Abstract T MP80: Comparison of ICD-9-CM and Clinical Diagnoses for Stroke Patients in the Paul Coverdell National Acute Stroke Registry, 2013

Background: Few large studies have examined the comparison of ICD-9-CM codes and clinical diagnoses (CDX) for acute stroke patients. We analyzed the concordance between these two coding systems for patients entered in the Paul Coverdell National Acute Stroke Registry (PCNASR). We hypothesized that discordance between patient-level ICD-9-CM codes and clinical diagnoses will result in differences in outcomes on stroke performance measures (PMs). Methods: Eleven states submitted 90,035 records between Jan 1 and Dec 31, 2013. In-hospital strokes and records with multiple or missing CDXs were excluded from the analysis, resulting in a final sample size of 85,024. Clinical diagnoses were defined as: ischemic stroke (IS), transient ischemic attack (TIA), subarachnoid hemorrhagic stroke (SAH), intracerebral hemorrhagic stroke (ICH), stroke not specified (SNS), admission for carotid intervention only (CI), and other. Data was analyzed by comparing the documented principal ICD-9-CM code and final CDX of each patient. Any discordance between ICD-9-CM codes and CDX was analyzed further by examining its effect on outcomes of ten stroke performance measures (PMs) collected by PCNASR. Results: The agreement rate between ICD-9-CM codes and clinical diagnoses differs by stroke type (94.1% (IS) to 1.6% (SNS)) (Table 1). Discordance in diagnosis led to differences in outcomes on stroke PMs (range: difference of 0.1-1.03 percentage points; p-value < 0.05 for 5 measures), with 8 out of the 10 PMs having improved performance when ICD-9-CM codes are used (data not shown). Using only cases where the CDX is the same as the ICD-9-CM results in even higher adherence to PMs. Conclusion: In conclusion, due to discrepancies between ICD-9-CM codes and clinical diagnoses, outcomes on stroke PMs can vary significantly, depending on the diagnosis method used.