Article Text

Download PDFPDF
Investigation of bias after data linkage of hospital admissions data to police road traffic crash reports
  1. P C Cryer1,
  2. S Westrup1,
  3. A C Cook1,
  4. V Ashwell2,
  5. P Bridger3,
  6. C Clarke4
  1. 1Centre for Health Services Studies, University of Kent, Canterbury, UK
  2. 2Health Promotion, West Sussex Health Authority, Worthing, West Sussex, UK
  3. 3Public Health, East Sussex Brighton and Hove Health Authority, Lewes, East Sussex, UK
  4. 4Highways and Transport, East Sussex County Council, Lewes, East Sussex, UK
  1. Correspondence to:
 Dr Colin Cryer, CHSS at Tunbridge Wells, University of Kent, Oak Lodge, David Salomons' Estate, Broomhill Road, Tunbridge Wells, Kent, TN3 0TG, UK


Research question—Does a database of hospital admission data linked to police road traffic accident (RTA) reports produce less biased information for the injury prevention policymaker, planner, and practitioner than police RTA reports alone?

Design—Data linkage study.

Study population—Non-fatal injury victims of road traffic crashes in southern England who were admitted to hospital.

Data sources—Hospital admissions and police RTA reports.

Main outcome measures—The estimated proportion of road traffic crashes admitted to hospital that were included on the linked database; distributions by age, sex, and road user groups: (A) for all RTA injury admissions and (B) for RTA serious injury admissions defined by length of stay or by nature of injury.

Results—An estimated 50% of RTA injury admissions were included on the linked database. When assessing bias, admissions data were regarded as the “gold standard”. The distributions of casualties by age, sex, and type of road user showed major differences between the admissions data and the police RTA injury data of comparable severity. The linked data showed smaller differences when compared with admissions data. For RTA serious injury admissions, the distributions by age and sex were approximately the same for the linked data compared with admissions data, and there were small but statistically significant differences between the distributions across road user group for the linked data compared with hospital admissions.

Conclusion—These results suggest that investigators could be misinformed if they base their analysis solely on police RTA data, and that information derived from the linked database is less biased than that from police RTA data alone. A national linked dataset of road traffic crash data should be produced from hospital admissions and police RTA data for use by policymakers, planners and practitioners.

  • data linkage
  • hospital records
  • police road traffic crash reports
  • bias (epidemiology)

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • * The injury must be sufficiently serious to require a hospital stay of four days or more.

  • Injuries resulting in death within 30 days of the accident are classified as “fatal”.

  • This estimate assumes that the proportion of cases with missing E codes was the same for RTAs as other causes of injury.

  • § The usual χ2 test for a contingency table could not be used when comparing the distribution of cases because of the lack of independence of the data from the data sources being compared. These include many of the same cases.