The data jungle – understanding casualty statistics

נשלח 17 במאי 2013, 11:54 על ידי Sustainability Org   [ עודכן 17 במאי 2013, 11:55 ]
Malcolm J. Wardlaw, 8/5/2013


Researchers involved with cycling say it is not a dangerous activity. But many people are still afraid to take their bike. One of the main reasons are misleading statistics. So how are accidents on bicycles being recorded? And how do they compare to pedestrians, for instance?

 “In the Netherlands, 45% of those seriously injured on the roads are injured while cycling with no motor vehicle involved….” (from European Transport Safety Council 5th Road safety PING Report 2011).

“Pedal cyclist casualties account for 17 per cent of HES [i.e. hospital records] road traffic accident casualties in England, but only 8 per cent in the STATS19 [i.e. police reports] dataset.” (from Road Casualties Great Britain 2006, the annual road casualty report of the UK)

 The casual reader would draw two conclusions from these quotes:

  1. cycling is obviously pretty hazardous, as it produces a disproportionate number of hospital admissions;
  2. cyclist casualties are under-reported by the police.

Both of these conclusions are in fact wrong, but in order to understand why, one has to be more than a casual reader!

The recording of casualties

In all EU countries, there are two sources of information on the number of people killed or seriously injured on the roads.

The oldest is hospital records. Fatal or serious casualties are recorded in a well-established system called the International Classification of Disease (ICD), currently in its tenth edition (hence ICD-10), this is common to most EU countries.

The other source is police reports which are collated into national databases completely separately from the ICD-10 system.

Comparing ICD-10 data with police reports

ICD-10 and police reports have different objectives. ICD-10 is a record of all deaths and hospital admissions, whatever the cause. In contrast, police reports deal only with injuries related to motor vehicles on public roads.

This leads to problems; firstly the term “traffic accident”. In police reports, traffic accident relates specifically to motor vehicles or those hit by them. In ICD-10 parlance, traffic accident means any incident on a public road. The incident could be a fall with no motor vehicle involved. The distinct definitions are generally not appreciated. This leads to comparison between incompatible data and as a result, the incorrect conclusion that the police under-report cyclist casualties.

Secondly, in ICD-10 all cyclist casualties are classed as “transport accidents”. This means a mountain-biker injured in the woods gets recorded under “transport accident”. In contrast, pedestrians only count as “traffic accidents” if injured by a motor vehicle. This means that cyclist “transport accidents” appear large relative to pedestrians, until the cyclist falls are removed (this is frequently not done, as already noted).

The ICD-10 has a coding system that enables the recording of the location and cause of injury. For instance the code “W10.4” means “Pedestrian fall on steps in the public highway”. So despite the above quirks, in theory it should be possible to extract from ICD-10 those casualties specifically relevant to compare with police reports.

In practice, this does not happen:

  • Pedestrian falls in the highway are never included in surveys of transport casualties. In the UK, there are more admissions due to pedestrian falls (30,000/a) than all road traffic collisions (22,000/a). Therefore the greatest cause of transport injuries has no official recognition.
  • There is a widespread myth that cyclist casualties are under-reported by the police, as already noted above. When the falls are removed, the two datasets agree closely (at least in UK). For instance, these are 2006 cyclist admissions data:

ICD-10                        Police reports

2,186                           1,899

Getting it right…

In the Netherlands, one third of serious injuries on the roads are cyclists, and one third pedestrians. However, 18% of time spent travelling in NL is by bike, as against 12% walking. Therefore in NL, the average risks of walking are higher than for cycling.

Pedal cyclist casualties account for 14% of serious injuries on GB roads, and pedestrians for 64%, when falls are included. However, in police reports, only 7% of serious casualties are cyclists, and 30% are pedestrians. It is clear that both cyclist and pedestrian casualties are under-reported by the police. This is because falls are not reported by the police at all for pedestrians, and only rarely for cyclists.

By including all relevant data, the quotes at the beginning of this note may be re-written[1],[2]:

ECF Road Safety Policy officer Ceri Woolsgrove says “What this sort of research does is open the debate on cycling risk and questions the idea that cycling is an overly risky activity. More specifically if we include falls and solo type accidents then the situation is that pedestrian injuries are hugely under reported, far more so than cyclist injuries. If we do not include falls and solo type accidents then cycling does not seem to be underreported at all and chimes perfectly well with the police transport data. What may also be interesting to see is how this relates to transport/traffic and solo accidents; how many cycling solo falls do we see on the road and how many are as a result of non-traffic activity?”

[1] NL reports 18,000 serious road casualties annually. Of these, 8,000 are cyclists (including falls) while 1,000 are pedestrians (not including falls). An figure of 7,000 is added for pedestrian falls (estimated pro rata from UK data).

[2] GB reports 22,000 serious road casualties annually. To this add 6,000 cyclist falls and 30,000 pedestrian falls, both figures taken from ICD-10 data.