Bédard Michel, Weaver Bruce, Man-Son-Hing Malcolm, Classen Sherrilene, Porter Michelle
Lakehead University, Thunder Bay, Canada.
J Prim Care Community Health. 2011 Apr;2(2):133-5. doi: 10.1177/2150131910397704.
Dobbs and Schopflocher published an article in which they introduced a tool to identify people who are unfit to drive because of cognitive impairment. In our view, their conclusion that this tool has ". . . a high degree of accuracy that can be used for immediate decisions in the clinical setting"(1(p119)) is too strongly stated, particularly given that the cut-points they used yield false positive (FP) and false negative (FN) percentages in the 6% to 11% range. We believe the reason for using dual cut-points is to ensure that FP and FN fractions are both controlled very stringently, and that it would be more appropriate to set cut-offs that maintain both of them closer to 1%. Using our own data, we constructed two pairs of dual cut-points-one pair that yielded FP and FN percentages similar to those from the Dobbs and Schopflocher article and another pair that yielded FP and FN percentages no greater than 1%. For the first pair of cut-points, 53% of test results were indeterminate (compared to 50% for Dobbs and Schopflocher). For the second pair of cut-points, 86% of test results were indeterminate. Presumably, the same pattern would be observed in Dobbs and Schopflocher's data if their current dual cut-points were replaced with cut-points that controlled the FP and FN percentages at more appropriate levels. We also plotted receiver operating characteristic curves, and calculated the area under the curve (AUC) for the Screen for the Identification of Cognitively Impaired Medically At-Risk Drivers, A Modification of the DemTect (SIMARD-MD) and for the combination of the Mini-Mental State Examination and Trail-Making Test A (using our data for the latter). The difference between them was trivial (AUC = 0.75 and 0.72, respectively). Taken together, the results of the two analytic approaches suggest that other tools currently in use by physicians perform at least as well as the SIMARD-MD, and that it does not represent a significant breakthrough.
多布斯和肖普夫洛彻发表了一篇文章，文中介绍了一种工具，用于识别因认知障碍而不适宜驾驶的人。在我们看来，他们关于该工具具有“……高度准确性，可用于临床环境中的即时决策”（1（第119页））的结论表述过于强硬，尤其是考虑到他们使用的切点产生的假阳性（FP）和假阴性（FN）百分比在6%至11%的范围内。我们认为使用双重切点的原因是为了确保FP和FN比例都得到非常严格的控制，并且设定使两者都更接近1%的切点会更合适。利用我们自己的数据，我们构建了两对双重切点——一对产生的FP和FN百分比与多布斯和肖普夫洛彻文章中的相似，另一对产生的FP和FN百分比不超过1%。对于第一对切点，53%的测试结果不确定（相比之下，多布斯和肖普夫洛彻为50%）。对于第二对切点，86%的测试结果不确定。据推测，如果用能将FP和FN百分比控制在更合适水平的切点取代多布斯和肖普夫洛彻目前的双重切点，在他们的数据中也会观察到相同的模式。我们还绘制了接收者操作特征曲线，并计算了用于识别有医学风险的认知障碍驾驶员的筛查工具（对DemTect的一种修改，即SIMARD - MD）以及简易精神状态检查表和连线测验A组合（使用我们的数据用于后者）的曲线下面积（AUC）。它们之间的差异微不足道（AUC分别为0.75和0.72）。综合来看，这两种分析方法的结果表明，医生目前使用的其他工具至少与SIMARD - MD表现相当，并且它并不代表一项重大突破。
