1 Predictive risk factors for increased direct medical cost of stroke in Tunisia Safa Aouinti 1 , Dhafer Malouche 1 , Hela Ouaili-Mallek 1 , Olfa Saidi 2 , Olfa Lassoued 2 , Faycel Hentati 3 , Habiba Ben Romdhane 2 1 Engineering School of Statistics and Information Analysis, University of Carthage, Tunisia 2 Laboratory of Epidemiology and Prevention of Cardiovascular Disease-Faculty of Medicine, Tunis, Tunisia 3 National Institute of Neurology of Tunis, Tunisia ABSTRACT Background and Purpose— The economic impact of stroke is considerable for the individual and society. The purpose of this study was to evaluate the cost of management for stroke and to identify predictive risk factors that influence its increase. Methods—Identification of these factors from our prospective study of 630 patients hospitalized for stroke in 2010 at the National Institute of Neurology of Tunis, was based on polytomous logistic regressions applied following a classification of the in-hospital, post-hospitalization and annual costs. This allowed to distinct classes of each cost and reasoning about those involving patients having the highest costs and to characterize the interactions existing between these classes versus the risk factors, graphical interaction models have been designed. Results—The total estimated direct medical cost of stroke was conducted, it appeared that 289 206.974 Tunisian Dinars (US$187984) was spent taking burden of stroke patients during their hospital stay is about 491.013 Tunisian Dinars (US$319) per patient. While the overall annual budget devoted to this pathology was 772 970,560 Tunisian Dinars (US$502431) with an average of 1370.515 Tunisian Dinars (US$891) per patient. Hypertension, diabetes and cigarette Smoking led to a remarkable rise in the cost of this pathology. Conclusions—The high cost of medical management of stroke in Tunisia, requires greater educational campaigns which should focus on controlling risk factors, with a well integrated way in order to reduce the cost of care for this disease. Keywords: Stroke, direct medical cost, EM-Algorithm, polytomous logistic regressions analysis, graphical interaction models.