Regional Variations in In-hospital Mortality after Endovascular Thrombectomy in the United States

Introduction: Endovascular thrombectomy (EVT) is emerging as a vital therapeutic option for acute strokes. However, regional variability of post-EVT outcomes have not been assessed despite existence of several other US stroke metrics that consider geographical variations. We pursue to measure regional variation in in-hospital mortality in the US during 2006-2014. The period 2006-2014 was chosen to establish a national reference of post-EVT in-hospital mortality before publication of pivotal trials in 2015 against which current real-life experience can be compared in future research.

Methods: We analyzed the National Inpatient Sample (NIS) database (2006-2014) for patients with primary diagnosis of ischemic stroke who received EVT. Patients were stratified based on the 4 US regions [Northeast (NE), Midwest (MW), South (S), & West (W)]. Chi-square test was used to compare baseline variables. Independent predictors of in-hospital mortality were calculated using multivariable logistic regression.

Results: From a weighted sample of 1,031,643 stroke patients, 6,049 (0.58%) received EVT (NE 19.1%, MW 25.2%, S 31.4%, W 24.3%, p=0.0001). Multivariate logistic regression showed lower mortality in MW (OR 0.72 ;95% CI 0.54-0.97) and W (OR 0.71; 95% CI 0.54-0.94) compared to NE. Other factors associated with greater in-hospital mortality include males (female OR 0.84; 95% CI 0.72-0.99) and older age [ORs for 45-64 y: 1.5 (95% CI 1.1-2.2); 65-84 y: 2.1 (95% CI 1.4-2.9); ≥85y: 2.9 (95% CI 1.9-4.5) (reference: < 45 y)]. Ethnicity, hospital size, and academic status of the treating hospital were not associated with in-hospital mortality. In-hospital mortality trend declined over study period with most significant decline noted in S (trend p=0.002).

Conclusion : Factors associated with lower in-hospital mortality included MW and W regions compared to NE, females and younger age group. Lack of patient-specific variables relevant for stroke outcomes are unavailable in the NIS database limiting hypothesis generation regarding the causation.