Purpose: To compare a new CMS pain screening quality indicator (QI) (visit-level cumulative average) with alternative specifications (visit-specific, patient-level cumulative average) of pain management QIs to evaluate the most efficient, accurate method for characterizing the quality of pain management.
Methods: We developed, validated, and piloted a QI to assess whether or not pain screening occurred at outpatient encounters. Trained nurses abstracted electronic clinical records using a computer-based abstraction tool for 118 veterans with metastatic cancer at the Veterans Administration Greater Los Angeles (GLA). Data were based on total number of outpatient visits (January 1, 2006 to December 31, 2007
). We calculated pain screening rates using three scoring mechanisms: visit-specific; visit-level cumulative average; and patient-level cumulative average. We used two-sided z-scores to compare proportions and to estimate the visit numbers to achieve maximum precision and a stable estimate of quality.Results: We calculated pain screening rates for 106 eligible patients (90% of cohort) receiving at least one outpatient visit during the study interval. Patients received on average 12 visits each during two years median follow-up. In 80% of patients, 8 visits were achieved in less than 6 months. At any one encounter, pain screening occurred for between 22% (visit #1; n=106; CI 14%, 30%) and 46% (visit #8; n=54; CI 33%, 60%) of patients. Rates varied non-monotonically between visits, and precision declined rapidly with increasing visit numbers because of cohort attrition due to death, with 95% confidence interval spread exceeding 35% beyond visit #12. Visit-level and patient-level cumulative average pass rates provided comparable results, with average pass rates of 34% (95% CI 31%; 37%) and 30% (22%; 40%), respectively. They differed, however, in number of visits required to reach maximum precision or steady state, occurring at visits #14 and #11, respectively, for visit-level, compared with visits #4 and #8 for patient-level.
Conclusion: We identified an efficient strategy for pain screening quality assessment, indicating that assessment cycles may be instituted in half the length required by the CMS indicator. Because patient dropout is serious issue in applying quality measurement to advanced illness populations, shorter cycles may be more feasible in many practices.
Candidate for the Lee B. Lusted Student Prize Competition