Timing of onset of persistent critical illness: a multi-centre retrospective cohort study.

PURPOSE: Persistent critical illness has been described as a subtype of chronic critical illness, characterized as a transition after ICU admission where primary diagnosis and illness acuity are no better at predicting outcome than pre-hospital characteristics. Herein we describe the occurrence and outcomes associated with persistent critical illness in a large Canadian health region.
METHODS: In this multi-center observational cohort study, all patients aged older than 14 years admitted to 12 ICUs in Alberta, Canada, between June 2012 and December 2014 were included. Primary outcome was in-hospital mortality. Predictors at ICU admission were separated into: (1) antecedent characteristics component (e.g., demographics, chronic health component of the APACHE II score, comorbid conditions); and (2) acute illness component (e.g., APACHE II score at admission, SOFA score, primary diagnostic category, surgical status, acute organ support). Using multiple statistical methods and randomly splitting the cohort into development and validation samples for risk scoring using logistic regression, we examined mortality prediction of each of these components to characterize the timing of transition to persistent critical illness.
RESULTS: We included 17,783 patients with a median (IQR) age 61 years (49-71), 62% were male, and mean APACHE II score was 19.0 (7.9). In-hospital mortality was 16.8%. Among patients alive and in ICU, the acute illness component, which accurately predicted outcome at the time of admission [area under the receiver operating characteristics curve (AUC) 0.861; 95% CI 0.860-0.862], progressively lost predictive ability and was no longer more predictive than antecedent characteristics after 9 days. This transition defined the onset of persistent critical illness and comprised 16.1% (n = 2856) of the cohort. Transition ranged between 5 and 21 days across subgroups. In-hospital mortality was greater for those with persistent critical illness [23.9% vs. 15.5%, odds ratio (OR) 1.54; 95% CI 1.43-1.67, p < 0.001]. Persistently critically ill patients accounted for 54.5% of 97844 ICU bed-days and 36.3% of 420119 hospital bed-days, respectively.
CONCLUSIONS: Persistent critical illness occurred in one in six patients admitted to Alberta ICUs and portended greater risk of death, prolonged ICU and hospital stay, and disproportionate use of health resources compared to patients without persistent critical illness.