Article Text

Download PDFPDF

316 The role of selektion bias in a case-crossover study on occupational injuries
  1. Anna H Oesterlund1,
  2. Flemming Lander2,
  3. Søren Rytter3,
  4. Jens M Lauritsen1
  1. 1Accident Analysis Group, Orthopaedic Department, Odense University Hospital and Institute of Clinical Medicine, University of Southern Denmark, Denmark
  2. 2Department of Occupational and Environmental Medicine, Odense University Hospital, Denmark
  3. 3Department of Orthopaedics, University Clinic for Hand, Hip, and Knee Surgery, Regional Hospital West Jutland, Denmark


Background The case-crossover studies in injury epidemiology research are gaining ground. The method compares transient exposure during intervals when an outcome occurs, to exposures during intervals without outcome for the same individual. Although non-responding is a concern in all interviewed based studies, previous studies have not involved and discussed the importance of selection as a source of bias that could influence the overall reliability of the risk outcomes. Due to the unique Danish identification number it is possible to examine data between responders and non-responders in our case-crossover study concerning injuries, industry, sex and age.

Methods The population base for the study was derived from two public Hospital Emergency Departments in Denmark. All contacts due to occupational injuries during 2013 were included. An occupational injury was based on asking the patient whether the injury occurred during paid work. All such injured patients were asked to participate. Those confirming participation (responders) were compared with those not responding or denying (non-responders).

Results Among all 4002 injured, 1693 were responders (42%). The overall tendency of sex, age and distribution of injury were the same between both responders and non-responders except for “other types of injuries”. When dividing into minor and major injuries there were no difference between responders and non-responders. Further investigation of potential bias for subgroups or industries awaits further analysis.

Conclusions So far, our results indicate no serious selection bias in sex, age and injury distribution in spite of low participation rate, and thus, provide good possibility for broader generalisation of the risk outcomes. This indicates a strengthening of the overall reliability of the risk outcomes from our and previous case-crossover studies.

  • Occupational injury
  • transient risk factors
  • selection bias and case-crossover studies

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.