Friday, May 23, 2008 - Morning: Dr. Keith Chambers, epidemiologist
Presentation (.pdf format) [LINK]
Wow.
Slide 11: Small sample sizes as low as 15. N= 66 is very small. Almost no power to detect adverse events, unless very high frequency. Why such small sample sizes? ... Most experiments are a limited number of shots (mostly single shot) and to the back which is not the real world.
Slide 17: There appears to be a serious chance of missing a very significant relative risk due to the use of Tasers given the existing data.
The primary point being made ever-so-gently above is that most of the studies were very badly designed if the goal was to prove safety (sample size too small to prove anything). On the other hand, if the goal was to produce a thick stack of useless and meaningless reports, then those studies were perfectly designed.
Slide 21: Recommendations
- Guidelines are badly needed. It appears reasonable to make these conservative with some form of use limitation, until appropriate population data comparing harms to benefits is available.
- Guidelines should be standardized (at the least, provincially).
- Condition of use and indications for use, by an agency, should require compliance with a standardized reporting mechanism.
- Reporting should be linked to an outcome database.
- Stakeholders should be encouraged to get together to find common ground to develop research strategies including the creation of a large, independent multi-centered database (North American wide?) to resolve unanswered questions.
Related news items: [LINK] [LINK] [LINK]
No comments:
Post a Comment