Thread by @iwashyna

#BigData #ICU #sepsis research I wish some1 would do

Optimal Duration of Time-Limited Trials in Critically Ill Patients

"If they are going to respond to this, we should see a response by ..."

What are median + 95th percentile for responses to common life-saving therapies?

I feel like we "know" this for lots of small things
- Lasix should make 'em pee within 20 minutes
- PEEP improves oxygenation in 20 minutes
- uroseptic shock should break within 24 hour

There are some scattered superb research projects:
- optimal durations of SBTs
- optimal time to neuroprognostication s/p anoxic injury
- Blood cultures will turn positive for MRSA w/i 48 hours if they are going to

(of course as I write this, I am quite dubious about the evidence base upon which most my "knowledge" rests)

Side-project: cross-site and international variation in clinician folk beliefs in time-to-response of common ICU therapies. How do folks decide when to redose?

But as we think about #TimeLimitedTrials, can we move beyond our experience-derived anecdotes to a more rigorous information basis?

Project 1: Time to physiologic response or (time profile of likelihood of death if no response for common treatments, purely observational

Project 2: Causal inference version: during what time periods is physiologic response a valid surrogate marker of ultimate response to therapy?

Project 3: Variation in folk-beliefs and practices (actual redosing intervals)