基于诊疗知识库和电子病历的整合推荐方法研究

A hybrid recommendation method based on medical knowledge base and electronic medical records

目的:探索一种基于临床诊疗知识库和电子病历为患者推荐诊疗相关知识的有效方法。方法:引入诊疗知识库的内容过滤结合电子病历建立整合评分矩阵,融合基于电子病历的患者协同过滤(PbCF)和疾病诊断的协同过滤(DbCF)结果,构建整合推荐方法,减少预测结果的误差。以平均绝对百分误差(MAPE)作为评价指标,对整合推荐方法与独立PbCF、DbCF预测结果进行评估比较。结果:整合推荐方法的MAPE要小于PbCF和DbCF。局限于所选诊疗知识库的高血压疾病种类覆盖面不及电子病历中高血压患者所涉及的疾病种类,扩大诊疗知识库疾病覆盖面对推荐结果的影响仍需进一步探讨。结论:整合推荐方法比独立的PbCF、DbCF效果更好,为患者推荐疾病诊疗知识,具有良好的应用前景。

Objective To explore an effective method for providing recommendation knowledge related to diagnosis and treatment for patients based on medical knowledge base and electronic medical records. Methods The content filtering method of medical knowledge base was combined with electronic medical records to establish an integrated scoring matrix. The fusion of the results obtained with patient-based collaborative filtering and diagnosis-based collaborative filtering was used to construct a hybrid recommendation method, so as to reduce the error of prediction results. With mean absolute percentage error as the evaluation index, the prediction results obtained with hybrid recommendation method, patient-based collaborative filtering and diagnosisbased diagnostic collaborative filtering were evaluated. Results The average absolute percent error of hybrid recommendation method was lower than that of patient-based collaborative filtering and diagnosis-based collaborative filtering. The limitation of this study is that the types of hypertensive diseases in medical knowledge base were less than those in electronic medical records, and that the effects of the disease coverage on the recommendation results remained to be further explored. Conclusion The hybrid recommendation method which is superior to patient-based collaborative filtering and diagnosis-based diagnostic collaborative filtering can be used to provide medical knowledge for patients, having a good application prospect.