Abstract
BACKGROUND: The PREdiction of DELIRium in ICu patients (PRE-DELIRIC), an intensive care unit (ICU) delirium prediction tool, was developed in an international collaboration. However, its predictive ability is not well established among Chinese ICU patients with severe pneumonia (SP). AIM: This study aimed to validate the PRE-DELIRIC model's predictive performance in SP patients in the ICU. STUDY DESIGN: This was a prospective cohort study. We consecutively enrolled SP patients in a tertiary hospital in Southwest China from 2 September 2024, to 30 September 2025. The PRE-DELIRIC model's 10 predictor variables were collected within 24 h of ICU admission. Additionally, delirium was assessed twice a day using the Confusion Assessment Method for the ICU (CAM-ICU). Our outcome measures included invasive mechanical ventilation usage, ICU stay duration and mortality rate. The model's discrimination was evaluated by the area under the receiver operating characteristic curve (AUROC), and its calibration was assessed using the calibration slope and intercept. RESULTS: Our study included 173 SP patients. The incidence of delirium was 47.40%. Although 92 (53.18%) SP patients received invasive mechanical ventilation, the median ICU length of stay was 9 days, and the mortality rate was 28.32%. Moreover, the AUROC was 0.578 (95% confidence interval [CI], 0.493-0.664). The optimal cut-off value identified by the maximum Youden index was 0.1663; the sensitivity, specificity, positive predictive value and negative predictive value were 0.793, 0.374, 0.533 and 0.667, respectively. The calibration plot of pooled data demonstrated a calibration slope and intercept of 0.143 and -1.990, respectively. CONCLUSION: The PRE-DELIRIC model is not effective for predicting delirium in SP patients. Hence, future efforts should focus on developing or refining those delirium risk prediction tools that are customised for such patients. RELEVANCE TO CLINICAL PRACTICE: Delirium is characterised by significant fluctuations and a high risk of underdiagnosis. As frontline primary clinical observers, nurses spend the maximum time with patients and can predict as well as identify delirium early. Although the PRE-DELIRIC model shows limited applicability for SP patients, we highlighted the necessity of developing disease-specific prediction tools for such patients. Additionally, such validated models might enable caregivers to identify high-risk individuals early, thereby facilitating targeted nursing interventions.