Waimei Tai, MD

  • Kwok CS
  • Potter JF
  • Dalton G
  • George A
  • Metcalf AK
  • Ngeh, J.
  • The SOAR Stroke Score Predicts Inpatient and 7-Day Mortality in Acute Stroke. Stroke. 2013


    Kwok et. al recently wrote an interesting article that externally validates a predictor score of near term mortality for stroke (ischemic and hemorrhagic). The same group designed the predictor score using another cohort (unable to review the original paper as it is in press) and now they have externally validated the predictor using routinely collected data in NHS hospitals. The strength of the score is its simplicity. SOAR (0-7 scale) uses simple clinical data: stroke type (ischemic/hemorrhagic), stroke distribution (Oxford Community Stroke Project Classification), age, and pre-morbid Rankin.  For readers unfamiliar with OCSP, it was originally designed for ischemic stroke and classifies stroke as:  total anterior (TACI), partial anterior (PACI), lacunar (LACI), and posterior (POCI) circulation infarcts.

    The trouble I see with the intrinsic design of the score are two-fold: (1) as far as I’m aware, OCSP is not a routinely recorded variable in the clinical data in North America and (2) pre-morbid Rankin is also not routinely collected for clinical purposes although one may attempt to extrapolate pre-morbid Rankin based on function if that is recorded. The OCSP score maybe derived from review of imaging, so it maybe easy to score post-hoc, but given that it was initially designed for ischemic stroke alone, I am not sure of its applicability on hemorrhagic stroke.

    The simplicity of the SOAR score itself belies the hidden complexity of inputs that themselves need extrapolation from routine clinical data. For now, it is simple relative to other predictor scores available.

    The low positive predictive value of 23% for score > 3 is also concerning, as the authors mentioned even though the sensitivity (80%) and specificity (72%) seem reasonable. Eventually a score derived from more routine clinically collected information (similar to APACHE score used in intensive care) would seem more widely adoptable.