1. QUESTIONS

Sports utility vehicles, or SUVs, have become increasingly popular in recent years: as of 2023, SUVs made up 48% of new car sales globally, up from 15% in 2010 (IEA 2024). These larger vehicles consume more resources to manufacture, take up more space when parked, and on average produce more pollution when driven. A growing number of cities across the world are therefore seeking to discourage the proliferation of larger vehicles, through measures such as parking surcharges (Goodman et al. 2025).

One potential rationale for seeking to discourage SUVs relates to road danger. A recent systematic review found that, in the case of a crash, being hit by an SUV was associated with more severe injuries to pedestrians and cyclists than being hit by a passenger car (Robinson et al. 2025). Most of this evidence came from the USA, however, where vehicle design standards are weaker than in Europe with regards to pedestrian safety. In the present study we examine for the first time whether being hit by an SUV, as opposed to a passenger car, is associated with more severe injuries for pedestrians and cyclists in Great Britain.

2. METHODS

We used police-reported Stats19 road traffic crash data for Great Britain for 2004-2023. Alongside the open access police data, we requested the data fields of vehicle make, model and registration year, which we used to look up the vehicle’s body style, kerb weight, height and length. This look-up used manufacturer-reported body style and dimension data compiled by Vehicle Data Global Ltd (https://vehicledataglobal.com).[1] Where vehicle dimensions varied within a given make/model/year, Vehicle Data Global Ltd provided us with the current average value (e.g. the mean kerb weight) amongst all vehicles registered in the UK in February 2025.[2]

We restricted our analysis to pedestrian and cyclist casualties in crashes that involved a single car and no other motor vehicles. As shown in Figure 1, we excluded casualties where a valid make and model of the car was not recorded in Stats19, where the body style of the striking vehicle was something other than a passenger car or an SUV, or where data was missing on one or more covariates. This left a study population of 359,003 casualties, who are described in Table 1.

Figure 1
Figure 1.Flow chart showing casualties included in the analysis
Table 1.Distribution of casualty, driver and crash characteristics in the study population
Variable type Variable Response levels Number of casualties % of casualties
Outcome Injury severity Slight 275,978 77%
Serious 78,433 22%
Fatal 4,592 1%
Exposures Car body style† Passenger car 323,529 90%
SUV 35,474 10%
Kerb weight‡ <1550kg 311,940 88%
1550-2049kg 35,100 10%
≥2050kg 6,216 2%
Height <1575mm 296,849 90%
1575-1749mm 25,248 8%
≥1750mm 7,375 2%
Length <4600mm 277,632 84%
4600-4899mm 46,147 14%
≥4900mm 6,113 2%
Casualty Travel mode Pedestrian 148,256 41%
characteristics Cyclist 210,747 59%
Age band of 0-9 years 38,324 11%
casualty 10-18 years 88,987 25%
19-35 years 91,556 26%
36-55 years 82,400 23%
56-75 years 40,946 11%
76+ years 16,790 5%
Sex of casualty Male 240,519 67%
Female 118,484 33%
Striking Age band of car ≤20 years 26,388 7%
vehicle driver 21-25 years 41,505 12%
characteristics 36-35 years 78,616 22%
36-55 years 130,625 36%
56-75 years 65,494 18%
76+ years 16,375 5%
Sex of car driver Male 218,466 61%
Female 140,537 39%
Age of vehicle 0-4 years 111,486 31%
5-9 years 128,256 36%
10-14 years 92,801 26%
15-19 years 23,695 7%
20+ years 2,765 1%
Crash location Region/country North East 15,104 4%
and conditions North West 48,668 14%
Yorkshire and The Humber 36,973 10%
East Midlands 23,061 6%
West Midlands 31,552 9%
East of England 30,427 8%
London 54,288 15%
South East 51,045 14%
South West 29,885 8%
Scotland 25,737 7%
Wales 12,263 3%
Rural/urban Rural road 53,632 15%
status Urban road 305,371 85%
Road class$ Motorway or A 125,594 35%
B 44,403 12%
C 28,705 8%
Unclassified 160,301 45%
Road speed limit 20mph or below 23,062 6%
30mph 299,214 83%
40mph 19,168 5%
50mph or above 17,559 5%
Light conditions Daylight 270,597 75%
and infrastructure Darkness, street lights lit 76,830 21%
Darkness, street lights unlit, absent, or status unknown 11,576 3%
Road surface Dry 270,273 75%
Wet or damp 84,729 24%
Snow, frost or other 4,001 1%
Year 2004-2008 104,669 29%
2009-2013 94,255 26%
2014-2018 89,264 25%
2019-2023 70,815 20%

†We classified as passenger cars vehicles with the following manufacturer-reported body styles: convertible, coupe, estate, hatchback, saloon. We classified as SUVs vehicles with the following manufacturer-reported body styles: SUV, SUV convertible, SUV coupe, SUV estate, SUV hatchback, SUV saloon. We excluded other body styles, e.g. vans, multi-purpose vehicles and taxis. ‡These distributions have been chosen to approximate an 80% / 17% / 3% split among striking vehicles as of 2023. $‘A roads’ in the UK are major roads that are designed and intended to accommodate substantial traffic volumes. ‘B roads’ are generally somewhat smaller and are intended to distribute traffic from A roads to minor roads. ‘C’ roads and ‘Unclassified’ roads are different types of minor roads, and are designed and intended for local traffic.

We used two binary outcomes: fatal vs non-fatal injury; and killed and seriously injured (KSI) vs slight injury. Our primary exposure of interest was whether the striking vehicle was a passenger car or an SUV. Additional analyses used the weight, height or length of the striking vehicle as an alternative exposure of interest. We did this to inform policy decisions around which metric to use to define ‘large cars’ (Goodman et al. 2025). We examined these alternative exposures in separate models, i.e. including either SUV status or height or weight or length, but never more than one of these.

We tested a priori for interactions between SUV status and a) casualty age (0-9 years, 10-18 years, 19+ years);[3] b) casualty travel mode (pedestrian versus cyclist); c) urban versus rural status; d) year (as a continuous variable).

We first present minimally-adjusted models that adjust only for casualty travel mode, region of Great Britain and calendar year. We then present fully adjusted models that adjust for all the covariates shown in Table 1. All analyses used Stata/SE 18.5.

3. FINDINGS

From 2004-2023 in Great Britain, the proportion of pedestrian and cyclist car crash casualties that involved an SUV increased from 4% to 26% (Figure 2). Across that period, there was likewise an increase in the mean kerb weight (+20%), height (+6%) and length (+3%) of the cars involved in these collisions.

Figure 2
Figure 2.Proportion of striking cars that are SUVs (versus passenger cars), for pedestrian and cyclist casualties injured in single-vehicle car crashes

Over this period, SUVs increased substantially as a share of the car fleet in Great Britain. The increase shown in this graph therefore does not imply that SUV collision rates increased per vehicle. However, to the extent that SUVs are more dangerous than passenger cars in the case of a crash, the fact that SUV collisions have become more common will increase their impact at the population level.

There was strong evidence for an interaction between SUV status and casualty age category (p<0.001 for interaction in adjusted models in relation to both KSI/slight and fatal/non-fatal outcomes). We therefore present all our analyses stratified by age, choosing the overlapping age strata 0-9 years; 0-18 years; and 19+ years, to facilitate comparisons with a recent systematic review (Robinson et al. 2025). There was also an interaction between casualty travel mode and SUV status in relation to KSI injuries (p<0.001 overall; p<0.001 in adults; but p=0.94 in children), and we therefore additionally stratify our adult KSI results by mode. There was no evidence of an interaction with urban/rural status or year (p>0.05).

As shown in Table 2, the effect estimates for the impact of being hit by an SUV versus a passenger car increased upon adjustment for the variables in Table 1. This increase in the effect estimates was primarily driven by adjustment for vehicle’s age and driver’s age. On average, SUVs have a more recent date of manufacture than passenger cars, and SUVs are also currently more likely to be driven by middle-aged drivers rather than very young or very old drivers. Both factors are associated with reduced injury risk, meaning they act as ‘negative confounders’ in the minimally-adjusted analyses.

Table 2.The association between injury severity to pedestrians and cyclists and being hit by an SUV versus a passenger car: results from Great Britain, 2004-2023
Outcome Casualty age No. casualties Odds ratio (95% CI) for being hit by an SUV vs passenger car Equivalent pooled odds ratio (95% CI) from Population Attributable
Minimally-adjusted Fully- adjusted recent systematic review [No. studies]‡ Fraction, as of 2023 $
Fatal vs Adults, 19+ years 231,692 1.07 (0.96, 1.18) 1.14 (1.02, 1.28)* 1.44 (1.33, 1.56)*** [N=15] 4%
non-fatal Children, 0-18 years 127,311 1.45 (1.10, 1.90)** 1.77 (1.33, 2.35)*** 1.82 (1.57, 2.11)*** [N=6] 17%
Children, 0-9 years 38,324 2.56 (1.63, 4.02)*** 3.09 (1.92, 4.97)*** 2.30 (2.09, 2.53)*** [N=2] 35%
KSI vs slight Adults, 19+ years: pedestrians¥ 118,978 0.91 (0.87, 0.95)*** 0.97 (0.93, 1.02) 1.28 (1.13, 1.44)*** [N=13] [not calculated]
Adults, 19+ years: cyclists¥ 112,714 1.14 (1.08, 1.19)*** 1.11 (1.06, 1.16)*** 1.19 (1.07, 1.33)*** [N=7] 3%
Children, 0-18 years 127,311 1.12 (1.06, 1.17)*** 1.18 (1.12, 1.24)*** 1.28 (1.19, 1.37)*** [N=4] 4%
Children, 0-9 years 38,324 1.29 (1.18, 1.40)*** 1.35 (1.23, 1.47)*** 1.48 (1.34, 1.63)*** [N=2] 8%

†p≤0.10, *p≤0.05, **p≤0.01, ***p≤0.001. CI = confidence interval, KSI = killed and seriously injured. Minimally-adjusted analyses adjust for travel mode, region of Great Britain and the year of the crash (in two-year bands). The fully-adjusted analyses adjust for all the ‘casualty characteristics’, ‘striking vehicle characteristics’, and ‘crash location and conditions’ variables shown in Table 1, as categorical variables (but using two-year not five-year age bands). See the Supplementary Material for tables presenting the raw numbers of casualties, and for a version of this table that presents results for urban areas only. ‡ Note that the studies in this systematic review were mostly from the USA (16/24 studies) and most relied on unadjusted odds ratios (20/24 studies). ¥p<0.001 for interaction between the impact on pedestrians and cyclists, in relation to KSIs among adults. $ Population attributable fractions calculated using the formula “(P * (OR - 1)) / (1 + P * (OR - 1))”, where OR is the fully-adjusted odds ratio, and P is 26% which was the prevalence of SUVs among car crashes in 2023. We only calculated these fractions where the odds ratio in question was statistically significant.

After adjusting for covariates, there was evidence of a modest increase in the odds of fatality for adult pedestrians and cyclists, if hit by an SUV versus a passenger car (odds ratio 1.14, p=0.02). There was also evidence of a modest increase in the odds of a KSI injury for adult cyclists (odds ratio 1.11, p<0.001), but no evidence of an effect for adult pedestrians (odds ratio 0.97, p=0.25). These effect sizes were all smaller than those obtained in a recent systematic review, which drew mainly on data from the USA (Robinson et al. 2025).

For child pedestrians and cyclists, being hit by an SUV versus a passenger car was associated with a larger increase in the odds of fatality and KSI in adjusted analyses (odds ratio 1.77, p<0.001, for fatality; odds ratio 1.18, p<0.001, for KSI). These effects were particularly large in young children (odds ratio 3.09, p<0.001, for fatality; odds ratio 1.35, p<0.001, for KSI), and were similar to those documented in the two comparable studies from the USA in the recent systematic review.

As of 2023, these effect sizes translated into population attributable fractions ranging from 3% to 35% (Table 2). These fractions capture the proportion of car crash fatalities/injuries that one would be expected to be averted in Great Britain if all SUVs were replaced with passenger cars.

Finally, we examined injury severity in relation to the dimensions of weight, height and length (Table 3). We found a strong, dose-response associations with weight and height, but at most a weak association with length.

Table 3.The association between injury severity to pedestrians and cyclists and the dimensions of the striking car: results from Great Britain, 2004-2023
Outcome Casualty age Fully-adjusted odds ratio (95% CI) for the effect of striking vehicle kerb weight‡ Fully-adjusted odds ratio (95% CI) for the effect of striking vehicle height Fully-adjusted odds ratio (95% CI) for the effect of striking vehicle length
N <1550 kg 1550-2049kg ≥2050kg N <1575 mm 1575-1749mm ≥1750mm N <4600 mm 4600-4899mm ≥4900mm
Fatal vs non-fatal Adults, 19+ years 228,360 1 1.17 (1.05, 1.30)** 1.38 (1.11, 1.71)** 215,095 1 1.07 (0.93, 1.22) 1.38 (1.13, 1.68)*** 215,332 1 1.10 (1.00, 1.21)† 1.12 (0.88, 1.43)
Children, 0-18 years 124,896 1 1.64 (1.25, 2.16)*** 1.34 (0.69, 2.61) 114,377 1 1.57 (1.11, 2.22)* 1.65 (0.95, 2.85)† 114,560 1 1.18 (0.91, 1.54) 0.86 (0.38, 1.98)
Children, 0-9 years 37,584 1 2.21 (1.33, 3.69)** 3.75 (1.49, 9.45)** 34,300 1 2.23 (1.24, 4.01)** 3.20 (1.37, 7.44)** 34,363 1 1.35 (0.81, 2.25) 1.85 (0.56, 6.08)
KSI vs slight Adults, 19+ years: pedestrians† 116,741 1 1.05 (1.00, 1.09)* 1.02 (0.92, 1.12) 108,942 1 0.96 (0.91, 1.01) 0.99 (0.90, 1.08) 109,073 1 1.00 (0.97, 1.05) 1.01 (0.92, 1.12)
Adults, 19+ years: cyclists† 111,619 1 1.07 (1.02, 1.12)** 1.11 (1.00, 1.23)* 106,153 1 1.06 (1.00, 1.12)† 1.27 (1.15, 1.39)*** 106,259 1 0.99 (0.94, 1.03) 1.00 (0.90, 1.11)
Children, 0-18 years 124,896 1 1.12 (1.06, 1.18)*** 1.26 (1.12, 1.41)*** 114,377 1 1.15 (1.09, 1.22)*** 1.28 (1.15, 1.41)*** 114,560 1 1.03 (0.99, 1.08) 1.02 (0.90, 1.16)
Children, 0-9 years 37,584 1 1.24 (1.13, 1.36)*** 1.40 (1.13, 1.73)** 34,300 1 1.27 (1.14, 1.40)*** 1.46 (1.23, 1.74)*** 34,363 1 1.11 (1.03, 1.21)** 1.01 (0.80, 1.28)

†p≤0.10, *p≤0.05, **p≤0.01, ***p≤0.001. All analyses in this table adjust for the ‘casualty characteristics’, ‘striking vehicle characteristics’, and ‘crash location and conditions’ variables shown in Table 1, as categorical variables (but using two-year not five-year age bands). ‡This table presents results for kerb weight categories. See the Supplementary Material for a version of this table that presents results for an ‘adjusted weight’ variable, that uses different weight thresholds for electric and plug-in hybrid vehicles.

We conclude that in Great Britain, being hit by an SUV as opposed to a passenger car increases injury severity among pedestrians and cyclists, with the strongest effect in children. This effect is well-captured by the dimensions of weight and height, but not by length.

ACKNOWLEDGEMENTS

This study received ethical approval from the London School of Hygiene and Tropical Medicine (ref 31606). We are grateful to Tony Percival and colleagues at Vehicle Data Global Ltd for performing look-ups of body styles and dimensions based on vehicle make, model and registration year. The researcher time on this work was unfunded, but we are grateful to Clean Cities and to the Rees Jeffreys Road fund for contributing towards the cost of data matching by Vehicle Data Global Ltd, and to Ruth Carlson in helping to arrange the Rees Jeffreys Road contribution.

  1. Vehicle Data Global Ltd also compiles manufacturer-reported information on vehicle width. We did not use this data in the present study, however, as we identified that the Vehicle Data Global Ltd width data was not of a consistent type. Instead, Vehicle Data Global Ltd typically recorded the width including wing-mirrors but sometimes recorded the width excluding wing-mirrors.

  2. The Stats19 dataset includes weight data on some vehicles, obtained via a look-up of their numberplate to the information held on the vehicle’s log book. The proportion of missing data is, however, very high: 63% missing for unladen weight in our sample, and 82% missing for gross weight. In addition, the weight data held in Stats19 included some clear errors (e.g. unladen weights ranging from 100kg to 44,000kg). We therefore instead examine the impacts of weight using the kerb weight data returned by Vehicle Data Global Ltd, which is only missing for 2% of vehicles. We established that these two datasets were highly correlated (r=0.92 for gross weight, the metric available in both datasets).

  3. We stratified into these three non-overlapping age groups when testing for interactions. In our main analysis, however, we present the overlapping age groups of 0-9 years and 0-18 years, to facilitate comparisons with a recent systematic review.