Accident Analysis and Prevention 130 (2019) 125–135 Contents lists available at ScienceDirect Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap Implications of estimating road traffic serious injuries from hospital data T a,b,c,⁎ d d e f g h K. Pérez , W. Weijermars , N. Bos , A.J. Filtness , R. Bauer , H. Johannsen , N. Nuyttens , L. Pascali, P. Thomase, M. Olabarriaa,b,c, The Working group of WP7 project a Agència de Salut Pública de Barcelona (ASPB), Spain b CIBER Epidemiología y Salud Pública (CIBERESP), Spain c Institut d’Investigació Biomédica (IIB Sant Pau), Spain d Institute for Road Safety Research (SWOV), Netherlands e Transport Safety Research Centre, Loughborough University (LOUGH), United Kingdom f Austrian Road Safety Board (KFV), Austria g Medical University of Hannover (MHH), Germany h Vias institute, Belgium i French Institute of Science and Technology for Transport, Development and Networks (IFSTTAR), France A R T I C LE I N FO A B S T R A C T Keywords: To determine accurately the number of serious injuries at EU level and to compare serious injury rates between Road traffic injury different countries it is essential to use a common definition. In January 2013, the High Level Group on Road MAIS Safety established the definition of serious injuries as patients with an injury level of MAIS3+(Maximum Data linkage Abbreviated Injury Scale). Whatever the method used for estimating the number or serious injuries, at some Injury severity point it is always necessary to use hospital records. The aim of this paper is to understand the implications for (1) in/exclusion criteria applied to case selection and (2) a methodological approach for converting ICD (International Classification of Diseases/Injuries) to MAIS codes, when estimating the number of road traffic serious injuries from hospital data. A descriptive analysis with hospital data from Spain and the Netherlands was carried out to examine the effect of certain choices concerning in- and exclusion criteria based on codes of the ICD9-CM and ICD10. The main parameters explored were: deaths before and after 30 days, readmissions, and external injury causes. Additionally, an analysis was done to explore the impact of using different conversion tools to derive MAIS3 + using data from Austria, Belgium, France, Germany, Netherlands, and Spain. Recommendations are given regarding the in/exclusion criteria and when there is incomplete data to ascertain a road injury, weighting factors could be used to correct data deviations and make more real estimations. 1. Introduction increasing (Berecki-Gisolf et al., 2013; OECD/ITF, 2011; Weijermars et al., 2015). In light of this trend, serious road injuries are more In most countries, road safety performance has traditionally been commonly being adopted by policy makers as an additional road safety measured by the reduction of fatalities. The total number of road indicator. Moreover, the Valletta Declaration on Road Safety estab- fatalities in Europe declined by 42% overall between 2000 and 2013 lished that the transport ministers will set a target of halving the within the 32 countries in the International Road Traffic and Accident number of serious injuries in the EU by 2030 from the 2020 baseline Database (IRTAD) for which data are consistently available (Road using this common definition and in the framework of an overall road Safety Annual Report 2015, 2015). However, crashes also cause nu- safety strategy for this period (Malta, March 2017) (“Valletta merous serious road injuries, resulting in considerable economic and Declaration on Road Safety,” 2017). human costs. For the first time, in 2016, the European Commission However, one difficulty with tackling the serious road injury pro- published a global estimate for the number of people seriously injured blem is how to define serious road injuries. In the past, different on Europe’s roads: 135,000 in 2014 (European Commission, 2016). It countries have applied different definitions which usually related to has been estimated that in some countries the number of serious road time of hospitalisation. To determine accurately the number of serious injuries has not been decreasing as fast as the number of fatalities, while injuries at EU level and to compare serious injury rates and develop- in other countries the number of serious road injuries has even been ments between different countries, it is essential to use a common ⁎ Corresponding author at: Agència de Salut Pública de Barcelona (ASPB), Plaça Lesseps 1, 08023 Barcelona, Spain E-mail address: cperez@aspb.cat (K. Pérez). https://doi.org/10.1016/j.aap.2018.04.005 Received 29 July 2016; Received in revised form 2 April 2018; Accepted 6 April 2018 Available online 19 April 2018 0001-4575/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/). K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 definition. In January 2013, the High Level Group on Road Safety re- are no clear guidelines on how to apply each method. It is re- presenting all EU Member States established the definition of serious commended that all EU countries provide data for serious injuries ac- injuries as patients with an injury level of MAIS (Maximum Abbreviated cording to this definition from 2014 onwards. Currently, the individual Injury Scale) ≥ 3 (Jeanne Breen, 2012, European Commission, 2013). EU Member States use different procedures to determine the number of The Abbreviated Injury Scale (AIS) and its derivative the MAIS MAIS3 + casualties. The procedure that is applied in a country is for a (Maximum AIS) is an anatomical-based consensus derived coding large part determined by the available data. Consequently, the quality system, created by the Association for the Advancement of Automotive of the data differs between Member States and therefore the numbers Medicine to classify and describe the severity of injuries. The AIS allows generated may not be fully comparable. The impact of this hetero- injuries to be ranked by severity. It classifies each injury using 7-digit geneity on final estimations is unknown. One of the work packages of number that describes the body region (1), type of anatomical structure the SafetyCube project7 (WP7) has the objective to assess and improve (2–3), specific anatomical structure (3–4), specific injuries (5–6), and the estimation of the number of serious road injuries by providing the severity scale (7). The severity is based on a 6-point ordinal scale, guidelines for Member States. All three of the recommended methods one being a minor injury and six being maximal, currently untreatable used for estimating the number of serious traffic injuries (MAIS3+) are (1 Minor, 2 Moderate, 3 Serious, 4 Severe, 5 Critical, 6 Maximum, 9 Not in one way or another based on a selection of hospital records. There- specified). An AIS- Severity Code of 6 is not the arbitrary code for a fore, before any guidelines or recommendations can be made addres- deceased patient or fatal injury, but the code for injuries specifically sing how the methods should be applied it is first necessary to have assigned an AIS 6 severity. An AIS-Code of 9 is used to describe injuries clear criteria for which hospital records should be included or excluded for which not enough information is available for assessing its severity. as a first step in the process. This paper considers how hospital data can The AIS scale is a measurement tool for single injuries. The MAIS be used in a harmonised way. The aim of this paper is to understand the (Maximum AIS) is the maximum of the AIS scores for each region of the implications for (1) in/exclusion criteria applied to case selection and body, and is frequently used for assessing overall severity. It does not (2) a methodological approach for converting ICD to AIS codes, when necessary have a linear relationship with the probability of death. AIS estimating the number of road traffic serious injuries from hospital can be coded directly or can be derived from other injury coding sys- data. tems, like the International Classification of Diseases in its several versions (ICD, ICD9-CM, ICD10, etc). When using large databases, 2. Data and method usually hospital databases, it is likely that AIS is not coded directly but derived using a convertor algorithm. In this section, we describe the methods used to identify the inclu- Various conversion tools are available for recoding ICD codes into sion and exclusion criteria and the factors that affect obtaining AIS codes: ICDmap90,1 ICDpic,2 DGT,3 ECIP,4 AGU5 or AAAM.6 The use MAIS3 + through conversion tools. The paper addresses the two ob- of any of these conversions tools leads to the so-called ICD-derived AIS jectives in two separate sections. Firstly, we explore the effect of in/ values. Some of these tools recode the ICD codes into the latest AIS© exclusion criteria using data from Spain and the Netherlands. Secondly 2005/update 2008 codes, but other older tools recode ICD data into AIS we analyse the impact of using different conversion tools to derive codes that are based on previous versions of the AIS coding (AIS2005, MAIS3 + using data from Austria (AT), Belgium (BE), France (FR), AIS1998 or AIS1990) (Ringdal et al., 2012). Recoding always has the Germany (GE), Netherlands (NL) and Spain (ES). disadvantage compared to direct coding, that some information gets lost or is not available so that a best match must be selected (in the 2.1. Inclusion and exclusion criteria recoding tool). This may influence the severity that is assigned to a casualty and therefore also on the estimated number of MAIS3 + ca- A descriptive analysis was carried out to examine the effect of cer- sualties. However, there is very limited amount of literature comparing tain choices concerning in- and exclusion criteria based on codes of the directly coded (M)AIS levels to ICD-derived (M)AIS levels, derived International Classification of Diseases/Injuries (ICD9-CM, ICD10) with using recoding tools (Bartolomeo et al., 2010; Greene et al., 2015). the aim to arrive at a common consensus as to which codes should be After establishing the definition of serious injuries, the High Level used to report road traffic serious injuries. Group recommended that all EU countries provide data to the European The main parameters explored were: Commission and identified three main ways which Member States can use to estimate the number of serious road traffic injuries: 1) by ap- • What to do with deaths before and after 30 days plying a correction to police data, 2) by using hospital data and 3) by • Whether scheduled admissions should be included using linked police and hospital data (Omran, 1971). However, there • Whether readmissions should be included • How to treat hospitalisations of 1-day treatment or less • ICD9-CM 1 90 Johns Hopkins University (1998). ICDmap90 and ICDMAP-90 user's guide. ○ Inclusion of E-codes E827-E829, E929.0 and E988.5 Baltimore. ○ Usefulness of the E-code E849. (Place of the occurrence of the 2 Clark, Osler, Hahn (2010). Stata module to provide methods for translating accident) International Classification of Diseases (Ninth Revision) diagnosis codes into standard ○ Inclusion of people without any traumatic injury (800–959) injury categories and/or scores. https://ideas.repec.org/c/boc/bocode/s457028.html. 3 Directorate General de Trafico Madrid, Spain. SAS-algorithm. Available for this study. ○ Inclusion of codes 905–909 and 959 ICD9cm (version 1996) to AIS1998. • Traumatic injury only in the main diagnosis or in any diagnosis 4 European Center for Injury Prevention, University of Navarra, Algorithm to transform ICD-10 codes AIS 90 (1998 update) and ISS, [version 1 for SPSS] [version 1.0 for STATA]. • ICD10 codes for external causes Pamplona, Spain 2006.with partial funding from the EU, DG SANCO Grant Agreement Nº 2004119 Project Apollo WP2. According to the International Classification of Injuries (ICD9-CM) 5 Schmitt KU, Baumgartner L, Muser M, Furter K, Scholz S, Lüber B, Thomas P, Simma traumatic injury definition includes codes from 800 to 959. These in- A (2014) Developing a scheme to report AIS‐coded injury severity for Swiss traffic accident clude fracture, dislocation, sprain, internal injury, open wound, injury data. IRCOBI Conference 2015. Berlin. 2014 Paper no. IRC-14-50. Schmitt KU, Baumgartner L, Muser M, Baudenbacher M, Simma A, (2015) Improving the Swiss National Accident Statistics by Providing AIS Data to Classify Injury Severity. 24th ESV Conference. 7 SafetyCube (Safety CaUsation, Benefits and Efficiency) (H2020 N.633485) is a re- Gothenburg. 2015. Paper No. 15-0323. search project funded by the European Commission under the Horizons 2020, the EU 6 AAAM9: AAAM (2015).” Copy of aaam_icd9map_v1 0_Feb2015 read only.xls”. Framework Programme for Research and Innovation, in the domain of Road Safety. The AAAM10: AAAM (2015).” Copy of aaam_icd10map_v1 0_Feb2015 read only.xls”. project started on May 1st, 2015 and will run for a period of three years. http://www. AAAM10-cm: AAAM (2015).” Copy of aaam_icd10map_v1 0_Feb2015 read only.xls”. safetycube-project.eu. 126 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 of physical trauma (905 to 909, 958 and 959). In addition, external codes E810-819 (“Motor vehicle traffic accident”) are clearly relevant to identifying road traffic casualties. However, there are codes that in some cases do not fulfil the police definition of “traffic accident” such E826 (“Pedal cycle accident”), E820-E825 “Motor Vehicle Nontraffic Accidents (not on a public road)”, E827 “Animal-drawn vehicle acci- dent”, E828 “Accident involving an animal being ridden”, E829 “Other road vehicle accident” and E988.5 “Injury by crashing of motor vehicle, undetermined whether accidentally or purposely inflicted”. We will include them in the sensitivity analysis. Fig. 1. Injury classification according to ICD9-CM (International Classification For the purposes of this study, we used as a core definition of traffic of Injuries, ninth revision clinical modification). injury one that has been considered based by previous studies (Pérez Adapted from (Smith et al., 1991). et al., 2014). Due to the frequency of missing information on the code of external cause (E code) we also considered using the compensation payer company to identify traffic injury cases. Traffic injuries were to blood vessel, superficial injury, contusion, crushing, foreign body defined as those injuries meeting either criteria 1 or criteria 2: entering through body orifice, burns, and injury to nerves and spinal cord. It also includes late effects of injury and complications of physical 1 E-code for external cause (ICD9-CM): E810-819, E826-829, E929, trauma (905 to 909, 958 and 959). Defining traumatic injuries derived E988.5. from a traffic accident is a key issue because injury severity is calcu- 2 Road Accident compensation payer. lated by the Maximum Abbreviated Injury Score (MAIS), which uses the traumatic injuries for its calculation. Not all injuries are traumatic, and Police registers include road traffic fatalities as those up to 30 days if they are not traumatic no AIS can be derived. Fig. 1 shows injury after the collision. Therefore, in hospital databases it is important to classification according to ICD9-CM. ensure that fatalities are not double counted (as injured and as fatality), In addition, as in many countries, frequently the availability of data and would produce an overestimation. This means that if a person is does not allow identification of all specific criteria to select cases from admitted to hospital but finally dies within 30 days after the admission hospital data. A sensitivity analysis has been carried out with the aim of they should be counted as a fatality (as in the police registers). deriving weighting factors to use when no data is available. Weighting However, if the person dies after 30 days, they should be recorded as factors were calculated based on the effect resulting from including/ injured according to their severity, since they are not included in the excluding each option. fatalities statistics. A descriptive analysis was carried out for each parameter using For ICD9-CM, AIS and MAIS were derived using the icdpic module hospital data from Spain and The Netherlands separately. of Stata (“STATA Data Analysis and Statistical Software,” n.d.) from the In Spain, two data sources were used. First, the Spanish Database 14 diagnoses reported in the Spanish Hospital Discharge Database for from Hospital Discharge Register (HDR) for 2011 was used to analyse 2011. criteria based on ICD9-CM. This database includes all hospitalisations For ICD10 it has not been possible to derive MAIS (apart from all for any injury (traffic and non-traffic) in Spain from all public hospitals fatal cases) because there is no information in the data set about in- and around 99% of private hospitals. The population of study is whole juries. There is only information about the external cause of injury. the Spanish population. The unit of analysis of the database are ad- Therefore, we used this information to show the distribution of ICD10 missions, not individuals. To identify individuals a personal identifi- codes stratified by traffic and non-traffic, in order to establish criteria cation code is used. The criteria for defining a readmission can vary for inclusion. across countries. In Spain, if the same person has different admissions In the Netherlands, a selection of the Dutch Database from the due to the same reason in the same hospital or another within less than Hospital Discharge Register (HDR) for 1993–2013 was used. Dutch 30 days of difference, they are classified as re-admissions (of the same hospitals code injuries with the ICD9-CM. The database includes up to individual). Due to data protection, frequently it is not possible to 10 diagnoses (1993-2009 including diagnosis of injuries, diseases and identify number of visits of the same person, therefore we need to rely codes of external causes). After 2012, hospitals gradually introduced on the hospital definition. In the same sense hospital admissions can ICD10 for coding. In this study only the ICD9-CM cases of 2012 and take place through emergency attendance or be scheduled. These 2013 have been considered. For this analysis, all ICD10 codes have scheduled admissions may be a second episode of a previous emergency been converted to ICD9-CM and treated likewise. The Dutch Hospital injury or may not, so we also studied them in the analyses. Data includes all hospitalisations for patients with a transport accident Spain was coding diseases and injuries using the ICD9-CM up until as the external cause (E-codes in: E800-E848), extended with a wide December 2015 when the process of coding changed to ICD10. The range of external causes where –in case of miscoding- other traffic ca- database includes up to 14 diagnoses (including diagnosis of injuries, sualties can be found (E-codes in: E880-E889 (falls), E890-E899 diseases and codes of external causes). To analyse criteria based on (burns), E928 (unknown accidental), E929 (late effects), E958 (suicide), ICD10, we used data from the National Register of Mortality for the E988 (unspecified)). All hospitals are included in the database. For years 2009–2013. We included cases with ICD10 codes V01 to V99. some of the years of interest some hospitals did not provide data, but Spain has coded mortality using ICD10 since the year 2000. because the number of patients treated is known an accurate estimate of In order to identify road traffic injuries in hospital data it is ne- the missing data can be given. The population of study is complete the cessary to know the police definition. The Spanish General Directorate Dutch population, including all persons –irrespective of their nation- of Traffic (DGT) in accordance with Eurostat defines a ‘Road traffic ality- having an accident in the Netherlands who are admitted to a collision with victims’ as “collision occurring or starting on a road Dutch hospital. At a first stage, the unit of analysis of the database is which is the object of motor vehicle traffic and road safety legislation admissions, not casualties of (traffic) incidents. Therefore, as with the (public road), involving at least one vehicle in motion, and which re- Spanish data, it is necessary to identify the individuals before any sults in the death and/or injury of one or more people”. analysis can be completed using the patient number. AIS and MAIS According to the ICD9-CM the definition of road traffic injuries in- were derived from the icdmap90 module of Johns Hopkins University cludes any traumatic injury including codes from 800 to 959. We will (1998). analyse the impact of including late effects of injury and complications In the Netherlands, for this study, to examine the sensitivity for the 127 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 Fig. 2. Issues related to deriving MAIS3 + that may influence the number of MAIS3 + casualties. inclusion criteria the following definition was used: all severities on the ES, BE where the ICD-codes were recoded to AIS with different tools: MAIS scale, including deaths, and all patients with an E-code in the ICDmap90, DGT, ICDpic, AAAM9, ECIP and AAAM10. range E810-E829 + linked cases for E929.0 and E988.5 were included. d The effect of using a limited number of injuries for the MAIS cal- According to this selection in the Hospital Discharge Register (HDR) culation was investigated by studying the effect on the estimated there were 433,077 admissions in the Netherlands in the period number of MAIS3 + casualties in case of leaving out injuries (NL, 1993–2013 due to traffic accidents. BE, ES). e The effect of truncation of injury codes was assessed by truncating 2.2. Conversion tools to derive MAIS3+ the ICD injury codes in countries where full codes were available (NL, BE, ES). When deriving MAIS3 + from ICD codes using conversion tools f To study the effect of the different tools to derive the AIS from ICD, there are a number of factors that might affect the final estimates such two or more tools were used on the same data set. The results are as: analysed to show the differences in the total number of MAIS3 + casualties, for combinations of ICD-version, tool and AIS- 1 AIS versions and ICD-derived AIS compared to direct AIS coding; version where the difference was substantial (DE, NL, BE, ES). 2 Conversion tools for converting ICD codes into AIS codes (this is a mixed effect of the ICD-version, the tool and the AIS version that it converts to) – in relation to the gold standard of directAIS coding; 3. Results 3 Using a limited number of injuries per casualty for the MAIS score; 4 Using 4 digits instead of the full ICD injury code when deriving AIS. 3.1. Inclusion and exclusion criteria for road traffic case selection from hospital data Fig. 2 summarises the issues related to deriving MAIS3+. In an ideal situation MAIS3 + casualties would be selected on the basis of directly The Hospital Discharge Register (HDR) of Spain included 258,432 coded AIS levels of all injuries of a casualty, based on the latest AIS episodes of hospitalisations during 2011 due to all causes of injuries. coding instructions. However, in practice this is often not possible. Out of them, there were 31,339 traffic admissions in Spain according to Using ICD conversion tools is often necessary to derive AIS, but, the the definition of criteria 1 or 2 (E-code or Accident compensation combination of differing ICD conversion tools and various versions of payer). In 25% of cases there was no E-code but the compensation payer the coding instructions may result in variations in the estimated was a road traffic insurance company. number of MAIS3 + casualties. In a similar way, including a limited Therefore, according to the Spanish Hospital Discharge Register number of injuries and the use of truncated injury codes has the po- (HDR), 31,339 persons were hospitalised due to traffic injury in 2011: tential to affect the estimated casualties. 583 people died (526 people died within 30 days after admission in We studied these effects using hospital data from Austria (AT), hospital and 57 died 30 days or more after admission), 21,835 were Belgium (BE), France (FR), Germany (DE), Netherlands (NL), and Spain slightly injured (MAIS 0–2) and 8888 were seriously injured (MAIS3+). (ES). Eighteen fatalities that died after 30 days of admission had slight or The implications for each of the factors (1–4) listed above were moderate injuries (MAIS = 0–2) (Table 1). (Obviously, there were considered in relation to the following: fatalities who died at the scene and were not admitted to hospital). From this point and for the purposes of this study, deaths within 30 a The effect of different AIS versions was studied using the GIDAS days (n = 526) are excluded for the sensitivity analyses. Deaths after 30 (DE) in-depth accident survey data set of accidents recorded since days are distributed according to their MAIS. Therefore, the database 2008 that have been coded in both AIS1998 and AIS2008. includes 30,813 admissions (Table 1). b To study the effect of direct coded AIS compared to ICD-derived AIS, According to data showed in Table 2, if re-admissions are excluded, a dataset from Germany (GIDAS) was available where AIS-codes and the percentage of severe road traffic injuries is similar (29.0% versus independent ICD10 codes were available for the same RTC (Road 29.4%). The percentage increases when scheduled admissions are ex- Traffic Casualty). ICD10 codes were recoded with AAAM10, ECIP cluded (from 29.0% to 33.8%). This is mainly due to the fact that the and AGU to rate the severity and compared to the direct AIS. total number of road traffic injuries dropped from 30,813 to 24,761. If Unfortunately, no other datasets were available to study other we exclude people injured who stay 0 days hospitalised (defined as no combinations of AIS versions. overnight stay), the percentage remains similar from 29.0% to 28.8%. c The effect of different mapping tools was studied with data from NL, On the other hand, if only people with traumatic injuries are se- lected as traffic injuries, the number of people injured will drop from 128 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 Table 1 Table 3 Hospital admissions in Spain according to severity of injury and outcome. Traffic hospital admissions in the Netherlands according to all the different Hospital Discharge Register, Spain 2011. definitions. (HDR, The Netherlands 1993–2014). Severity Traffic Other Injuries Total MAIS1/ MAIS2 MAIS3+ index Overestimation Injuries unknown Factor Deaths Death within MAIS 0-2 73 1086 1159 MAIS3 + definition 88,555 196,112 97,138 100% 30 days MAIS3+ 453 4694 5147 Including E828 4018 12,680 3394 3.5% 0.966 Unknown 0 84 84 Including E929 47 1 0 0.0% 1.000 Total Deaths 526 5864 6390 Including E820-825 2248 5946 2176 2.2% 0.978 Injured Death after 30 MAIS 0-2 18 84 102 Including re- 1991 8194 3656 3.8% 0.964 days MAIS3+ 39 317 356 admissions Unknown 0 25 25 Including deaths 540 403 4623 4.8% 0.955 Alive MAIS 0-2 21,835 143,120 164,955 within 30d MAIS3+ 8888 74,548 83,436 Including 172 809 373 0.4% Unknown 33 3135 3,168 combinations Total Injured 30,813 221,229 251,559 Total 97,571 224,145 111,361 Total 31,339 227,093 258,432 (12.1%) (87.9%) ICD10: “A transport accident (V01-V99) is any accident involving a device designed primarily for or being used at the time primarily for, 30,813 to 27,140 and consequently the percentage of severely injured conveying persons or goods from one place to another”. ICD10 distin- people change from 29.0% to 32.9%, although total number of severely guishes between “Traffic accident” (any vehicle accident on a public injured will not change (8927). Moreover, if traumatic injuries are se- road) and “Non-traffic accident” (any vehicle accident occurring en- lected only from the main diagnosis, the number of people severely tirely somewhere other than on a public road). Table A1 (see Appendix injured falls to 11,885 and the percentage of seriously injured changes A) shows the distribution of cases of fatalities according to ICD10 to 25.7%. classification of road user by traffic and non-traffic for the mortality Finally, the sensitivity analysis has been done by varying the in- database in Spain. For all road users, there are a notable number of clusion of the E-codes (E826-E829, E929.0, E988.5). The numbers are cases with unspecified information. Tables A2 and A3 in the appendix so small that the percentage of severely injured people will not change present the list of codes of external causes for ICD9-CM and for ICD10. in most of the cases (Table 2). According to the Dutch Hospital Discharge Register (HDR), 433,077 hospital admissions took place because of a traffic accident, of which 3.1.1. Use of weighting factors to make data more comparable with other 6308 people died (1.5%). Most of them (92%) died within 30 days after countries admission in hospital and 8% died after 30 days of admission. Table 3 Sometimes it is not possible to have complete hospital data for a shows what would be the effect of using different criteria. country. In that case weighting factors could be used to correct for data Regarding the external causes for ICD10 (V codes), according to deviations to generate estimations which are more comparable to Table 2 Traffic hospital admissions according factors and injury severity. Hospital discharge Register of Spain, 2011. MAIS0-2 MAIS3+ Totala Re-admissions (n) 978 204 1183 Re-admissions included 21,853 8927 (29.0%) 30,813 (100%) Re-admissions excluded 20,875 8723 (29.4%) 29,630 (100%) Scheduled admissions (n) 5491 551 6052 Scheduled admissions included 21,853 8927 (29.0%) 30,813 (100%) Scheduled admissions excluded 16,362 8376 (33.8%) 24,761 (100%) People with 0 days of hospitalisation (n) 140 119 259 0 days hospitalisation included 21,853 8927 (29.0%) 30,813 (100%) 0 days hospitalisation excluded 21,713 8808 (28.8%) 30,554 (100%) Traumatic Injuries (ICD9:800-959) (n) 18,180 0 27,140 All injuries included 21,853 8927 (29.0%) 30,813 (100%) Selecting only traumatic Injuries 18,180 8927 (32.9%) 27,140 (100%) Traumatic injuries (Number of diagnosis) Traumatic injury in any diagnosis 21,853 8927 (29.0%) 30,813(100%) Traumatic injury only in the main diagnosis 9584 2295 (25.7%) 11,885(100%) External causes (ICD9)b Total traffic injuries 21,853 8927 (29.0%) 30,813(100%) E826 Pedal cycle accident (n) 2716 1018 3734 Traffic injuries excluding E826 19,137 7909 (29.2%) 27,079 (100%) E827 Animal-drawn vehicle accident (n) 18 7 25 Traffic injuries excluding E827 21,835 8920 (29.0%) 30,788 (100%) E828 Accident involving an animal being ridden (n) 450 193 644 Traffic injuries excluding E828 21,403 8734 (29.0%) 30,169 (100%) E829 Other road vehicle accident (n) 80 28 108 Traffic injuries excluding E829 21,773 8899 (29.0%) 30,705(100%) E929.0 Late effects (n) 228 8 238 Traffic injuries excluding E929.0 21,625 8919 (29.2%) 30,575 (100%) E988.5 Injury by crashing of motor vehicle, undetermined whether accidentally or purposely inflicted (n) 1 3 4 Traffic injuries excluding E988.5 21,852 8924 (29.0%) 30,809 (100%) a Traffic injuries with unknown severity are not shown. b All the sensitivity analysis has been done assuming that excluded cases are not selected as road traffic injury by any other criteria. 129 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 Table 4 Weighting factors among selection criteria for serious injuries (MAIS3+) traffic hospital admissions in Spain and in The Netherlands. Spanish Hospital Discharge Register 2011 and Dutch Database from Hospital Discharge Register, 1993-2013. Spain Netherlands Average Effect of not meeting criteria Including deaths within 30 days 0.95 0.96 0.95 If fatalities within 30 days cannot be excluded, the results is an overestimation of 5% Including Re-admissions 0.98 0.96 0.97 If re-admissions cannot be excluded the result is an overestimation of 3% Including E929 0.99 1.00 1.00 Inclusion of E929 does not have an effect Including E828 0.98 0.97 0.97 Inclusion of E828 results in an overestimation of 3% Including E820-825 0.99 0.98 0.99 Inclusion of E820-E825 results in an overestimation of 1 to 2%. estimates from other countries. Table 4 shows weighting factors cal- difference can only be partly explained by the difference in AIS version. culated for road traffic serious injuries (MAIS3+) with the Spanish As differences between tools that use ICD9 and tools that use ICD10 can Hospital Discharge and with the Dutch Database from Hospital Dis- only be investigated by converting ICD9CM to ICD10 and vice versa and charge Register. For each ‘deviation’ from the ideal situation, we cal- the effect of these conversions is unknown, it was not possible to make a culated the effect on the estimated number of serious road injuries. good comparison between tools that use ICD9 and tools that use ICD10 Based on these effects, weighting factors were derived. When these to recode to AIS. weighting factors are very close to 1, will not have a large impact on the estimated number of MAIS3 + casualties, therefore it would be better 3.2.2. Consequence of using a limited number of injuries per casualty not to apply any weighting factor in this instance. This is for example In some cases, only a limited number of injuries were coded or the case regarding the inclusion of E-codes E820-E825 and E929 in available for analysis. The consequence of this limitation was in- Spain and the Netherlands. Inclusion of fatalities within 30 days has the vestigated by running the conversion tools considering the primary largest effect on the estimated number of MAIS3 + casualties of the diagnosis, 1, 2, 3 and all injuries and comparing the results. This ana- analysed factors and results in an overestimation of 5%. lysis is done using Belgium, Spanish and Dutch data. On average, taking into account only 1 injury results in an estimated number of MAIS3 + casualties which is only 78% of the number that is estimated 3.2. How to derive MAIS3+ based on all injuries. When 2 injuries are taken into account, 90% of the serious road injuries are identified and when 3 injuries are taken into 3.2.1. Direct coding versus various conversion tools account, on average 95% of the number of MAIS3 + casualties are Directly coded AIS levels were compared with ICD derived AIS le- identified (Table 6). vels, different ICD recoding tools and different AIS versions were con- sidered. The results of those analyses are summarised in Table 5. In the German GIDAS database, 16,695 casualties were coded in both AIS1998 3.2.3. Consequence of truncated injury codes and AIS2008. Line 1 in Table 5 shows that the AIS1998 version results Due to hospital practice or privacy regulations, some countries use in roughly 12% more MAIS3 + casualties compared to the AIS2008 4-digit injury codes instead of 5-digit codes. ICD conversion tools differ version. Some of the data from the German GIDAS database could be on how they deal with these so called truncated codes. Some tools matched to patient information from the Medical School in Hannover. simply say that the injury is not detailed enough to assess the severity For 209 trauma casualties, ICD10GM (ICD10 German modification) and return MAIS = 0 or MAIS = 9. Other tools, like AAAM, also pro- injury codes were available and recoded to AIS using various ICD10 vide a severity for the aggregated level, by considering the severities of recoding tools. Line 2 of Table 5 shows that using ECIP results in a 13% the injuries underneath. higher number of MAIS3 + casualties compared to direct coding, Table 7 shows the effect on the number of MAIS3 + casualties whereas both AAAM10 and AGU result in an underestimation. Note caused by truncation when using various conversion tools. Most con- that the difference between direct coding and ECIP can be fully ex- version tools, except from ICDpic and AAAM10 appear to be quite plained by a difference in AIS version. capable of dealing with truncated codes. ICDpic shows a large decrease Lines 2–6 in Table 5 show comparisons between different recoding in the number of MAIS3 + casualties when injuries are truncated. The tools, in combination with different AIS and ICD versions. As it was not AAAM10 tool shows a considerable increase in the number of possible to compare ICD9 derived AIS with direct coding, it was not MAIS3 + casualties in cases of truncated injury codes compared to possible to determine which tool provides the best result. Considering a using full codes. 12% difference between AIS1990/1998 and AIS2005/2008, ICDpic appears to result in the highest number of MAIS3 + casualties in all 4. Discussion three countries. Besides, the difference between the ICD9-CM recoding tools is at most 7%. Also, these analyses show that the AAAM10 tool Hospital data is an important data source for all three EC re- results in a much lower number of MAIS3 + casualties than ECIP. This commended methods for estimating MAIS3 + serious injuries. These Table 5 Estimated number of MAIS3 + casualties when applying different AIS versions and different ICD to AIS recoding tools. In italic bold: ‘true number’. N Country AIS 1998 AIS 2008 ICD9-CM ICD10 ICDmap90 DGT ICDpic AAAM9 ECIP AAAM10 agu 1990 1998 1985 2008 1998 2008 2008 1 Germany 1 019 (112%) 909 100% 2 Germany 103 (100%) 116 (113%) 82 (80%) 89 (86%) 3 Netherlands (1993–2013) 107,738 109,605 103,747 102,900 4 Netherlands (2012–2014) 14,384 8391 5 Belgium 19,143 18,381 6 Spain 8274 7656 130 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 Table 6 Estimated numbers of serious road injuries when taking 1, 2, 3, or all injuries • Include all traffic injury hospitalisations even those with short length of stay into account and applying different conversion tools (upper half), and pro- portion of MAIS3 + casualties that is included when taking account 1, 2 or 3 • Include all cases with any injury diagnosis (ICD9CM: 800–999; ICD10: S00-T88) injuries (all injuries is 100%). • Include external causes for road traffic injuries: (ICD9CM: E810- BE NL NL (ICD10 ES SUM Average E819, E826, E827, E829, E988.5; ICD10: V01-89 for those codes for converted to ICD9cm traffic injuries and/or weighting -correcting for non-public road- for ICD9cm) non-traffic injury codes) • ICD9cm ICD9cm ICD9cm ICD9cm ICD9 (BE+NL ICDpic ICDmap90 ICDmap90 ICDpic +ES)/3 If it is not possible to have complete data of these hospital data, weighting factors could be used to correct data deviations and make All 19,142 107,735 15,078 8274 135,151 more real estimations. 3 17,900 105,728 14,766 7753 131,381 2 16,654 102,392 14,258 7315 126,361 1 13,678 91,159 12,489 6357 111,194 Recommendations for using different conversion tools to derive MAIS3+ are: All 100% 100% 100% 100% 100% • In order to make data from different countries more comparable to 3 94% 98% 98% 94% 97% 95% 2 87% 95% 95% 88% 93% 90% 1 71% 85% 83% 77% 82% 78% each other, the number of MAIS3 + casualties should be multiplied by a factor 0.89 when injuries are coded in AIS1990 or AIS1998 instead of AIS2005 or AIS2008. Table 7 • The ECIP recoding tool for ICD10 seems to result in the most reliable numbers of serious road injuries. Estimated number of MAIS3 + casualties when using truncated codes com- pared to using full codes, % of underreporting when using truncated codes and • The difference in the estimated number of MAIS3 + casualties be- tween the ICD9CM recoding tools is at most 7%. It was not possible weighted factors to correct for truncated codes. to investigate the difference between ICD9 tools and direct coding. Using full codes Using truncated codes % Factor • Current version of the AAAM10 tool (2016) is not recommended until it is updated to better-fit European needs. • ES ICDpic 8274 2108 25% 3.9 BE ICDpic 19,143 3949 21% 4.8 The following weighting factors could be applied in cases where less NL ICDmap90 107,735 101,549 94% 1.06 than 4 diagnoses of injuries are taken into account for the de- DGT 115,380 109,039 95% 1.06 termination of the number of MAIS3 + casualties: ICDpic 109,373 17,454 16% 6.3 ○ 1.28 in cases of 1 injury AAAM9 108,509 97,660 90% 1.11 ○ 1.11 in cases of 2 injuries NL ECIP 14,519 14,071 97% 1.03 AAAM10 8480 12,123 143% 0.70 ○ 1.05 in cases of 3 injuries • Do not use the ICDpic tool when codes are truncated. • The following weighting factors could be used to correct for trun- are the types of injuries that cause significant or long-term damage and cated codes: consequences and this is where efforts should be focused (European • 1.06 in case of ICDmap90 or DGT Commission, 2013). This paper emphasises the impact that different • 1.03 in case of ECIP inclusion and exclusion criteria for selecting casualties from hospital • 1.11 in case of AAAM9 data has on overall estimated number of serious injuries. Additionally, the methods for deriving MAIS using conversion tools from ICD diag- 4.1. General criteria for case selection/inclusion noses also influence end estimates of MAIS3 + casualties. It is hoped that by harmonising the way hospital data is selected for inclusion there As it is shown in the results, it is necessary to exclude re-admissions will be increased comparability of MAIS3 + injuries across countries. as it would lead to overestimates. The time established to consider a re- Finally, a definition of road traffic injuries for use with hospital data is admission might be different across countries preferably during a 12- recommended. month period from the date of the hospitalisation. However, it is not When using hospital data, missing E-codes, the AIS version applied always possible to identify them due to data protection. As it is ex- and considering only a limited number of injuries particularly influ- plained, in Spain a re-admission is defined as “within 30 days in the ences the estimated number of MAIS3 + casualties. A number of same hospital” due to hospital procedures (Pérez et al., 2014). Re-ad- weighting factors are proposed to deal with some of these differences. missions are identified by the hospitals using information not available However, it was not possible to derive weighting factors for all meth- in the data file made available for road accident research. In other odological issues. The accuracy of estimates will vary if the criteria for countries, as in the Netherlands, re-admissions can be identified in the processing hospital data differ between countries. Failure to obtain other hospitals as well over a period of a full calendar year, as the HDR reliable and valid figures will result in under or over reporting and will of the previous year is available and readmissions of road traffic ca- compromise the comparisons across countries. sualties can be excluded in a de-duplication process where a patient In summary, following exploratory analysis with data sets from with the same date of birth, gender, township of residency and main several European countries, the following criteria for identifying road diagnosis can be found. traffic cases from hospital data are recommended in order to obtain the The way scheduled admissions (contrary to acute admissions) are most accurate estimation of MAIS3 + casualties possible: treated can also vary from one country to the other. If a scheduled admission is interpreted as the second medical examination after a first • Exclude fatalities within 30 days after admission visit to the hospital (probably in the accident & emergency department) • Fatalities after 30 days should be counted as injured according to and the diagnosis information is specific enough to identify a his/her MAIS MAIS3 + injury and this case is present in the hospital data file, then • Exclude re-admissions related to the same crash to avoid duplicates the scheduled admission can be left out as information of the first en- • If it is possible to avoid duplicates by just excluding re-admissions, it counter is present. If re-admissions can be excluded from the database is not necessary to exclude scheduled admissions. If not, scheduled (as these are explicitly recorded or can be identified in a de-duplication admissions should be excluded. process), it is not necessary to exclude scheduled admissions as the 131 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 process of de-duplication takes care of that. If re-admissions are not between countries is missing external codes and it is not possible to specifically recorded, as in the case of Spain, then scheduled admissions calculate weighting factors for that because it is very country specific. they should be also excluded as they are likely to include many re- admissions. 4.2. Methods to derive MAIS3+ Ideally MAIS3 + casualties that are not admitted to the hospital should also be taken into account as a serious road injury. However, as There is very limited amount of literature comparing directly coded in most countries, hospital discharge data is the only database available (M)AIS levels to ICD-derived (M)AIS levels, derived using recoding to estimate the number of serious road injuries, it is not possible to tools. Both Di Bartolomeo et al (2010) and Greene et al (2015) com- include non-hospitalised casualties. The Rhone register in France esti- pared severity levels generated by ICDpic with severity levels based on mates that exclusion of non-hospitalised MAIS3 + casualties results in direct AIS coding. De Bartolomeo et al only compared ISS (Injury Se- an underestimation of the number of serious road injuries of roughly verity Score, Injury scale based on AIS (Baker et al., 1974)) scores and 5%. The actual inclusion or not of outpatients within the definition of conclude on the basis of 289 cases that agreement between scores based serious road traffic injuries varies depending on the country; however, on ICDpic and scores based on direct coded AIS is poor. The main cause it is convenient to include them in the definition all traffic injury hos- for this poor agreement was incomplete ICD9-CM coding. The study of pitalisations, even if they only generate short stays in the hospital. Greene et al. (2015) had a much larger sample of over 40,000 patients In the case of the inclusion of late effects of injury and complications and compared both AIS levels and ISS for ICDpic and direct AIS coding. of physical trauma (905–909, 958 and 959) it should be noted that they They found that the performance of the ICDpic tool differs by body automatically are excluded when obtaining severity. As it is previously region; injury severity is reasonably well classified for thoracic and explained some E-Codes do not fulfil the police definition of “traffic abdominal injuries, moderately well for head and neck injuries, but accident” such as E826, E820-E825, E827, E828, E829 and E988.5. It is only fair for face and extremity injuries. However, ICDpic performs suggested to include the E-codes: E810-E19, E826, E827, E829 and quite well in classifying AIS3 + injuries for each body region. Greene E988.5 and exclude E828 because this is not supposed to fulfilled et al conclude that ICDpic may be a preferred tool in determining injury Eurostat definition of motor vehicle traffic in the sense that maybe there severity for large trauma datasets, but caution needs to be taken when is not at least one vehicle in motion involved in a public road. examining smaller trauma sets. Frequently external causes of injury are underreported. For ex- AAAM convertors were developed by the Association for the ample, in Belgium, the registration of E-codes in hospitals has been Advancement of Automotive Medicine for the European Commission, compulsory since 2003. Yet, despite this obligation, they are still not and are the official tool recommended and freely distributed by DG consistently recorded. However, registration improves year after year: MOVE (Directorate-General for Mobility and Transport). The tool pro- in 2004 35% of hospitalised patients with a principal diagnosis within vides for any specific diagnosis whether its MAIS is < or ≥ 3. The the range 800–959.9 did not receive any E-code. Since 2008 this per- European Commission acknowledges that the AIS (in all its versions) is centage has always been lower than 20% and in 2010 this figure was no the property of the Association for the Advancement of Automotive more than 16%. This percentage covers all patients visiting a hospital Medicine (AAAM), owner of the Copyright (European Commission, because of an external cause. It is uncertain to what extent this overall 2013). Results from our study show that AAAM for ICD9-CM provides percentage can be extrapolated to the whole subgroup of road victims. acceptable figures, while for ICD10 results in an underestimation of In Spain, 17% of records are missing the E-code. Due to the lack of a number of serious injuries. complete database with E-codes it is not possible to derive a weighting The consequences of truncated injury codes and of including a factor adjusting for missing E-codes. Some countries look for other limited number of injuries when determining MAIS were investigated variables to identify traffic injury cases. In Spain for example, “com- using data from Spain, Belgium and the Netherlands. The effect of pensation Payer Company” is used as an additional variable to select truncation appears to depend on the recoding tool that is applied; road traffic casualties. The use of this variable enabled identifying 25% AAAM10 and ICDpic do not seem able to deal with truncated codes. of all road traffic casualties that were not identified by the selection of ICDpic shows a large decrease in the number of MAIS3 + casualties E-codes. when injuries are truncated while the AAAM10 tool shows a con- ICD10 distinguishes between “Traffic accident” (any vehicle acci- siderable increase in the number of MAIS3 + casualties. We re- dent on a public road) and “Non-traffic accident” (any vehicle accident commend repeating this analysis when the conversion tables are occurring entirely somewhere other than on a public road). To address adapted for the European needs to have a closer look at how the tool difficulties in knowing whether a crash is or is not a traffic accident, deals with truncated codes. recommendations are to include codes V01-89 and/or weighting -cor- Moreover, it is recommended that the AAAM10 tool is adapted to recting for non-public road- for non-traffic injury codes. E-codes from better-fit European needs. Other tools – ICDmap90, DGT, and AAAM9- ICD9-CM do not allow identification of whether the collision occurred result in an underestimation of MAIS3 + casualties between 3% and in a public road or not. The E-code 849 allows identification of this but 10%. Including less than 4 injuries when determining MAIS results may not be accurately collected. However, if the proportion of cases leads to an underestimation in serious injuries. On average, taking that occur away from a public road is known through any other data into account only 1 injury results in an estimated number of source, a weighting factor can be applied to avoid overestimation. MAIS3 + casualties which is 78% of the number that is estimated on Inclusion criteria for ICD10 in the Netherlands, ‘non traffic’ is not ac- the basis of all injuries. When 2 injuries are taken into account, 90% of curately coded, so therefore these cases are included using a weighting the serious road injuries are identified and when 3 injuries are taken factor (Bos et al., 2014). into account, on average 95% of the number of MAIS3 + casualties are To improve the accuracy of serious injury estimates it is re- identified. commended to obtain weighting factors from at least a sample of that This study has a number of limitations. In general, most results are country’s hospital data. If this is not possible it may then be appropriate based on data from only a few countries so are likely not to be re- to apply weighting factors from another country. It is recommended to presentative of all countries, particularly those that differ in motorisa- only include weighting factors that are 0.97 or smaller, as higher tion and share of modes of transport. Also, the weighting factors that weighting factors only have a very limited effect. Moreover, we should are proposed are based on in a limited set of data. More specifically, the note that the weighting factors generated from this work are based on comparison of MAIS3 + casualties based on direct coding with ICD two countries only and therefore should be applied with caution. derived codes, are based on a very small sample and concern trauma Nonetheless applying weighting factors makes countries slightly more cases with many and relatively serious injuries that are coded in the comparable. One of the biggest obstacles weakening the comparability German version of ICD10 (ICD10GM). We recommend repeating the 132 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 exercise with a larger sample and an ICD version that is used within the Funding recoding tools. In conclusion, the current paper has considered the implications for This work has been funded by the SafetyCube (Safety CaUsation, deciding which hospital cases should be included as part of Benefits and Efficiency) (H2020 N.633485). MAIS3 + calculations and how to derive AIS from ICD conversions. The most suitable hospital data should provide at least 4 diagnosis of in- Acknowledgments juries, no truncation of ICD codes, registration of E-codes, and use the latest version of AIS (2008). The definition of road traffic casualty Working group of WP7 SafetyCube project: Pérez, K., Olabarria, M. should: exclude fatalities within 30 days after admission, readmissions, (ASPB, Agència de Salut Pública de Barcelona), Spain; Weijermars, W., and scheduled admissions; and include all cases with any injury diag- Bos, N. (SWOV Institute for Road Safety Research), Netherlands; nosis (ICD9CM: 800–999; ICD10: S00-T88), and external causes for Machata, K., Bauer, R. (KFV, Austrian Road Safety Board), Austria; road traffic injuries: (ICD9CM: E810-E819, E826, E827, E829, E988.5; Amoros, E., Martin, JL., Pascal, L. (IFSTTAR, French Institute of Science ICD10: V01-89 for those codes for traffic injuries and/or weighting and Technology for Transport, development and Networks), France; -correcting for non-public road- for non-traffic injury codes). When Filtness, A. (LOUGH, Loughborough University), United Kingdom; there is incomplete data to ascertain a road injury, weighting factors Dupont, E., Nuyttens, N.,Van den Berghe, W. (Vias institute); could be used to correct data deviations and make more real estima- Johannsen, H. (MHH, Medical University of Hannover), Germany; tions. It is essential that maximum effort be applied to obtaining hos- Leskovsek, B. (AVP, Slovenian Traffic Safety Agency), Slovenia. pital data of high quality which best allows estimates of serious injuries, We thank Laurie Brown from Loughborough University for re- to monitor trends and allow comparisons between countries. viewing the manuscript. Appendix A Table A1 Distribution of external causes for ICD10 (V codes) among traffic injuries (Mortality Register, Spain 2009–2013). CODE Traffic Injuries Non-traffic Injuries Total V01 Pedestrian injured in collisions w pedal cycle 7 0 7 V02 Pedestrian injured in collision w 2-3PW 61 3 64 V03 Pedestrian injured in collision w car, pick-up truck or van 923 14 937 V04 Pedestrian injured in collision w heavy transport vehicle or bus 186 13 199 V05 Pedestrian injured in collision w railway train or railway vehicle 41 217 258 V06 Pedestrian injured in collision w other nonmotor 1 1 2 V09 Pedestrian injured in other and unspecified transport accidents 1010 118 1128 V10 Cyclist injured in coll w pedest or animal 0 0 0 V11 Cyclist injured in coll w other cycle 0 0 0 V12 Cyclist injured in coll w 2-3PW 8 0 8 V13 Cyclist injured in coll w car, pick-up truck or van 146 3 149 V14 Cyclist injured in coll w heavy transport veh or bus 31 0 31 V15 Cyclist injured in coll w railway train or railway veh 1 1 2 V16 Cyclist injured in coll w other nonmotorveh 0 0 0 V17 Cyclist injured in coll w fixed or stationary object 13 0 13 V18 Cyclist injured in noncoll transport acid 52 15 67 V19 Driver cyclist injured in coll w other and unspecif motor veh 82 5 87 V20 Motorcycle rider inj in coll w pedest or animal 13 1 14 V21 Motorcycle rider inj in coll w cycle 2 0 2 V22 Motorcycle rider inj in coll w 2-3PW 30 1 31 V23 Motorcycle rider inj in coll w car, pick-up truck or van 509 4 513 V24 Motorcycle rider inj in coll w heavy transport veh or bus 91 1 92 V25 Motorcycle rider inj in coll w railway train or railway veh 3 0 3 V26 Motorcycle rider inj in coll w other nonmotorveh 1 0 1 V27 Motorcycle rider inj in coll w fixed or stationary object 224 8 232 V28 Motorcycle rider inj in noncoll transport accid 328 7 335 V29 Motorcycle rider inj in coll w other and unspecif motor veh 427 20 447 V30 Occupant of 3PW inj in coll w pedest or animal 0 0 0 V31 Occupant of 3PW inj in coll w cycle 0 0 0 V32 Occupant of 3PW inj in coll w 2-3PW 2 0 2 V33 Occupant of 3PW inj in coll w car, pick-up truck or van 0 0 0 V34 Occupant of 3PW inj in coll w heavy transport veh or bus 3 0 3 V35 Occupant of 3PW inj in coll w railway train or railway veh 0 0 0 V36 Occupant of 3PW inj in noncoll transport accid 0 0 0 V37 Occupant of 3PW inj in coll w fixed or stationary object 1 0 1 V38 Occupant of 3PW inj in noncoll transport accid 1 0 1 V39 Occupant of 3PW inj in coll w other and unspecif motor veh 0 0 0 V40 Car occupant inj in coll w pedest or animal 13 0 13 V41 Car occupant inj in coll w cycle 1 0 1 V42 Car occupant inj in coll w 2-3PW 8 2 10 V43 Car occupant inj in coll w car, pick-up truck or van 1239 7 1246 V44 Car occupant inj in coll w heavy transport veh or bus 567 5 572 V45 Car occupant inj in coll w railway train or railway veh 17 0 17 V46 Car occupant inj in coll w other nonmotorveh 7 0 7 V47 Car occupant inj in coll w fixed or stationary object 554 13 567 V48 Car occupant inj in noncoll transport accid 1103 20 1123 (continued on next page) 133 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 Table A1 (continued) CODE Traffic Injuries Non-traffic Injuries Total V49 Car occupant inj in coll w other and unspecif motor veh 702 25 727 V50 Occupant of pick-up truck or van inj in coll w pedest or animal 6 0 6 V51 Occupant of pick-up truck or van inj in coll w cycle 0 0 0 V52 Occupant of pick-up truck or van inj in coll w 2-3PW 1 0 1 V53 Occupant of pick-up truck or van inj in coll w car, pick-up truck or van 42 1 43 V54 Occupant of pick-up truck or van inj in coll w heavy transport veh or bus 73 0 73 V55 Occupant of pick-up truck or van inj in coll w railway train or railway veh 0 1 1 V56 Occupant of pick-up truck or van inj in coll w other nonmotorveh V57 Occupant of pick-up truck or van inj in coll w fixed or stationary object 13 0 13 V58 Occupant of pick-up truck or van inj in noncoll transport accid 60 1 61 V59 Occupant of pick-up truck or van inj in coll w other and unspecif motor veh 7 7 14 V60 Occupant of heavy transpvehinj in coll w pedest or animal 0 0 0 V61 Occupant of heavy transpvehinj in coll w cycle 0 0 0 V62 Occupant of heavy transpvehinj in coll w 2-3PW 0 0 0 V63 Occupant of heavy transpvehinj in coll w car, pick-up truck or van 19 0 19 V64 Occupant of heavy transpvehinj in coll w heavy transport veh or bus 78 0 78 V65 Occupant of heavy transpvehinj in coll w railway train or railway veh 1 0 1 V66 Occupant of heavy transpvehinj in coll w other nonmotorveh 0 0 0 V67 Occupant of heavy transpvehinj in coll w fixed or stationary object 14 0 14 V68 Occupant of heavy transpvehinj in noncoll transport accid 125 6 131 V69 Occupant of heavy transpvehinj in coll w other and unspecif motor veh 36 11 47 V70 Bus occupant inj in coll w pedest or animal 0 0 0 V71 Bus occupant inj in coll w cycle 0 0 0 V72 Bus occupant inj in coll w 2-3PW 0 0 0 V73 Bus occupant inj in coll w car, pick-up truck or van 6 0 6 V74 Bus occupant inj in coll w heavy transport veh or bus 8 0 8 V75 Bus occupant inj in coll w railway train or railway veh 1 0 1 V76 Bus occupant inj in coll w other nonmotorveh 0 1 1 V77 Bus occupant inj in coll w fixed or stationary object 0 0 0 V78 Bus occupant inj in noncoll transport accid 25 1 26 V79 Bus occupant inj in coll w other and unspecif motor veh 11 3 14 V80 Animal-rider or occupant of animal-drawn vehicle injured in transport accident 0 41 41 V81 Occupant of railway train or railway vehicle injured in transport accident 1 95 96 V82 Occupant of streetcar injured in transport accident 1 0 1 V83 Occupant of special vehicle mainly used on industrial premises injured in transport accident 1 6 7 V84 Occupant of special vehicle mainly used in agriculture injured in transport accident 131 297 428 V85 Occupant of special construction vehicle injured in transport accident 2 11 13 V86 Occup of special all-terrain or other motor veh designed primarily for off-road use, inj in transpaccident 9 10 19 V87 Traffic accident of specified type but victim's mode of transport unknown 14 0 14 V88 Non-traffic accident of specified type but victim's mode of transport unknown 0 2 2 V89 Motor- or nonmotor-vehicle accident, type of vehicle unspecified 1,671 89 1,760 V90-V99 0 214 214 Total 10,770 1295 12,065 Table A2 External causes for ICD9-CM (E-codes) and inclusion recommendations. ICD9 Codes Traffic Injuries E810 Motor vehicle traffic accident involving collision with train All E811 Motor vehicle traffic accident involving re-entrant collision with another All motor vehicle E812 Other motor vehicle traffic accident involving collision with motor vehicle All E813 Motor vehicle traffic accident involving collision with other vehicle All E814 Motor vehicle traffic accident involving collision with pedestrian All E815 Other motor vehicle traffic accident involving collision on the highway All E816 Motor vehicle traffic accident due to loss of control, without accident while All boarding or alighting E817 Non collision motor vehicle traffic accident while boarding or alighting No injuring passenger in motor vehicle other than motorcycle E818 Other no collision motor vehicle traffic accident All E819 Motor vehicle traffic accident of unspecified nature All E826 Pedal cycle accident Weighted E827 Animal-drawn vehicle accident Weighted E828 Accident involving an animal being ridden No E829 Other road vehicle accident Weighted E988.5 Injury by crashing of motor vehicle, undetermined whether accidentally All or purposely inflicted For patients in collisions without involvement of motorvehicles (external cause E826-E829) it is not known if the crash occurred on a public road. Therefore it is recommended that only a fraction is counted as traffic casualties. This can be achieved by a weighting factor or a random selection. For motor vehicle crashes the range E820-E825 indicates that the crash did not occur on a public road. 134 K. Pérez et al. Accident Analysis and Prevention 130 (2019) 125–135 Table A3 Externals causes for ICD10 (V codes) and inclusion recommendations. ICD10 - codes Traffic injuries Non-traffic transport injuries (Did not occur on public road) Pedestrian: V01 - V06 .1, .9 All .0 Weighted V09 .2, .3 All .0, .1, .9 Weighted Pedal cyclist: V10 - V18 .4, .5, .9 All .0, .1, .2, .3 Weighted V19 .4, .5, .6, .9 All .0, .1, .2, .3, .8 Weighted Motorcycle rider: V20 - V28 .4, .5, .9 All .0, .1, .2, .3 Weighted V29 .4, .5, .6, .9 All .0, .1, .2, .3, .8 Weighted Occupant of three-wheeled motor vehicle: V30 - V38 .5, .6, .7, .9 All .0, .1, .2, .3, .4 Weighted V39 .4, .5, .6, .9 All .0, .1, .2, .3, .8 Weighted Car occupant: V40 - V48 .5, .6, .7, .9 All .0, .1, .2, .3, .4 Weighted V49 .4, .5, .6, .9 All .0, .1, .2, .3, .8 Weighted Occupant of pick-up truck or van: V50 - V58 .5, .6, .7, .9 All .0, .1, .2, .3, .4 Weighted V59 .4 .5, .6, .9 All .0, .1, .2, .3, .8 Weighted Occupant of heavy transport vehicle: V60 - V68 .5, .6, .7, .9 All .0, .1, .2, .3, .4 Weighted V69 .4, .5, .6, .9 All .0, .1, .2, .3, .8 Weighted Bus occupant: V70 - V78 .5, .6, .7, .9 All .0, .1, .2, .3, .4 Weighted V79 .4, .5, .6, .9 All .0, .1, .2, .3,.8 Weighted Animal-rider or occupant of animal-drawn vehicle V80 – .0.1, .2, .3, .4, .5, .7, .9 Weighted Occupant of railway train or railway vehicle: V81 .0, .1 All .0, .2, .3, .4, .5, .6, .7, .8, .9 Weighted Occupant of streetcar: V82 .1, .9 All .0, .2, .3, .4, .5, .6, .7, .8 Weighted Occupant of special industrial vehicle: V83 .0, .1, .2, .3 All .4, .5, .6, .7, .9 Weighted Occupant of special agricultural vehicle: V84 .0, .1, .2, .3 All .4, .5, .6, .7, .9 Weighted Other: V85 - V86 .0, .1, .2, .3 All .4, .5, .6, .7, .9 Weighted V87 .0, .1, .2, .3, .4, .5, .6, .7, .8, .9 All – V88 – .0, .1, .2, .3, .4, .5, .6, .7, .8, .9 Weighted Type of vehicle not specified: V89 .2, .3 All .0, .1, .9 Weighted References Jeanne Breen, 2012. High-Level Group on Road Safety Consultation on the Development of the Injuries Strategy 2 nd Working Document. OECD/ITF, 2011. Reporting on Serious Road Traffic Casualties: Combining and Using Baker, S., O’Neill, B., Haddon, W., Long, W.B., 1974. The injury severity score: a method Different Data Sources to Improve Understanding of Non-Fatal Road Traffic Crashes. for describing patients with multiple injuries and evaluation of emergency care. J. Omran, A.R., 1971. The epidemiologic transition: a theory of the epidemiology of po- Trauma 14, 187–196. pulation change. Milbank Mem. Fund Q. Health Soc. 49, 509–538. Bartolomeo, S., Di, Tillati, S., Valent, F., Zanier, L., Barbone, F., 2010. ISS mapped from Pérez, K., Seguí-Gómez, M., Arrufat, V., Barberia, E., Cabeza, E., Cirera, E., Gil, M., ICD-9-CM by a novel freeware versus tra ditional coding: a comparative study. Scand. Martín, C., Novoa, A.M., Olabarría, M., Lardelli, P., Suelves, J.M., Santamariña- J. Trauma Resusc. Emerg. Med. 18, 1–7. Rubio, E., 2014. Definition of hospital discharge, serious injury and death from traffic Berecki-Gisolf, J., Collie, A., McClure, R., 2013. Work disability after road traffic injury in injuries. Gac. Sanit. 28, 242–245. http://dx.doi.org/10.1016/j.gaceta.2013.10.006. a mixed population with and without hospitalisation. Accid. Anal. Prev. 51, 129–134. Ringdal, K.G., Hestnes, M., Palmer, C.S., 2012. Differences and discrepancies between http://dx.doi.org/10.1016/j.aap.2012.11.010. 2005 and 2008 abbreviated injury scale versions - time to standardise. Scand. J. Bos, N.M., Houwing, S., Stipdonk, H.L., 2014. Ernstig Verkeersgewonden 2013. SWOV. R- Trauma Resusc. Emerg. Med. 20, 11. http://dx.doi.org/10.1186/1757-7241-20-11. 2014-31, https://www.swov.nl/en/publication/ernstig-verkeersgewonden-2013. Road Safety Annual Report 2015, 2015. European Commission, 2016. Road Safety: New Statistics Call for Fresh Efforts to Save Smith, G.S., Langlois, J.A., Buechner, J.S., 1991. Methodological issues in using hospital Lives on EU Roads. discharge data to determine the incidence of hospitalized injuries. Am. J. Epidemiol. European Commission, 2013. On the Implementation of Objective 6 of the European 134, 1146–1158. Commission’S Policy Orientations on Road Safety 2011–2020 – First Milestone Valletta Declaration on Road Safety, 2017. Towards an Injury Strategy (Commission Staff Working Document). Weijermars, W., Bos, N., Stipdonk, H.L., 2015. Serious road injuries in the Netherlands Greene, N.H., Kernic, M.A., Vavilala, M.S., Rivara, F.P., 2015. Validation of ICDPIC dissected. Traffic Inj. Prev. 37–41. http://dx.doi.org/10.1080/15389588.2015. software injury severity scores using a large regional trauma registry. Inj. Prev. 1042577. 325–330. http://dx.doi.org/10.1136/injuryprev-2014-041524. 135