Question
Public health practitioners wanted to evaluate the impact of a community-wide bicycle safety intervention on the incidence of head concussions. Community A was exposed to an intensive bicycle helmet use campaign, and community B, which did not receive the intervention, was chosen as a comparison population. Rates of injury associated with bicycle accidents and the use of helmets were measured in both communities.
Table 1 presents data on the “true” injury status of study participants. Although generally the extent of misclassification is unknown to the investigators, in this example we know both the “true” injury classification status and the proportions of injured and not injured individuals that are misclassified with respect to head injury status.
Table 1. True injury status (injured vs. not injured)
A. Calculate the relative risk of injured participants in Community A vs. Community B
B. Suppose that 20% of participants with head injuries are incorrectly classified as not injured,
and that 5% of participants without head injuries were incorrectly classified as injured. This
misclassification of injury occurs irrespective of community of residence (the “exposure”).
Fill in the cells for the “observed” data in Table 2, below and calculate the relative risk based
on the data in Table 2 (as if misclassification of injury had occurred)
Table 2. “Observed” injury status (injured vs. not injured)
Injured
|
Not injured
|
Total
| |
Community A
|
100
| ||
Community B
|
100
|
C. What type of misclassification is apparent?
D. Compare this “observed” relative risk to the true relative risk that we found with no misclassification present (calculated in part A above).
E. Describe what particular concerns in regards to selection bias are associated with case-control studies and cohort studies.
Source: 250A Problem Set
