Development and Validation of Subsyndromal Delirium Prediction Model in Patients With Advanced Malignant Tumor

Abstract

Background Subsyndromal delirium (SSD) is a clinical manifestation between delirium and nondelirium. There is no established guideline for diagnosing SSD, with a few different tools used for diagnosis. Objectives To construct and verify the risk prediction model for subdelirium syndrome in patients with advanced malignant tumors and explore its application value in risk prediction. Methods A total of 455 patients admitted to the Oncology Department in a tertiary grade A hospital in Hengyang City were recruited from December 2020 to May 2021. They were selected as the modeling group. The model was constructed by logistic regression. A total of 195 patients with advanced malignant tumors from June 2021 to July 2021 were selected to validate the developed model. Results The predictors incorporated into the model were opioids (odds ratio [OR], 1.818), sleep disorders (OR, 1.783), daily living ability score (OR, 0.969), and pain (OR, 1.810). In the modeling group, the Hosmer-Lemeshow goodness-of-fit test was P = .113, the area under the receiver operating characteristic curve was 0.884, the sensitivity was 0.820, and the specificity was 0.893. In the validation group, the Hosmer-Lemeshow goodness-of-fit test P = .108, the area under the receiver operating characteristic curve was 0.843, the Yuden index was 0.670, the sensitivity was 0.804, and the specificity was 0.866. Conclusions This model has excellent precision in the risk prediction of subdelirium in patients with advanced malignant tumors. Implications for Practice The model we developed has a guiding significance for specialized tumor nurses to care for patients with advanced malignant tumors and improve their quality of life.