University of Naples Federico II, Department of Transportation Engineering Luigi Tocchetti, Via Claudio 21, 80125 Naples, Italy.
Accid Anal Prev. 2010 Mar;42(2):571-81. doi: 10.1016/j.aap.2009.09.025. Epub 2009 Oct 30.
The identification of crash hotspots is the first step of the highway safety management process. Errors in hotspot identification may result in the inefficient use of resources for safety improvements and may reduce the global effectiveness of the safety management process. Despite the importance of using effective hotspot identification (HSID) methods, only a few researchers have compared the performance of various methods. In this research, seven commonly applied HSID methods were compared against four robust and informative quantitative evaluation criteria. The following HSID methods were compared: crash frequency (CF), equivalent property damage only (EPDO) crash frequency, crash rate (CR), proportion method (P), empirical Bayes estimate of total-crash frequency (EB), empirical Bayes estimate of severe-crash frequency (EBs), and potential for improvement (PFI). The HSID methods were compared using the site consistency test, the method consistency test, the total rank differences test, and the total score test. These tests evaluate each HSID method's performance in a variety of areas, such as efficiency in identifying sites that show consistently poor safety performance, reliability in identifying the same hotspots in subsequent time periods, and ranking consistency. To evaluate the HSID methods, five years of crash data from the Italian motorway A16 were used. The quantitative evaluation tests showed that the EB method performs better than the other HSID methods. Test results highlight that the EB method is the most consistent and reliable method for identifying priority investigation locations. The EB expected frequency of total-crashes (EB) performed better than the EB expected frequency of severe-crashes (EBs), although the results differed only slightly when the number of identified hotspots increased. The CF method performed better than other HSID methods with more appealing theoretical arguments. In particular, the CF method performed better than the CR method. This result is quite alarming, since many agencies use the CR method. The PFI and EPDO methods were largely inconsistent. The proportion method performed worst in all of the tests. Overall, these results are consistent with the results of previous studies. The identification of engineering countermeasures that may reduce crashes was successful in all of the hotspots identified with the EB method; this finding shows that the identified hotspots can also be corrected. The advantages associated with the EB method were based on crash data from one Italian motorway, and the relative performances of HSID methods may change when using other crash data. However, the study results are very significant and are consistent with earlier findings. To further clarify the benefits of the EB method, this study should be replicated in other countries. Nevertheless, the study results, combined with previous research results, strongly suggest that the EB method should be the standard in the identification of hotspots.
事故多发点的识别是公路安全管理过程的第一步。热点识别错误可能导致安全改进资源的低效利用，并降低安全管理过程的整体效果。尽管有效识别热点（HSID）方法很重要，但只有少数研究人员比较了各种方法的性能。在这项研究中，比较了七种常用的 HSID 方法与四种强大而有信息量的定量评价标准。以下是比较的 HSID 方法：碰撞频率（CF）、等效财产损失仅（EPDO）碰撞频率、碰撞率（CR）、比例法（P）、总碰撞频率的经验贝叶斯估计（EB）、严重碰撞频率的经验贝叶斯估计（EBs）和改进潜力（PFI）。使用地点一致性测试、方法一致性测试、总秩差测试和总得分测试比较了 HSID 方法。这些测试评估了每种 HSID 方法在多个领域的性能，例如识别持续表现不佳的安全性能的地点的效率、在后续时间段识别相同热点的可靠性以及排名一致性。为了评估 HSID 方法，使用了意大利高速公路 A16 的五年碰撞数据。定量评估测试表明，EB 方法的性能优于其他 HSID 方法。测试结果突出表明，EB 方法是识别优先调查地点最一致和可靠的方法。EB 总碰撞期望频率（EB）的性能优于 EB 严重碰撞期望频率（EBs），尽管当识别出的热点数量增加时，结果仅略有不同。CF 方法的理论论据更有吸引力，性能优于其他 HSID 方法。特别是，CF 方法比 CR 方法表现更好。这一结果令人震惊，因为许多机构都在使用 CR 方法。PFI 和 EPDO 方法在很大程度上不一致。比例法在所有测试中表现最差。总体而言，这些结果与先前的研究结果一致。使用 EB 方法识别的所有热点都成功地确定了可能减少碰撞的工程对策；这一发现表明，识别出的热点也可以得到纠正。EB 方法的优势基于意大利一条高速公路的碰撞数据，HSID 方法的相对性能在使用其他碰撞数据时可能会发生变化。然而，研究结果非常显著，与早期研究结果一致。为了进一步阐明 EB 方法的优势，应该在其他国家复制这项研究。尽管如此，研究结果与先前的研究结果相结合，强烈表明 EB 方法应该成为识别热点的标准。
