The accuracy of hospital discharge coding for hemorrhagic stroke

Background: Hospital discharge data is used in monitoring stroke epidemiology, and ensuring adequate resource allocation to treatment programs. Previous studies have reported variable accuracy levels for such data. We present the first study assessing the accuracy of International Classification of Diseases 10^th Edition (ICD-10) discharge coding for hemorrhagic stroke in England. Methods: We identified all patients with a primary diagnosis of intracerebral hemorrhage (ICH; ICD-10 code: 161.x) and subarachnoid hemorrhage (SAH; 160.x) admitted to the Newcastle upon Tyne Hospitals from 2002-2007. Positive predictive values (PPV) were calculated through validation with patient notes. Results: Hospital discharge coding identified 978 ICH and 1169 SAH admissions over the six years. The number of diagnoses verified by patient notes was excellent for both ICH (n = 938) and SAH (n = 1123), with a PPV of 95.9% for ICH (95% confidence interval, CI = 94.5-97.0%) and 96.1% for SAH (95% CI = 94.8-97.0%). The coding errors observed were largely expected, with different types of stroke miscoded as ICH and SAH. Conclusions: The accuracy of ICD-10 hospital discharge coding for hemorrhagic stroke was excellent. However, further research is needed to find ways to further improve its accuracy.