Neuroepidemiology 2010 29;35(1):53-8. Epub 2010 Apr 29.
Department of Neurology, Liverpool Hospital and South Western Sydney Clinical School, University of New South Wales, Liverpool, N.S.W., Australia.
Background: It is important to establish the validity of diagnostic coding in administrative datasets used in stroke and transient ischemic attack (TIA) research. This study examines the accuracy of emergency department (ED) TIA diagnosis and final diagnostic coding after hospital admission.
Methods: Using administrative datasets, we identified all patients with an ED TIA diagnosis (435.9; ICD-9) admitted to Liverpool Hospital from January 2003 to December 2007. ED and hospital admission records were matched and final diagnosis codes (ICD-10-AM) recorded. All records were expertly reviewed to determine coding validity.
Results: 570 patients were admitted with an ED TIA diagnosis. According to ICD-10-AM coding, 46% had TIA, 29% stroke and 25% TIA mimic diagnoses. Expert review determined final diagnoses of TIA in 51.4%, stroke in 26.1% and TIA mimic in 22.5% of the patients. The positive predictive value of a final TIA diagnosis (ICD-10-AM) was 88.2% when subjected to expert review. TIA mimic disorders diagnosed after admission included serious conditions.
Conclusions: Half of the emergency diagnoses retained a TIA diagnosis after hospital admission. In the setting of neurological admission there were small percentage differences between coded final diagnosis for TIA, stroke and mimic and diagnoses at expert review. Admission of ED TIA cases permitted identification of TIA mimics with serious conditions requiring non-TIA management.