Spinal imaging with prevalence information does not reduce burden but decreases opioid prescription


  • Including prevalence information about features on lumbar spine images did not affect subsequent health care use but did reduce likelihood of opioid prescription.

Why this matters

  • The authors argue that this intervention of benchmarking imaging results is simple, inexpensive, and easily implemented and that it may have relevance for other imaging and genetic tests.

Study design

  • Objective: to evaluate the effect of benchmarking with prevalence data from people without pain as part of routine spinal imaging for back pain on subsequent health care utilization.
  • People whose physician had ordered lumbar spinal imaging were randomized to a control group (n=117,455; 57.8% were women) or an intervention group (n=121,431; 57.2% were women).
  • The imaging reports for the intervention group included age- and modality-appropriate benchmarks for common findings based on prevalence.
  • The primary endpoint was health care use relating to the spine within a year of imaging.

Key results

  • There was no difference between the control and intervention groups for the primary endpoint of health care use (difference -0.7%, 95% confidence interval [CI] -2.9–1.5%; P=0.54).
  • There was a small decrease in the secondary endpoint of likelihood of opioid prescription within a year of imaging (odds ratio 0.95; 95% CI 0.91–1.00; P=0.04).
  • The authors concluded that benchmark prevalence information as part of spine imaging reports did not decrease spine-related health care use but did reduce opioid prescriptions.


  • Opioid prescribing decreased generally during the course of the study – despite accounting for this in modelling, it remains a potential confound.
  • No data collected on functional status, pain, or psychosocial functioning.
  • Generalizability of results to healthcare systems with more restrictions on imaging and for-profit hospitals not known.