摘要
目的以研究医保住院费用报销的合理性为目的,分析慢性传染性疾病门诊费用结构、门诊治疗情况及患者静态特征等因素对住院费用的影响,并通过数据挖掘算法构建住院费用的预测模型。方法在构建模型的过程中,分别使用了决策树、Naive Bayes、Adaboost等算法,并比较了这些算法的准确度。结果以上采用的模型算法可以使得预测模型的准确度达到93%以上,能够较好地预测住院费用的合理性。肾衰竭/尿毒症的门诊特病费用与住院费用呈负相关趋势(r=-0.10,P<0.05),高血压、冠心病、糖尿病、脑血管病均呈正相关趋势(r=0.11、0.10、0.12、0.11,P<0.05)。结论根据不同类别疾病的特点,可以调整部分住院治疗项目到门诊,以实现对住院费用的调控。
Objective In order to analyze the rationale of the hospitalization cost,investigating the impact of outpatient cost structure,outpatient treatment situation,and static characteristics of patients on the hospitalization cost by establishing a data-mining model to predict the hospitalization cost.Methods In the process of building the model,the algorithms of decision tree,Naive Bayes,Adaboost were applied,and the accuracy of these above-mentioned algorithms were compared.Results The model could achieve an accuracy above 93%in the prediction of the hospitalization cost.The negative correlation existed between the outpatient fees for special diseases of renal failure/uremia and the hospitalization cost (r=-0.10,P<0.05),hypertension,coronary heart disease,diabetes,cerebrovascular disease showed positive correlation trend (r=0.11,0.10,0.12,0.11,P<0.05).Conclusion According to the characteristics of different diseases types,some inpatient treatment items can be adjusted to outpatient service to realize the regulation and control of the hospitalization cost.
引文
[1]阎正民,毛正中.城市慢病费用灰色预测及分析[J].中国卫生事业管理,1997,13(2):93-95.
[2]王敏,张开金,姜丽,等.BP神经网络技术在慢性病患者住院费用研究中的应用[J].中国卫生经济,2010,29(11):86-88.
[3]宋蕾,张开金,包思敏,等.不同支付方式下慢性病患者住院费用对比分析[J].中国全科医学,2013,16(4):440-441.
[4]陶阳红,张开金,陈积瑞,等.参加城镇职工基本医疗保险的慢性病患者住院费用的影响因素调查[J].中国全科医学,2015,18(7):811-813.
[5]余震.杭州市基本医疗保险住院费用分析[D].杭州:浙江大学,2013.
[6]陶红兵.基于单病种管理的医院住院病人费用控制的管理策略研究[D].武汉:华中科技大学,2008.
[7]何倩,吴骋,王志勇,等.成分数据预测模型在住院费用构成中的应用[J].中国卫生经济,2013(3):27-29.
[8]高金金,周伟,高咏,等.子宫肌瘤单病种住院费用[J].协和医学杂志,2016,7(2):152-155.
[9]杨金娟,高建民,许永建,等.恶性肿瘤化疗患者住院费用构成及影响因素分析[J].中国卫生经济,2014,33(9):35-38.