1、BSI Standards Publication PD ISO/TS 18506:2014 Procedure to construct injury risk curves for the evaluation of road user protection in crash testsPD ISO/TS 18506:2014 PUBLISHED DOCUMENT National foreword This Published Document is the UK implementation of ISO/TS 18506:2014. The UK participation in i
2、ts preparation was entrusted to Technical Committee AUE/7, Automobile occupant restraint systems. A list of organizations represented on this committee can be obtained on request to its secretary. This publication does not purport to include all the necessary provisions of a contract. Users are resp
3、onsible for its correct application. The British Standards Institution 2014. Published by BSI Standards Limited 2014 ISBN 978 0 580 83771 5 ICS 43.020 Compliance with a British Standard cannot confer immunity from legal obligations. This Published Document was published under the authority of the St
4、andards Policy and Strategy Committee on 30 September 2014. Amendments issued since publication Date Text affectedPD ISO/TS 18506:2014 ISO 2014 Procedure to construct injury risk curves for the evaluation of road user protection in crash tests Procdure de construction des courbes de risques pour lva
5、luation de la protection des usagers de la route dans les essais de choc TECHNICAL SPECIFICATION ISO/TS 18506 First edition 2014-03-01 Reference number ISO/TS 18506:2014(E)PD ISO/TS 18506:2014ISO/TS 18506:2014(E)ii ISO 2014 All rights reserved COPYRIGHT PROTECTED DOCUMENT ISO 2014 All rights reserve
6、d. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior written permission. Permission can be requested from either ISO a
7、t the address below or ISOs member body in the country of the requester. ISO copyright office Case postale 56 CH-1211 Geneva 20 Tel. + 41 22 749 01 11 Fax + 41 22 749 09 47 E-mail copyrightiso.org Web www.iso.org Published in SwitzerlandPD ISO/TS 18506:2014ISO/TS 18506:2014(E) ISO 2014 All rights re
8、served iii Contents Page Foreword iv 1 Scope . 1 2 T erms and definitions . 1 3 Methodology . 2 3.1 General . 2 3.2 Collect the relevant data . 2 3.3 Assign the censoring status . 2 3.4 Check for single injury mechanism . 2 3.5 Estimate the coefficients 3 3.6 Identify overly influential observations
9、 3 3.7 Check the distribution assumption . 3 3.8 Choose the best distribution . 3 3.9 Check the validity of the prediction 3 3.10 95 % confidence intervals and its relative size . 4 3.11 Quality index . 4 3.12 Recommendation of the injury risk curves 4 4 Related electronic documents 5 Bibliography 6
10、PD ISO/TS 18506:2014ISO/TS 18506:2014(E) Foreword ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies (ISO member bodies). The work of preparing International Standards is normally carried out through ISO technical committees. Each member b
11、ody interested in a subject for which a technical committee has been established has the right to be represented on that committee. International organizations, governmental and non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the International Electro
12、technical Commission (IEC) on all matters of electrotechnical standardization. The procedures used to develop this document and those intended for its further maintenance are described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for the different types of
13、ISO documents should be noted. This document was drafted in accordance with the editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives). Attention is drawn to the possibility that some of the elements of this document may be the subject of patent rights. ISO shall not be held
14、responsible for identifying any or all such patent rights. Details of any patent rights identified during the development of the document will be in the Introduction and/or on the ISO list of patent declarations received (see www.iso.org/patents). Any trade name used in this document is information
15、given for the convenience of users and does not constitute an endorsement. For an explanation on the meaning of ISO specific terms and expressions related to conformity assessment, as well as information about ISOs adherence to the WTO principles in the Technical Barriers to Trade (TBT) see the foll
16、owing URL: Foreword - Supplementary information The committee responsible for this document is ISO/TC 22, Road vehicles, Subcommittee SC 12, Passive safety crash protection systems.iv ISO 2014 All rights reservedPD ISO/TS 18506:2014TECHNICAL SPECIFICATION ISO/TS 18506:2014(E) Procedure to construct
17、injury risk curves for the evaluation of road user protection in crash tests 1 Scope The aim of this Technical Specification is to provide a procedure to develop injury risk curves for biomechanical samples. These samples are often heavily censored and limited in size, such that specific steps and c
18、hecks are required when developing injury risk curves. Moreover, several statistical methods were historically used to build injury risk curves. The curves resulting from the different methods could be rather close or fairly different, depending on the biomechanical samples. It is therefore of major
19、 importance to recommend a consensual method to be used. Otherwise, injury thresholds candidates to be included into regulations could be highly influenced by the statistical method used. These different methods were compared in a statistical simulation study 10 . The conclusions of this study were
20、used to recommend the survival analysis. This Technical Specification identifies steps to be followed to develop injury risk curves. It recommends preliminary checks, statistical method, and recommendation process of the injury risk curve. Guidelines on the type of the samples to be used (cadaver, a
21、nimal, or dummy), as well as the variables to be used, and the injury severity level are beyond the scope of this Technical Specification. 2 T erms an d definiti ons For the purposes of this document, the following terms and definitions apply. 2.1 injury risk curve curve giving the probability, for
22、a defined population and for a given input, to sustain a specified severity of injury 2.2 injury risk function mathematical function that relates a value of an injury criterion and possible additional factors (variables) to a risk of sustaining an injury of a certain level 2.3 injury criterion physi
23、cal parameter which correlates well with a scale of injury severity of the body region under consideration 2.4 injury value value of an injury criterion 2.5 injury mechanism mechanical behavior leading to an injury 2.6 abbreviated injury scale AIS scale allowing for the classification of injury seve
24、rity 2.7 variable parameters which contribute to the prediction of the injury severity ISO 2014 All rights reserved 1PD ISO/TS 18506:2014ISO/TS 18506:2014(E) 3 Methodology 3.1 General An injury risk curve corresponds to a statistical modelling of biomechanical data in order to predict the risk of in
25、jury. Some general steps are required, as for any statistical model. The modelling includes the choice of the variables to be used in the model. The choice of the model is also of major importance. Once done, the variables of the model are estimated. Essential further steps are to check the assumpti
26、ons of the model and evaluate the model fit relative to the data used. The prediction is finally made possible when the checks are done. 3.2 Collect the relevant data The first step is to collect the relevant data, including injury types and severities, and injury values measured during the tests. T
27、here is no specification for the injury severity level (AIS 2, AIS 3) or the type of injuries evaluated (fracture, soft tissue contusion, etc.). The data could also include variables such as the Post Mortem Human Subject (PHMS) characteristics (age, bone mineral content, gender, etc.). The choice of
28、 the variables to be included in the construction of the injury risk curves is out of the scope of the guidelines. The researchers expertise should guide this decision and is expected to vary depending on the body region and injury mechanism considered. 3.3 Assign the censoring status Once the biome
29、chanical data are available, the censoring status must be assigned (left, right, interval censored, exact). Data are left-censored or right-censored if the measured injury value is greater than or less than the injury threshold, respectively. The injury threshold is the minimum value that is associa
30、ted with the occurrence of injury for that subject. Data are interval-censored when the injury threshold is bounded within a known range of injury values. The censoring status can be provided by a coding which might differ from one statistical software to another and should be checked before the use
31、r analyses the data. It can also be provided using a mathematical formulation for censoring without coding, giving the interval (e.g. right-censored data are an interval from the injury value to eternity). The correct censoring according to the used software or function is to be used. After this ste
32、p, the data set includes the measured injury values, and possibly, the associated coding for the censoring status indicated. 3.4 Check for single injury mechanism The complexity of the human body might lead to different injury mechanisms 3 . This could lead to injury risk curves with changes of slop
33、e and discontinuities. The next step, before applying a parametric method to build an injury risk curve, is to check that the biomechanical sample only included a single injury mechanism. This is achieved by building the injury risk curve with the Consistent Threshold Estimate (CTE) 3for doubly cens
34、ored data and extended CTE for a mix of actual and censored data 4 .Visually inspect the resulting step function for changes in slope which can be indicative of different injury mechanisms. If there is evidence of multiple injury mechanisms, separate the biomechanical sample into samples with single
35、 injury mechanism before starting again the guidelines given at 3.2. If there is no evidence of 2 ISO 2014 All rights reservedPD ISO/TS 18506:2014ISO/TS 18506:2014(E) multiple injury mechanisms, use the survival analysis to build the injury risk curve 67810according to the following steps. 3.5 Estim
36、at e the c oeffi cients The coefficients should be estimated with the survival analysis method. Substantially different results can be obtained when extrapolating outside the range of observed data. Several distributions should be evaluated in order to recommend the one that best predicts the injury
37、 risk function. Consider Weibull, log-normal and log-logistic distributions prior to using a more complicated spline formulation in case one of the other distributions is not confirmed in 3.7. 3.6 Identify ove r ly influential observ ations Overly influential observations can be identified using the
38、 dfbetas statistics. The dfbetas statistic gives an indication of the change of each parameter estimate when deleting one observation of the sample at a time. The dfbetas corresponds to the difference between the regression coefficient when all the data is included and the regression coefficient cal
39、culated with one observation is deleted, scaled by the standard error calculated with the observation deleted. The dfbetas is calculated for each coefficient and for each observation. Choose a cut-off value close to the cut-off value recommended in Reference 1 given the size of the biomechanical sam
40、ples 2/sqrt (sample size). For the biomechanical sample generally available in the biomechanical field, an absolute value of the dfbetas statistic higher than 0,3 give an indication that the associated observation might be overly influential. These observations are checked for any specificity. If th
41、ere is no evidence of difference between these observations and the others included in the sample, keep these observations in the construction of the injury risk curve. 3.7 Check the distribution assumption Check the assumed distribution. One way is to check graphically using a Q-Q plot (“Q” stands
42、for quantile) 511 . The plot of the percentiles of the distribution against the corresponding percentiles of the biomechanical sample is presented in the Q-Q plot. If the plot follows a line through the origin with slope equal to one, then the chosen distribution is appropriate. Another way to check
43、 is to graphically plot the cumulated risk calculated with the survival analysis, with a given distribution against the cumulated risk calculated with the CTE method. If the cumulated risks lie close one to the other, then the chosen distribution is appropriate. Note that this is not possible if sur
44、vival analysis is performed with more than one variable. 3.8 Choose the best distribution Choose the distribution with the best fit based on the Akaike information criterion (AIC). The AIC criterion assesses the likelihood that the model takes into account the number of variables used in the model (
45、AIC = 2*log likelihood+2*number of variables). The lowest AIC indicates the best fit of the model with the test data. Of the three distributions, select the one with the lowest AIC. 3.9 Check the validity of the prediction Check the validity of the injury risk curve predictions against existing resu
46、lts (such as accidentology outcome), if available. ISO 2014 All rights reserved 3PD ISO/TS 18506:2014ISO/TS 18506:2014(E) 3.10 95 % c onfidenc e int erv als and its r elati v e size Calculate the 95 % confidence interval of the injury risk curve with the normal approximation of the error. The relati
47、ve size of the confidence interval is defined as the width of the 95 % confidence interval (upper limit minus lower limit of 95 % confidence interval) at a given injury risk, relative to the value of the stimulus at this same injury risk. Perform the calculation at the 5 %, 25 %, and 50 % risks of i
48、njury. 3.11 Quality index Four categories of the quality index are defined, based on the relative size of the 95 % confidence interval (“good” from 0 to 0,5, “fair” from 0,5 to 1, “marginal” from 1 to 1,5, “unacceptable” over 1,5). The ranges of the quality indices were established to distribute inj
49、ury risk curves from biomechanical samples into four categories. Provide the quality index along with each injury risk curve. 3.12 Recommendation of the injury risk curves 3.12.1 Principle It is recommended to select only one injury risk curve per body region, injury type, and injury level. The injury risk curve recommended should be a function of the more relevant injury measure associated with the single injury mechanism of the sample. This recommendation is intended to avoid inconsistent injury risk predictions due to contradic
