Saturday 16 June 2012

Despite what I said before, maybe cycling casualty rates aren't strictly comparable between Britain and the Netherlands. How about fatality rates then?

Back in April I posted an analysis of fatal and serious casualty rates for cyclists in Britain and the Netherlands, which reached the eye-catching conclusion that "just over 500 British cyclists are killed or seriously injured in collisions with motor vehicles for every billion km cycled, over eight times the rate in the Netherlands". I used data from the British Department for Transport and the Dutch road safety institute SWOV, but as a couple of commenters pointed out these do not in fact seem to use completely consistent definitions of what constitutes a 'serious' injury, so the comparison might be misleading. For example, snigbo said,
On comparing serious injuries: as discussed in DfT rrcgb2010-6.pdf, GB "serious" includes all admissions to hospitals, many of whom will have a MAIS score of 1. Table 11 suggests that 28% of hospital-admitted cyclists were MAIS 1, so reducing the GB number by that amount would be appropriate.
There is also the problem that for both countries the casualty figures are based on what is reported to police, and there may be differential rates of under-reporting, particularly for incidents not involving motor vehicles.

These are valid concerns, so I've updated that post with a health warning over the data. I'm grateful to the commenters for pointing these issues out, and I apologise if anyone was misled.

Can we then make any useful comparisons between the two countries? Well, it may be tempting fate to go back to the same data sources, but as was also pointed out the data on fatality rates should in principle be more comparable. After all, a fatality is a fatality wherever you are, and you would also expect minimal problems of under-reporting.

The problem here is that the SWOV data on contributory factors in cycling fatalities includes a large number (a majority, in fact) of cases described as 'Not matched', i.e. they don't say whether a motor vehicle was involved or not [Update: Note, this doesn't affect the total fatality rate, which is known - it's just that not every case is allocated to a particular type of incident]. You can just exclude all these unmatched cases and look only at the breakdown of the remainder, which is what I have done in the chart below. Having been burned before I would be cautious about putting too much weight on the results, mainly because of the large number of 'not matched' cases in the Netherlands data, but they do at least seem to be telling a similar story to the previous data, i.e. that the relative risk posed by motor vehicles to cyclists is higher in Britain than in the Netherlands. 



6 comments:

  1. Jim, thanks for this update. Two questions: (1) Do you mean that the actual fatality rate in NL is not 9 but closer to 20 per bn km? (2) Would you mind giving the dates which the data refers to? Cheers

    ReplyDelete
  2. (1) No. Why do you think that?
    (2) 2009.

    ReplyDelete
  3. I was just confused by this bit wher you seem to say that the majority of records were excluded from the chart:

    "a large number (a majority, in fact) of cases described as 'Not matched', i.e. they don't say whether a motor vehicle was involved or not. You can just exclude all these unmatched cases and look only at the breakdown of the remainder, which is what I have done in the chart below"

    ReplyDelete
  4. Ah, I see. No, that just means that a full breakdown of the overall fatality rate isn't provided. The total rate is still the same. I'll try to clarify the paragraph - thanks for pointing this out.

    ReplyDelete
  5. "that the relative risk posed by motor vehicles to cyclists is higher in the Netherlands than in Britain"

    Surely the other way around?

    ReplyDelete
  6. D'oh. Fixed, thanks!

    ReplyDelete

Note: only a member of this blog may post a comment.