Abstract WP195: ICD9 Ischemic Stroke Code Validity Biased By Demographics
Background: Collecting and reporting stroke incidence is a time intensive and costly endeavor, so many researchers prefer to utilize administrative databases to ascertain rates and trends. Incidence rates of stroke and TIAs reported in the literature varies based on differing methodologies and populations used for estimation. We sought to examine the demographic differences between the validated cases and non-cases of ischemic stroke (IS) in a population-based study.
Methods: The Greater Cincinnati/ N. Kentucky Stroke Study, a population-based epidemiologic study, collected data on all potential stroke and TIA cases among residents of a 5-county region in 2005. All medical records with discharge ICD-9 codes 430 through 436 were reviewed by study nurses, and those classified as potential cases were reviewed by stroke team physicians, who determined which events were actual verified cases and classified them by stroke subtype. We specifically examined 433.x1 and 434.x1 which are the ICD-9 codes most commonly assigned for ischemic stroke.
Results: Of the original 2335 individual medical records assigned ICD9 codes 433.x1 and 434.x1, 80% were verified as ischemic strokes. Of the 2206 verified IS cases in adults (≥18 years old) in 2005, 1873 (85%) had an ICD9 discharge code of 433.x1 or 434.x1. Potential strokes found to be non-cases were more likely to be patients who were older or living in a nursing home (Table).
Conclusions: Ischemic stroke incidence rates that are based solely on subtype-defined ICD9 codes from administrative databases are likely to be biased with respect to population demographic and thus may not accurately reflect the population incidence. Meaningful population-based epidemiology requires both ICD9 codes and medical record review to ensure accurate incidence reporting.
- © 2012 by American Heart Association, Inc.