|
Activity Number:
|
177
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #301070 |
|
Title:
|
Misclassification Bias in the Case-crossover Design Applied to Wireless Telephones and the Risk of Road Crashes
|
|
Author(s):
|
Francois Bellavance*+ and Mustapha Bourhattas and Sophie Lapierre and Claire Laberge-Nadeau and Stephane Messier
|
|
Companies:
|
HEC Montreal and HEC Montreal and Ecole Polythecnique De and Universite De Montreal and Universite De Montreal
|
|
Address:
|
3000 Chemin De La Cote-Sainte-Catherine, Montreal, PQ, H3T 2A7, Canada
|
|
Keywords:
|
simulations ; relative risk ; hazard interval
|
|
Abstract:
|
The case-crossover design was proposed a decade ago to avoid control selection bias in a study to estimate the risk of myocardial infarction after exposure to an intermittent event such as a physical exertion. More recently, this design was used to study the association between wireless telephones calls while driving and motor vehicle collisions. One important possible source of bias in this latter study is the misclassification of phone calls due to reporting errors in the exact time of the collision. Indeed, the time written in the police report is often a multiple of 5 minutes and tends to be larger than the exact time of the collision. Results of simulation studies showed that the case-crossover design gives unbiased estimates of the relative risk when there is no misclassification bias and the hazard interval is less than the average duration of phone calls. However, when we introduced random errors between the exact time of the collision and the time in the police report, we obtained estimates that were up to three times larger than the true relative risk. The bias due to exposure misclassification was larger for smaller values of the true relative risk.
|
