Incremental Variance in Low-Value Test Ordering Explained by Specific Physician Communication Behaviors
| Variables, by Model | Adjusted Probability of Test Ordering* (95% CI) | P Value | Pseudo-R2 |
|---|---|---|---|
| Base model† | — | — | 9.3% |
| Series 1‡ | |||
| Normalization | 4.7% (−23.6 to 32.9%) | .75 | 9.3% |
| Risks outweigh benefits | −22.6% (−33.1 to 12.0) | <.001 | 15.8% |
| Reassurance | −19.2% (−37.7 to −0.8) | .04 | 20.3% |
| Watchful waiting | −38.6% (−43.6 to −33.6) | <.001 | 53.1% |
| Evidence-based recommendations | −15.7% (−40.3 to 8.9) | .21 | 9.6% |
| Series 2§ | 16.5% | ||
| Normalization | 6.4% (−23.4 to 36.2%) | .67 | |
| Risks outweigh benefits | −21.9% (−32.5 to −11.2%) | <.001 | |
| Reassurance | −14.7% (−31.9 to 2.6%) | .10 | |
| Evidence-based recommendations | 1.8% (−19.9% to 23.5%) | .87 | |
| Series 3‖ | 55.7% | ||
| Normalization | 9.8% (−11.1 to 30.6%) | .67 | |
| Risks outweigh benefits | −6.7% (−15.2 to 1.7%) | .12 | |
| Reassurance | 5.4 (−12.9 to 23.6) | .57 | |
| Watchful waiting | −38.0% (−44.3 to −31.7%) | <.001 | |
| Evidence-based recommendations | 10.3% (−2.2 to 22.7%) | 0.11 |
↵* Adjusted for the randomized controlled trial intervention.
↵† The base model included study arm and standardized patient (patient with back pain requesting magnetic resonance imaging, woman requesting dual-energy x-ray absorptiometry, patient with headache requesting magnetic resonance imaging). Headache was associated with significantly less test ordering than back pain.
↵‡ In series 1, each communication behavior was included individually in separate models with base model variables.
↵§ In series 2, all communication behaviors, except watchful waiting, were included simultaneously together with base model variables.
↵‖ In series 3, all communication behaviors were included simultaneously together with the base model.
CI, confidence interval.