Brian King

Powering a trial would be difficult, and it depends on the outcome we are looking to prevent. The data we have is on "serious" events (death, readmission). That's are exceedingly rare, requiring a sample size in the tens of thousands. A cluster-randomized design may be feasible. Perhaps something with stepwise deimplementation. However - we also should ask ourselves whether it is a good investment of money to conduct a large trial to prove to us to stop something that we never proved to help in the first place.

Hi Stefan, Yes this is a predominantly a uniquely US issue which stems from a clinical practice guidelines that has been in place since the 90's, and was (at the time) based on two small cohort studies (~<50) that suggested preterm infants have more events in a car seat compared to term infants. Practice is changing - in addition to the Kaiser centers in California, several other networks there have started to stop. There are likely pockets elsewhere. Hopefully later in the year you'll see the second part of this project in print, which was a decision analytic model of the impact of car seat testing on NICU days and post-discharge outcomes. As you can imagine - the probability of benefit regarding post-discharge outcomes based on available evidence is small (David Braun's study, which they are also updating to double the sample size with Northern California). Conservative estimates of excess NICU days (ignoring other resource-related costs from additional testing) end up in the tens of millions of dollars annually (we have ~350,000 preterm infants discharged annually in the USA who would qualify for a car seat test, not including the term infants who also qualify at many places). The AAP is holding practice change back, when in reality the CPG itself is based on no real evidence that the test prevents any adverse outcomes. Evidence should support doing a test in the first place, as opposed to proving ourselves we should stop.