust in Physician Scale | Day 7 | Likert scale (1-5) and composite score for 11 items (11-55) | Analysis of variance or Kruskal-Wallis test |
|
|
Follow-up visits for pain | Self-report of additional provider visits | Day 14 and at 3 months | Dichotomous (yes or no visits; yes or no provide pain pills) | χ2 test and general linear model with log-linear link |
|
|
Patient-reported measure of shared decision making | CollaboRATE | Day 1 | Likert scale (1-10) | Analysis of variance or Kruskal-Wallis test |
|
|
Opioid misuse | Current Opioid Misuse Measure | At 3 months | Dichotomous score (≥9 vs <9) | χ2 test and general linear model with log-linear link |
| Covariates (subgroup analyses) |
|
|
|
|
|
|
|
Demographics | GENACISa and BRFSSb 2011 | Baseline | Varies by question | N/Ac |
|
|
Overall health | Self-rated health | Baseline and at 3 months | Likert scale (1-5) | N/A |
|
|
Medical condition (back vs renal colic) | ICD-9d codes and primary complaint | Baseline | ICD-9 codes | N/A |
|
|
Medical history | Electronic health record | Baseline | ICD-9 codes | N/A |
|
|
Risk for opioid dependency | Opioid Risk Tool | Baseline | Continuous score (0-26) risk level (high, medium, or low) | N/A |
aGENACIS: Gender, Alcohol, and Culture: An International Study.
bBRFSS: Behavioral Risk Factor Surveillance System.
cN/A: not applicable. As these were not outcomes, we do not report statistical tests; these are listed as covariates.
dICD-9: International Classification of Diseases, Ninth Revision.