To determine the prevalence of persistent critical illness (PerCI) after cardiac surgery, describe the characteristics, trajectory, and outcomes of these patients, identify risk factors, and develop predictive risk scores.

Retrospective observational study between 2008 and 2021. In keeping with the literature, PerCI was defined as an intensive care unit admission for longer than six consecutive days. Fifty-two (52) hospitals across Australia and New Zealand that submitted data to the Australia New Zealand Society of Cardiac and Thoracic Surgeons database. Participants were all patients who underwent cardiac surgery or thoracic surgery requiring cardiopulmonary bypass, in participating units.

We analysed 155,819 patients, of which 14,940 (9.6%) met the criteria for PerCI. Prognostic prediction models with out-of-sample area under the curve ≥0.70 and 0.75 for preoperative and immediately postoperative time points, respectively, were developed for early identification of patients at risk of developing PerCI. Patients affected by comorbidity, salvage or emergent surgery, compromised cardiac function, and those requiring intra-aortic balloon pump were more likely to develop PerCI. Patients with PerCI had more postoperative complications and required more intensive care unit resources (extracorporeal membrane oxygenation, blood products, inotropes, haemofiltration). Patients with PerCI were less likely to die from cardiac causes (21.9% vs 36.5% of deaths; p<0.001) but had a higher in-hospital mortality (9.8% vs 1.3%; p<0.001) and mortality at 1 year (7.5% vs 1.9%; p<0.001), with no change in such rates over time.

Nearly one in every 10 patients undergoing cardiac surgery will develop PerCI, with frequent non-cardiac complications and death from non-cardiac causes. With prognostic prediction models, most patients at risk of developing PerCI can be identified early, thus enabling the use of targeted preventive interventions.