Length of hospital stay for liver cancer patients: risk factors and predictive model construction

Abstract

Purpose: In the context of rapid rehabilitation, the length of hospital stay can serve as a reflection of the overall diagnosis and quality of nursing care provided. However, the factors influencing the duration of hospitalization in patients with primary liver cancer are currently not well understood. This research aimed to develop a predictive model for evaluating the length of hospital stay of patients. Methods: Admitting information of patients with liver cancer and undergoing surgical treatment were included in this study. This research analyzed 31 indicators. A binary logistic regression model was constructed with the length of hospital stay greater than the median as the dependent variable and presented in a line chart format. The performance of the line chart was tested through ROC curve, calibration plot, and decision curve analysis. Furthermore, the model underwent internal validation by utilizing the validation dataset. Results: This study included a total of 966 patient data. The research cohort was randomly divided into a training set and a validation set in a 7:3 ratio. Multifactor logistic regression analysis showed that factors such as Cholinesterase are predictive factors for prolonged hospital stay in patients with primary liver cancer. The nomogram model constructed using these factors demonstrated good consistency and accuracy. The AUC of the prediction model and internal validation set were .66 (95% CI .61-.70) and .56 (95% CI .49-.63) respectively. Hosmer-Lemeshow test values were p = .857 and p = .590. The calibration curve showed significant consistency between the nomogram model and actual observations. ROC and DCA indicated that the nomogram has good predictive performance. Conclusion: The model serves as a valuable tool for healthcare professionals to anticipate the risk factors associated with extended hospital stays in patients diagnosed with primary liver cancer.