神经系统疾病诊断相关分组研究——以河南省某三甲医院为例

目的 分析神经系统疾病患者的住院费用影响因素，建立该疾病的诊断相关分组（DRGs）模型，为DRGs实施和细化分组提供参考。方法 以河南省某三甲医院2013—2019年43 198例神经系统疾病患者住院病案首页数据为研究对象，分别用非参数检验、人工神经网络进行单因素、多因素分析，用决策树CHAID算法构建分组模型，用变异系数（CV）、ROC曲线和非参数检验对分组结果评价。结果 以是否手术（0.5090）、病例分型（0.2799）、有无其他诊断（0.1702）、入院途径（0.0217）4个重要因素作为分类截点，分成15个DRG组；模型评价结果是ROC曲线下面积均＞0.5，变异系数（CV）最大为0.98， P值均＜0.05，分组效果良好。结论 是否手术、病情危重、有无其他诊断、入院途径是神经系统疾病患者分组的重要因素，人工神经网络和CHAID算法相结合，建立的分组方案合理。

Diagnosis related groups of nervous system disease: a case study on a tertiary hospital in Henan Province

Objective To analyze the influencing factors of inpatient costs of patients with neurological diseases and to establish a diagnosis-related grouping(DRGs) model for this disease to provide reference for DRGs implementation and refinement of grouping. Methods This study collected the medical record front sheet data of 43 198 patients with nervous system disease from 2013 to 2019 in a tertiary hospital in Henan province. With non-parametric test, artificial neural network for single factor and multiple factors analysis, grouping model was constructed using CHAID decision tree algorithm. Nonparametric test method, ROC and coefficient of variation were used to evaluate result of grouping. Results Four important factors, whether to operate(0.5090), case typing(0.2799), presence of other diagnoses(0.1702), and route of admission(0.0217),were used as classification cutoffs and divided into 15 DRG groups. The results of model evaluation were that the area under the ROC curve was >0.5, and the maximum coefficient of variation(CV) was 0.98, with P values <0.05, and the grouping effect was good. Conclusion Surgery or not, critical condition, presence of other diagnoses, and admission route are important factors in grouping patients with neurological diseases, and the combination of artificial neural network and CHAID algorithm establishes a reasonable grouping scheme.