Serious injuries in road accidents
What are we measuring, exactly?


Government road safety targets are set in terms of persons killed or seriously injured (KSI). Since fatalities run at under 10% of the number of serious injuries, road safety performance is effectively being measured in terms of serious injuries. Several anomalies have led us to look rather closely at the serious injuries figures, and we present some interesting findings. This investigation started when we noticed something rather odd in the lethality ratios for serious accidents in September 2003.

They key question has become: "Why doesn't the serious injury trend match up with the trends in slight injuries and fatalities?"


In the speed camera era we have seen recorded road casualty figures change as follows:

GB figures 1994 2002 change
slight casualties 265,178 263,198 -0.75%
serious injuries 46,531 35,976 -22.7%
fatalities 3,650 3,431 -6.0%
traffic (bvkm) 421.45 485.95 +15.3%

Clearly the best news is in the serious injury figures. The improvement in fatalities is extremely disappointing in relation to earlier decades, and we literally haven't seen the fatalities figure fall for the last five years.

So what does this reduction in serious injuries represent? A success story for speed cameras? Or something else? We want to look at the serious injury statistics in some detail.

Casualty trends compared

With this graph we have scaled official casualty figures for five categories so that 1994 = 100%. This allows us to compare trends in the five groups of figures on the same graph despite the large differences in the actual numbers.

The green trace shows that there has been little change in the number of slight injury casualties over the period.

The blue trace shows that there has been very little difference in the total number of road injuries reported over the period. 

The red trace shows a useful fall in deaths, but the fall was somehow concentrated in a single year (1997-1998).

The yellow trace shows a sustained improvement in serious injuries and the light blue KSI trace closely follows the trend. The KSI trace follows the serious trace closely because the number of killed is less than 10% of the number seriously injured.

This graph raises a couple of key questions:

1) Was the sudden and isolated drop in fatalities in 1998 a real road safety gain, or was it caused by a change in definition or reporting practice? It looks very out of place to us in an extremely consistent series. 

2) How could it be possible that casualties and fatalities would remain more or less the same, yet serious injuries fall by almost 23%? One tries to imagine alterations in road accident severity or frequency that might yield such trends, but all things being equal, we would have to see low energy accidents and high energy accidents remain about the same but medium energy accidents reduce in frequency or severity. It seems unnatural and unlikely that the middle group would change in such a way while the groups either side are more or less unaffected. Items like vehicle safety improvements or speed reductions should have their greatest effects on fatal accidents.


This graph (figure 2) shows how accident severity ratios based on official figures have changed over the recent period. 

The green trace shows that in 1994 14.8% of injury casualties were serious casualties, but this proportion steadily reduced to 11.9% in 2002.

The red trace shows that the proportion of injury casualties varied around 1.1% from 1994 to 2002. (note that the "killed / all" ratio is multiplied by 10 so that it can be shown on the same scale.) No apparent trend is visible over the period in question.

The yellow trace shows that in 1994 7.8% of serious casualties were fatal. This proportion rose steadily and in 2002 9.5% of serious injuries were fatal. That's an increase of 22%.

We find it quite bizarre that the severity ratios involving serious injuries have changed in this way. In fact it is so bizarre that we doubt that it represents reality and went looking for alternative explanations.


In the UK, but only for the roads, a serious injury is defined as follows:

"Serious injury: An injury for which a person is detained in hospital as an “in-patient”, or any of the following injuries whether or not they are detained in hospital: fractures, concussion, internal injuries, crushings, burns (excluding friction burns), severe cuts and lacerations, severe general shock requiring medical treatment and injuries causing death 30 or more days after the accident. An injured casualty is recorded as seriously or slightly injured by the police on the basis of information available within a short time of the accident. This generally will not reflect the results of a medical examination, but may be influenced according to whether the casualty is hospitalised or not. Hospitalisation procedures will vary regionally." (from DfT publication: RCGB notes (click here))

An interesting definitions document from IRTAD contains the following snippet:

"Several OECD countries are not able to provide hospitalised data for the time being. For example, in the UK data on those hospitalised are included within the seriously injured category and are not available separately although research indicates that about one-half of serious injuries are hospitalised."

We're concerned about many things that could influence the recording of a serious casualty. For example:

Reporting practice. Different instructions may be given to those that fill in the forms or those who enter forms data into databases. Different auditing methods with different degrees of checking may also result in different figures being recorded.

Improved Paramedic care. We expect that improvements in paramedic care reduce the need for hospital admission by three mechanisms. 

  • Better medical information would lead to fewer precautionary hospital admissions.
  • Better casualty stabilization at the scene would lead to fewer consequential and "in transit" injuries (especially neck and back injuries). 
  • Effective early treatment will help patients to need less medical attention on arrival at hospital, and they may not be admitted as a consequence.
Changes in hospital admissions policy. Any alteration in a hospital admissions policy will alter the numbers who are admitted in the sensitive (and probably large) middle region.

Environmental factors such as flu epidemics and bad weather increase pressure on hospital bed availability, and might discourage marginal and precautionary admissions.

Pressure to return results and meet targets. We fear that there might be cases where pressure from government and partners may encourage the regrading of casualties from serious to slight in order to maintain performance indicators and to meet targets.

Sorting the figures

In order to understand the situation we have looked at what it means if the casualty figures are sorted by severity. In the two graphs that follow, the most serious casualties are placed to the left and the least serious towards the right. The X scale refers to casualty number. Casualty number 300,000 was at the very mild end of casualty severity, while casualty number 1 was the most seriously injured of all and probably died instantly in an extremely serious collision. 

It is interesting to note that collision severity must rise very rapidly at the low (number) end to account for the deaths and serious injuries, while 100 times as many slight injuries are caused by far less severe collisions. See this Safe Speed page for a different (and experimental) view of the same analysis. (ten

Notice that we have provided two additional categories within serious injury. The "hospitalised" category represents 50% of serious injuries according the the IRTAD document quoted above. The "long term injuries" category is supported only by common sense and anecdotal evidence. We made it 10% of serious injuries, and we are confident that the estimate is in the correct range. It would be most interesting to determine an accurate "long term injuries" figure.

We think it is obvious that tens of thousands of casualties must exist on the borders between "serious" and "slight" and it is easy to see how such cases may end up classified in either category. 

Is this the answer?

We experimented with reclassifying slight injuries as serious according to the following table:

adjustment 1994 1995 1996 1997 1998 1999 2000 2001 2002
percentage of serious reclassified 0% 3% 6% 9% 12% 15% 18% 21% 24%
number reclassified 0 1,366 2,670 3,867 4,900 5,868 6,868 7,790 8,634
percentage of slight reclassified 0% 0.5% 1.0% 1.4% 1.7% 2.1% 2.5% 2.9% 3.3%

The percentage (3% per annum of serious accident figures) was chosen experimentally to bring the trends "back into alignment". At the very least, this tells us with reasonable precision the size of the effect we need to find to explain the trends. It is perhaps illuminating that the percentage of slight injuries requiring reclassification is really very small.

Then we recreated the first two graphs above, but this time from the adjusted data. (compare figures 5 and 6 below with figures 1 and 2 above) The anomalies in the accident data have disappeared. Unfortunately so has the last decade's claimed road safety benefit. "All" and "fatality" figures were unaffected by this process.

In these two adjusted graphs everything makes sense.


In this adjusted graph there are no strange changes in the severity ratios. This is very much what we would have expected to find instead of the bizarre trends in figure 2.

new A little bit of anecdotal evidence:

This from an impeccable source:

"We had an officer from ***** police at our local road safety meeting this evening. 

Apart from stating quite clearly that she was "not the biggest fan of cameras" she also made the following comments in response to some questioning from the audience (!)

i) The ***** SI figures are up in 2003 compared to 2002. 

ii) The national SI figures are up for the same period. 

iii) "People" (which she later qualified as "statisticians") are pressurizing the police officers attending scenes of accidents to classify injuries as slight rather than serious. She gave two examples:

Broken bones of a minor nature (the example she used was fingers) were suggested as being "slight". Concussion (a "serious" injury according to the official definition) can no longer be logged unless a doctor has diagnosed the patient as having concussion - an unlikely event at the scene, IMO, which is where the Stats 19 form is completed, as far as I know. 

She stated that "she" (couldn't be sure whether she meant herself personally or ***** generally) was scrupulous in putting down the correct definition for each injury severity.

I will try to obtain more details in due course."

We have removed the county name to protect the innocent.

new Does the effect apply across road user groups and road type groups?

Yes it does. See the 9 graphs on this page (serious2).


On the basis of our investigation so far it looks highly likely that the apparent reduction in serious injuries over the last ten years or so does not represent any change in the number of serious injuries, but instead represents improvements in paramedic care and alterations in reporting practice.

Even if the last 8 years change in serious casualties did represent a reduction in serious road accidents, it would be extremely difficult to explain why the lethality ratios have behaved so strangely.

The serious accident statistics are not a reliable series. There are too many marginal cases and too many external factors that can affect the numbers.

The government's stated objective to reduce KSI by 40% by 2010 is laudable, but with figures like these it is highly unlikely that we will actually know if any improvement at all has taken place even if the target is met.

While we would very much like to see a firmer definition for a "serious" road casualty, we would also like to see continuity of data. Perhaps a good start would be to publish "hospitalization" figures alongside "serious" figures.

On the other hand, the fatality figures suffer little ambiguity. In the fatality figures we are seeing clearly the bad effects of an ill founded and hopelessly flawed road safety policy.

These concerns about the usefulness of the serious injury figures reflect strongly on the benefits claimed for speed camera policy. All the speed camera benefits claimed depend entirely on reductions in serious injuries.

This page is analysis and theory. No actual data exists to verify that the serious accident classification has drifted. However, there must be an explanation for the strange trend in the severity statistics, and this analysis provides a perfect fit and a rational explanation.

We will continue to investigate.

(all data used from official DfT sources)

new Comments

Safe Speed encourages comments, further information and participation from our visitors. See our (forums).

Read about our comments policy (here).

The "serious2"  page provides data to support the observations on this page. Consequently the comments topic includes both pages. You can (view) or (add comment) to the forum topic for these pages. Posting in the forum requires simple registration.

Calling for real road safety, based on truth

We have a strict editorial policy regarding factual content. If any fact anywhere on this web site can be shown to be incorrect we promise to remove it or correct it as soon as possible.
Copyright © SafeSpeed 2004
Created 12/02/2004. Last update 9/03/2004