Scientists have developed a new machine learning-based model to predict delirium in patients with advanced cancer undergoing palliative care. This groundbreaking study involved a multicenter, patient-based registry cohort from four hospitals in South Korea.
The research aimed to explore various machine learning models to accurately predict delirium in patients with advanced cancer admitted to the acute palliative care unit (APCU) and identify the key features influencing the predictive model. This study is crucial as delirium can significantly impact patient outcomes and quality of life.
A total of 2328 patients with advanced cancer who met specific eligibility criteria were included in the study. After exclusions, the final sample consisted of 2314 patients admitted to the APCU at the four participating centers between January 2019 and December 2020.
The study utilized 39 variables, including general information, clinical risk factors, laboratory results, and a history of diseases. The primary objective was to predict delirium onset in patients with advanced cancer at the time of APCU admission using machine learning models.
Seven machine learning algorithms were evaluated, including extreme gradient boosting, adaptive boosting, and logistic regression. These models were optimized using input parameters and hyperparameters to enhance performance. An ensemble approach combining multiple models was also employed to improve prediction accuracy and robustness.
The researchers identified the most optimal hyperparameters through a rigorous validation process, calculating metrics such as area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, accuracy, and balanced accuracy. These metrics were used to evaluate the overall performance of the machine learning models.
The study also investigated the feature importance of the best-performing models, highlighting the most influential factors in predicting delirium in patients with advanced cancer. Popular software tools such as Python, TensorFlow, NumPy, and Scikit-learn were utilized to implement the machine learning models.
Moreover, the machine learning model was deployed on a public website, enabling healthcare professionals to predict delirium in patients based on the provided information. The website ensured data privacy by encoding and promptly deleting entered information after delivering the prediction result.
Overall, this innovative research provides valuable insights into predicting delirium in patients with advanced cancer using machine learning models. By leveraging advanced technology and data analysis, healthcare providers can improve patient care and outcomes in palliative settings.
