基于关联规则的病案首页数据挖掘

Data Mining of Front Pages of Medical Records Based on Association Rules

目的利用Apriori算法找到出院患者信息各个指标中的关联规则,为医院管理和决策提供理论依据。方法利用R软件中的arules包对2015年某院出院患者做Apriori关联分析,探索出院科室与性别,费别、出院科室、住院天数与总费用,费别、出院科室与是否手术的关联规则,并分析其原因。结果对2015年出院患者共计49 737条病案首页记录抽取字段进行分析后,得到以下的关联规则:呼吸病区、消化病区、普外病区男性出院人数要多于女性出院人数,其强关联规则的置信度分别为0.621、0.531、0.518;神内病区、眼科病区女性出院人数要多于男性出院人数,其强关联规则的置信度分别为0.565、0.561;不同科室的医保出院患者的住院费用与住院天数有密切关系,其强关联规则的置信度分别为0.731、0.649、0.745、0.545;是否采用手术治疗与科室存在密切关系,其中强关联规则的置信度分别为0.951、0.748、0.985、0.974、0.735。结论关联规则方法可以探索不同指标的潜在关联规则,为医院的管理决策和方针制定提供依据。

Objectives To find the association rules of each index of discharged patients’information in the use of Apriori algorithm, provide a theoretical basis for hospital management and decision making. Methods Apriori correlation analysis was conducted on discharged patients in 2015 with the application of R software, to explore gender department and hospital, medical treatment, hospital departments, hospitalization days and total expenses, medical treatment, hospital departments and association rules whether the operation, and analyzed its causes. Results After the field analysis on the front pages of medical records of 49737 cases of patients discharged in 2015, we found the rules below：the discharged number in respiratory ward, digestion ward, general surgery ward, male were more than female patients, and the confidence of the strong association rules were 0.621, 0.531,0.518;in neurology ward and ophthalmology ward, female were more than male in discharged patients, and the confidence of the strong association rules were 0.565, 0.561;health care hospital hospitalization expenses was closely related with the duration of hospitalization, and the confidence of the strong association rules were 0.731、0.649、0.745、0.545;whether to adopt surgical treatment and there was a close relationship between departments, and the confidence of the strong association rules were 0.951、0.748、0.985、0.974、0.735. Conclusions The potential association rules of association rules could explore different indicators, and provide the basis for hospital management and policy decision.