Abstract| Volume 14, ISSUE 4, SUPPLEMENT , S15, April 2013

Prediction of persistent clinically important back pain among older adults beginning a new episode of care for back pain

      Despite the high prevalence of back pain among older adults, little is known concerning prognosis in this population. The objectives of this large, longitudinal observational study were to describe back pain outcomes over time and to develop a model for predicting outcomes among older patients initiating a new episode of care for back pain. We hypothesized that greater back pain duration, back pain intensity, leg pain intensity, physical disability (Roland score), pain interference (Brief Pain Inventory), depression (PHQ-2), and anxiety (GAD-2), and lower recovery expectations, at baseline would be risk factors for moderate to severe back pain intensity (5-10 on 0-10 scale of average back pain in the past week) 6 months later. 2625 patients aged 65 years or older with a new visit for back pain (i.e., none in the past 6 months) at 3 geographically diverse large U.S. health care systems enrolled and completed baseline measures. Six months later, 2225 of these patients (85%; median age 73 years, 66% female, and 72% non-Hispanic Caucasian) completed outcome measures; 961 (43%) rated their average back pain as 5 or greater. In univariate analyses, all hypothesized baseline predictors were associated significantly (P < 0.05) with 6-month back pain intensity, in the expected directions. Also, older, black, and less educated patients were more likely to have moderate to severe back pain at 6 months. A multivariable model that included all hypothesized baseline predictors plus age, gender, race/ethnicity, and education predicted 6-month back pain intensity (area under the receiver operating characteristic curve = 0.78; 95% CI, 0.76-0.80). In sum, 43% of older adults initiating a new episode of care for back pain reported moderate to severe back pain 6 months later, and a combination of baseline sociodemographic and self-report measures predicted this outcome with acceptable accuracy. Supported by AHRQ grant 1R01 HS019222.