Abstract 5: National Institutes of Health Stroke Scale Reporting in Administrative Claims Data: Incomplete and Biased Reporting in the First Three Years

Abstract

Background: Since 2016, hospitals have been able to document International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes for National Institutes of Health Stroke Scale (NIHSS). As of 2023, Center for Medicare & Medicaid Services uses NIHSS as a risk adjustment variable. We assessed associations between patient- and hospital-level variables and contemporary NIHSS reporting using Medicare datasets. Methods: Retrospective cross-sectional analysis of acute ischemic stroke (AIS) admissions from October 1, 2016-December 31, 2019 using de-identified, national 100% inpatient Medicare datasets. Index AIS admissions were identified using the ICD-10-CM code of I63.x. Demographic information, medical comorbidities, hospital characteristics, and NIHSS were abstracted. Medicare and Mount Sinai Health System registry data were linked using a matching algorithm. We calculated documentation of NIHSS at the patient and hospital level, predictors of NIHSS documentation, change over time, and concordance with local registry data. Results: There were 770104 index AIS admissions. NIHSS was documented in 40.1% of AIS admissions and by 62.5% of hospitals. Hospitals documenting ≥1 NIHSS were more commonly teaching hospitals, in metropolitan areas, stroke certified, higher-volume, and had ICU availability. Adjusted odds of documentation were lower for patients of black race compared to white and with inpatient mortality, increasing Charlson comorbidity score, discharge home, and treatment in non-metropolitan areas. NIHSS was documented for 52.9% of Medicare cases even though it was documented in 93.1% of registry cases, and 74.7% of Medicare NIHSS scores equaled registry admission NIHSS. Conclusions: Hospitals face financial implications for risk-standardized mortality rate adjustment, and administrative data play an increasingly important in stroke health services and outcomes research. Yet, missing ICD-10-CM NIHSS data remains widespread three years after introduction of the ICD-10-CM NIHSS code, and there are systematic differences in reporting at the patient and hospital levels.