Abstract W MP51: ICD-9 Code Positive Predictive Value for Stroke Subtype in a Population-Based Epidemiology Study

Abstract

Introduction: Population-based epidemiological studies contribute substantially to the knowledge of stroke incidence, temporal trends, stroke subtype distribution, and risk factor prevalence. Often data are gathered from ICD-9 codes in administrative databases. We instead gathered data through labor-intensive inspection of medical records to examine the positive predictive values (PPV) of stroke ICD-9 codes. Methods: The Greater Cincinnati/Northern Kentucky Stroke Study measures temporal trends in the incidence rates in a biracial population of 1.3 million. Discharge lists with primary and secondary ICD-9 codes 430-436 from the 15 area hospitals for 2010 were obtained. Detailed information from medical records of stroke and TIA cases was abstracted by trained research nurses and reviewed by a stroke physician, who determined if the event was a case and, if so, the stroke subtype (SAH, ICH, infarct, or TIA). Results: A total of 7691events were reviewed. Positive predictive values (PPV) for the primary ICD-9 codes for specific subtypes varied from 48.6% to 90.1%. All primary and secondary ICD-9 codes, stroke subtypes, and percent yield are presented in the Table. Interestingly, 293 cases were found coded under a different primary ICD-9 subtype. An additional 497 cases were found under secondary ICD-9 codes. Most events coded under 433.x0 were non-cases, as expected. Many cases coded 435 were not found to be true TIAs, and code 436 yielded very few cases. Conclusions: Stroke cases captured through codes 430-436 were not necessarily in the primary position or coded to the correct stroke subtype. Complete medical record review revealed varying degrees of accuracy for each code. This could cause stroke incidence rates to be significantly misreported if based solely upon primary subtype-defined ICD-9 codes. Population-based epidemiology studies generate the most reliable incidence estimates by combining stroke ICD-9 codes with medical record review.