Diogo C. Haussen, MD
In medicine, scales are mostly used to classify extend of a variable related to a disease process aiming to define uniformization for a specific relation. This process is ubiquitous to all specialties and diseases, and turns out to be extremely important not only in clinical practice but also in research grounds. Therefore, understanding the interrater reliability and its convergent validity (the degree to which two measures that should be theoretically correlated are in fact related) of a scale is paramount.
The ABCD2 score has been increasingly utilized to triage patients with transient ischemic attack. Ishida et al report a straightforward and interesting report that indicates that retrospective estimation of ABCD2 is considerably inaccurate. The authors compared the prospectively determined ABCD2 scoring (established at the time of the emergency room visit) with retrospective scoring through chart review in 102 TIA patients. The correlation coefficient between the two raters that performed retrospective analysis was very good. However, the agreement between retrospective and prospective ABCD2 score rating was only 44%. This is solely slightly better than two times chance alone. The suboptimal convergent validity may, in part, explain the “mixed results” of the previous observational ABCD2 validation studies described in the manuscript. Although the study did not evaluate the predictive value of retrospective ABCD2 scoring for stroke risk, this report exemplifies the major limitations of retrospective analyses, and draws attention to the importance of validation studies for scales utilized in practice or research.