Research in context
Evidence before this study
Damuth and colleagues published a systematic review on patients with prolonged mechanical ventilation, results of which showed wide variability in the length of time used to define “prolonged”. Before this study, we published a narrative review of different concepts of critical illness. We also did a survey of Australian and New Zealand intensive care unit (ICU) clinicians that revealed that persistent critical illness was believed to develop after a median of 10 days (IQR 7–14) and to be somewhat uncommon (occurring in 10% of all ICU patients [5–15]). From our engagement with relevant literature in the previous decade, we knew that few data existed providing an empirical test of the persistent critical illness hypothesis: that a day exists during critical illness beyond which admission diagnosis and physiological illness severity cease to predict outcome more accurately than do simple antecedent patient characteristics. Likewise, no data existed measuring the date of such transition if it occurred.
Added value of this study
This study uses binational data from many ICUs and patients and separate derivation and validation samples to minimise the potential for overfitting during risk score development. These data show that the transition point hypothesised to define persistent critical illness does in fact exist and occurred in this population as a whole after 10 days in the ICU. This study showed that patients staying in the ICU for 10 days or more, while accounting for only 5% of all ICU patients, use a very disproportionate amount of ICU resources and have outcomes that are markedly worse than are those for patients staying in the ICU for less than 10 days.
Implications of all the available evidence
Persistent critical illness can be empirically defined at a population level with a clinically relevant definition. In patients with persistent critical illness, the prognostic significance of admission diagnosis and severity of illness is markedly attenuated; by contrast, pre-ICU factors are more associated with their likelihood of in-hospital risk of death than are admission diagnosis and severity of illness. Clinicians should be aware of the risk of anchoring bias in their decision making and prognostication in the face of such rapidly changing prognostic factors. Future research should examine the mechanisms of persistent critical illness, develop and validate patient-level case definitions, and test strategies for prevention and salvage of persistent critical illness.