Delirium During Critical Illness and Subsequent Change of Treatment in Patients With Cancer: A Mediation Analysis*

Abstract

OBJECTIVES: To assess whether delirium during ICU stay is associated with subsequent change in treatment of cancer after discharge. DESIGN: Retrospective cohort study. SETTING: A 50-bed ICU in a dedicated cancer center. PATIENTS: Patients greater than or equal to 18 years old with a previous proposal of cancer treatment (chemotherapy, target therapy, hormone therapy, immunotherapy, radiotherapy, oncologic surgery, and bone marrow transplantation). INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: We considered delirium present if Confusion Assessment Method for the ICU was positive. We assessed the association between delirium and modification of the treatment after discharge. We also performed a mediation analysis to assess both the direct and indirect (i.e., mediated by the development of functional dependence after discharge) of delirium on modification of cancer treatment and whether the modification of cancer treatment was associated with mortality at 1 year. We included 1,134 patients, of whom, 189 (16.7%) had delirium. Delirium was associated with the change in cancer treatment (adjusted odds ratio [OR], 3.80; 95% CI, 2.72–5.35). The association between delirium in ICU and change of treatment was both direct and mediated by the development of functional dependence after discharge. The proportion of the total effect of delirium on change of treatment mediated by the development of functional dependence after discharge was 33.0% (95% CI, 21.7–46.0%). Change in treatment was associated with increased mortality at 1 year (adjusted OR, 2.68; 95% CI, 2.01–3.60). CONCLUSIONS: Patients who had delirium during ICU stay had a higher rate of modification of cancer treatment after discharge. The effect of delirium on change in cancer treatment was only partially mediated by the development of functional dependence after discharge. Change in cancer treatment was associated with increased 1-year mortality.