Safety Risk Model: Risk Profile Report, version 7.5

Transcription

1 Safety Risk Model: Risk Profile Report, version 7.5 This report is issued by RSSB. If you would like to give feedback on any of the material contained in this report, please contact: Marcus Dacre RSSB Block 2, Angel Square 1 Torrens Street London EC1V 1NY risk@rssb.co.uk RAIL SAFETY AND STANDARDS BOARD LTD ALL RIGHTS RESERVED This publication may be reproduced free of charge for research, private study or for internal circulation within an organisation. This is subject to it being reproduced and referenced accurately and not being used in a misleading context. The material must be acknowledged as the copyright of RSSB and the title of the publication specified accordingly. For any other use of the material please apply to RSSB for permission. This publication can be accessed via the RSSB Rail Risk Portal at This is Issue 1.1 of the report. It supersedes Issue 1.0, which was published in March Issue 1.1 updates the High Level Output Specification safety metrics, primarily to incorporate information from the 2012 RSSB survey of workforce hours, which has become available since the report was initially published. The SRM risk estimates are unchanged. Issue 1.1 Published June 2013

2 This page has been intentionally left blank ii Version 7.5 June 2013

3 Table of Contents Executive Summary... vii 1 Introduction SRM Objectives and Overview Objectives Overview Key assumptions and exclusions SRMv7.5 Update Strategy Overview of the SRMv7.5 update Modelling and scope changes from SRMv6.5 to SRMv6.6 and SRMv7 to SRMv Update from SRMv7.1 to SRMv Total Risk on the Mainline Railway Overall profile Risk by ASPR hazard categorisation Discussion HLOS Safety Metrics Background Changes to the baseline HLOS safety metrics Progress against the HLOS safety metrics Uncertainty in the SRM risk estimates Introduction Uncertainty Methodology SRM Model Uncertainty Results Next steps SRM Governance and SRM Updates SRM Practitioners Working Group Update history Updates to the SRM during CP RSSB Rail Risk Portal SRM Risk Profile Bulletins/ Risk Profile Report Yards, Depots and Sidings SRM Risk Profile Report Risk assessment guidance Version 7.5 June 2013 iii

4 Contents 8.4 SRM Risk Profile Tool Taking Safe Decisions Analysis Tool Fixed Lineside Telephony Analysis Tool SPAD Risk Ranking Tool Risk Management Forum Assistance and training Injury Weightings Contributors Acronyms and Glossary Acronyms Glossary References iv Version 7.5 June 2013

5 Contents List of Tables Table 1: Table 2: Risk, normalisers and normalised risk by person type... viii Risk by accident type... viii Table 3: System boundaries... 4 Table 4: Total risk by accident category... 8 Table 5: Total risk by person category... 9 Table 6: Total risk to each person category from each accident category Table 7: ASPR hazardous event groupings Table 8: Summary of the revised estimates for the baseline HLOS safety metrics Table 9: Summary of the progress against the HLOS safety metrics Table 10: Summary of risk by person type Table 11: SRM update history Table 12: Injury degrees and weightings Version 7.5 June 2013 v

6 Contents List of Charts & Figures Figure 1: Summary of the SRM modelling and data updates... 5 Chart 1: Total risk profile for passengers, the workforce and members of the public % of total FWI/year Chart 2: Combined risk profile (FWI/year) includes % change from SRMv Chart 3: Combined risk profile (fatalities/year) includes % change from SRMv Chart 4: Chart 5: Chart 6: Chart showing risk distribution and the 95% confidence interval for the total SRMv7.5 HEM/HEN risk Charts showing the risk distribution and the 95% confidence interval for the SRMv7.5 HEM and HEN risk separately Charts showing the risk distribution and the 95% confidence interval for the SRMv7.5 passenger, workforce and public risk separately Chart 7: Timeline for updates of the SRM and the SRM-RPB for CP vi Version 7.5 June 2013

7 Executive Summary Executive Summary The Safety Risk Model (SRM) consists of a series of fault tree and event tree models representing 121 hazardous events (HEs) that collectively define the overall level of risk on the mainline railway. It provides a structured representation of the causes and consequences of potential accidents arising from railway operations and maintenance. The reported risk estimates relate to the network-wide risk and they indicate the current level of residual risk (i.e. the level of risk remaining with the current risk control measures in place and with their current degree of effectiveness). The Department for Transport and the Office of Rail Regulation are using outputs from the SRM as the primary means of measuring the performance of the industry against the High Level Output Specification (HLOS) safety metrics. The risk estimates in version 6 of the SRM (SRMv6) provided the initial baseline against which safety performance through Control Period 4 (CP4, April 2009 to March 2014) will be compared. SRMv7.5 provides the second comparison against the baseline HLOS safety metrics. To enable this comparison to be meaningful, the update of the SRM to version 7.5 has been split into two distinct stages. The first stage was to revise earlier risk estimates by incorporating all modelling changes and error corrections: SRMv6.5 (a previous update to the SRMv6 figures) was thus updated to SRMv6.6 and SRMv7 was updated to SRMv7.1. These interim versions represent the risk as it would have been calculated for SRMv6 and SRMv7, had the modelling changes in SRMv7.5 been implemented at the time these versions were produced. The second stage was then a refresh of selected HE models using data up to the 30 June Train accidents HEs have not been updated in SRMv7.5. In addition to presenting the results for SRMv7.5 and the updated HLOS safety metrics, this report outlines the methodology being developed for quantifying uncertainty in the SRM results. Headlines Overall Risk SRMv7.5 estimates that the overall level of risk (excluding the direct risk from suicide events) for the railway is FWI/year. This represents a decrease of 1.3% from the figure of FWI/year estimated by SRMv7.1 (the revised SRMv7 overall risk). vii Version 7.5 June 2013

8 Contents Risk by Person Type The overall figures for SRMv7.5 and SRMv7.1 can be broken down by each exposed group and compared with their normalisers as shown in Table 1. Table 1: Risk, normalisers and normalised risk by person type Risk (FWI/year) Normaliser Risk (FWI/unit of exposure) v7.5 v7.1 Unit of exposure v7.5 v7.1 v7.5 v7.1 billion passenger km billion passenger journeys million workforce hours [No normaliser used] Absolute passenger risk has increased by 4.9% since SRMv7.1. However, over the same period passenger journeys have increased by 19.2% and passenger kilometres have increased by 8.6% (from 53.1 billion passenger km in SRMv7.1 to 57.7 billion passenger km in SRMv7.5). The main reason for the increase in passenger risk is that there has been a real increase in the risk from HEN-14: Passenger slips, trips and falls. This is due to an increase in the frequency of events seen for this HE in the last two years. However, the risk per passenger journey has decreased, from 42.1 FWI/billion passenger journeys in SRMv7.1 to 37.1 FWI/billion passenger journeys in SRMv7.5. Workforce risk is now 26.9 FWI/year, representing a 0.3% reduction since SRMv7.1. Risk to the public has decreased by 6.9% when compared with the SRMv7.1 figure. Given the uncertainty associated with this risk estimate (which is dominated by trespasser fatalities), this is not considered to indicate any significant change in the underlying risk. Risk by Accident Type Alternatively, the overall risk figure of FWI/year can be broken down by accident type as shown in Table 2. Table 2: Risk by accident type v7.5 Risk (FWI/year) v7.1 Risk (FWI/year) Train accidents (HETs) Movement accidents (HEMs) Non-movement accidents (HENs) HEMs have seen a reduction of 6.5% and HENs have seen an increase of 2.8%. The reduction in estimated trespasser fatality risk explains much of the fall in HEM risk, and the increase in the risk from slips, trips and falls influences the change in HEN risk. The risk from HETs has not changed from SRMv7.1 because none of the HET models were updated with new incident data for SRMv7.5. viii Version 7.5 June 2013

9 Contents HLOS Safety Metrics The SRMv7.5 figures have been used to demonstrate progress against the HLOS safety metrics. These have been calculated as follows: SRMv7.5 passenger safety metric FWI per billion passenger kilometres. SRMv7.5 workforce safety metric FWI per million worker hours. When compared to SRMv6.6, representing the start of CP4, the SRMv7.5 results represent a 5.7% decrease in the HLOS passenger safety metric and an 11.6% decrease in the HLOS workforce safety metric. These should be compared against the target of at least a 3% reduction in both of these safety metrics over CP4. The risk results from SRMv7.5 are presented in this report as a measure of the absolute risk on the mainline railway. As with any quantified risk assessment, the results are estimates and are dependent on modelling assumptions and limitations of the available data. Quantified risk estimates can be a useful input to the decision making process, but should not be the only input, and their inherent uncertainty must be taken into account. Version 7.5 June 2013 ix

10

11 1 Introduction RSSB works with its members to support the development of safety strategies, develop standards and monitor and report on the safety performance of the industry. An understanding of the overall risk level and risk profile of the railway is a key foundation for this role. RSSB supports its members who comprise the railway industry by providing risk information to help them understand their own risk profile and benchmark their performance. This in turn helps them formulate their own safety policies, plans and measures. The Safety Risk Model (SRM) provides the network-wide risk profile and this information is communicated to the industry in a range of ways, the primary one being the SRM Risk Profile Bulletin (SRM-RPB). Version 7.5 of the SRM consists of a series of fault tree and event tree models representing 121 hazardous events (HEs), which collectively define the overall level of risk on the mainline railway. It provides a structured representation of the causes and consequences of potential accidents arising from railway operations and maintenance on railway infrastructure as well as other areas where the industry has a commitment to record and report accidents. The SRM has been designed to take account of both high-frequency, low-consequence events (occurring routinely, and for which there is a significant quantity of recorded data) and low-frequency, high-consequence events (occurring rarely, and for which there is little recorded data). The results for each HE are presented in terms of the frequency of occurrence (number of events per year) and the risk (number of fatalities and weighted injuries (FWI) per year). The FWI weightings equate injuries of differing degree with a fatality event, which allows all of the risk on the railway to be totalled and contrasted in comparable units. These weightings are shown in Section 9. The risk estimates presented can be used to support risk assessments and for judging how the risk relating to particular operations compares with and contributes to the network-wide risk. The information contained in this document relates to the network-wide risk on railway infrastructure covering all running lines, rolling stock types and stations currently in use. Risk associated with areas away from the operational railway, such as yards, depots, sidings (YD&S), or station car parks, is not included (with the exception of workforce involved in road traffic accidents). Work to extend the scope to include YD&S is ongoing and initial risk estimates for depots, yards and sidings are available in the Yards, Depots and Sidings Risk Profile Report [Ref. 1]. The system boundaries for SRMv7.5 are detailed in Section 2.2. The risk estimates in SRMv7.5 are for the current level of residual risk on the mainline railway, which is the level of risk remaining with the current risk control measures in place and with their current degree of effectiveness. The cut-off date for incident data used to inform SRMv7.5 was 30 June Because of the network-wide nature of the SRM, it is necessary to make average assumptions that represent the general characteristics of the network. The model also hinges on the definitions of the HEs and precursors by which risk estimates are reported. 1 There are four hazardous event models that are exceptions to this data cut-off: HEM-12, HEM-25, HEM-31 and HEN-77. The mapping of incidents to these HEs is influenced by coroners reports, which may not be available until sometime after the event. Therefore, to ensure there is confidence in the data used to analyse them, an earlier cut-off date (30 June 2011) was used. Version 7.5 June

12 Introduction These definitions will soon be provided on the RSSB Rail Risk Portal at (see Section 8.10), and a thorough understanding of them is essential to the correct interpretation and use of the risk information reported here. The SRM does not provide risk profiles for specific lines of route and train operating companies (TOCs), although a Risk Profile Tool is also available from the Rail Risk Portal to help scale the results for this purpose. The information in this report should not be considered to be representative of the risk for any particular line of route or TOC, without further localised analysis. The SRM Practitioners Working Group (SRM-PWG) is the industry governance body for the SRM. It was formed under the authority of the Safety Policy Group (SPG) to engage stakeholders in the development and control of future versions of the SRM and its related outputs. Section 7.1 contains more information regarding this group and its aims. The modelling changes implemented as part of the update of the SRM to version 7.5 have been endorsed by the SRM-PWG. The revised version 6 figures (SRMv6.6) and the revised version 7 figures (SRMv7.1) were presented to the group and approved in January The Department for Transport (DfT) is using the outputs from the SRM as the primary means of measuring the performance of the industry against the High Level Output Specification (HLOS) safety metrics, rather than using a measure of safety performance based solely on accident statistics. The risk estimates derived from SRMv6 provided the initial baseline against which to compare safety performance through Control Period 4 (CP4, April 2009 to March 2014). This is achieved by comparing the risk metrics derived from SRMv7.5 and future versions against the baseline safety metrics determined from SRMv6 (and subsequently updated to SRMv6.6). The main part of the SRM-RPR sets out: The objectives of the SRM (Section 2.1) System boundaries (Section 2.2) Overview of the SRMv7.5 update (Section 3) Total risk on the mainline railway (Section 4) Details of progress against the HLOS safety metrics (Section 5) Additional appendices for this document may be downloaded from the RSSB Rail Risk Portal at (in Excel format). Appendix A contains frequency, consequence and risk estimates for each HE (Table A1), and describes the changes from SRMv6.5 to SRMv6.6 (Table A2), from SRMv7 to SRMv7.1 (Table A3), and from SRMv7.1 to SRMv7.5 (Table A4) in detail. Appendix B contains frequency and risk contributions for all precursors leading to each HE (Table B1). 2 Version 7.5 June 2013

13 2 SRM Objectives and Overview 2.1 Objectives The primary objectives of the SRM are: To provide an estimate of the extent of the current risk on the mainline railway. To provide risk information and risk profiles relating to the mainline railway. The SRM has been developed and published to support RSSB members. Its specific purpose is to provide risk estimates for use in risk assessments, appraisals and decisions throughout the railway industry. This includes: To enable risk-informed assessments and cost-benefit analyses (CBAs) to be carried out to support decisions taken about: Whether changes to the railway can be made safely Which control measures should be applied on the railway Where current risk control measures can be relaxed or changed. Technical modifications and upgrades such as new infrastructure investment. Revision of Railway Group Standards (RGS), in terms of their contribution to risk mitigation (including development of impact assessments for proposed changes to the RGS). To provide risk information to support: The development of priorities for the Industry Strategic Business Plans [Ref. 2]. Prioritisation of areas for research on the railway. Transport operator risk assessments, as required by The Railways and Other Guided Transport Systems (Safety) Regulations Significant changes which require application of the Common Safety Method on Risk Evaluation and Assessment. Identification and prioritisation of issues for audit. To provide an understanding about the contribution of a particular item of equipment or failure mode to the overall risk. To provide risk estimates to be used as the basis of the HLOS safety metrics. 2.2 Overview The SRM includes the safety risk from incidents which could occur during the operation and maintenance of the mainline railway within the boundaries defined in Table 3. For SRMv7.5 there have been no changes to the system boundary scope, however one new HE within this scope has been identified and this is discussed further in Section 3. Appendix G of [Ref. 3] contains a more detailed discussion of the SRM scope and the system boundaries and gives specific guidance as to what aspects of the operation and maintenance of the railway are within scope of the SRM. 3 Version 7.5 June 2013

14 SRM Overview Table 3: System boundaries People Pasengers on trains In SRM Scope Pasengers at stations within areas to which they have legitimate access. Railway workers on trains Railway workers in public areas at stations Railway workers working on or near the line Railway workers in signal boxes, signalling centres, or electrical control offices Railway workers involved in road traffic accidents while on duty. Members of Public (not passengers) outside the mainline railway or legitimately crossing the mainline railway (i.e. on level crossings). Members of Public who enter the mainline railway with no legitimate purpose (e.g. tresspassers including passengers who enter areas for which they have no legitimate access). Events associated with vandalism and Members of Public falling or trespassing on the mainline railway are also included. Injuries directly associated with suicides or attempted suicide are quantified but not included in the overall results discussion. On trains All on-train events. Events on the mainline railway which affect trains including level crossings. All accidents related to the movement of OTP that occur within possessions. In stations All public areas associated with the movement of passengers and staff inside the physical boundaries of stations. Not in SRM Scope People The SRM does not quantify the risk to staff due to long-term occupational health issues. Risk associated with terrorist activity is excluded. Yards, sidings and depots Events occurring within yards, sidings and depots are not included within the SRM (this is being developed). However, those events relating to the movement of trains entering and leaving yards, sidings and depots, and events relating to the condition of trains joining the system from the depots have been included. In stations Non-public areas at stations, i.e. the work side of a ticket office (however, where a member of the workforce is assaulted by a Member of Public who is on the public side of the office, this has been included). Retail outlets within stations. Station toilets. Everything roadside of a station e.g. car parks, access roads, forecourts, taxi ranks etc. Offices. 2.3 Key assumptions and exclusions Appendix F of version 7 of SRM-RPB [Ref. 3] lists key assumptions that are applicable to the SRM. Further clarity on the definitions and assumptions applicable to individual HEs can be provided on request. 4 Version 7.5 June 2013

15 3 SRMv7.5 Update Strategy 3.1 Overview of the SRMv7.5 update The SRM is being used by the DfT as the primary means of measuring the performance of the industry against the HLOS safety metrics. As a result of this it is now necessary to be able to distinguish between changes in risk arising from genuine changes in the underlying data and changes due to refined modelling of HEs. To enable this comparison to be meaningful, the update of the SRM to version 7.5 has been split into two distinct stages. The first stage was to incorporate all changes and error corrections into the model and produce revised versions of the previous models SRMv6.6 and SRMv7.1. These interim versions represent the risk as would have been calculated for SRMv6 and SRMv7, had the modelling changes implemented in version 7.5 been implemented at the time the previous versions were created. The second stage was then a data refresh of selected HE models with data up to 30 June Risk estimates from SRMv7.1 to SRMv7.5 can therefore be meaningfully compared; the difference between them represents the estimated change in risk due to a refresh of the data up to June 2012 (compared to September 2010 for version 7.1). This is the second time the SRM has been updated in this fashion, the other occasion being SRMv7. The aim is to be able to provide a framework that is compatible with the requirements for monitoring the HLOS metrics and to provide risk estimates for different points in time that are calculated on a like-for-like basis. Figure 1 below summarises the different SRM versions and how they have been derived. For HLOS comparative purposes, the horizontal arrows show valid comparisons based solely on changes in data between the different versions of model. Figure 1: Summary of the SRM modelling and data updates SRMv6 Baseline Remodelled to SRMv6.5 Data refreshed to SRMv7 First comparison Remodelled Remodelled to to SRMv6.6 Data refreshed to SRMv7.1 Data refreshed to SRMv7.5 Second comparison 5 Version 7.5 June 2013

16 SRMv7.5 Update 3.2 Modelling and scope changes from SRMv6.5 to SRMv6.6 and SRMv7 to SRMv7.1 The significant changes from SRMv6.5 to SRMv6.6 can be split into two main categories the first is the introduction of new hazardous events (usually to provide better event classification/understanding) and the second is modelling changes. Changes that fall under the first category include: The creation of a new HE in SRMv6.6 and SRMv7.1, namely: Member of public (nontrespasser) fall from platform and struck by train (HEM-49). This was introduced in response to a recent incident. In the second category a number of modelling changes were identified for the update from SRMv6 to SRMv6.5. The main changes are: A number of modelling changes have been made to the derailment models (HET-12, HET-13) for SRMv6.6 and v7.1. The main change is a remodelling of derailments on bridges to account for the fact that a train will not always fall from the bridge (as had previously been assumed). The estimates for HET-04 (collisions of trains with objects not resulting in derailment) in SRMv7.1 have been corrected to account for an error identified in the analysis. The frequency and consequences for a number of HEs in SRMv6.6 and SRMv7.1 have been re-examined in light of the version 7.5 update in order to incorporate improved modelling assumptions and make them comparable between versions. See Table A2 (for SRMv6.5 to SRMv6.6) and Table A3 (for SRMv7 to SRMv7.1) for a full discussion of these. 3.3 Update from SRMv7.1 to SRMv7.5 The update from SRMv7.1 to SRMv7.5 comprises a partial update of the model. None of the train accidents (HETs) were updated. A total of 53 movement accidents (HEMs) and non-movement accidents (HENs) were identified to be updated based on an analysis of the most recent injury data and a comparison with the SRMv7 estimates. The criteria for updating a HE were: Where overall risk for a HE had significantly changed from the SRMv7 estimate. This was determined by looking at the 95% confidence interval for the SRMv7 estimates and testing if the most recent data would result in a new risk estimate outside these limits. Where the overall risk for a HE has changed appreciably. There is a trade-off to be made in defining what is meant by appreciable. If the limit is set low, then every HE will change appreciably. If it is set too high, then very few HEs will have changed appreciably and therefore warrant updating. An absolute risk change of 0.1 FWI/year was selected as being an appropriate level to set as being an appreciable change in risk. In addition to this, a review of the HEs selected for update was also carried out to identify further HEs for update, either because similar HEs were being updated and it made sense to update them as well or because it was felt that the HE was of particular significance and should be updated. 6 Version 7.5 June 2013

17 SRMv7.5 Update The aim of these criteria was to identify HEs that warranted updating in order to accurately reflect the overall change in the risk profile while ensuring that the update could be carried out as efficiently as possible. The updated HEs accounted for around 85% of the overall risk profile. The 53 HEs identified were updated using data from incidents occurring up to and including 30 June For a full discussion and explanation of the significant and appreciable differences between SRMv7.1 and SRMv7.5 see Table A4. 2 There are four hazardous event models that are exceptions to this data cut-off: HEM-12, HEM-25, HEM-31 and HEN-77. The mapping of incidents to these HEs is influenced by coroners reports, which may not be available until some time after the event. Therefore, to ensure there is confidence in the data used to analyse them, an earlier cut-off date was used (30 June 2011). Version 7.5 June

18 4 Total Risk on the Mainline Railway 4.1 Overall profile This section presents the overall risk for the 121 HEs on the mainline railway which are considered within the SRM. Risk is presented in terms of: injury severity by accident category (see Table 4); injury severity by person category (see Table 5); and person injured by accident category (see Table 6). It should be noted that the totals presented exclude the direct risk due to suicide and attempted suicide. However, all secondary risk (e.g. the shock/trauma that can arise when drivers witness suicides) associated with these events has been included. The total risk from the 121 HEs is assessed to be FWI/year. approximately: 67 fatalities per year 484 major injuries per year 2107 Class 1 reportable minor injuries per year Class 2 reportable minor injuries per year 1764 cases of shock/trauma per year This is made up of This compares to FWI/year as calculated in SRMv7.1 (reported in SRM-RPB version 7 as FWI/year). These total risk estimates are broken down by accident category and injury type in Table 4: Total risk by accident category below. Table 4: Total risk by accident category Accident category (POS = inside possession) FWI / year Fatalities / year Major injuries / year Class 1 minor injuries / year Class 2 minor injuries / year Class 1 shock/trauma / year Class 2 shock/trauma / year Train accidents (excl. POS) Movement accidents (excl. POS and trespass) Non-movement accidents (excl. POS and trespass) Inside possession (POS) Trespass Total Note 1: The direct risk from suicide and attempted suicide has been excluded, however all secondary risk associated with suicide has been included. Note 2: Some totals may not appear to add up correctly within the table due to the effects of rounding. Version 7.5 June

19 Total Risk on the Mainline Railway Table 5 shows the risk to each person category on the railway. Risk to the public forms the greatest proportion of the total risk, at 57.7 FWI/year (a decrease of 6.9% from SRMv7.1). This is mainly due to a large number of fatalities from trespassing events. Given the uncertainty associated with this risk estimate, the 6.9% decrease is not considered to indicate a significant change in the underlying risk. Absolute passenger risk has increased by 4.9% since SRMv7.1. However, over the same period passenger journeys have increased by 19.2% and passenger kilometres have increased by 8.6% (from 53.1 billion to 57.7 billion passenger km). The main reason for the increase in passenger risk is from HEN-14: Passenger slips, trips and falls. This is due to an increase in the frequency of this HE in the last two years. However, after the passenger risk is normalised by passenger journeys, it can be seen that risk has actually decreased, as the absolute risk increase (4.9%) is less than the increase in the normaliser (19.2%). The risk to the workforce is now 26.9 FWI/year which represents a 0.3% reduction from SRMv7.1. This small change is the aggregated effect of reductions in the risk from some HEs and increases in the risk from others. Notable reductions were seen for the following event types: Workforce struck by / contact with / trapped by object not at a station Workforce electric shock (conductor rail) Workforce assault The following event types showed increases: Witnessing a traumatic event (movement) Workforce electric shock (overhead line equipment) Witnessing a traumatic event (non-movement) Table 5: Total risk by person category SRMv7.5 SRMv7.1 Person category FWI/year Fatalities / year Major injuries / year Class 1 minor injuries / year Class 2 minor injuries / year Class 1 shock/trauma / year Class 2 shock/trauma / year FWI/year % Change from SRMv7.1 to SRMv7.5 Passenger % Workforce % Public % Total % Note: The direct risk from suicide and attempted suicide has been excluded, however all secondary risk associated with suicide has been included. Note 2: Some totals may not appear to add up correctly within the table due to effects of rounding. Version 7.5 June

20 Total Risk on the Mainline Railway Table 6 presents the risk broken down into person category and accident categories. The table shows that risk due to train accidents (HETs) has not changed as it has not been updated for SRMv7.5. Movement accidents (HEMs) have seen a 6.5% decrease in risk, whereas non-movement accidents (HENs) have seen an increase of 2.8%. Overall, the total decrease in risk from SRMv7.1 to SRMv7.5 is 1.3%. Table 6: Total risk to each person category from each accident category SRMv7.5 SRMv7.1 Hazardous event Total FWI/year Passenger FWI/year Workforce FWI/year Public FWI/year Total FWI/year % Change from v7.1 to v7.5 HET % HEM % HEN % Total % Note: The direct risk from suicide and attempted suicide has been excluded, however all secondary risk associated with suicide has been included. 10 Version 7.5 June 2013

21 Total Risk on the Mainline Railway Chart 1: Total risk profile for passengers, the workforce and members of the public % of total FWI/year Note: The direct risk from suicide and attempted suicide has been excluded, however all secondary risk associated with suicide has been included. Version 7.5 June

22 Total Risk on the Mainline Railway Chart 1 presents the total risk profile for passengers, the workforce and the public. It shows that the bulk of the risk is split between passengers and the public, with 39.2% and 41.4% respectively the remaining proportion (19.4%) is attributed to workforce incidents. The profile of injury across person categories has remained similar to SRMv7, with approximately 50% of the risk to passengers and the workforce resulting from major injuries. Fatalities still dominate the risk to the public, comprising 90.9% of the overall risk to the public per year. Most of these fatalities are trespassers (41.8 fatalities per year). 4.2 Risk by ASPR hazard categorisation In this section, the HEs have been grouped into 22 accident types, 3 consistent with the groupings used in the Annual Safety Performance Report (ASPR) [Ref. 4]. Combining the HEs in this manner allows identification of the types of accidents that contribute the greatest proportion of risk to the overall figure. The HEs have been grouped as shown in Table 7. Table 7: ASPR hazardous event groupings Assault and abuse Contact with object Contact with person Falls from height Event type Fires and explosions (not involving trains) Lean or fall from train in running Machinery/tool operation Manual handling / awkward movement On-board injuries Platform-train interface (boarding/alighting) Platform edge incidents (not boarding/alighting) Road traffic accident Slips, trips and falls Struck/crushed by train Suicide HEN-64, HEN-65, HEN-66 Hazardous events HEM-20, HEM-32, HEM-42, HEN-21, HEN-23, HEN-26, HEN-44, HEN-55, HEN-56, HEN-59, HEN-76 HEN-55, HEN-56 HEN-15, HEN-25, HEN-45 HEN-01, HEN-02, HEN-03, HEN-04, HEN-05, HEN-48, HEN-49 HEM-03, HEM-07, HEM-15, HEM-17 HEN-22, HEN-27, HEN-56 HEN-73, HEN-74, HEN-82, HEN-83 HEM-38, HEM-39, HEN-62, HEN-63 HEM-05, HEM-06, HEM-09, HEM-16, HEM-21, HEM-22, HEM-23, HEM-43 HEM-06, HEM-08, HEM-10, HEM-21, HEM-40, HEM-41, HEM-49, HEN-09, HEN-10, HEN-13, HEN-52, HEN-67 HEN-35 HEN-14, HEN-16, HEN-24, HEN-25, HEN-46, HEN-68 HEM-11, HEM-14, HEM-19, HEM-27 HEM-31, HEN-77 3 Some HEs have been split across two or more accident types. 12 Version 7.5 June 2013

23 Total Risk on the Mainline Railway Table 7: ASPR hazardous event groupings (cntd) Event type Train accidents: collisions and derailments Train accidents: collisions with objects Train accidents: collisions with road vehicles at level crossings Train accidents: other Trespass Workforce electric shock Other Hazardous events HET-01, HET-02NP, HET-02P, HET-03, HET-06, HET-09, HET-12, HET-13, HET-26 HET-04 HET-10, HET-11 HET-17, HET-20, HET-21, HET-22, HET-23, HET-24, HET-25 HEM-12, HEM-25, HEM-30, HEM-44, HEN-36, HEN-37, HEN-38, HEN-39, HEN-40, HEN-41, HEN-42, HEN-43, HEN-71, HEN-72 HEN-27, HEN-30, HEN-31, HEN-32 HEM-01, HEM-50, HEN-07, HEN-08, HEN-11, HEN-27, HEN-28, HEN-29, HEN-33, HEN-50, HEN-51, HEN-53, HEN-54, HEN-57, HEN-58, HEN-60, HEN-61, HEN-70, HEN Discussion Chart 2 presents the risk profile in FWI/year and indicates the percentage change in risk between SRMv7.1 and SRMv7.5 for each of the 22 HE categories listed above. The greatest overall risk contribution results from Trespass with 44.4 FWI/year, which is dominated by fatality risk. The next-highest risk contribution results from Slips, trips and falls with 33.1 FWI/year, an increase of 8.1% compared with SRMv7.1. The majority of risk from Slips, trips and falls occurs to passengers, contributing 25.6 FWI/year, which represents 46.8% of the overall risk to passengers. After Slips, trips and falls, the category which contributes most to the overall risk to passengers is Assault and abuse, representing 8.4 FWI/year, followed by Platform edge incidents (both boarding/alighting and non-boarding/alighting). Considered together, these four categories account for over 80% of the risk to passengers. The greatest workforce risk also comes from the Slips, trips and falls category (6.3 FWI/year), with the second-highest contribution coming from Contact with object (4.2 FWI/year). Together these categories represent 38.9% of the risk to the workforce. A large proportion of the risk to the public results from Trespass (44.3 FWI/year), followed by Struck/crushed by train (not trespass) with 5.9 FWI/year. Together they represent 86.9% of the risk to the public. Version 7.5 June

24 Total Risk on the Mainline Railway Chart 2: Combined risk profile (FWI/year) includes % change from SRMv Version 7.5 June 2013

25 Total Risk on the Mainline Railway Chart 3 shows the combined risk by event type in fatalities per year (excluding the contribution from non-fatal injuries, shock and trauma). Fatality risk is dominated by Trespass accidents, accounting for more than half, with 41.8 fatalities per year down 8.4% on the SRMv7.1 figure. The accident type contributing the second-highest number of fatalities is Struck/crushed by train 4 with 8.3 fatalities per year (the majority of these occurring at level crossings), suggesting a decrease of 0.6% compared with SRMv7.1. Together, these two categories account for 75.0% of fatalities. The highest contribution to passenger fatalities is Platform edge incidents (excluding boarding/alighting), which accounts for 3.6 fatalities per year (representing 34.5% of passenger fatality risk). The most significant contributor to workforce fatalities is Struck/crushed by train, accounting for 1.9 fatalities per year (48.4% of the workforce fatality risk total). The 41.8 fatalities per year due to public Trespass represent 79.8% of public fatality risk. 4 This excludes trespassers struck by trains, as well as people struck by trains at the platform edge or as a result of boarding or alighting accidents Version 7.5 June

26 Total Risk on the Mainline Railway Chart 3: Combined risk profile (fatalities/year) includes % change from SRMv Version 7.5 June 2013

27 5 HLOS Safety Metrics 5.1 Background The government s white paper Delivering a Sustainable Railway [Ref. 5] sets out the HLOS. This describes the improvements in safety, reliability and capacity that the industry is committed to deliver during CP4 (April 2009 to March 2014) and the Statement of Funds Available to secure these improvements. The improvements in safety are quoted in terms of a reduction in two safety metrics. These state that there should be a 3% reduction in the national level of risk for both passengers and the workforce over CP4. The passenger risk is expressed as FWI per billion passenger kilometres, whilst the workforce risk is expressed as FWI per million employee hours. The DfT is using the SRM as the primary means of measuring the performance of the industry against these safety metrics, rather than using a measure of safety performance based on accident statistics. This is because, for rare high-consequence events, the rate of occurrence of accidents over any given period does not provide a good measure of underlying safety performance. The risk estimates from SRMv6 were used in order to calculate the baseline risk from which the 3% reduction will be measured. 5.2 Changes to the baseline HLOS safety metrics As discussed in Section 3, for the purposes of calculating progress against the HLOS safety metrics, the update for SRMv7.5 has been split into two separate and distinct stages. The first stage was to incorporate all changes and error corrections into the model and produce revised versions of the previous models SRMv6.6 and SRMv7.1. These versions represent the risk as would have been calculated for SRMv6 and SRMv7, had the modelling changes implemented in version 7.5 been implemented at the time the previous versions were created. This creates a revised and more accurate baseline for HLOS monitoring. The second stage was then a data refresh of selected HE models with data up to 30 June The changes from SRMv6.6 to SRMv7.5 represent the latest estimate of risk changes since the beginning of CP4. In 2010 RSSB was commissioned by Network Rail to undertake an independent review of compliance with The Reporting of Injuries, Diseases and Dangerous Occurrences Regulations 1995 (RIDDOR) by Network Rail staff and its contractors [Ref. 6]. The review concluded that a number of minor injury events had not been reported into the industry s Safety Management Information System (and hence are missing from the SRM data). Based on the estimated level of under-reporting agreed with Network Rail, for SRMv6.6 an additional contribution of FWI per year has been added to the overall workforce risk to account for the under-reported RIDDOR-reportable minor injury events. For SRMv7.1 and SRMv7.5 additional contributions of and per year FWI, respectively, have been added to the overall workforce risk to account for the under-reported RIDDOR-reportable minor injury events during the period used to calculate the risk estimates. Table 8 below summarises the revised baseline HLOS safety metrics and compares them with the previously calculated values from version 6.5 of the SRM. Note that there are two workforce figures: one with the under-reported RIDDOR reportable minor injuries and one 17 Version 7.5 June 2013

28 HLOS Safety Metrics without. These revised safety metrics also incorporate the revised normaliser figures for the baseline period. Table 8: Summary of the revised estimates for the baseline HLOS safety metrics Passenger FWI / billion passenger km Workforce FWI / million workforce hours Excluding underreported minor injuries Including underreported minor injuries v6.5 HLOS safety metric v6.6 HLOS safety metric % change +0.9% -0.2% -0.2% It should be noted that the baseline HLOS safety metric figures are subject to change as modelling refinements are identified which necessitate a recalculation of the SRM figures. There are therefore likely to be further changes to these numbers in the future when SRMv8 is completed in Progress against the HLOS safety metrics The passenger and workforce risk figures have been used along with the relevant normalisation data to calculate the progress HLOS safety metrics and a comparison is then made against the baseline figures to determine progress against the HLOS target 5. As agreed with the DfT and ORR the HLOS metrics exclude three areas of SRM risk because of concerns about the reliability and consistency of reporting. The exclusions are: Non-physical assaults Workforce involved in road traffic accident whilst on duty 6 Witnessing a traumatic event The passenger km normalisation figure has been taken from the ORR s National Rail Trends (Jul 2011 to Jun 2012) [Ref. 7]. This figure aligns with the data cut-off for SRMv7.5 (30 June 2012). The total number of passenger km for this period is billion, which is an increase of 15% from the passenger km figure used for SRMv6.6 (50.4 billion). The workforce hours normalisation data has been collected as agreed by SPG. The workforce hours estimate has been based on industry responses received by RSSB. This issue of the report uses the 2012 numbers. It updates Issue 1.0, which was based on 2011 figures. The total number of workforce hours in the relevant period is million, which is an increase of 1.6% from the workforce hours estimate used for SRMv6.6 (209.6 million). Table 9 summarises the progress HLOS safety metrics and the comparison of them with the baseline safety metrics. 5 The risk figures and normalisation data used to calculate the metrics exclude the contributions from HS1 which are outside the scope of the HLOS safety metrics but are inside the scope of the SRM. 6 An initiative is currently underway to enable better data collection of these events. 18 Version 7.5 June 2013

29 HLOS Safety Metrics Table 9: Summary of the progress against the HLOS safety metrics 7 Passenger FWI / billion passenger km Workforce FWI / million workforce hours Excluding underreported minor injuries Including underreported minor injuries v6.6 HLOS safety metric v7.5 HLOS safety metric % change -5.7% -10.8% -11.6% From Table 9 it can be seen that there has been a decrease in the passenger safety metric of around 5.7% since the start of Control Period 4 and a larger decrease of around 11.6% in the workforce metric. Table 10 provides a summary of the estimated risk for SRMv6.6 and v7.5 and the differences between them. It also shows this risk broken down by person type (excluding the under-reported RIDDOR reportable minor injuries). Table 10: Summary of risk by person type Person category SRMv7.5 (FWI/year) SRMv6.6 (FWI/year) % Change Passenger % Workforce % Public % Total % Table 10 shows the change in absolute risk (FWI per year) between the HLOS baseline (SRMv6.6) and the latest risk estimates (SRMv7.5). The passenger metric is showing a decrease despite the increase in absolute risk (shown in Table 10) due to the number of passenger km increasing by a greater percentage than the absolute risk increase. The workforce safety metric shows a decrease that is broadly in line with the observed decrease in absolute workforce risk shown in Table Note that the normalised risk figures presented in Table 9 do not match those in Table 1 because some elements of SRM risk are excluded from the HLOS calculation, as described at the start of Section 5.3. Version 7.5 June

30 6 Uncertainty in the SRM risk estimates 6.1 Introduction The SRM risk estimates represent the underlying level of risk on the GB rail network. They are based on an analysis of data reported to SMIS, which is classified according to the HE and precursor structure of the SRM. The amount of data for each precursor in the SRM can vary quite considerably. In some cases, there is a lot of data to base a precursor estimate on, in other cases there is little or very little, and in some cases there is none. The SRM precursor/he estimates represent the best estimate of risk based on the available data for that precursor/he. Where there are a lot of data, there is a high level of confidence in the estimate, however where there are few data there is less confidence in the estimate. It has been a long standing aim to quantify this level of uncertainty and to develop a framework that can determine selected confidence intervals around the SRM estimates. This section will give some background to the work that has been undertaken in this area, outline the methodology that is currently being developed and finally present some results based on the version 7.5 update. 6.2 Uncertainty Methodology In 2003, RSSB funded research at Strathclyde University to investigate quantification of the uncertainty in a risk assessment model, using the SRM as an example. This research formed R&D project T306 [Ref. 8]. While methods investigated apply primarily to the train accident (HET) models (as they are built using fault and event tree models), the principles can be equally applied to the other models of the SRM. The main difference between the HET models and the HEM/HEN models is that in general the HEM/HEN models are based on significant amounts of actual injury-related data, while the HET models are not. This means that the HET uncertainty methodology is mainly aimed at quantifying model uncertainty in conjunction with statistical uncertainty due to the lack of data. The HEM/HEN methodology focus, however, can be aimed more at quantifying statistical uncertainty, as there is ample data. The main idea behind the methodology for quantifying the uncertainty in the HEM/HEN estimates is to construct a framework whereby the uncertainty in the frequency and consequence estimates for each HE can be quantified and combined to give an overall confidence interval for the risk for each precursor. This will involve constructing distributions to model the frequency and the average consequence estimates of each precursor. These distributions could be constructed either from data or expert judgement, or a mixture of the two. Once these have been set, a simulation can be run, where frequency and consequence estimates are sampled from the distributions many thousands of times and combined to give a distribution of the risk estimates for each precursor. These can then in turn be sampled from to give distributions of the risk at HE level, HEM/HEN level or at the overall SRM system risk level. 20 Version 7.5 June 2013

31 Taking Safe Decisions 6.3 SRM Model Uncertainty Results The uncertainty methodology currently in development has been applied to the version 7.5 HEM/HEN risk estimates. The HET risk has been excluded from this analysis along with the additional workforce risk from the under-reported RIDDOR injuries. Excluding these contributions gives an overall SRM combined HEM/HEN risk of 131 FWIyear. Chart 4 below shows the 95% confidence interval 8 around the overall HEM/HEN risk of 131 FW/year. The boxes above the chart show some properties of the distribution of the risk that has been constructed. The central white box gives the mean (average) value of the distribution and it can be seen that this corresponds 9 with the overall SRM HEM/HEN risk estimate. The other two white boxes either side show the extent of the 95% confidence interval, with a lower limit (LL) of FWI/year and an upper limit (UL) of FWI/year. The blue boxes show the absolute risk difference between the LL/UL and the mean. In this case it can be seen that the UL is 6.7 FWI/year above the mean risk, while the LL is 6.4 FWI/year below the mean risk. Chart 4: Chart showing risk distribution and the 95% confidence interval for the total SRMv7.5 HEM/HEN risk. SRM LL Mean UL Chart 5 below shows the 95% confidence interval for HEM risk and HEN risk considered separately, while Chart 6 below shows the 95% confidence interval for passenger, workforce and public risk considered separately. Figures for the mean risk of each of these distributions along with the LL and UL of the 95% confidence interval can be read from the white and blue boxes as before. 8 A confidence interval indicates the range of values an estimate is likely to lie in given a specified level of confidence. In this case a 95% confidence interval means that if we were to rerun (if this was possible) the whole operation of the railway again, as it was over the SRMv7.5 data period, collect the data and calculate a 95% confidence interval for the overall HEM/HEN risk, we would expect in 19 out of 20 occasions (ie 95% of the time) the confidence interval would contain the true underlying level of risk. 9 This is to be expected from the way the HEM/HEN uncertainty model has been constructed. It is not necessarily always the case and depends on the assumptions made to construct the model. Version 7.5 June