Machine learning to predict 30-day quality-adjusted survival in critically ill patients with cancer

• Published in: Journal of critical care Vol. 55; pp. 73 - 78
• Main Authors: Santos, Hellen Geremias dos, Zampieri, Fernando Godinho, Normilio-Silva, Karina, Silva, Gisela Tunes da, Lima, Antonio Carlos Pedroso de, Cavalcanti, Alexandre Biasi, Chiavegatto Filho, Alexandre Dias Porto
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.02.2020; Elsevier Limited
• Subjects: Adult; Aged; Algorithms; Area Under Curve; Brazil - epidemiology; Cancer; Chemotherapy; Clinical decision making; Critical care; Critical Illness; Critically ill patients; Decision Trees; False Positive Reactions; Female; Hospitalization; Hospitals, Public; Humans; Intensive care unit; Intensive Care Units; Leukemia; Machine Learning; Male; Medical prognosis; Middle Aged; Neoplasms - diagnosis; Neoplasms - mortality; Neoplasms - therapy; Palliative care; Pattern Recognition, Automated; Probability; Prognosis; Prospective Studies; Quality of Life; ROC Curve; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Brazil; Prognosis; Clinical decision-making; Intensive care unit; Machine learning; Critically ill patients; Quality of life
• ISSN: 0883-9441; 1557-8615; 1557-8615
• DOI: 10.1016/j.jcrc.2019.10.015

To develop and compare the predictive performance of machine-learning algorithms to estimate the risk of quality-adjusted life year (QALY) lower than or equal to 30 days (30-day QALY). Six machine-learning algorithms were applied to predict 30-day QALY for 777 patients admitted in a prospective cohort study conducted in Intensive Care Units (ICUs) of two public Brazilian hospitals specialized in cancer care. The predictors were 37 characteristics collected at ICU admission. Discrimination was evaluated using the area under the receiver operating characteristic (AUROC) curve. Sensitivity, 1-specificity, true/false positive and negative cases were measured for different estimated probability cutoff points (30%, 20% and 10%). Calibration was evaluated with GiViTI calibration belt and test. Except for basic decision trees, the adjusted predictive models were nearly equivalent, presenting good results for discrimination (AUROC curves over 0.80). Artificial neural networks and gradient boosted trees achieved the overall best calibration, implying an accurately predicted probability for 30-day QALY. Except for basic decision trees, predictive models derived from different machine-learning algorithms discriminated the QALY risk at 30 days well. Regarding calibration, artificial neural network model presented the best ability to estimate 30-day QALY in critically ill oncologic patients admitted to ICUs. •Five of the six machine-learning algorithms presented good discrimination to stratify QALY risk at 30 days.•Tree based ensemble methods offered substantially greater predictive performance compared to basic decision trees.•Neural networks and gradient boosted trees achieved the best calibration for predicted 30-day QALY probability.