THE EFFECTS OF MOBILE SPEED CAMERAS ON ROAD SAFETY (main report on separate page)
8.11 Acknowledgements, references and supplementary information
The author would like to thank TVSRP (Thames Valley Safer Roads Partnership) for supplying the verified database of collisions at all speed camera sites in Thames Valley along with all supplementary information that was requested and received.
The author would also like to thank the mathematicians and engineers who assisted in the production of this report.
1. Effectiveness of speed cameras in preventing road traffic collisions and related casualties: systematic review. Article cited as: BMJ, doi:10.1136/bmj.38324.646574.AE (published 14 Jan 2005). Paul Pilkington, Sanjay Kinra. Download BMJ article.
Regarding the quality of evidence presented in speed camera reports, the BMJ systematic review finds:
BMJ review p1: "Results
14 observational studies met the inclusion criteria; no randomised controlled trials were found."
BMJ review p1: "Conclusions
The level of evidence is relatively poor, however, as most studies did not have satisfactory comparison groups or adequate control for potential confounders. Controlled introduction of speed cameras with careful data collection may offer improved evidence of their effectiveness in the future."
BMJ review p3: "Implications of the research
This review has highlighted the limited nature of the evidence base underpinning the large scale introduction of speed cameras and the need for further robust evidence. Two possibilities exist for improving this evidence base. Randomised controlled trials offer the highest form of evidence. In countries where a large scale introduction of speed cameras is planned and the subject is not politicised, speed cameras could be introduced in a controlled fashion, randomising the allocation of cameras within a larger sampling framework of high risk sites (with remaining sites serving as controls)."
BMJ review p3: "the research needs to be conducted as soon as possible, before the widespread introduction of cameras results in a permanent loss of such opportunities."
2. Speed cameras for the prevention of road traffic injuries and deaths (Review). 2010 (Issue 10). Wilson C, Willis C, Hendrikz JK, Le Brocque R, Bellamy N. The Cochrane Collaboration, the Cochrane Library. Download Cochrane article.
Regarding the quality of evidence presented in speed camera reports, the Cochrane Collaboration systematic review finds:
Cochrane review p4: "PLAIN LANGUAGE SUMMARY
The quality of the included studies in this review was judged as being of overall moderate quality at best ... higher quality studies, using well designed controlled trials where possible, ...are needed." and "There is a greater need for consistency in methods ...and agreed methods for controlling bias in studies."
Cochrane review p15: "Stage 4: Quality Assessment
The assessment of the quality of non-randomised trials is problematic. Controlled before-after studies (CBAs) and interrupted time series studies (ITSs) are acknowledged as methodologically weaker than randomised controlled trials, and few if any validated quality assessment instruments exist for non-randomised studies."
Cochrane review p36: "DISCUSSION
Whilst randomised controlled trials offer the highest level of evidence, we found no studies using this study design."
Note: Issue 1 of the Cochrane Collaboration systematic review (2009) stated (p4) "...the quality of the included studies being judged to be weak..." but had failed to include the UK government's largest speed camera report, the 4YE (4 Year Evaluation, see 3a). The latest Cochrane review, issue 10, did include the 4YE but failed to notice that the estimate for regression to the mean (RTM) in the report meant that the report presented no real evidence that the speed cameras all across Britain had saved any lives, nor prevented any serious injuries. This oversight is understandable considering that the 4YE did not apply its estimates of RTM to the overall results. Speed camera reports at the time of issue 10 (2010) were of no higher quality than they were the year before for issue 1 (2009), so it is unclear why the authors changed their assessment of the quality of the evidence from "weak" to "overall moderate quality at best".
3. Two prominent UK reports have indicated that the number of people killed or seriously injured (KSI) at speed camera sites would have reduced substantially anyway (without the speed cameras) largely as a result of regression to the mean (RTM). Both of these reports estimated that RTM had a greater influence on KSI casualty reduction than the actual speed cameras (or any other factor).
3a The UK's largest speed camera report is: The national safety camera programme Four-year evaluation report Dec 2005: UCL, PA Consulting, Adrian Gains, Michael Nordstrom, Benjamin Heydecker, John Shrewsbury, Linda Mountain, Mike Maher.
To estimate RTM, the 4YE uses an EB (Empirical Bayes) approach, as described in (p143) "Appendix H: Estimates of regression-to-mean effects at safety cameras". The EB method requires a large amount of data not normally collected, so only a small number of sites were analysed. RTM was estimated to have produced a larger reduction in KSI collisions than all other effects combined, but this estimate was not applied to the overall results. When this is done, and after taking into account the possible effects of diversion of traffic, the report presents no real evidence that speed cameras saved any lives, nor prevented any serious injuries (government reports).
3b The Effectiveness of Speed Cameras A review of evidence, Richard Allsop. November 2010 RAC Foundation.
The RAC report uses incomplete data obtained from 6 unidentified partnerships in table 5 (p31) and assumes that the pre-baseline data was outside the SSP. If the sites in table 5 were selected according to the national guidelines (after around 2000) then the pre-baseline data may well have been outside the SSP and could therefore be used to determine the mean. Unfortunately, the RAC report provides no evidence of when the SSP was for these sites and, since the data sources are not identified (and are being withheld when requested), it is not possible to investigate further. The report goes on to use the relationship between the pre-baseline and baseline data to give (p32) an "...indication of the possible scale of the contribution of RTM". The report goes on to indicate that RTM might be the largest influence on KSI casualty reductions at speed camera sites (p44, Figure 2), but goes no further than providing indications of RTM levels.
3c Note: Some reports compare speed camera sites to sites that were considered to be unsuitable for speed cameras.
Some reports compare speed camera sites to other sites that were similar in nature, but where speed cameras were considered unsuitable as an intervention. The fact that it was decided not to use speed cameras in the comparison sites means that those sites were considered to be different to the locations actually chosen to have speed cameras. Therefore, comparisons between the two groups would contain the effects of those selection choices. Only when sites are randomly selected to have speed cameras in an RCT, with remaining sites used as control sites, are comparisons free of RTM effects.
4. The standard method for assessing road safety interventions is to compare a period after the start of the intervention (often the first 3 years), to a baseline period before (typically the 3 years used to select the site). This is the method generally used for assessing speed cameras as detailed in: The National Safety Camera Programme NSCP Data Monitoring Handbook. The handbook shows the authorities (such as local councils and partnerships) how to produce a (p31) "Camera Site Effect report" (in "F. SITE EFFECT") which, as the screen shot shows, compares a period after speed cameras were deployed to a 3 year baseline period before. The results columns ("effect" and "Total # avoided") show the changes that occurred due to the combination of random variation at individual sites, RTM, general influences, diversion of traffic and the speed cameras. There is no analysis to separate the effect of the speed cameras from all the other influences. (Note, there are other obvious issues, such as a change from 0 to 2 is not a 50% increase).
5. The proportion of motorists that might cause a KSI collision while speeding in their lifetime is estimated using 2010 figures on spreadsheet motorists causing KSI in lifetime.xls.
Summary of spreadsheet: In 2010, there were around 33,600 KSI collisions and Police found that 7.1% of those that were investigated involved a vehicle exceeding a speed limit (speeding) (table 1.2). This suggests that around 2,400 KSI collisions involved speeding in 2010 and therefore nearly 140,000 KSI collisions might involve speeding during the 58 years a person might be a motorist. As there are around 30 million motorists, these collisions could be caused by up to 0.5% of them. Therefore, the other 99.5% of motorists would not cause a KSI collision while speeding in their lifetime.
6. The national safety camera programme Four-year evaluation report Dec 2005: UCL, PA Consulting, Adrian Gains, Michael Nordstrom, Benjamin Heydecker, John Shrewsbury, Linda Mountain, Mike Maher. (see 3a above)
On page 156 "...part of the reduction in collisions attributable to the cameras may be due to diversion of traffic away from routes with cameras."
7. Almost all of the mobile speed camera sites in Thames Valley were installed by TVSRP from 2002 onwards. See Figure 7.1 from report on Thames Valley speed cameras, or spreadsheet mobile cameras TVSRP.xls.
8. Proposed mobile speed camera sites were expected to have a minimum of 4 collisions, with at least 2 of them resulting in KSI casualties, per km in the three years prior to the proposal being submitted to the DfT. See (p91) Table A1.
9. Analysis by the DfT has found that Police only record around a third of all the casualties that occur (and therefore around a third of all collisions as well). In "Reported Road Casualties in Great Britain: 2010 Annual Report" the Police recorded 208,648 casualties but the DfT estimate there were actually around 730,000 casualties in total (3.5 times the Police figure).
The database of collisions on which this report is based was supplied by TVSRP (Thames Valley Safer Roads Partnership).
During the first quarter of 2009, TVSRP undertook a full check and correction procedure on their collision records to produce a verified database of collisions at all speed camera sites in Thames Valley. This report is based on the new verified database supplied by TVSRP containing the number and severity of collisions and casualties at each speed camera site within each calendar month from Apr 1990 to Mar 2009 inclusive (a 19-year period consisting of 228 individual months of data for each site). Only data before Jan 2009 is used in this report because it was considered that there may have been insufficient time to submit, verify and enter the data from the very latest police collision reports during the final 3 month period.
The complete database contains data for a total of 517 speed camera sites. Of these, 261 are mobile speed camera sites but 89 of these have been decommissioned. A further 97 of the active sites were commissioned after the start of 2006 and therefore had insufficient data for this report (they had been operating for under 3 years by the start of 2009). This report therefore includes all 75 active mobile speed camera sites that had been operating for 3 or more years at the start of 2009.
Excel spreadsheet with mobile speed camera site information: mobile cameras TVSRP.xls.