Clinical checklist: sensitivity and specificity

Tags: Poster
Celani M, Cantisani T, Cusi C, Filippini G, Munari L, Taus C, Candelise L

Abstract: A number of "checklists" have been published to assess the validity of the methodological quality of a review (Oxam, Jadad). Conversely, few "checklists" have been published to assess the validity of its clinical interest. The primary aim of this study is to compare the value of a clinical checklist in the neurological field to a methodological quality checklist considered as "gold standard". Two pairs of neurologists from different clinical backgrounds (consultants and senior consultants) have assessed forty reviews, selected from the database of the Cochrane Neurological Network where 587 reviews, without quality selection, were classified. The methodological quality checklist, according to Oxman, includes nine specific items and one overall evaluation. The clinical checklist, previously validated by two independent readers, includes seven items and one overall evaluation. Both overall evaluations were done using a seven point scale. Each reader received short training and a legend to help him/her to complete the form. For the present communication the overall evaluation of both scales were dichotomised in two levels (good=5,6,7/ poor=1,2,3,4). The mean agreement between the two pairs of readers for both the overall methodological quality evaluation and overall clinical evaluation was moderate (K= 0.51 and K=0.40, respectively). The comparison between clinical overall items versus methodological quality overall items showed 36 true positive, 86 true negative, 4 false negative and 34 false positive judgements. Thus, 90% of the neurological review articles with a good overall methodological quality evaluation also had a good overall clinical quality evaluation (high sensitivity), while the 72% of the reviews that did not have a good methodological had a poor clinical quality (moderate specificity ). We conclude that if a review has a good methodological quality, important clinical information may be detected. On the other hand, the review may be of informative clinical value even if the methodological quality is poor. In our opinion a simple clinical checklist could be used to help neurologists choose the reviews most suitable for their individual needs.