Short- and Long-Term Mortality Prediction in Critically Ill Cancer Patients Admitted to the Intensive Care Unit (CanICU) Using Machine Learning

Abstract

Background: Although cancer patients are increasingly admitted to the intensive care unit (ICU) for cancer- or treatment-related complications, improved mortality prediction remains a big challenge. Current prognostic models for a general ICU population underestimate the mortality risk in cancer patients. Methods: We developed CanICU, a machine learning-based 28-day and 1-year mortality prediction program in adult cancer patients admitted to ICU from Medical Information Mart for Intensive Care (MIMIC) database in the USA (n=766), Yonsei Cancer Center (YCC, n=3,571), and Samsung Medical Center in Korea (SMC, n=2,563). We constructed the classifier and compared it with a general prognostic model. Findings: A total of 6,900 patients were included, with a 28-day mortality of 10.2%/12.7%/36.6% and 1-year mortality of 30.0%/31.4%/58.5% in the MIMIC, YCC, and SMC cohort, respectively. Nine clinical and laboratory factors were used to construct the classifier using a random forest machine-learning algorithm. CanICU had the area under the receiver operating characteristic (AUROC) of 0.95 (95% CI 0.93-0.97) for 28-day and 0.79 (95% CI 0.76-0.82) for 1-year mortality, showing better performance than a current prognostic model (Sequential Organ Failure Assessment [SOFA], 0.77 [95% CI 0.72-0.81] for 28-day and 0.67 [95% CI 0.64-0.70] for 1-year mortality). Application of CanICU in external data sets across the countries yielded better AUROC for 28-day mortality than SOFA (95% CI, 0.75-0.80 vs. 0.57-0.63 with SOFA in SMC; 95% CI, 0.72-0.79 vs. 0.64-0.72 with SOFA in MIMIC). Interpretation: CanICU offers improved performance for predicting short- and long-term mortality in critically ill cancer patients admitted to ICU. A user-friendly online implementation is available and should be valuable for better mortality risk stratification. This might help physicians to determine to allocate ICU care for patients with cancer. Funding Information: This study was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. 2019R1C1C1006709, 018R1A5A2025079, 2020M3F7A1094093), a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, the Republic of Korea (No. KHIDIHI19C1015010020), and was also supported by the ational R&D Program for Cancer Control through the National Cancer Center (NCC) funded by the Ministry of Health & Welfare, Republic of Korea (HA21C0065). Declaration of Interests: The authors have no potential conflicts of interest to disclose. Ethics Approval Statement: The institutional review boards of all participating hospitals approved this study and waived the requirement for informed consent because of the observational nature of the research. All patient records and data were anonymized and de-identified before analysis.