Validation of Stroke Diagnosis in the National Registry of Hospitalized Patients in the Czech Republic

Background Stroke is a common cause of mortality and morbidity in Eastern Europe. However, detailed epidemiological data are not available. The National Registry of Hospitalized Patients (NRHOSP) is a nationwide registry of prospectively collected data regarding each hospitalization in the Czech Republic since 1998. As a first step in the evaluation of stroke epidemiology in the Czech Republic, we validated stroke cases in NRHOSP. Methods Any hospital in the Czech Republic with a sufficient number of cases was included. We randomly selected 10 of all 72 hospitals and then 50 patients from each hospital in 2011 stratified according to stroke diagnosis (International Classification of Diseases Tenth Revision [ICD-10] cerebrovascular codes I60, I61, I63, I64, and G45). Discharge summaries from hospitalization were reviewed independently by 2 reviewers and compared with NRHOSP for accuracy of discharge diagnosis. Any disagreements were adjudicated by a third reviewer. Results Of 500 requested discharge summaries, 484 (97%) were available. Validators confirmed diagnosis in NRHOSP as follows: transient ischemic attack (TIA) or any stroke type in 82% (95% confidence interval [CI], 79-86), any stroke type in 85% (95% CI, 81-88), I63/cerebral infarction in 82% (95% CI, 74-89), I60/subarachnoid hemorrhage in 91% (95% CI, 85-97), I61/intracerebral hemorrhage in 91% (95% CI, 85-96), and G45/TIA in 49% (95% CI, 39-58). The most important reason for disagreement was use of I64/stroke, not specified for patients with I63. Conclusions The accuracy of coding of the stroke ICD-10 codes for subarachnoid hemorrhage (I60) and intracerebral hemorrhage (I61) included in a Czech Republic national registry was high. The accuracy of coding for I63/cerebral infarction was somewhat lower than for ICH and SAH.