BACKGROUND: The Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classification has been widely used to determine etiology of ischemic stroke. However, interrater reliability is known to be modest. The complexity of abstraction and the interpretation of various clinical and laboratory data might limit the accuracy of the TOAST classification. In this study, we developed a computerized clinical decision support system for stroke classification that can be used in a handheld device and tested whether this system can improve diagnostic accuracy and reliability. METHODS: Based on the TOAST classification, a logical algorithm was developed and implemented on a handheld device, named iTOAST. After answering six questions using the touch interface, the stroke subtype result is displayed on the screen. Four neurology residents were randomly assigned to classify stroke subtypes using iTOAST or the conventional method (cTOAST). Using a crossover design, they classified the stroke subtypes of 70 patients. The standard subtypes were determined by three stroke experts. Correlated kappa coefficients using iTOAST compared with cTOAST were determined. RESULTS: The kappa (SE) value of iTOAST [0.790 (0.041), 95% CI: 0.707-0.870] was higher than that of cTOAST [0.692 (0.046), 95% CI: 0.600-0.782] (P<0.001). Neither sequence (P=0.857) nor period effect (P=0.999) was observed. CONCLUSIONS: The stroke classification tool using a handheld, computerized device was easy, accurate, and reliable over the conventional method. It may have additional benefit because a handheld, computerized device is accessible anytime and anywhere.
RCT Entities:
BACKGROUND: The Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classification has been widely used to determine etiology of ischemic stroke. However, interrater reliability is known to be modest. The complexity of abstraction and the interpretation of various clinical and laboratory data might limit the accuracy of the TOAST classification. In this study, we developed a computerized clinical decision support system for stroke classification that can be used in a handheld device and tested whether this system can improve diagnostic accuracy and reliability. METHODS: Based on the TOAST classification, a logical algorithm was developed and implemented on a handheld device, named iTOAST. After answering six questions using the touch interface, the stroke subtype result is displayed on the screen. Four neurology residents were randomly assigned to classify stroke subtypes using iTOAST or the conventional method (cTOAST). Using a crossover design, they classified the stroke subtypes of 70 patients. The standard subtypes were determined by three stroke experts. Correlated kappa coefficients using iTOAST compared with cTOAST were determined. RESULTS: The kappa (SE) value of iTOAST [0.790 (0.041), 95% CI: 0.707-0.870] was higher than that of cTOAST [0.692 (0.046), 95% CI: 0.600-0.782] (P<0.001). Neither sequence (P=0.857) nor period effect (P=0.999) was observed. CONCLUSIONS: The stroke classification tool using a handheld, computerized device was easy, accurate, and reliable over the conventional method. It may have additional benefit because a handheld, computerized device is accessible anytime and anywhere.
Authors: Azmil H Abdul-Rahim; David Alexander Dickie; Johann R Selvarajah; Kennedy R Lees; Terence J Quinn Journal: Trials Date: 2019-02-08 Impact factor: 2.279