SHM Converge Scientific Abstract ePoster Gallery-Validation of the Mayo Delirium Prediction Tool in Individuals With Cancer

Pdf Summary

This study validated the Mayo Delirium Prediction (MDP) tool in hospitalized cancer patients, addressing a gap because delirium risk models had not previously been tested in this population. Delirium is common in cancer inpatients and is often difficult to screen for with standard nursing-based tools like CAM, so an automated EHR-based model could reduce burden and improve care.<br /><br />Researchers conducted a retrospective study of 930 hospitalized patients age 50 or older at MD Anderson Cancer Center between September 1, 2022 and September 1, 2023, excluding patients with acute substance use disorder. MDP variables were extracted from electronic health records within 24 hours of admission, and the tool was applied to generate delirium risk predictions. Performance was evaluated using ROC curves and classification metrics.<br /><br />Overall, the MDP tool showed strong predictive accuracy, with an AUC of 0.8047 across all patients. Performance was good in medical patients (AUC 0.7867) and especially strong in surgical patients (AUC 0.9488). Delirium incidence increased with predicted risk: 2.5% in the low-risk group, 10.4% in the medium-risk group, and 47.1% in the high-risk group. At a 30% risk cutoff, specificity was high (94.6%) but sensitivity was lower (42.6%). At a 5% cutoff, sensitivity improved markedly (88.3%) but specificity declined (50.5%).<br /><br />The authors concluded that the MDP tool performs well in hospitalized cancer patients, particularly surgical patients, and may help identify those at highest delirium risk. They suggest that using an automated prediction model could support targeted prevention efforts and better allocation of clinical resources in oncology settings.

Asset Subtitle

Noman Ali

Meta Tag

Author List Deepti Chopra, Heather Lin, Josiah Halm, Lara Magnabosco, Maria Susan Gaeta, Noman Ali, Sandeep R. Pagali, Shant Ayanian, Tacara Soones

Category Research

Concept Mayo Delirium Prediction Tool

Concept Delirium Risk Assessment

Concept Hospitalized Cancer Patient

Concept AUC

Concept Retrospective Study

Distinguished Non-Finalist

Presenter Organization UT MD Anderson

Presenting Author Noman Ali

Track Technology in Hospital Medicine

Keywords

Mayo Delirium Prediction

MDP tool

delirium risk

hospitalized cancer patients

electronic health records

predictive accuracy

AUC

oncology inpatients

retrospective study

delirium prevention

Mayo Delirium Prediction Tool

Delirium Risk Assessment

Hospitalized Cancer Patient

AUC

Retrospective Study