A Multiple Imputation Model for Ascertaining Chronic Disease Cases in Administrative Data: Application to Osteoporosis

• Activity Number: 351
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Lisa M. Lix*+ and Heather J. Prior and Chel Hee Lee and Marina Yogendran
• Companies: University of Saskatchewan and Manitoba Centre for Health Policy and University of Saskatchewan and Manitoba Centre for Health Policy
• Keywords: misclassification ; missing data ; administrative health databases ; chronic disease ; bias

Abstract

Previous research shows that administrative health databases, including physician and hospital data, have low sensitivity for ascertaining cases of many chronic diseases. A measurement error (ME) model with multiple imputation (MI) was used to correct for under-ascertainment bias in physician billing claims from Manitoba, Canada when estimating osteoporosis prevalence. The validation data were bone mineral density tests from fiscal years 1999/00 to 2000/01. The unadjusted osteoporosis prevalence estimate was 3.3% (95% confidence interval [CI]=3.2, 3.3). A model that included covariates of patient sociodemographic, drug treatment, fracture, and physician characteristics resulted in good discriminative performance (c-statistic = 0.81; 95% CI=0.80, 0.83) and the estimate of prevalence using MI (13.4%; 95% CI=12.2, 14.6) was comparable to a previously published Canadian estimate (15.8%).