Abstract 13274: Frailty is an Independent Prognostic Factor when Added to Existing Acute Ischemic Stroke Mortality Models
Introduction: Frailty is a state of decreased reserve and cumulative decline in multiple physiologic systems. Several frailty measures have been developed but have not been considered in existing stroke mortality prediction models. We sought to determine whether frailty is an independent predictor when added to existing stroke mortality models.
Methods: Clinical data from incident ischemic stroke admissions in the Cincinnati/ Northern Kentucky Stroke Study during 2005 were abstracted by research nurses. Using established methods, we developed a frailty index using 35 age-related deficits that included pre-stroke function, comorbidities, symptoms, and clinical and lab values at admission. The index was specified as the total number of deficits or frailty score (range 0 to 35). Two established stroke mortality prediction models - one for in-hospital mortality and one for 1-year mortality, were identified and applied to the data. The independent contribution of the frailty score (expressed as a 1 deficit increase) to these existing models was determined from the logistic regression models.
Results: A total of 2092 ischemic strokes were included. The median age was 72 years, 22% black, 56% female, median NIHSS 4 (IQR 2, 7). In-hospital and 1-year mortality were 8.8% and 26%, respectively. The median frailty score was 6 (IQR 4, 9). Both existing models fit the data well; the model c-statistics were both high (0.877 and 0.808, respectively) (Table). The frailty score was independently associated with higher mortality in both models; a 1 unit increase in age-related deficits was associated with a 20% and 17% increase in the adjusted odds ratio of in-hospital and 1-year mortality, respectively.
Conclusions: When added to existing mortality prediction models, a frailty score showed strong independent associations with mortality. These data suggest that adding frailty to stroke mortality prediction models could improve their predictive accuracy and clinical utility.
Author Disclosures: M.J. Reeves: None. H. Sucharew: None. J. Khoury: None. K. Alwell: None. C. Moomaw: None. B. Kissela: None. D. Woo: None. M. Flaherty: None. O. Adeoye: None. P. Khatri: None. S. Ferioli: None. D. Kleindorfer: None.
- © 2014 by American Heart Association, Inc.