Deliverable 5.6 “Evaluation Tools” represents the essential outcome of Sub-Task 5.3.2 “Methods and Tools” of the DaCoTA project. The intention of the Deliverable is to give an overview on the state of art of evaluation tools and by this providing some kind of reference book for the application of these tools. It provides methodology and examples for the evaluation of (mainly) vehicle safety systems with regard to • data collection methods; • data analysis methods; • socio-economic methods; • pitfalls and difficulties. The Deliverable is structured into 3 main parts. The first part (Chapter 2) is dedicated to the assessment of safety benefits of systems in terms of measuring the effects on accident involvement (and injury) risk. After some general remarks on non-statistical considerations crucial for the evaluation of safety systems the following study designs are presented: • accident involvement survey; • cohort study; • case-control study; • comparative accident study based on the concept of induced exposure; • matched pairs design; • matched case-control design: one accident involved vehicle (case) and one not involved vehicle (control) are paired; • matched cohort design: pairing a fitted (e.g. ESP) vehicle (protective factor present) with an unequipped one (protective factor absent). Subsequently epidemiological data analysis techniques suitable for the assessment of accident involvement risk and injury risk like: • relative risk; • odds ratio; • incidence density; • statistical models for different types of risk measures are introduced and illustrated by selected empirical examples. Statistical analyses are provided for a cohort study, the induced exposure technique, and the matched pairs design (matched cohort as well as matched case-control design). Up to now, matched pairs designs have relatively seldom been applied in accidentology. Both the analysis of the cohort study and the matched pairs analysis are based on a Germany study where (among other things) the combined effects of three different eSafety systems fitted in heavy trucks (ESP, ACC, Lane Guard System (LGS)) were investigated (1 250 heavy goods vehicles; n=715 vehicles fitted with the three systems; n=535 vehicles without the systems). It can be said that heavy goods vehicles fitted with ESP, ACC and LGS show a considerably lower accident involvement risk compared to vehicles without these systems. The observed group differences with respect to accident involvement (all accident types) are statistically significant. For the matched pairs design special statistical analysis techniques particularly matched odds ratio and conditional logistic regression) are introduced and explained. The induced exposure example is based on GIDAS data, the sample consists of n=10 270 accident involved passenger cars (study period 1995 to 2011) with and without ESP, i.e. the system to be evaluated is ESP. For the evaluation of ESP the accident characteristic “skidding” has been chosen to distinguish between systemspecific accidents (car was skidding) and neutral accidents (no skidding). It appears that the chance of skidding is approx. two-thirds lower for cars with ESP compared to cars without this feature. At the end of Chapter 2 some additional methodological hints on the evaluation of infrastructure measures are given. The second part (Chapter 3) is dealing with the evaluation of systems in terms of socio-economic benefits. In this part methods for assessing potential benefits of safety applications are presented. After the description of potential impacts of safety systems with respect to costs, benefits, and factors influencing the effects in different countries, the assessment of the impact of safety systems on the number and severity of accidents is dealt with. Subsequently, a short overview on Efficiency Assessment Tools (Cost Benefit Analysis, Cost Effectiveness Analysis) is given. The main focus of this chapter is on the application of a Cost Benefit Analysis and the provision of standard values for accident costs. Part 3 (Chapter 4) contains some considerations on the expansion of evaluation results from selected regions or countries to the EU27-level by using the so-called iterative proportional fitting procedure. The method is especially relevant for expanding results on the a priori evaluation of vehicle safety systems by using a simulation tool, for instance. Here, for each case it can be determined whether or not the presence of the system would have avoided or mitigated the accident. Thus, the distributions (regarding e.g. injury severity, lighting conditions, etc.) of both the affected and unaffected accidents are known and can be expanded to a wider accident population. A crucial prerequisite for the application of this method is, of course, that the relevant marginal distributions (accidents, casualties) are available at EU-27 level. (Author/publisher)
Abstract