Methods

Digital road and traffic calming information in England and Wales is captured on the Ordnance Survey's MasterMap integrated transport network (ITN) layer.26 This is a digital dataset of the road network for Great Britain. Linked to the spatial data are tables containing information, including height restrictions, traffic calming and one-way roads. Traffic calming is defined as all speed humps (round or flat top), raised junctions and sections of road and speed cushions. The ITN layer also contains information on road segments across the UK, including 12-digit geo-references for junctions at the beginning and end of each segment.

The ITN files were obtained for substantial areas of England and Wales, including south Wales, south-west England, and parts of the east Midlands and Surrey, for analyses of traffic calming related to the Advocacy in Action cluster randomised trial.27 All road segments containing the label ‘traffic calming’ were identified within ArcMap28 for each small area. Analysis of traffic-calming distribution was carried out at the road segment level with the latest updates received in 2007.

Lower super output areas (LSOA) are the smallest geographical boundary areas developed for use in small area analysis in the UK and have an average population of 1500.29 LSOA are a consistent size and dimension and, unlike electoral divisions, are not subject to boundary changes. Electoral divisions are the geographical units used previously in these types of analyses. There is a total of 34 378 LSOA in England and Wales (32 482 in England, 1896 in Wales). Road segments were mapped to LSOA using the ITN files available to us, with 6647 LSOA within four separate areas being covered. All LSOA within the four areas were contiguous.

The total length of road was calculated for each LSOA along with the total length of traffic-calmed road segments. The percentage of total road length traffic calmed per LSOA was calculated. Population data from the 2001 census were obtained from the Office for National Statistics for each LSOA allowing population density to be calculated per LSOA.30 The Townsend index, derived from the 2001 census at LSOA level for England and Wales, was used as an area measure of deprivation. The Townsend deprivation index is widely used in health research, particularly in academic research and is considered to be valid and reliable.31–33 LSOA were divided into five groups of equal numbers based on the Townsend index score, and will be referred to as deprivation fifths.

Summary statistics of the distribution of traffic calming by deprivation fifth were calculated. The likelihood of traffic calming per road segment was compared by deprivation fifth, adjusting for population density using weighted logistic regression in the statistical package R.34 35 Total road lengths were used to weight the traffic-calming lengths for each LSOA. Predicted probabilities were calculated and included as part of the results because these are often more easily understood than odds ratios. Represented as percentages, they are simply the sum of the likelihood of the event happening over the corresponding inverse; the likelihood of the event not happening (that is probability/1−probability).

Results

This study analysed 124 249 road segments in 6647 LSOA, which amounted to 95 791 km of road, of which 3525 km comprised segments that were traffic calmed (3.7%). Table 1 shows the summary descriptive analyses by deprivation fifth. The mean proportion of roads traffic calmed within each LSOA was 6.6%, but across all LSOA the proportion of traffic-calmed road ranged from 0% to 98.1%. The data demonstrated considerable positive skewness. Several transformations were attempted but the transformed data remained skewed. Variations by deprivation fifth were substantial. The most affluent fifth had both the lowest median traffic-calmed roads by LSOA (1.0%) and the narrowest range of road calmed across all LSOA in the fifth category (69.0%). In contrast, the most deprived fifth had a median proportion of roads traffic calmed by LSOA of 8.4% and a range of 95.1%. Aggregating the data across all LSOA in a fifth showed a sixfold difference between the most affluent and deprived fifths (2.4%, 12.2%).

Road segment lengths varied slightly between deprivation fifths, with median lengths of 59.1 m and 54.9 m in the most and least affluent fifths (Kruskal–Wallis test, χ²=174.8, 4 df, p<0.001). For all LSOA in this study, there was a mean population of 1507 people per LSOA, with a minimum of 1004 and a maximum of 4497 people per LSOA. The percentage of traffic calming per road segment was correlated with population density (Spearman's r=0.23, p<0.001) and population density increased with deprivation (Kruskal–Wallis test, χ²=895.1, 4 df, p<0.001).

Due to the highly skewed data and potential confounding by population density logistic regression was carried out to determine whether the presence of traffic calming in road segments was influenced by deprivation, adjusting for population density. Table 2 shows the results of the logistic regression analysis (Hosmer–Lemeshow goodness of fit test; p=0.09). Townsend fifth category and population density were both significant independent predictors of traffic calming. The odds of a traffic-calmed road segment was 12% lower in the next affluent compared with the most affluent. Traffic calming was most prevalent in the most deprived fifth, with 2.8 times the amount of traffic calming compared with that in the most affluent. In addition, for each increase of 10 people per hectare, the odds of traffic calming increased by a factor of 1.22.

These findings can also be interpreted using predicted probabilities derived by holding the population per hectare constant at the mean (33.3 people per hectare). With population density held constant, the probability is that 2.91% of roads in the most affluent areas would be traffic calmed, compared with 7.63% of roads in the most deprived areas (figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Percentage traffic calming for each deprivation fifth.

Discussion

The results of this study show that traffic-calming distribution heavily favours the most deprived areas. This is a positive finding because it indicates that there is a degree of social equity in traffic calming, given the overwhelming evidence of substantially higher pedestrian injuries among deprived groups.2 3 15–17 In recent years there has been a narrowing of the inequality gap in child pedestrian injuries in the UK.36 It is not possible with this cross-sectional research design to attribute a particular reduction of pedestrian injury inequality to the distribution of traffic calming. Other factors may also be involved, such as changes in modal transport, for example, reduced walking activity, and any other factor that reduces traffic speed. Nevertheless, given that empirical evidence suggests traffic calming is effective in curtailing the frequency and severity of road collisions, some reduction in injuries will have occurred in deprived areas.7–11 There are very few examples of health factors that run contrary to the inverse care law, and it is pleasing to see a population-based public health intervention that bucks the trend.22

It has been commented recently that the scope for making improvements in road safety using road engineering is reducing because they are believed to be at saturation, and that there should be a greater shift towards enforcement and education.36 However, our analyses suggest that there is still considerable potential for traffic-calming use in all areas; both affluent and deprived. The ranges of percentage traffic calming indicate that some LSOA are very heavily traffic calmed, but the overall figures of 3.7% and 12.2% in the most deprived areas, suggest that even in these areas, there is still relatively little traffic calming. Obviously, 100% coverage is impractical because speed humps are only suited to certain types of road, but the Audit Commission suggestion that there is little further scope for engineering is not supported by our findings.37

Although our study has several strengths it also suffers from a few limitations that must be considered when appraising the research findings. One major strength lies in the enormous amount of data collected in a standardised method over large parts of the UK. The study includes some 124 249 road segments amounting to 95 791 km of roads, distributed across 6647 LSOA (19.4% of total in England and Wales). These large numbers ensure that the study had sufficient statistical power to detect important differences. The areas covered were chosen to be within 50 km of the five Advocacies in Action Study centres (Swansea, Cardiff, Bristol, Nottingham and Guildford) rather than representing any particular feature of road safety, and there is no obvious reason why the findings from this study should not be generalisable across England and Wales.27

One weakness of the study is that we were only able to obtain information on vertical traffic calming (speed bumps). Whereas Ordnance Survey's ITN data contain information on vertical traffic calming the equivalent horizontal features, such as chicanes and road narrowings, are not captured. In a previous study covering a smaller area, some of the authors of this investigation carried out a street-by-street audit allowing a broader view of traffic calming to be documented, including narrowings and chicanes as well as road humps.24 However, it is not logistically possible to carry out such audits on nearly 100 000 km of road, as would have been the case had this methodology been adopted as part of this current study. Nevertheless, in the previous study, carried out in 2002, vertical calming was much more prevalent than horizontal calming. Given that we have not included horizontal calming measures, our results underestimate the full extent of traffic calming present in the areas studied. However, the size of any bias is probably small and there is unlikely to be significant variation in the method used by deprivation fifth (review of data from the study by Jones et al24 suggested that the proportions of vertical and horizontal deflectors were similar within each fifth). Our findings on the relative distribution of traffic calming by deprivation should thus be robust. Furthermore, it has been suggested that vertical deflectors are most effective at reducing speeds and pedestrian casualty rates.38 This study therefore provides a profile of the distribution of the single engineering approach with the greatest potential benefit for child pedestrian casualty prevention.

Another limitation of our method is that the MasterMap ITN layer data record whether there is any or no traffic calming on a road segment rather than displaying each specific feature.26 It is thus possible that a long segment of road may have just one speed hump on it, or three or four. Obviously, the impact on vehicle speeds and therefore collision risk is likely to be very different in these two scenarios. As a consequence, the extent of effective traffic calming within a traffic-calmed segment may be overestimated by our approach and be somewhat lower than the 3.7% of total length that we report here.

Although there is a considerable literature on social inequalities in pedestrian injuries, there is much less on the social distribution of countermeasures.2 3 15–17 24 The mere existence of strategy and policy documents is insufficient to ensure that effective countermeasures are implemented.39 Neither the public nor policy makers have access to information on the distribution of countermeasures. Without such information it is difficult for groups to lobby effectively for further investment in prevention. We consider that there is a need to put such information into the public domain. Analyses such as those that form the basis of this research could be made available to the public. Information could be disseminated through government websites or that of the Injury Observatory for Britain and Ireland website.40 Whether such information leads to subsequent increased investment in preventive countermeasures and further reductions in injury inequalities are testable hypotheses.

Implications for prevention

This research study indicates that it is possible to use high-resolution map data within a GIS to measure and monitor the equitable distribution of traffic calming. Traffic calming in the UK is distributed in a manner that is likely to be reducing inequalities in child pedestrian injury rates, although the magnitude of this effect cannot be measured without further in-depth studies. There still appear to be very few traffic-calmed roads in most areas of England and Wales, whether deprived or affluent and thus, contrary to the suggestions of the Audit Commission, the scope for the introduction of additional road safety features across the UK continues to exist. Making local traffic-calming data publicly available, allowing it to be tested as an advocacy approach, represents an important recommendation for future research.