LGRAS (00) 10 - THE POSSIBLE AVAILABILITY, AND THE LIKELY RELIABILITY, OF SOME INFORMATION ABOUT ROAD ACCIDENTS
1. This paper asks Members of the Group to comment upon the possible availability of information about mobile phone usage and road accidents, and about work-related road accidents. It also asks about the likely reliability of the "STATS 19" information on the Road Class and Road Number of non-trunk roads.
Information about mobile phone usage
2. Earlier in the year, a Member of the Scottish Parliament asked for information about mobile phone-related accidents in recent years, whether the numbers of such accidents would be monitored, and whether the Scottish Executive would make representations to the UK Government requesting the introduction of legislation preventing the use of mobile phones whilst driving. There have been calls for a ban on the use of mobile phones while driving, due to concerns that such use increases the chances of an accident.
3. At present, the "STATS 19" does not collect information about mobile phone-related accidents. Nor can the proposed national system of Contributory Factors identify specifically accidents thought to have been caused by mobile phones: the code for "Distraction - physical in/on vehicle" would include distraction by (e.g.) children fighting in the back of a car, or a dog suddenly starting to bark, as well as a mobile phone starting to ring; similarly, the code for "Distraction - stress / emotional state of mind" could cover a range of situations as well as the use of a mobile phone.
4. The first question, therefore, is whether it is feasible to collect information about what might be considered to be mobile phone-related accidents in Scotland. One would have to define what kinds of accidents would be counted. For example, the use, at the time of the accident, by the driver of one of the vehicles that was involved does not necessarily mean that the accident was related to the use of a mobile phone (for example, the person using the mobile phone could have been in a parked car that was hit by another vehicle). There is also the question of whether the fact that a person was using a mobile phone would be made known to the Police: presumably, some accidents which involved the use of mobile phone would not be identifiable as such. What might be the likely scale of such under-counting? Would it reduce greatly the value of any information that was collected? Incomplete figures could still be of use, for example to indicate the likely lower limit to the number of such accidents, or perhaps trends in their numbers (if it can be assumed that the "reporting rate" remains constant).
5. If it were feasible to collect the information, the second question would be how might it best be done? For example, one possibility would be to for Police Forces to use the "Special Project" field on the Attendant Circumstances record to identify any accidents that were known to be "mobile phone-related", and/or to use the "Special Projects" field on the Vehicle record to identify any drivers who were known to have been using a mobile phone at the time of the accident.
6. Could Police Forces provide a reasonably reliable indication of the overall broad level of such accidents by looking through the "accident stories" for a sample of accidents, in order to identify any references to mobile phones? Might Police Forces which record Contributory Factors (whether using the proposed national system, or a local system) be able to speed up the search by using those fields to identify accidents with codes like the "Distraction ..." ones mentioned above? However, would police officers generally mention any use of a mobile phone in the accident story if there is no requirement to record it for "STATS 19" or other purposes?
7. Finally, if it is feasible to collect such information, one would have to consider the likely value of the information relative to the likely cost of collecting it.
8. Members of the Group are invited to comment upon the desirability and feasibility (or otherwise) of collecting information about mobile phone-related accidents in Scotland.
Information about work-related road accidents
9. The Health and Safety Commission has proposed that a group be set up to look at work-related road traffic accidents, and ways of preventing these. Two of the main tasks of the group would be to establish accurate accident statistics for work-related road journeys, and to establish the main causes of such accidents. The question therefore arises: could "STATS 19" be used for these purposes?
10.The "STATS 19" data do not, at present, provide any way of identifying work-related road accidents. One would have to define what kinds of accident should be counted. From a limited understanding of the Health and Safety interest, a suitable definition might be along these lines: "accidents in which one or more of the drivers or casualties is someone who making a road journey in the course of his or her work". Therefore, "commuting" journeys between home and work would not be counted: the person would have to be travelling as part of his/her work.
11. The "STATS 19" database maintained by the Department of the Environment, Transport and the Regions is augmented with Driver and Vehicle Licensing Agency data for the vehicles which were involved in accidents. These details include a field which identifies company cars. However, that information does not identify "work-related" accidents: at the time, a company car may have been being used for private purposes, and there will be accidents involving private cars being used for work-related journeys. So, it is not possible to estimate the numbers of "work-related" accidents from the current data.
12. One possibility, if an estimate were required quickly, might be for the Scottish Executive to pick a sample of accidents from its road accident statistics database, and ask Police Forces to look at the accident stories for the chosen accidents to see if any of the drivers or casualties involved in the accident had been recorded as travelling in the course of their work. The sample could be focussed on the most likely candidates for "work-related" accidents, by using in the selection process information about (e.g.) the time of day at which the accident occurred, the types of vehicles involved and the characteristics of the people involved. This would reduce the amount of time that might be spent looking at accidents which turn out to involve only (say) pensioners, and so should enable a smaller sample to be used than would be the case with a simple random sample of all accidents. The results for the sample could then be grossed-up to estimate the national totals. However, the question arises: how likely is it that the Police would have recorded in the accident story that a driver or a casualty was travelling in the course of his/her work? Perhaps it is recorded in some cases where it is particularly relevant to the circumstances of the accident, and it may be obvious in some other cases (e.g. if a bus or a lorry is involved), but could there be many cases in which someone travelling in the course of work is not identified as such in the accident story?
13. Another possibility would be to use the "Special Projects" boxes to identify such accidents over a specified period (say three months). The question then is: how easy (or otherwise) would it be for the Police to establish whether drivers and casualties were travelling in the course of their work. In many cases, the only way to find out whether someone was travelling in the course of his/her work would be to ask him/her - and this may not be possible (e.g. in the case of an accident reported to the police by a member of the public who does not know all the other people involved in the accident).
14. Again, if it is feasible to collect such information, one would have to consider the likely value of the information relative to the likely cost of collecting it.
15. Members of the Group are invited to comment upon the desirability and feasibility (or otherwise) of collecting information about work-related accidents in Scotland.
Information about the Road Class and Road Number of the location of accidents
16. The "STATS 19" information about the location of each accident includes the Road Class [i.e. Motorway, A(M), A, B, C or unclassified], the Road Number (eg 720) and the Ordnance Survey Grid Reference (Easting and Northing: in theory, these should be recorded to the nearest 10 metres, but it is recognised that in some areas it is possible only to specify the location to the nearest 100 metres). In the case of accidents which occur at a junction, the Road Class and Road Number of the "second" road are recorded.
17. The contractors responsible for the Scottish Executive's trunk road database of information about trunk roads use these fields to identify accidents which appear to have occurred on a trunk road. Usually, the Road Class and Number are consistent with the Easting and Northing, and it is clear whether or not the accident on a trunk road. However, there are cases where (e.g.) the Road Class and Number suggest that it is a trunk road (eg the M8) but the Easting and Northing are for a location which is far away from that road. And, there are cases where the Easting and Northing suggest that an accident occurred on a trunk road, but the Road Class and Number do not. In such cases, the contractors may change the data if they think that it is obvious what is wrong (e.g. if the Road Class and Number say "M8" and interchanging the Easting and Northing values gives a location which is on the M8, the contractors may conclude that the co-ordinates were put in the wrong way round by someone who was unfamiliar with the grid referencing system). All such changes are listed in a printout, which the Transport Statistics branch passes on to the Police Force concerned, so that the latter can check the correction and amend its copy of the data. In other cases, it is unclear how an inconsistency in the data should be resolved, so the contractors include the accident in a "queries" printout, which the Transport Statistics branch sends to the Police Force concerned, so that the latter can specify which values should be changed.
18. The Scottish Executive does not check the consistency of the Road Class and Number and the Eastings and Northings for accidents which occur on local (i.e. non-trunk) roads, because it is not responsible for these roads. However, it understands that some local authorities find similar errors in the data for accidents on local roads, and that these authorities may correct the details held in the databases that they use themselves, and may inform the Police Force of the corrections.
19. Earlier in the year, a Scottish Parliamentary Question asked for statistics of the numbers of accidents and casualties for the past five years for a "local" (non-trunk) road that passes through the areas of six different local authorities. The Scottish Executive answered on the basis of the information held in its central database on the accidents which were selected on the basis of the (first) Road Class and Road Number identifying the specified road. The answer was accompanied by a warning that the figures in the answer might differ from those which the relevant local authorities might provide, because they did not take account of any changes or corrections that local authorities might have made to the data for use at local level.
20. The question therefore arises: how reliable are the Road Class and Road Number details held in the central database? Informal discussions with some people involved in the supply of the data have produced differing views, which may reflect differences across Scotland in the arrangements for recording Road Class and Number, allocating Eastings and Northings, and checking the consistency of the information.
21. A "lower bound" for the level of such errors can be obtained from the Scottish Executive's road accident statistics database. This holds two versions of these fields: the values that were on the most recently-received records for each accident set, and the "final" values, which take account of any subsequent corrections made by the contractors. The estimate will be on the low side because it is not possible to identify all the accidents for which there has been a correction (as the "most recently-received" and "final" values are the same in cases where a Police Force resubmitted an accident set incorporating corrections). Looking at the data for 1998 (because there are some as yet unresolved queries on the 1999 data), at least 5% or so of trunk road accidents had corrections made to one or more of the following fields: first Road Class, first Road Number, Easting and Northing (for the last two, only a difference of 200 metres or more is counted, in order to avoid counting slight repositionings). The Road Class had been changed in at least 1.2% of cases, and the Road Number had been changed in at least 0.7% of cases.
22. All trunk roads have a road class and number. For those non-trunk roads which are classified roads, it seems likely that "Road Class" and "Road Number" will have an error rate that is higher than the error rate for trunk roads, because the latter are more likely to be known by their road class and number. There will be many accidents occurring on classified non-trunk roads within cities and towns, which are roads that may well be much better known by their local names than by their road class and number. It seems more likely that the Road Class and/or Road Number details could be wrong for such accidents than in the case of accidents occurring on trunk roads (particularly on long stretches of trunk road in rural areas, which may not have any local name). Using examples from within Edinburgh, it may not be widely known that (for example) Ferry Road is the B9085 west of Crewe Toll but is the A902 east of Crewe Toll; Broughton Street is the B901; Great Junction Street is the A901; Portobello Road is the A1140; a road running through some southern suburbs, which has different names at different points along its route (e.g. Wester Hailes Road, Gillespie Road, Redford Road, Oxgangs Road, Frogston Road, Captain's Road, ...., Ferniehill Road) is the B701; Gorgie Road is part of the A71 but Dalry Road is part of the A70; and so on. Further problems could arise in areas where roads have been reclassified and/or renumbered, as people may well remember the old class and number. Therefore, given the complexities, it would not be surprising if more than a few percent of accidents occurring on classified non-trunk roads were given the wrong Road Class and/or Road Number. For example, there may be times when an accident which occurred on, say, Gorgie Road is described in the statistical returns as having occurred on an "unclassified" road because the person providing the information did not realise that Gorgie Road was part of the A71.
23. Therefore, it is uncertain whether the Scottish Executive's statistical database can provide precise figures for a particular local road. The local authorities who are responsible for the road, and are familiar with it and the area through which it runs, may use their local knowledge to change the statistical information about the location of the accidents on or around its route. After studying the data, perhaps using Geographic Information Systems, they may allocate special codes to identify the routes on which accidents occurred. In addition, local authorities can use their knowledge of the sites concerned to decide whether or not to count against a particular road any accidents which occurred at junctions between that road and other roads. (The numbers given in answer to the parliamentary Question were produced simply by extracting from the statistical database those accidents for which the "First Road Class" and the "First Road Number" identified the required road.) So, it seems likely that local authorities' figures for the numbers of accidents on the road would be more reliable than those produced from the Scottish Executive central statistical database.
24. Members of the Group are asked for their views on these matters. For example:
- what appears to be the likely level of error in the "Road Class" and "Road Number" details for classified non-trunk roads?
- in areas where the error rate is low, what are the main reasons for this? Might there be recommendations, based on these areas' experience, that could help to reduce the number of errors in other areas?
- if the error rate is thought to be high in some areas, are there ways in which their procedures could be changed, without great difficulty, to reduce the number of errors?
- do local authorities check the locational details, and (if appropriate) correct them in their copies of the data? If so, do they then inform the Police of the changes that they have made?
- if asked for the number of accidents on a particular classified road, would they count only accidents which had the specified road class and number in the first road number and class fields, or would they count accidents which had the specified road class and number in both the first and the second road number and class fields?