The impact of frailty on death, discharge destination and modelling accuracy in patients receiving organ support on the intensive care unit.

Abstract

This study aims to identify any effect of frailty in altering the risk of death or poor outcome already associated with receipt of organ support on ICU. It also aims to assess the performance of mortality prediction models in frail patients. All admissions to a single ICU over 1-year were prospectively allocated a Clinical Frailty Score (CFS). Logistic regression analysis was used to investigate the effect of frailty on death or poor outcome (death/discharge to a medical facility). Logistic regression analysis, area under the Receiver Operator Curve (AUROC) and Brier scores were used to investigate the ability of two mortality prediction models, ICNARC and APACHE II, to predict mortality in frail patients. Of 849 patients, 700 (82%) patients were not frail, and 149 (18%) were frail. Frailty was associated with a stepwise increase in the odds of death or poor outcome (OR for each point rise of CFS = 1.23 ([1.03-1.47]; p = .024) and 1.32 ([1.17-1.48]; p = <.001) respectively). Renal support conferred the greatest odds of death and poor outcome, followed by respiratory support, then cardiovascular support (which increased the odds of death but not poor outcome). Frailty did not modify the odds already associated with organ support. The mortality prediction models were not modified by frailty (AUROC p = .220 and .437 respectively). Inclusion of frailty into both models improved their accuracy. Frailty was associated with increased odds of death and poor outcome, but did not modify the risk already associated with organ support. Inclusion of frailty improved mortality prediction models.