The study of cancer patients hospital costs based on principal component analysis and BP neural network combination modelAuthor(s): Wu Jianhui, Xue Ling, Hu Bo, Yin Sufeng, Wang Guoli
To collect totally 2340 cases of patients’ hospital costs and related information in June 2009-March 2011 in a 3A-grade hospital’s Surgical Oncology of Tangshan City. Gender, age, occupation, marital status, number of admission, admission illness, payment methods, surgical cases, secondary diagnosis, length of stay and treatment outcome are reduced dimensionality and denoising by Principal component analysis, BP neural network model was built between the selected principal component score matrix which is as input variables and hospital costs which is as output variables, and on the basis of the built model, the factors of hospital costs were analyzed by sensitivity analysis. The results showed that 8 Principal components were selected, and the cumulative contribution rate reached to 82.48%, Using a Bayesian algorithm, optimal BP neural network model was built basing on that the number of hidden layer neurons is 5, sensitivity analysis results showed that the top three influence factors on the costs of hospitalization were age, the number of days in hospital and treatment results. By this study, it was founded that using principal component analysis and BP neural network model to analyze the influence factors on the costs of hospitalization is feasible, and the hospital costs may be controlled by improving hospital efficiency, strengthening medical quality management and shortening the number of days in hospital appropriately.