The study in brief
Crash severity during changing mobility.
Using linked STATS19 records, this study examined 384,472 police-reported injury crashes in England during 2019-2022 across 14 COVID-19 policy periods. Adjusted logistic regression compared fatal-or-serious with slight outcomes. Selected adjusted probabilities were 20.5% before the pandemic, 22.0% during the first national lockdown and 22.6% during postrestriction follow-up. Policy-period differences were modest and sensitivity-dependent; vehicle involvement and crash context showed stronger associations. These are observational findings about severity among reported injury crashes, not causal policy effects or risk per journey.
Selected published estimates
Compare policy periods.
Adjusted probability of a fatal or serious outcome among reported injury crashes. Three selected periods from the 14-period analysis.
Dots: adjusted probabilities. Lines: 95% confidence intervals.
Reference period. Estimates are adjusted over the full analytical sample.
Source: published article, Table 3. Differences use the paper's unrounded estimates; they need not equal subtraction of displayed rounded percentages.
Reading the evidence
Context and uncertainty matter.
Only the first-lockdown association remained statistically significant with police-force-by-year controls. The follow-up result is sensitivity-dependent. Vehicle involvement and crash context had stronger associations than policy period.
These probabilities are conditional on a reported injury crash. They are not crash rates, risk per mile or causal estimates of policy effects.
Study method and scope
Linked collision, vehicle and casualty records; complete-sample unweighted logistic regression; calendar-week cluster-robust intervals; marginal standardisation over the full analytical sample. Consult the article for all adjustments and sensitivity tests.
Source and access
Read the version of record.
This page is an independently written summary and interactive presentation of reported estimates. The journal PDF and original figures are not hosted here. Read the published article at BMJ; publisher access and reuse conditions apply.