Delirium occurs in as many as 29% of patients in the ICU and can have significant adverse consequences. A retrospective, observational cohort study was undertaken to compare the prognostic accuracies of three validated delirium prediction models-the PRE-DELIRIC, E-PRE-DELIRIC, and Lanzhou models-in ICU patients and to investigate the difference in the prognostic accuracy of the PRE-DELIRIC model in patients receiving mechanical ventilation compared with those who were not.

All patients ages 18 years or older admitted to the ICU over a two-year period were eligible for the study. Delirium was defined as having at least one positive score on the Confusion Assessment Method for the Intensive Care Unit or at least one administered dose of haloperidol or quetiapine. The prognostic accuracy of the three models was expressed as the area under the receiver operating characteristic curve (AUC).

Of the 1,353 patients included in the study, 410 experienced delirium. The AUC was 0.716 for the PRE-DELIRIC model, 0.681 for the E-PRE-DELIRIC model, and 0.660 for the Lanzhou model. The PRE-DELIRIC model outperformed the other two. The difference in model discrimination was statistically significant for comparison of the PRE-DELIRIC with the E-PRE-DELIRIC and Lanzhou models. The AUC of the PRE-DELIRIC model was 0.711 in patients receiving mechanical ventilation and 0.664 in those not receiving mechanical ventilation.

The authors note that the applicability of the three prediction models in clinical practice may be limited because their discriminative ability was poor to moderate. The study also had several limitations, including difficulty in using the same definition for some variables (such as coma), the possibility of missed diagnoses, and the fact that the E-PRE-DELIRIC and Lanzhou models might be less accurate in a population in which many patients are receiving mechanical ventilation.