Hypothesizing that the determinants of death within 30 days of admission to ICU are different to those for death after 90 days, Garland et al looked at a Canadian 11 ICU registry. We’re familiar with survival curves levelling to population baseline rate after a few months (if you can make it 3 months it’s almost like it never happened!). Examining the curve between 1 and 90 days the section before 30 days is considerably steeper. So unsurprisingly early deaths relate to type and severity of the critical illness whereas by about day 90 co-morbidities have become the stronger predictor. Age was again a weak marker.
Perhaps sometime we forget that for patients with multiple concurrent medical problems their ‘pre-ICU’ Kaplan-Meier was already pretty steep. If we get them through 3 months, they rejoin their potentially ill-fated subgroup curve.
Does this suggest we should stop looking at 30 day mortality rates and make 90 day survival our default primary outcome when choosing our evidence?