Answer to Final Questions Regarding Death Rates in Elderly Drivers
- Are there any factors that might confound the association between being al elderly driver and the risk of death after a motor vehicle collision? If so, what factors would you consider? How would you deal with these potential confounders?
- The figure below summarizes some of the data obtained from the Boston Medical Center Trauma registry. The upper contingency table shows deaths among the 74 elderly drivers hospitalized after an MVC and the 960 younger drivers who had been hospitalized after an MVC. The lower two tables summarize the findings after stratifying based on whether the drivers had the benefit of safety devices (seat belt buckled and/or air bag in the vehicle. Do these findings suggest the presence of effect modification? Why or why not?
Crude Analysis:
|
|
Died |
Lived |
Total |
|---|---|---|---|
|
Age ≥70 |
13 |
61 |
74 |
|
Age‹70 |
25 |
935 |
960 |
Crude risk ratio = 6.75
Stratified by Use of a Safety Restraint:
Unrestrained (no seatbelt or air bag):
|
|
Died |
Lived |
Total |
|---|---|---|---|
|
Age ≥70 |
8 |
16 |
24 |
|
Age‹70 |
13 |
359 |
372 |
Stratum-specific risk ratio= 9.54
Restrained with Seatbelt, Air Bag, or Both
|
|
Died |
Lived |
Total |
|---|---|---|---|
|
Age ≥70 |
5 |
45 |
50 |
|
Age‹70 |
12 |
576 |
588 |
Stratum-specific risk ratio= 4.90
Answers:
- One can think of a number potential confounding factors such as speed of the two vehicles, type of vehicle (e.g., a small light car versus a sturdy vehicle with good protection, site of impact (driver's side versus passenger side), severity of injuries, general health of the the driver prior to the accident, etc. One could explore confounding by these factors by first performing a series of stratified analyses. One might then use multiple logistic regression to simultaneously adjust for several confounding factors.
- The risk of death was substantially greater in elderly drivers regardless or restratint use. However, unrestrained elderly drivers had almost a ten-fold increase in death rate compared to younger drivers, whereas restrained elderly drivers had a five fold increase in death rates compared to younger drivers. Therefore the effect of age on risk of death after a car crash was different depending on wether restraints were in use. There is good evidence for effect modification here because the stratum-specific measures of association are substantially different.
.