A M~icrocomputer Accident Analysis Package and its use In Develoiping Countries. by B.L. Hills (Transport' end Road Research Laboratory) and G.J. Elliott (Transportation Research Group, University of Southampton) ABSTRACT The programs and facilities currently available In the TREL Overseas Unit Microcomputer Accident Analysis Package for, Developing Countries are described and examples of its various outputs given. Since it Is intended to be part of a wider Accident Investigation and Prevention process, the different objectives and techniques of accident analysis at National, Regional and Local levels are br~iefly summarised. The'need to identify the location of accident sites accurately from an accident report form is given a high priority and the system. adopted Is described. The Package itself is de- siVgied for users with no previous computing experience; hard disc and bilingual versions have nox~ been developed for both the Concurrent CP/W4and MSDOS operating systems. At present, It is under evaluation in six countries.,, Paper presented at Indian Road Congress Road Safety Seminar, * Srinagar, September l7th/19th '1986. Hills, B L and G J Elliott, 1986. A. Microcomputer Accident Analysis Package and its use in Developing 'Countries. In: Indian Road safety Seminar. Srinagar. 17 -19 September 1986. CONTENTS Pg 1. Introduction 2 2. The TRRL. Microcomputer Accident Analysis Package 3 2.1 Programs in the Accident Analysis Package 3 2.1.1 Inputting New Accident Records 3 2.1.2 Creating Master Accident Files 3 2.1.3 Adding 'New Accidents' to a Master File 5 2.1.4 Making Backup Copies of a Master File 5 2.1.5 Identifying accident records satisfying particular conditions 6 2.1.6 Cross-tabulations 6 2.1.7 Graphics 7 2.1.8 Hard Disc Menu of Files 9 2.2 Locating accidents 9 2.3 Future Software Developments 10 2.4 Hardware Requirements 11 3. National and Regional Accident Statistics 11 3.1 Introduction 11 3.2 The objectives 11 3.3 Techniques of analysis 12 4. Local Authority Accident Investigation 12 4.1 The objectives 12 4.2 The Accident Investigation Process at a Local level 13 5. Conclusions 16 REFERENCES 18 1 Page 1. INTRODUCTION Experience gained in the 1970's by the Overseas Unit of TRRL. had shown that hand analysis of accident records in developing countries was a very slow, inconsistent and error-prone process. Computer analysis was clearly required; mainframe computers however, have had a chequered history In developing countries and for many years there has been a widespread concern among International Aid Agencies-about the-appropriateness of mainframe computing to developing countries. The rapid development of microcomputers in recent yearstherefore, has offered an exciting opportunity to overcome some of these difficulties. Microcomputers potentially offer a number of advantages to developing countries: (i) They are cheap; (ii) Relative to mainframes, they are physically robust and tolerant of high temperatures, humidity and dust (although air conditioning is generally to be recommended); (iii) They are relatively easy to maintain, with even do-it-yourself maintenance possible at a module level; (iv) They are readily accessible, ideally being in or close to, the user's office; (v) They are not Intimidating. In addition to these physical advantages, the style of programming that developed with microcomputers, with its emphasis on interaction and 'user friendliness', would seem to be particularly helpful to developing country users. It was against this background that the Overseas Unit of TRRL began development of the Microcomputer Accident Analysis Package in about 1981. ~The package was originally developed as part of a co-operative programme of road safety research between the Egyptian Government and the Overseas Unit of the Transport and Road Research Laboratory, and was developed in parallel with an easy;-to-use Police Accident Report Booklet (Gaber & Yerrell, 1985). The package was first described in Hills and Kassabgi (1984) and since then, the microcomputer system has been further developed, partly under contract to the Transpottation Research Grnoup of Southampton University. The main enhancements have Included the introduction of 16-bit and 16/32- bit versions using either MSDOS or Concurrent CP/M operating systems, and the introduction of Hard Disc and bilingual options. In addition to Egypt, the system is being evaluated on a trial basis in Saudi Arabia (Arabic version), Ethiopia, Botswana, 2 Pakistan and Papua New Guinea and is currently under consideration for trial in a number of other developing countries. French and Spanish version are now available. The purpose of Section 2 of this paper is to give an up-to-date description of the facilities now available with the system. Sections 3 and 4 give a brief description of how these types of facilities may be used In analysing accidents at three distinct levels: National, Regional and Local. The package itself has been designed primarily for use at a Regional and Local level and for use by staff with little or no previous computing experience. 2. THE TRRL MICROCOMPUTER ACCIDENT ANALYSIS PACKAGE 2.1 Programs in the Accident Analysis Package Table 1 shows the various programs so far developed which fall into two main groups: (i) Those for establishing and updating files of accident records - the accident database; and (ii) Those for analysing the accidents. 2.1.1 Inputting New Accident Records New accident records are entered directly from the police accident report using the first program option (NEWACCS). This work is normally carried out by a junior member of the office. Before doing so, each police report is checked for missing data and various items given code numbers. An example of an accident record is shown in Figure 1. The user In prompted for each item in the accident report in turn, and after entering a value for an item, various checks for valid values and logical consistency are carried out. With experience, 10-20 accidents an hour can be entered. This creates a small file of 'New Accidentsl which will later be checked and added to a 'master file' by the supervisor of the system. 2.1.2 Creating Master Accident Files To enable rapid retrieval of an accident, the master files are of a type known as 'Random Access'. To start a new file of accidents, say, at the beginning of the year, it is necessary to create a large, but empty, file using program option 3 (NEWFILE). 3 4- TABLE 1. PROGRAM(S IN MICRO ACCIDENT ANALYSIS PACKAGE A. PROGRAMS FOR CREATING ACCIDENT FILES e 1. NEW ACCIDENTS (NEWACCS) 2. FILE HANDLING (FILEHAND) 3. NEW FILE Used by typists to enter New Accidents .Used to ADD new accidents to Master File, and to INSPECT, PRINT, CHANGE or DELETE accident records Creates new (but empty) 4.COPY FILE M~akciest bafui opeloease File B. PROGRAMS FOR ANALYSING ACCIDENTS 5. SELECT 6. STANDARD CROSS- TABULATIONS 7. NON-STANDARD CROSS-TABULATIONS 8. GRAPHICS (i) (ii) (iii) Lists code numbers accidents eg Km 306 Highlands Highway For either ACCIDENT CASUALTY numbers eg by TIME OF DAY of SELECTED on the numbers or DAY OF WEEK For ACCIDENT numbers, CASUALTY numbers or VEHICLE numbers Histograms of accidents on a major city or Inter-city route Accident map of a town or city Stick Diagram Analysis - a technique for identifying patterns in accidents 4 This may be thought of as the equivalent of bringing into the office a new, but empty, filing cabinet in which to store the new year's accident reports. The size of this file is determined by the storage capacity of the floppy discs used by the microcomputer. For example, in the case of the Apricot or ICL Quattro, the floppy disc capacity is such that about 2,500 accidents can be stored in one file; for the standard IBM-PC format, about 1,200 accidents can be stored on a floppy disc. If, say, about 5,500 accidents per year needed to be recorded, three files would be required for one year's data under the Apricot or ICL format. In this case, the hard disc version of the package would be recommended, which would enable these three files to be selected f or analysis together. The master versions of each file should have TWO floppy disc backup copies, a 'Father' and 'Grandfather' copy. These two discs are inserted in the floppy disc drive as part of the NEWFILE file creation process. This file structure is illustrated in Figure 2. 2.1.3 'Adding 'New Accidents' to a master file Section 2.1.1 described how a file of New Accidents is created. To add these accidents to a Master ('Son') file, the second program option (FILEHAND) is used. This enables the supervisor of the accident analysis system (or his deputy) to check and, if necessary, correct the work of the person who has entered the new accidents. Thus, each accident record is checked and then added to the Master file and deleted from the New Accident file. It should be appreciated that the file of New Accidents would normally contain of the order of 1-50 new accident records, whereas the Master file will accumulate up to 2.500 accident records. The program FILEHAND also allows the user to find rapidly any particular accident record in the Master file (defined by Year, Police Region and Accident Serial Number). The user can then Print, Change or Delete that record. 2.1.4 Making Backup Copies of a Master File The fourth~prograx --Option -(COPYFILE) 'is uei'id to' copy the Master (Son' file to the backup 'Father' or 'Grandfather' floppy discs. The need to be meticulous in making backup copies of the master file on a regular basis cannot be overstressed. 5 Many man-weeks of time can be wasted if the master file is corrupted and no backup copy is available. It is suggested that the first backup ('Father') copy should be updated daily and the second backup ('Grandfather') weekly. Dependent upon the version, the 'Son' file can either beaon hard or floppy disc. However, the backups are always on floppy disc. 2.1.5 Identifying accident records satisfying particular conditions The fifth program option, SELECT, is the first of the programs that allows the user to analyse the accident files. It enables the accident Investigator to identify those records that meet some particular criterion. Examples might include: (i) A particular length of a selected road e.g. Km& 100-105 on the Highlands Highway (ii) A particular area in a city region e.g. Map 2; X-coordinates 25-30 Y-coordinates 5-10 (iii) A particular time on a specified day on a selected road e.g. 1500-1530 hrs on 12/1/186 on Waigani Drive. The user can choose to have printouts of either the code number only of each record found or the full details of each record. The first option, 'code number only', might be required if it were intended to go back to the original police accident reports, perhaps to look at the collision sketches. A third option now available is that the records can be stored on disc as a Working File. Working Files can be most useful ---and save considerable time when investigating a specific accident problem, by drawing together relevant data from a number of files. 2.1.6 Cross-tabulations ,Cross-tabulations are one of the most important tools of accident analysis, particularly in establish-ing-,the-,general-.patterrrof ~accidentts. Thret types are available: ACCIDENT, CASUALTY and VEHICLE. The difference between these types are perhaps best understood by a specific example. Suppose cross-tabulations of POLICE REGION by MONTH were of interest. The three different cross-tabulation types would give: 6 1. ACCIDENT CROSS-TABULATION - Counts of the number of accident in each month and police region 2 * CASUALTY CROSS-TABULATION - Counts of the number of casualties in each month and police region 3. VEHICLE CROSS-TABULATION - Counts of the number of vehicles involved in accidents for each month and police region. An example-of an Accident Cross-tabulation is shown in Figure 3. There are a very large number of~ possible cross-tabulations. For example, if there are 85 items in an Accident Report Form, for Casualty and Vehicle cross-tabulations there are 85 x 84 possible cross-tabulations. In practice, only a small subset of these is normally of interest. Therefore, both STANDARD and NON-STANDARD cross-tabulation programs are available. (Standard Vehicle cross-tabulations have yet to be defined but Non-standard Vehicle tabulations are available). If a particular cross-tabulation required is one of the Standard ones, then it is quicker and easier to select it through the 'menu' presented in the Standard version of the program rather than to have to specify the two axes in the Non-standard version. If required, CONDITIONS can be applied to all of these crops-tabulations. Examples might be:- (i) Night-time only (ii) Fatal only (ii) Kms 256-262 on the Highlands Highway 2.1.7 Graphics Until recently, there was little standardisation of the graphics systems available on microcomputers and printers. To enable the graphics programs of the TRRL Microcomputer Accident Analysis Package to run on different micros and printers, these programs were therefore written to produce 'pseudo-graphics', using ordinary character printing, rather than the true graphics facilities of the micros. (With the. adoption of the internationally-agreed Graphics'"Kernil, 'this situation is now changing). The three programs developed are: (i) ACCROUTE - a histrogram of accidents along an urban or inter-urban route; (ii) ACCMAP - an accident map for a town or city; (iii) STICK - Stick diagram analysis of accidents, a technique for finding patterns In accidents, usually at black-spots. 7 ACCROUTE is designed to analyse accidents in two situations: (a) Where a major route through a city or town can be defined as a sequence of Nodes (major junctions - aee Section 4); (b) Where locations along an Inter-city highway are defined in the accident records by kilometerage. For the second type, a summary histogram in 20 or 50 Km intervals is first generated, with each Fatal accident represented with the symbol * and each non-fatal accident represented with the symbol x. The Investigator can them request a printout giving a histogram of accidents In 1 Km sections for the entire road. A section of such a printout is shown in Figure 4. Alternatively, the user can 'zoom in' on a particular section of road with any size of KCm Interval stipulated (e.g. 5 Km sections grouped together). ACOIAP is primarily intended for analysing accidents In a town or city. The user is first which map is to be analysed and displayed, thus producing a summar map. Normally, this summary map provides accident counts over unit areas of 400 metres X 200 metres. The user can then 'zoom in' on a particular region of the map. The unit area at this level is 100 metres x 100 metres. Examples of each type are shown in Figures 5 and 6. A problem arises when accidents In a unit area, which covers one printable character, becomes larger than a single digit ie. 9. 10 accidents are therefore represented with the letter A, 11 by B, 12 by C and so on. More than 35 accidents are represented by the symbol *. A facility has been developed in ACCMAP whereby the worst locations on any particular map can be listed (see Figure 7), the length of the list being chosen by the user. STICK Is primarily Intended for more detailed analysis of accidents at a high accident location (accident blackspot). This location can be defined by kilomeiterage, by map coordinates, or by specifying a node or link. Each accident ~is..represented by the vertical listing (or "satick"1) of key information - each Item occupying the same place in the stick. An example of a format (see Figure 5) Is (from top to bottom): DD - Day of Week HH - Hour of Day Sev - Severity of accident Ped - Pedestrian accident NoV - Number of vehicles Involved P-U - Pick-up involved Alc - Alcohol involvement suspected a I - Head-on accident - Rear-end accident Sid - Side collision Rol - Vehicle rolled-over 0b2 - Vehicle hit obstacle off the road p - Parked vehicled involved ** - Nighttime accident Wet - Road surface was wet Only one format is available to the user at present, although in the near future it is hoped that the user will be able to select which of a number of stick formats he wishes to use. The value of Stick Diagram Analysis lies In arranging the sticks In groups; with skill, the accident investigator can find if there Is any pattern to the accidents. By hand, this is a rather tedious process; it is very much simpler by computer. In Figure 8, all the sticks for nighttime accidents have been grouped together; then, within both daytime and nighttime accidents, the various collision types have been grouped together. Before the sticks were sorted in this way, the pattern of the accidents was much more difficult to comprehend. This technique has been found to be a valuable tool by accident investigators in the United Kingdom. 2.1.8 Hard Disc Menu of Files The storage capacity of a hard disc unit is normally at least 15 times that of a floppy disc unit in the microcomputer. Consequently, a number of accident files can be stored on the hard disc at the same time. For all the analysis programs in the hard disc version of the package, the user is first asked whether the next accident file to be analysed Is on the hard disc or on floppy disc. If on the hard disc, he is then offered a menu of the files stored on the hard disc, from which either a single file or group of files can be selected. The list of files is normally updated auto- matically as files ar created or deleted from the hard disc. However, the user also has the option to amend this list of files himself. 2.2 Locating accidents The accurate location of accidents is a most important aspect of the accident investigation process. In the TRRL system, a 'belt-and-braces' approach is normally adopted In which accident locations In towns are defined by both of two standard methods: 9 (a) a version of the normal grid coordinate system; and (b) a NODE-LINK-CELL system, which gives reference numbers to: (i) NODES - the main junctions, each node being given a unique code number; (ii) LINKS - road sections between these main junctions. These are referenced by the code numbers for the nodes at each end of the road section; and (iii) CELLS - areas of the town bounded by these Links, each cell being given a unique code number. These reference numbers are entered as part of the accident record in the TREL system. With this system, special coded maps would eventually be required for each major town in a country. These have both the grid coordinates and the NODE-LINK-CELL system superimposed. An extract from such a map commissioned by the TRRL for Islamabad is shown in Figure 9. For the major inter-city highways, accident locations are defined by kilometre distance f om the beginning of the road. This almost inevitably requires Kilometre Posts to have been installed on the highways concerned. These already exist in a number of developing countries. Kilometre Posts are invaluable for both highway maintenance and accident recording on major highways, in addition to their obvious benefits to the motorist in helping locate turn-offs, buildings etc. The office investigating accidents may find it useful to develop a series of strip maps for the major highways of the country. In addition to locating towns and villages, these strip maps will locate bridges and other prominent features or buildings by kilometerage. Wall maps are also useful in showing accident locations in major towns; rather than using pins, it Is recommended that accidents be located with small coloured spots of adhesive paper on an acetate overlay that can be stored and replaced at the end of a year. 2.3 Future Developments Programs now under development Include: (i) a program that will enable the user to select from a range of 'stick' formats and allow him to modify these; (ii) a program that will enable the supervisor of the system to modify the accident record format himself without TREL assistance. The use of true 'graphics' to represent accident pattenra and accident maps will be the next major stage of development. 10 ---- I 2.4 Hardware requirements Versions 2.0 and later of the Package will run on moat of the. current generation of 16- or 16/32-bit microcomputers. The earlier 8-bit version (1.0) is no longer fully supported. Versions 2 will run under Concurrent CP/M-86 or MS-DOS; however a small incompatibility between MS-DOS and PC-DOS necessitates the Package on the IBM-PC and compatibles to be run under Concurrent DOS. Either twin floppy or hard disk versions are available; a hard disk system would normally be recommended where more than 5000 accidents per year are recorded. A printer is required and an Uninterruptible Power Supply (UPS) is strongly recommended. Where good local agents are not available for maintenance support, it is recommended that two identical microcomputers and a spare disc drive unit be purchased. Allowance for good stocks of printer paper and floppy disks should also be made. The hardware costs are therefore in the range of £3000- £7000 (UK prices). dependent upon the configuration required. 3. ANALYSIS OF ACCIDENTS AT NATIONAL, REGIONAL AND LOCAL LEVELS 3.1 Introduction Road accident Investigation can be carried out at three levels: National, Regional and Local. The techniquesof analysis at a National and Regional level are, in practice, broadly similar, relying heavily upon the use of cross-tabulations. These will be discussed in this section. However, the strategies and techniques of analysis at a local level are very different, and these will be discussed separately in Section 4. The differences between the two types of analysis are perhaps best illustrated by the fact that at a National or Regional level, it is only practical to analyse. those details of accidents stored on computer; but at the Local level, the policeman's accident sketch and the witnesses' statements are frequently examined in trying to identify factors associated with the accidents. 3.2 The objectives The general objectives of analysis of accidents at a national or region level are: (i) To give a general overview of the accident situation; (ii) To identify groups within the population most at risk eg. children or motor- cycle riders; (iii) To identify physical circumstances with high levels of accidents eg. nighttime or junctions,, 11 (iv) To monitor trends with time of the various accident types, particularly with a view to identifying as soon as possible a developing accident problem eg. an increase in motor cycle accidents; (v) To monitor the effectiveness of any national or regional road safety measure eg. introduction of seat-belts; (vi) To make national and international comparisons. Having statistical analyses of accident records available enables the country or region to determine Its road safety strategy on a rational, scientific basis. They will provide the foundation from which the cost-effectiveness of Its programme of measures can be maximised. 3.3 Techniques of analysis The cross-tabulation, such as that shown in Fi gurec 3, Is the most common technique of accident analysis used at a National or Regional level. These can be visually illustrated wit h the use of graphs, histograms or pie-charts. Most developed countries publish sets of such tables annually. The interpretation of these tables Is not always straightforward due to the need in some way to take account of exposure. Thus, it might be found that most accidents occur on Friday afternoon; but then that may prove to be when most vehicles are on the road. The most dangerous time to be on the road in terms of vehicle-miles driven might prove to be late Saturday night. Both findings - the most frequent and the most dangerous - are of significance in planning a road safety programme. It remains debateable where the best value for money lies, in tackling high accident frequencies or high accident rates. In many circumstances, this debate becomes academic since accurate vehicle-mileage data are not available - but the problem of exposure still needs to be taken into consideration in deciding priorities. 4. LOCAL AUTHORITY ACCIDENT-~INVESTIGATION 4.1. The objectives Common experience suggests, and accident investigation confirtms, that road accidents are not randomly scattered over the road network, but tend to cluster at certain locations. As a consequence of this, local authorities in a number of countries have established so-called 'Accident Blackapot' teams to tackle these problem sites. The United Kingdom has probably been the leading country in this particular field of Road Safety activity, and the results over the past ten years have shown that the techniques 12 developed have been highly cost-effective. For example, the Greater London Council's Accident Blackspot Team after well over 1000 schemes has achieved an overall First Year Rate of Return of 350%, including all overheads (Ridley, 1979). During the development of the techniques, the objectives of the Accident Blackspot teams in the UK have broadened to the extent that there are now four distinct strategies adopted: (i) BLACKSPOT PLANS: Measures to tackle accidents at specific sites; (ii) ROUTE ACTION PLANS: Comprehensive treatment of routes or roads with high accident densities; (iii) AREA ACTION PLANS: Comprehensive treatment of small areas of a town or city with high accident densities; (iv) MASS ACTION PLANS: Mass application of established counter-measures eg. application of high skid resis ting surfaces. In the UK, there has been a heavy emphasis on using low-cost counter-measures, so that major road works are only considered as an option as a last resort. The counter-measures frequently involve low-cost traffic engineering measures but when appropriate, Publicity, Education and Training at a local level are all options considered, the RIDE BRIGHT campaign for motor-cyclists in London being a successful example (Lalani and Holden, 1978). 4.2 The Accident Investigation Process at a Local level This section is intended to be a brief outline of the various procedures involved. More detailed accounts are available, such as the Accident Investigation Manual issued by the UK's Department of the Environment (1974). The overall process is summarised in Figure 10. In the schematic diamagram, the process is divided Into activities at the accident site and those in the Accident Investigation office. After an accident has occurred, a police report is filled-in and after a certain elapsed time, data from this report will be added to the computer accident database. As discussed In Section 2.2, being able to identify the accident's location from the computer record Is a basic necessity of any accident Investigation system operating at the Local Authority level. Ideally, it should be possible to identify the accident location to within 100. or better. Whilst this should be achievable In a town or city, It Is more difficult on Inter-city roads where distances are defined 13 using kilometre posts. In this case, locating accidents to within 1 Km may be a more achievable goal; in some circumstances, however, 1 Km can be a long distance in which to pinpoint a problem In the road during a site study. The 'Accident Location Sketch' (Figure 11) introduced In revised police report forms being tested in a number of developing countries will hopefully minimise this problem. The Graphics option 'Accidents on a Route' (ACCROUTE) and 'Accident Map' (ACCMAP) are the two programs in the Microcomputer Package used for identifying locations with high accident rates. (See Section 2.1.7). As Figure 10 indicates, the worst of the accident sites are them selected for more detailed analysis in the accident investigation office: (i) Cross-tabulations can be produced (using program options 6 and 7 in the Microcomputer Package) with limitations defined to restrict the analyses to just those accidents occurring at an accident site under investigation. These can be useful at an accident site under investigation. These can be useful in establishing the overall nature of accidents at the site. (ii) A Stick Diagram Analysis, or similar Factor Analysis, can be carried out (using option 8 in the Microcomputer Package) to try and establish patterns in the accidents. (iii) The serial numbers of the accidents occurring at the site can be listed from the computer (using option 5 in the Package) and the original police reports then retrieved from the police files. It is then valuable to read the statements of those involved in the accidents and the statements of any witnesses. These sometimes reveal new factors that should be included in the earlier Stick Diagram or Factor Analysis eg. a high percentage of those Involved may have been strangers to the area. (iv) Another valuable source of information in the police report is the policeman's collision sketch. For a junction site, it is often useful to plot the location of each accident on a plan of the junction in an Accident Plot (see Figure 12); more useful still is to represent diagramatically the movements of all the vehicles In each accident in a Collision Diagram for the junction (see Figure 13). The next phase of the procedure is to make a site study, armed with the analyses made at the office. A photographic record of the site can be most useful for future reference back at the office. The first task is to record the key physical characteristic~of the site eg. road width, shoulder width, existe nce of street lighting etc. A checklist is available in the UK Accident Investigation Manual; such checklists 14 may not always be relevant to the types of site being investigated, in which case the development of a locally applicable checklist should be considered. It is frequently useful to make preliminary surveys of traffic and pedestrian flows. A short survey of traffic approach speeds on different arms of a junction can sometimes be revealing; one simple method to do this is to use one person with a stopwatch stationed 50m from the junction and a second stationed 150. from the junction flagging when a vehicle passes hi.. A study of Vehicle Conflicts eg. ranked on a scale of 1 to 5 an illustrated in Figure 14, can often reveal hidden problems at the site. Observations of drivers' head turning movements or observing that drivers have to stretch to see approaching vehicles can sometime. lead to identifying a visibility problem. Note any other problem. of visibility, eg. shallow hillcrests or vegetation blocking the view of approaching vehicles. Talking to local residents can also be helpful, although opinions about causes should always be treated with a good deal of caution. If there is a clustering of accidents at a particular time of day or a particular period of the week, then the site study should be made at these times. Again, the UK Accident Investigation Manual offers extensive checklists of possible factors to look for. Having identified the probable causes of the accidents, the possible counter-measures should then be. listed and ranked according to cost. Normally, the lowest cost solution would be adopted unless there is good reason to believe that it would be unsuccessful. In implementing the counter-measure, due consideration should be given to whether or not advanced warning to the public is necessary, either in pre-publicity or in advance of the site with signs such as 'Experimental Junction Layout Ahead' (white-on-red). The first days of operation of a scheme should be carefully watched in case an unexpected problem has been incorporated Into the design. After a suitable elapsed period of time, the long term benefits of the scheme should be evaluated with before-and-after studies. For details of this, manuals such as the UK's Accident Investigation Manual should be referred to. It is most important to use 'control data' In such an analysis, comparing the before-and-after accident levels at the test site, either with those at similar but unchanged sites in the locality or with some regional data. Comparing th.before-and-,after...accidents at the test site with those at the control sites eliminates 'global' factors that may occur In the after period, such an a particularly good or bad period of weather period, a change in oil price affecting vehicle speeds or a change in traffic law. If this evaluation indicates that the scheme has been unsuccessful then further analyses and site studies may be necessary before an alternative scheme is tried. The Chi-squared statistical test with Yates' correction is the most common statistical test used in evaluating Before- and-After studies. i5 5. CONCLUSIONS The most encouraging finding of this research to-date is that developing country staff with no previous experience of computers are entirely capable of running the Microcomputer Accident Analysis Package. This has already led to a far greater insight into the accident problems of the study areas In which the system has been tested. One of the central purposes of the system Is to evaluate the cost-effectiveness of counter-measures introduced as a result of these accident analyses and this is the main thrust of the current phase of the research programm. Progress in this has to some extent been hampered by the need for 'institution building': In getting adequate staff, funds and facilities allocated to the work and in forging close links between the Traffic Police, Highway Engineers and other departments concerned with road safety. This has been the most significant problem encountered. Some minor hardware maintenance problems have occurred, but with the rapid influx of micros into virtually all developing countries, local maintenance expertise is growing and these problems should eventually become insignificant. Data gathered using the Package will be discussed in detail in a future paper. The data so far collected have shown, as in developed countries, that accident locations are not randomly scattered but tend to cluster at 'hazardous locations', and that the nature of the accidents differs markedly from site to site. Rural accidents have been found to be considerably more severe than urban accidents, most probably due to the combination of higher speeds and problems of medical recovery. Analyses of the accidents at hazardous locations Identified with the Package combined with site studies have suggested a wide range of problems, including speeding, close following, poor lane discipline, poor planning (especially direct frontal access from commercial buildings onto major highways and excessive use of crossroads), poor or non-existent junction design, the need for climbing lanes, steep side slopes resulting in rollovers, narrow shoulders and recovery. zones,-.poor s.kid-~resistance-, poor -delineation of curves, poor pedestrian behaviour, lack of footpaths, the special problems of animal-drawn vehicles, lack of facilities for local traders selling to passing traffic, lack of lay-bys for stopping or resting, lack of bus stop bays, faulty and misused vehicle lighting, poor tyre condition, etc. Correcting some of these problems is clearly a long term process, but some low-cost counter-measures have been implemented and are being evaluated. 16 The high severity of rural accidents combined with the clustering effect indicate that a high priority should be given to establishing a system of kilometre posts on major rural highways where these do not exist. The need for theft-proof road signs and kilometre posts has become apparent during the studies. Local Authority Accident Investigation has been found to be highly cost-effective in the UK (see Section 4.1). partly as a result of the heavy emphasis put on low-cost counter-measures. The Overseas Unit of TREL considers that this approach to Accident Investigation has considerable potentia for Developing Countries and it believes that the Microcomputer Accident Analysis Package now makes the establishment of effective local Accident Investigation Units an achievable goal. 17 REFERENCES DEPARTMENT OF THE ENVIRONMENT (UK): Accident Investigation and Prevention Manual. 1974. GABER M.A & YERRELL S.: Analysing the Road Safety Problems in Egypt. 10th Congress of the International Association for Accident and Traffic Medicine, Tokyo, 1985. HILLS B.L. & KASSABGI M.: A microcomputer road accident analysis package for developing countries. Proceedings of Annual Summer-Meeting of PTRC, Brighton. 1984. LALANI N & HOLDEN E.J.: The Greater London 'Ride Bright' campaign - Its effects on motorcyclist conspicuity and casualties. Traffic Engineering & Control, August/ September. 1978. RIDLEY G.: Engineering techniques for accident prevention. Paper presented to seminar on 'Road Safety - Remedial Action and the Local Authorities', London, 31st October 1979. SPICER B.: TRRL Report Laboratory, A traffic conflict study at an intersection on the Andoversford by-pass. LR 520, Department of the Environment, Transport and Road Research 1972. ACKNOWLEDGEMENTS The work described in this paper forms part of the programme of the Transport and Road Research Laboratory and the paper Is published by permission of the Director. Crown Copyright. The views expressed in this paper are not necessarily those of the Department of Transport. Extracts from the text may be reproduced, except for commercial purposes, provided the source is acknowledged. 18 ****** ACCIDENT RECORD 5*5**5 **** ISLAMABADM1/85/00205 **55 YEAR 85 CASE NUMBER 00205 POLICE DISTRICT 01. CITY 01 POLICE STATION NO: 02 F.I.R. Number 0069 TOTAL NO OF VEHICLES 01 NO. VEHICLES DAMAGED 01 NO. DRIVERS/RIDERS KILLED 00 NO.DRIVERS/RIDERS INJD 00 NO.PASSENGERS KILLED 00 NO.PASSENGERS iNJURED 01 NO-PEDESTRIANS KILLED 01 NO.PEDESTRIANS INJURED 00 MONTH 02 DATE IN MONTH 23 DAY OF WEEK 4 HOUR 02 MINUTES 30 VEHICLE MAKE VEHICLE YEAR REGISTRATION TYPE VEHICLE TYPE VEHICLE MANOEUVRE VEHICLE DAMAGE NOSE-TO-TAILVEHICLE LOADING 06178 I0301 1I1 *S**S**a5****a*aS* 202122232425262728293031323334353637 AM/PMACCIDENT SEVERITY WEATHERLIGHT CONDITIONS ROAD GEOMETRY ONE- OR TWO-WAY MEDIANROAD SURFACE TYPE RIDING QUALITY ROAD WIDTH SHOULDER WIDTH SHOULDER TYPE SURFACE CONDITION JUNCTION TYPE JUNCTION CONTROL HIT & RUN COLLISION TYPE ROADWORKS VEHICLE 1 sass ******* 46474849505152 TYRE BURST VEHICLE LIGHTING LENGTH SKID MARKS DRIVER SEX DRIVER/RIDER AGE DRIVER/RIDER INJURY LICENCE NUMBER **s* PASSENGER 1 sass PASSENGERPASSENGERPASSENGER: VEHICLE COr SEX(M/F)AGE 01F * 56 PASSENGER: INJURY * 57 POSITION IN VEHICLE 05 **s* PEDESTRIAN 1 **** PEDESTRIANPEDESTRIANPEDESTRI AN SEXAGEINJURY m67 F 5*5 61 PEDESTRIAN MANOEUVRE 62 PEDESTRIAN LOCATION 63 SCHOOL 5*** ACCIDENT LOCATION as's 5a5*5*S5S555S*5* 64 URBAN/RURAL 65 TOWN/VILLAGE CODE 66 KILOMETRE POST 67 100 METRES 68 MAP SERIES 69 MAP CODE 70 X CO-ORDINATE 71 Y CO-ORDINATE 72 MAJOR ROAD CODE 73 MINOR ROAD CODE 74 ~NODE -1 75 NODE 2 76 SECTOR 77 SUB-SECTOR 78 DIRECTION OF TRAVEL 79 CAUSE OF ACCIDENT 1 00010000 00016 220065 00 -207000Fl2N23 .Fig.1 Example of an accident record printout from the Micro Accident Analysis Package. The record format varies considerably in the different countries in which the Package is being evaluated (Fictitious data) Transport and Road Research LaboratoryPA39. 1234567891 01 11 21 314is1617181 9 3839404142434445 2F1122y21.1021242151 *** 535455 I205m24U K3042 55a 585960 H1 2) 1I )E PA 1395.1 A: MAIN ACCIDENT FILE Police Accident Booklet 'NEW ACCI DENTS' File Type: 'Son' 'Father' 'Grandfather' Copy: Master 1 st Backup 2nd Backup. Disc Drive: A: 8: B: Updated: Daily Daily Weekly File system used in the Accident Analysis Package The main accident file can be given any name up to S characters long (beginning with an alphabetic character; the full stop character is not permitted) Disc drives are for the Apricot hard disc microcomputer Figume 2 Transport and Road Reseac LaboratoryPA192 PA 1395.2 STANDARD ACCIDENT TABLE 8 TYPE A: ALL ACCIDENTS ACCIDENT RECORD FILES; CONDITIONS.SET: NONE HOUR DAY OF WEEK *Sun Mon Tue*Wad Thu Fri Sat *Total* 00/01 43 52 54 59 52 63 66 w 389* 02/03 40 10 8 4 14 21 60 * 157* 04/O5 25 4 9 9 9 20 58 * 1348 06/07 30 45 61 39 54 56 89 * 374* 08/09 .45 70 76 90 70 83 118 * 552* 10/11 28 s0 48 38 44 58 49 * 315* 12/13 25 42 42 41 51 64 43 * 309* 14/15 56 ill 96 113 112 161 113 * 762* 16/17 73 99 120 107 120 169 128 * 816* 18/19 53 45 52 53 56 102 168 * 529* 2Q/21 33 33 24 34 43 94 99 * 360* 22/23 25 31 33 40 67 86 81 * 363* Total 477 592 623 627 692 977 1072 * 5060* (Total Number of Accidents on File 5422 Fig.3 Examiple of a Standard Accident Cross-tabulation Transport and Road Research LaboratoryPA39. PA 1395.3 ACCIDENTS ON A ROUTE ACCIDENT FILE: CONDITIONS SET: K ILOMETRES 280 281 282 283 284 285 286 287 258 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 LOCATION CODE 1= HHY ACCIDENTS 0 0 0 0 I X 0 0 0 0 2 xx 0 0 0 0 0 5 xxxXX 0 0 0 2 xx 0 1 X 0 0 0 0 14 xxxxxxxxxxxxXX 0 0 0 0 1 X 312 0 313 13 xx'xxxxxxxxxxx 314 0 315 0 316 4 xxxx 317 . 318 0 319 0 320 0 Total= 43 * Fatal x Injury or Damage Fig.4 Example of Section of a 1 Km interval histogram for a major inter-urban highway, generated by the program ACCROUTE Transport and Road Reseac LaboratoryPA19. PA 1395.4 ACCIDENT MAP ACCIDENT F*LE:7* * MAP SERIES 12 HAP CODE 013 CONDITIONS SET: NOME 80: Area of 3~~~ 1 1 12 I 1 * ~ 16 2 I:I Character ~ 31 1 1 32 1 43 Area of : 3 1 1 1 112D :31 200mX400 21 133 4A 12 40 -- 2311-1-1-13---12 -A---------------------420 1 2 1622 1 11 1 1332 1 1 *1 1 21 21 11131 2 1 1 3 212 3 1 1 0 W-3------I---1---1----------------- 0 Characters uned to Cod* Accident Numbers : 1=1 2=2 3:3 4:4 5=5 6=6 7=7 8=8 9=9 A=10 B:1.1 C=12 Dz13 E=14 F=15 G=16 H=17 1:18 3:19 K=20 L=21 M=22 N=23 0:24 P=25 Q:26 R:27 5=28 T=29 U=30 V=31 W:32 X:33 Yz34 Z:35 435=* F ig.5 Example of an Accident Map generated by the program ACCMAP (see text) Transport and Road Research Laboratory PA19. 1 PA 1395.5 ACCIDENT MAP ACCIDENT FILE: HAP SERIES MAP CODE CONDITIONS SET: EachCharacterCovers an Area of loom X loo0. 20 --. _.25. -, -.... 30 :2. z 1 10 1-- 2- 11 1: -- .35. - :1 212:1: * : 15 --- - -- -- -10 1 22. .1 5 ----------- 11 11 1 11 1 …~~~11 2 -------- 5 1 0--- -- - - -- - -- - - -- - -- - --0 Characters used to Code Accident Numbers : 1=1 2=2 3=3 4=4 Sz5 6=6 7=7 8=8 9:9 A=10 9:11 C=12 D=13 E=14 F=15 0:16 H:17 1:18 3:19 K=20 Lz21 M=22 Nz23 0=24 P:25 Q=26 R:27 S:28 1:29 U=30 V:31 W=32 Xz33 Y:34 Z=35 >35=* Fig.6 An enlarged secion of the map in Figure 5, generated by the program ACCMAP Transpor and Road Research LaboratoryPA15. 12013NONE I_ ----- 1 11 PA 1395.6 LISTING WORST ACCIDENT SITES ACCIDENT FILE: MAP SERIES 12 MAP CODE 013 CONDITIONS SET: NONE 1 .... X= 31 Y= 49 12 Accidents 2 ..... X= 34 Y= 13 8 Accidents 3 ..... X= 35 Y= 41 8 Accidents 4 ..... X= 20 Y= 44 8 Accidents 5 ..... X= 14 Y= 55 6 Accidents 6 ..... X= 23 Y= 38 5 Accidents 7 ..... X= 23 Y= 45 5 Accidents 8 ..... X= 9 Y= 40 4 Accidents 9 ..... X= 18 Y= 26 3 Accidents 10.....X= 27 Y= 33 3 Accidents 11.....X= 25 Y= 34 3 Accidents 12.....X= 19 Y= 44 3 Accidents 13.....X= 10 Y= 53 3 Accidents 14.....X= 2 Y= 58 3 Accidents 15.....X= 45 Y= 59 3 Accidents 16.....X= 4 Y= 2 2 Accidents 17.....X= 26 Y= 4 2 Accidents 18.....X= 34 Y= 7 2 Accidents 19.....X= 22 Y= 10 2 Accidents 20.....X= 33 Y= 13 2 Accidents 21.....X= 33 Y= 14 2 Accidents 22.....X= 20 Y= 15 2 Accidents 23.....X= 20 Y= 16 2 Accidents 24.....X= 10 Y= 18 2 Accidents 25.....X= 6 Y= 22 2 Accidents Fig.7 A listing of the worst 25 accidents sites for the map in Figure 5 for lO0m x lO00m areas Transpor and Road Research LabcratoryPA39. PA 1395.7 STICK DIAGRAM ANALYSIS ACCIDENT FILE: CONDITIONS SET: NODE 1 =237 NODE 2 0 STICK:1 : : 11: 5 ' 10: 7 3 :4 :2 6 :, 9 13: 12: 0 0 : 0 0 : 0 MM :01 :02 :02 :02 :02 :02 :02 :02 :01 :02 :02 o :03 :03 DD : 1:'2 :56:1 :3 :6 :2:'.5: 4 :3 :2:'6 :2: HH :05 :06 :06 :07 :10 :11 :13 :13 :14 !15 !15 i16 :17 SEV D N MNDD DDH D HD D D D 0/T : :O/T: DIR :4 :4 :4 :3 :4 :3 :.1 :,3 :1 :4 :,2 :.4 :,1: : MM =Month DD =Day HR =Hour SEV=Severity PED=Pedestrian <- =Left turn 1-= Riaht turn U =U-turn O/T=Overtake = Nose-to-Tail = Junction *N*=Niaht. DIR=Direction of travel ACCIDENT SEVERITY: F =Fatal H =Hospital N =Not-H D =Damage Stick Number = Accident Code Number:- 1 :00023 2 =00076 3 :00116 4 =00167 5 :00196 6 =00224 7 =00259 8 :00303 9 =00311 10=00323 11=00427 12=00507 13=00542 Fig.8 Example of a Stick Diagram Analysis generae by the program STICK. Each 'stick' contains some key details of information about an accident. The accidents have been sorted by HH (Hour of day). It can be seen that the left-turn accidents all occur between 13.00 and 16.00 hours at this junction Transport and Road Research LaboratoryPA19. PA 1395.8 Fig.9 Extract from coded map commissioned for Islarnabad by TRRL Overseas Unit Both the grid coordinate and Node-Link.CeII systems are used iransport and Road Research Laboratory PA 1395.9 1 .Identify accident locations to within 1O0m 2. Identify locations with high accident rates 3. Select the worst for detailed analysis 4. Make site study 5. Identify probable causes 6. Identify possible counter-mneasures 7. Implement lowest-cost counter-measure 8. EVALUATE with before-and-after studies Fig.10 The process of accident blackspot investigation Transport and Road Research Laboratory P 351 1 PA 1395.10 Fig. 11 Examples of 'Accident Location Sketches' used for training purposes Tranh~f ard Road Res~ac Laboratory PA 1395.11 (3) 0 Fig.12 Example of an Accident Plot (from DoE, 1974) Fig.13 Example of a Collision Diagram (from DoE, 1974) Transport and Road Research LaboratoryPA19.23 (4) 0 0 PA 1395.12M3 1 I a - Diagram of occurence Conf lict severity ad ~~~~~~~~~~Precautionary conflict jr stow 1 i*e. braking for vehicle waiting to rA436 emerge, precautionary tone change, or anticipatory braking)I Controlled braking or lane change to 2 avoid collision but with ample time for the manoevre Rapid deceleration, lane change or 3 stopping to avoid collision, resulting 3 in a near miss situation. No time for steady controlled manoeuvre Emergency braking or violent swerve 4 to avoid collision resulting in very near miss situation, or occurence of a minor collision WJote on diagram direction, colour. Ee~nyato.floe oehicle type end evasive action taken by Emrcl action .oloe Example of conflict report dhet for one seton of junction Fig.14 Conflict study rating scale adopted by the TIRRL (from Spicer, 1972) Transport and Road Rese~rc Laboratory P 351 1 1 -7 DATE:- CONFLICT REPORT SHEET PA 1395.14