CONTEXT: Numerous models have been presented for the prognosis in acute stroke; however they have been criticized for being difficult to use, and few have been validated in independent samples. OBJECTIVES: To develop simple risk score models for 1-year mortality in acute stroke in patients > 60 years old and validate the models. DESIGN: From a cohort of 2321 consecutive patients > 60 years of age with acute stroke in one hospital, we randomly selected 800 patients for chart review. Among 737 patients with validated acute stroke, we randomly split the sample into (1) a derivation (60%; n = 442) and (2) a validation sample (40%; n=295). We used logistic regression to develop three models with 2-4 covariates and a corresponding risk score from the derivation sample. The models were validated using area under the receiver operating curves. RESULTS: Three risk score models for 1-year mortality after stroke were developed using combinations of age, Canadian Neurological Scale score (CNSscore) (< or = 3.5 = 0, >3.5 = 1), Charlson comorbidity index and stroke type (ischemic = 0, hemorrhagic = 1). Both 2-variable (Age - 60 + (30*CNSscore)), 3-variable (Age - 60 + (30*CNSscore) + 4*Charlson)) and 4-variable (Age - 60 + (25*CNSscore) + (5*Charlson) + (18*Stroke type)) models reliably predicted the outcome with an area under the receiver operating curve ranging 0.71 to 0.72. CONCLUSIONS: Simple models incorporating two to four covariates reliably predicted 1-year mortality. Such models can be used to stratify prognosis in clinical practice, research or intervention trials.
CONTEXT: Numerous models have been presented for the prognosis in acute stroke; however they have been criticized for being difficult to use, and few have been validated in independent samples. OBJECTIVES: To develop simple risk score models for 1-year mortality in acute stroke in patients > 60 years old and validate the models. DESIGN: From a cohort of 2321 consecutive patients > 60 years of age with acute stroke in one hospital, we randomly selected 800 patients for chart review. Among 737 patients with validated acute stroke, we randomly split the sample into (1) a derivation (60%; n = 442) and (2) a validation sample (40%; n=295). We used logistic regression to develop three models with 2-4 covariates and a corresponding risk score from the derivation sample. The models were validated using area under the receiver operating curves. RESULTS: Three risk score models for 1-year mortality after stroke were developed using combinations of age, Canadian Neurological Scale score (CNSscore) (< or = 3.5 = 0, >3.5 = 1), Charlson comorbidity index and stroke type (ischemic = 0, hemorrhagic = 1). Both 2-variable (Age - 60 + (30*CNSscore)), 3-variable (Age - 60 + (30*CNSscore) + 4*Charlson)) and 4-variable (Age - 60 + (25*CNSscore) + (5*Charlson) + (18*Stroke type)) models reliably predicted the outcome with an area under the receiver operating curve ranging 0.71 to 0.72. CONCLUSIONS: Simple models incorporating two to four covariates reliably predicted 1-year mortality. Such models can be used to stratify prognosis in clinical practice, research or intervention trials.
Authors: L P Kammersgaard; H S Jørgensen; J A Rungby; J Reith; H Nakayama; U J Weber; J Houth; T S Olsen Journal: Stroke Date: 2002-07 Impact factor: 7.914
Authors: Gustavo Saposnik; Robert Cote; Muhammad Mamdani; Stavroula Raptis; Kevin E Thorpe; Jiming Fang; Donald A Redelmeier; Larry B Goldstein Journal: Neurology Date: 2013-06-28 Impact factor: 9.910