Discussion
Discussion
This study was the first prospective study in Ethiopian hospitals with under-resourced settings . The sample size of our research was larger than the Siriraj and Greek scores . The proportion of haemorrhagic and ischemic strokes ( 70 vs . 68 ) respectively , which is consistent with the original study wherein the Siriraj score was validated ( 142 vs . 64 )[ 11 ] and also with the study conducted in Ethiopia for validation of Siriraj and Greek stroke score ( 29 vs . 20 )[ 12 ] and ( 49 vs . 42 )[ 13 ], respectively . However , it is inconsistent with the original validation study of Greek and Guys where its ischemic proportion was higher than the finding of this study ( 31 vs . 137 )[ 9 ] and ( 186 vs . 42 )[ 14 ] respectively .
The sensitivity of the Siriraj score in the diagnosis of haemorrhagic stroke was 83.9 %, which is comparable with the study conducted in India at 84 %[ 15 ] and it is slightly comparable with the original validation of Siriraj score in Thailand at 89.3 %[ 11 ], but lower than the study conducted in North-East Nigeria and South-East Nigeria which were 94.4 %[ 16 ] and 94 [ 17 ] respectively for differentiation of haemorrhagic stroke from ischemic stroke . However , it was higher than the study in Pakistan at 71.4 %[ 18 ], South Africa at 60 %[ 19 ], Tunisia at 60 %[ 20 ], and Ethiopia at 77 %[ 12 ]. This implies that the specificity of this study was 68.6 % which is slightly similar to the study conducted in India which showed 71 % of specificity for haemorrhagic stroke [ 18 ], but it is much lower than the study conducted in different parts of the world such as in India , Pakistan , South Africa , Tunisia 85 %[ 20 ] and South-East Nigeria which was 89 %[ 15 ], 81 %[ 18 ], 88 %[ 19 ] and 92 %[ 17 ] respectively .
The overall accuracy of this validation study was 82 % which is slightly comparable with the study conducted in North-East Nigeria at 84.6 %[ 16 ] and Tunisia at 81 %[ 20 ]. However it is higher than the study conducted in TikurAnbesa Teaching Hospital , Ethiopia 69.2 % [ 12 ], and the study conducted in Pakistan at 76.3 % [ 18 ], but lower than the original validation study in Thailand at 90.3 %[ 11 ], and the study conducted in South-East Nigeria 93 %[ 17 ]. Siriraj ' s stroke score showed a 22 % equivocal category in this study which is slightly comparable with the study conducted in Tunisia at 17 %[ 20 ], in Thailand at 20 % [ 11 ], and in North-East Nigeria at 22 %[ 16 ]. Our findings show possibilities of variability among clinical validation score diagnostic tools in different hospital settings within or outside any country of interest unless otherwise , quantitative data is also compared with qualitative information . Therefore , the validation scores can still be applied to categorize stroke subtypes in very low-equipped hospitals provided that both the structural and functional MRI will coincide .
Definitely , the Siriraj validation score seems to be more reliable in under-resourced Ethiopian hospital settings , in order to make differential diagnoses of acute stroke patients . The discrepancy in results could be due to differences in settings , variations in sample size , and differences in patients ’ transcultural and methodological variations . The ideal score to predict stroke type should have a good balance between validity , simplicity , and utility [ 20 ]. Since the Siriraj score is simple to collect and calculate , and available immediately at the bedside , it is also used to save costs in low-income countries in Africa like Ethiopia . So , it is better to use the Siriraj scoring system for a clinical and bedside diagnosis of haemorrhagic stroke in the setting where a brain CT scan is not available , especially in rural health centres or hospitals .
The Allen score validation study was the first prospective study in the Ethiopian context . The sensitivity of Guy ' s score of this study for differentiation of haemorrhagic stroke from ischemic was 47.8 % which is higher than the study conducted in South Africa ( 34 %) among populations that have high haemorrhagic stroke prevalence [ 19 ]. However , it is much lower than the study conducted in Oxford at 81 % and in London at 88 % [ 14 ], and also from Guy ' s hospital at 93 % [ 21 ]. It was lower than the study conducted in different parts of India at 63 %[ 15 ], 81 %[ 22 ], and 88.23 % for haemorrhage [ 23 ]. The specificity of this study was 89.7 % which is higher than the study in India among 130 stroke patients 76 %[ 22 ], but lower than the study conducted in other parts of India among 200 stroke patients which were 95 % [ 15 ] and the study in South Africa 95 % [ 19 ].
The overall accuracy of Guy ’ s score from this study was 74.5 % which is slightly comparable with the study from Oxford with an overall predictive accuracy of 78 %[ 14 ] and from London with an overall predictive accuracy of 82 %[ 14 ], but lower than from the study of Guys hospital 90 % [ 21 ] and from south India 91 %[ 23 ]. The overall accuracy variation of different studies may have occurred