Background: ICD-9-CM codes recorded in administrative databases are often used to identify patients with specific clinical conditions. We determined if there are variations in the accuracy of stroke and transient ischemic attack (TIA) ICD-9-CM codes based on hospital characteristics and stroke severity.
Methods: We used the records of patients discharged from hospitals in the Paul Coverdell National Acute Stroke Program in 2013. Diagnosis categories included ischemic stroke, TIA, subarachnoid hemorrhage, and intracerebral hemorrhage. We compared the agreement between the principal ICD-9-CM code and attending physician’s clinical diagnosis. The effects of hospital characteristics and stroke severity (National Institutes of Health Stroke Scale score, NIHSS) on percent agreement were assessed.
Results: Among 67,442 patient records with hospital characteristic data, agreement between ICD-9-CM codes and the clinical diagnosis for ischemic stroke was higher for hospitals with stroke units, stroke teams, larger numbers of beds, and locations in metropolitan areas (P<0.05) (Table). For 55,373 records with a documented NIHSS at admission, agreement was lower for mild ischemic strokes (NIHSS 0-7) compared with more severe strokes (P<0.001); disagreements were commonly due to the patient having a carotid endarterectomy (potentially reflecting continuity of care for a stroke event) or TIA rather than a stroke.
Conclusion: Systematic variations in the accuracy of ICD-9-CM codes by hospital characteristics and stroke severity may affect case identification in epidemiologic studies and have implications for hospital-level quality metrics.