Improving the selection of true incident cases of low back pain by screening retrospective administrative data

Abstract

The difficulty in identifying the onset of low back pain (LBP) limits the capacity to determine the incidence of LBP at the population level and, further, to identify risk factors. In the literature, incidence cohorts have been built with patients initially considered LBP-free for 6-12 months prior to their selection. This 'clearance period' might not be sufficient to exclude recurrent patients having experienced previous LBP episodes and might result in a misclassification bias. Using the Canadian province of Quebec's medical administrative physicians' claims database, a cohort of prevalent claims-based recurrent LBP patients was built for 2007. The medical history of 81,329 patients was screened for a period of 11 years. Positive predictive values (PPVs), kappa statistics and a survival function were calculated to determine the optimal clearance period for capturing first-time events. The 2007 annual incidence of adult claims-based recurrent LBP was estimated at 242 per 100,000 persons. Men between 18 and 34 years of age were found to be 1.18 times more at risk than their counterparts. Altogether, the elderly (over 80 years) had 52% more new cases than the 18-34 age group. A very good convergence for PPV and kappa was found for a 7-year clearance period. This allowed determining the annual incidence from 2000 to 2007, showing a decrease of 26%. Screening the medical history of LBP patients can provide more accurate incidence estimates by limiting the over-ascertainment of first-time LBP patients. A 4- to 7-year clearance period should be considered.