Conclusion
due to the exclusion or inclusion of equivocal results from the analysis and due to the prevalence of the stroke types in different settings and environmental conditions . The cost of achieving this accuracy in the diagnosis of haemorrhage was to misdiagnose 22 %, 22 %, and 18 %, respectively , of patients with infarction in Oxford , London , and Guy ' s hospitals respectively [ 14 ], which were comparable with the misdiagnosis of 26.7 % of infarction as haemorrhagic stroke in our study . But this score was uncertain for 76 ( 54.3 %) patients for identification of either haemorrhagic or ischemic stroke in this study and some variables such as level of consciousness and diastolic blood pressure can only be calculated 24 hours after the stroke and it is also difficult to calculate at the bedside , so it cannot be used in acute stroke management so that this tool is inadequate for recommending as a diagnostic screening tool at the bedside in our area of investigation even if it has good specificity for haemorrhagic stroke .
The sensitivity of Greek score of this study was 58.8 % which is much lower than the original validation study from Athens , Greek 97 % [ 9 ], West Bengal , India 80 %[ 15 ], from Cairo , Egypt 87.5 % [ 24 ] and from Addis Ababa , Ethiopia 77.8 %[ 25 ], but it is higher than the study conducted in a tertiary hospital of India 42 % [ 26 ]. The specify of this study was 88.5 % which is consistent with the study conducted in Cairo , t 89.3 % [ 24 ] Addis Ababa . 89.3 %[ 25 ], but it is much lower than the original validation study from Athens , Greek 99 % [ 9 ] and from the study conducted in West Bengal , India 99 %[ 15 ].
The overall accuracy of the Greek score of this study was 71.6 % which is slightly comparable with a study from a single Tertiary care Hospital in India 79.5 % [ 26 ], but lower than the study in Addis Ababa , Ethiopia 83.6 %[ 25 ]. The percentage of equivocal categories of Greek stroke score in this study was 55.7 % which is comparable with the study from India 51 % [ 26 ], but much higher than the study conducted in Addis Ababa , Ethiopia 39.6 % [ 25 ], and from the original validation study of Greek 7.8 % [ 9 ]. The validation study of the Greek stroke score did not show adequate accuracy which 71.6 % to identify haemorrhage from infarct . The high uncertainty cases ( 78 ) were the main limitation of this stroke score to be applied to guide the physician in the management of stroke , even if the Greek stroke score had good specificity in diagnosing haemorrhage stroke ( 88.5 %).
The original Besson score was devised to identify the non-haemorrhagic stroke from haemorrhagic stroke with a high positive predictive value ( 100 %) which means all patients with a score below 1 had a non-haemorrhagic stroke [ 10 ] which was much higher than the study conducted in different countries and in particulars of this study which was 75 %. The positive predictive value of this is also lower than the study conducted in India 98.8 %[ 15 ] and Nepal 95.2 % [ 27 ] positive predictive value of ischemic stroke . But this study is slightly comparable with the study conducted in Massachusetts , USA , 82 %[ 28 ], and the study conducted in Athens , Greece 82 % of positive predictive value for the score below 1 [ 9 ].
The sensitivity of Besson stroke score for identification of non-haemorrhagic stroke from a haemorrhagic stroke in this study was 35.3 % which is slightly comparable with the study conducted in Massachusetts , United States 38 %[ 28 ], Nepal 45.45 %[ 27 ], but lower than the study conducted in India 65 %[ 15 ] and the study conducted in Athens , Greek 82 %[ 9 ]. The specificity of Besson ' s score in this study was 88.9 % which is lower than the study conducted in Nepal at 96.3 %[ 27 ], in India at 98 %[ 15 ], and in Athens , Greece 96 %[ 9 ]. The variation of the results could be explained by sample size difference , subject selection bias , and prevalence of stroke subtype . For example , the prevalence of ischemic stroke was 68 % in Athens , Greece [ 9 ], and 58 % of ischemic stroke in Nepal [ 27 ] which is higher than this study ' s 49.3 %. Even if the Besson score is simple for calculation of the score at the bedside and no need for laboratory investigation , this study showed poor positive predictive value ( 75 %) and very poor sensitivity for ischemic stroke ( 35.3 %), and a large proportion of ischemic stroke ( 52.9 %) was not identified as ischemic stroke by using this tool in our research findings , so it is too difficult to take a risk of starting ischemic stroke management immediately during admission .
Conclusion
According to our clinical score validation study , none of the clinical score methods were adequate to identify haemorrhagic stroke from ischemic stroke . However , the Siriraj stroke score showed good sensitivity and fair overall accuracy for haemorrhagic stroke even if it had poor specificity . Therefore , Siriraj score might be used for the diagnosis of haemorrhagic stroke in areas where