Purpose: To develop and validate a user-friendly model to predict the risk of in-hospital mortality in solid cancer patients admitted to the ICU with sepsis.
Methods: Clinical data of critically ill patients with solid cancer and sepsis were obtained from Medical Information Mart for Intensive Care-IV database and randomly assigned to the training cohort and validation cohort. The primary outcome was in-hospital mortality. The least absolute shrinkage and selection operator (LASSO) regression and logistic regression analysis were used to feature selection and model development. The performance of the model was validated and a dynamic nomogram was developed to visualize the model.
Results: A total of 1584 patients were included in this study, of whom 1108 were assigned to the training cohort and 476 to the validation cohort. The LASSO regression and logistic multivariable analysis showed that nine clinical features were associated with in-hospital mortality and enrolled in the model. The area under the curve of the model was 0.809 (95% CI 0.782-0.837) in the training cohort and 0.770 (95% CI 0.722-0.819) in the validation cohort. The model exhibited satisfactory calibration curves and Brier scores in the training set and validation set were 0.149 and 0.152, respectively. The decision curve analysis and clinical impact curve of the model presented good clinical practicability in both the two cohorts.
Conclusion: This predictive model could be used to assess the in-hospital mortality of solid cancer patients with sepsis in the ICU, and a dynamic online nomogram could facilitate the sharing of the model.
Keywords: ICU; Mortality; Prediction; Prognosis; Sepsis; Solid cancer.
© 2023. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.