We aim to facilitate the development of a high performing clinical score for stroke classification to enable early and accurate diagnosis of stroke type. We present a systematic review of validation studies on Siriraj(SS), Guys Hospital/Allen(GHS/AS), Greek(GS) and, Besson(BS) Scores from 2001 to 2021. We further perform a comparative analysis of the performance of these scores.

We follow PRISMA guidelines, to identify a total of 22 studies for the defined scope and period. For all the included studies we derive a more accurate confusion matrix by including the equivocal predictions and calculate the corresponding performance metrics. We further compare them with results reported by the studies. 

Figure: PRISMA flow-diagram.

SS is most evaluated score, by 21 studies, followed by GHS/AS, GS, and BS by 12, 3, and 1 respectively.
Comparatively lower values of calculated performance metrics than the ones reported by the studies reflect over estimated performance of scores. SS performs better than the rest. However, GS and BS haven't been widely evaluated.

Table: Range of Reported and Calculated values of performance metrics.

SS perform better than others. However, inconsistency and high variability in performance across
different demographics makes its reliability questionable. Also, the equivocal interval in SS, GHS/AS and GS, result in poor classification capability. We emphasize the need of a global dataset with statistically important clinical features for developing a clinical score which avoids equivocals and generalizes globally.