Am Fam Physician. 2008;77(10):1457-1458
Background: The Joint Commission on Accreditation of Healthcare Organizations requires routine assessment of pain in patients in hospitals and clinics. A commonly used assessment tool is a numeric rating scale (NRS) that allows patients to rate their pain from 0 to 10, with 10 being the worst possible pain. However, it is unclear whether this scale is accurate enough to detect pain earlier than with routine care. Krebs and colleagues evaluated the accuracy of the NRS for pain assessment.
The Study: The study included adult patients presenting at a general internal medicine clinic who received routine pain assessment using the NRS. Recruitment continued until 20 percent of patients had a pain rating of 0; thereafter, only patients with a rating of 1 or more were enrolled. Patients were not told the specific reason for the study. Pain was categorized as mild (a rating of 1 to 3), moderate (4 to 6), and severe (7 to 10). Patients completed the Brief Pain Inventory (BPI) to assess mood and function, and were asked about the reasons for their visit. Pain that interfered with functioning (defined as a score of 5 or more on the BPI) and pain as a reason for the visit made up the reference standard against which the NRS was compared.
Results: The analysis included 275 of the 357 patients who originally agreed to participate in the study. The mean age of participants was 55 years. Most patients with pain reported musculoskeletal pain, and 40 percent of participants said that pain was a reason for the visit. Chronic pain (pain lasting more than six months) was reported by 55 percent of participants overall and by 77 percent of participants with pain at the time of the visit.
Pain interfering with function occurred in 37 percent of patients; the area under the curve was 0.76 compared with the reference standard, representing fair accuracy. Using a cutoff rating of 1, the NRS was 69 percent sensitive for identifying pain that interferes with function (95% confidence interval [CI], 60 to 78) and 78 percent specific (95% CI, 71 to 83). A cutoff of 4 lowered the sensitivity to 64 percent (95% CI, 54 to 72) and raised the specificity to 83 percent (95% CI, 77 to 88). Pain as a reason for the visit occurred in 40 percent of patients; the area under the curve was 0.78, representing fair accuracy. Using a cutoff rating of 1, the NRS was 71 percent sensitive for identifying pain that motivates a visit (95% CI, 62 to 79) and 81 percent specific (95% CI, 74 to 86). A cutoff of 4 lowered the sensitivity to 63 percent (95% CI, 54 to 72) and raised the specificity to 85 percent (95% CI, 79 to 90).
Conclusion: The authors conclude that the NRS has only modest accuracy in pain assessment, and it would have missed almost one third of patients with clinically important pain. This could be because the NRS does not identify patients with symptoms synonymous with, but not identical to, pain—and because patients might refer their answer only to the moment of questioning. The authors suggest that universal pain screening may have limited use in identifying patients with clinically important pain.