Case-Crossover Design

Description

Compare each case's exposures just before an incident with that same person's exposures in normal periods, so each person serves as their own control for transient risk factors.

The case-crossover design asks what was different in the hazard period immediately before an incident compared with normal periods for the same person. Because each case is their own control, there is no need to find matched controls, which is hard in small populations, and stable traits cancel out. The hazard period depends on the exposure: seconds for a falling object, hours for a sedating drug; with several factors the longest relevant period is used. In the hand-injury studies, injured workers were interviewed by phone, ideally within a day, about seven transient exposures; the main risks were a machine, tool or material behaving differently than usual, unusual work methods and unusual tasks. Later studies found time pressure and distraction mattered, but non-routine tasks and broken equipment mattered more.

Attributes

Loss type
Unmigrated
Stage
Post-launch
Effort
Weeks

Steps

  1. Define the discrete outcome (injury, incident, critical error) and recruit cases soon after it occurs.
  2. Choose candidate transient exposures and set a hazard period suited to each.
  3. Interview each case about exposures in the hazard period and in comparison periods of the same length, for example the same time the previous day or week, or usual frequency.
  4. Compute within-person relative risk for each exposure, for example with conditional logistic regression or matched-pair odds ratios.
  5. Report the transient factors that raise risk and design against them.

Sources

  • Handbook of Human Factors and Ergonomics (Salvendy & Karwowski, 2021)

Also known as

  • Case-crossover method
  • Self-matched case-crossover
  • Case-crossover design for transient risk factors

Categories

  • Epidemiology & Population Health