论文部分内容阅读
[目的]运用人工神经网络数学模型,设定合适的参数,建立基于BP神经网络的住院费用拟合模型,并在已建立的BP神经网络模型的基础上,进行影响因素的敏感度分析,探讨各因素对住院费用的影响程度。[方法]收集唐山市某三级甲等医院提供的2007～2008年脑梗死患者病案中的住院费用及其相关信息,建立Excel数据库,应用SAS8.2软件进行单因素分析,采用MathWorks公司的Matlab7.1软件建立BP神经网络模型并进行敏感度分析。[结果]单因素分析结果表明:性别、抢救与否、付费方式、住院天数等对住院费用都有明显影响(P﹤0.05)。敏感度分析显示:各影响因素的敏感度依次为住院天数(0.805 05)、治疗结果(0.761 36)、抢救与否(0.726 83)、年龄(0.474 49)、首次住院(0.471 05)、婚姻状况(0.436 65)、付费方式(0.33753)、性别(0.069 56),其中对住院费用影响最大的因素是住院天数,最小的是性别。[结论]影响住院费用最主要的因素是住院天数,应从各方面努力以缩短患者的住院天数,从而降低患者的住院费用。
[Objective] The research aimed to establish a fitting model of hospital expenses based on BP neural network by setting up appropriate parameters by using artificial neural network mathematical model and analyzing the sensitivity of influencing factors on the basis of established BP neural network model The impact of various factors on the cost of hospitalization. [Methods] The hospitalization expenses and related information of patients with cerebral infarction in 2007 ~ 2008 provided by a tertiary level Hospitals in Tangshan City were collected. Excel database was set up and uni-factor analysis was performed with SAS8.2 software. MathWorks Matlab7 .1 software to build BP neural network model and sensitivity analysis. [Results] The results of univariate analysis showed that gender, salvage or not, payment method, hospitalization days and so on had a significant effect on hospitalization costs (P <0.05). Sensitivity analysis showed that the sensitivities of each factor were hospitalization days (0.805 05), treatment outcome (0.761 36), rescue or not (0.726 83), age (0.474 49), first hospitalization (0.471 05), marital status (0.436 65), payment method (0.33753) and sex (0.069 56). Among them, the most significant factor affecting the hospitalization expenses was the hospitalization days and the smallest was gender. [Conclusion] The most important factor influencing hospitalization costs is the length of hospitalization. Various efforts should be made to shorten the hospitalization days and reduce the hospitalization costs of patients.