Yamamoto Yasuharu, Hirano Jinichi, Yoshitake Hiroshi, Negishi Kazuno, Mimura Masaru, Shino Motoki, Yamagata Bun
Department of Neuropsychiatry, Keio University School of Medicine, Tokyo, Japan.
Department of Human and Engineered Environmental Studies, The University of Tokyo, Tokyo, Japan.
Psychiatry Clin Neurosci. 2020 Sep;74(9):488-495. doi: 10.1111/pcn.13084. Epub 2020 Jul 20.
In Japan, fatal traffic accidents due to older drivers are on the rise. Considering that approximately half the older drivers who have caused fatal accidents are cognitively normal healthy people, it has been required to detect older drivers who are cognitively normal but at high risk of having fatal traffic accidents. However, a standardized method for assessing the driving ability of older drivers has not yet been established. We thus aimed to identify a new sensing method for the evaluation of the on-road driving ability of healthy older people on the basis of vehicle behaviors.
We enrolled 33 healthy older individuals aged over 65 years and utilized a machine-learning approach to dissociate unsafe drivers from safe drivers based on cognitive assessments and a functional visual acuity test.
The linear support vector machine classifier successfully dissociated unsafe drivers from safe drivers with accuracy of 84.8% (sensitivity of 66.7% and specificity of 95.2%). Five clinical parameters, namely age, the first trial of the Rey Auditory Verbal Learning Test immediate recall, the delayed recall of the Rey-Osterrieth Complex Figure Test, the result of the free-drawn Clock Drawing Test, and maximal visual acuity, were consistently selected as essential features for the best classification model.
Our findings improve our understanding of clinical risk factors leading to unsafe driving and may provide insight into a new intervention that prevents fatal traffic accidents caused by healthy older people.
在日本，老年驾驶员导致的致命交通事故呈上升趋势。鉴于约一半造成致命事故的老年驾驶员认知功能正常且身体健康，因此需要检测出认知功能正常但有发生致命交通事故高风险的老年驾驶员。然而，尚未建立评估老年驾驶员驾驶能力的标准化方法。因此，我们旨在基于车辆行为识别一种用于评估健康老年人道路驾驶能力的新传感方法。
我们招募了33名65岁以上的健康老年人，并采用机器学习方法，根据认知评估和功能性视力测试将不安全驾驶员与安全驾驶员区分开来。
线性支持向量机分类器成功地将不安全驾驶员与安全驾驶员区分开来，准确率为84.8%（敏感性为66.7%，特异性为95.2%）。五个临床参数，即年龄、雷伊听觉词语学习测验首次试验即时回忆、雷-奥斯特里思复杂图形测验延迟回忆、自由绘制钟表测验结果和最大视力，始终被选为最佳分类模型的基本特征。
我们的研究结果增进了我们对导致不安全驾驶的临床风险因素的理解，并可能为预防健康老年人导致的致命交通事故的新干预措施提供见解。
