Research on Basic Clinical Treatment Pattern Mining Based on Electronic Medical Record Big Data

This paper analyzed the basic clinical treatment pattern by exploring the relationship between diseases, symptoms and drugs, which help non-medical people understand the basic clinical treatment pattern, so as to better carry out medical and health big data mining and eliminate public prejudice against symptomatic treatment. The FP growth algorithm was used to mine the association rules from EMR big data. Combined with intersection analysis, the basic clinical treatment pattern was summarized. 507 disease-drug rules and 2141 symptom-drug rules were obtained, indicating that both diseases and symptoms are strongly associated with drugs. Intersection analysis showed that 33.7% of the disease-drug rules were symptom-independent, while 34.6% of the symptom-drug rules were disease-independent. The basic clinical treatment pattern consists of three parts: (1) The combination of disease and symptomatic medication pattern. (2) Independent disease medication pattern. (3) Independent symptomatic medication pattern.

I. Introduction

The advent of the big data era has brought great changes to traditional medicine [1]. The hidden rules mined in medical big data provide support for the diagnosis and treatment of diseases, and promote the development of medical health [2]. Data-driven artificial intelligence technology has a far-reaching impact on the medical system. Mining information from data can assist clinicians to conduct image analysis, assist diagnosis and predict disease risks, which is of great help to reduce medical errors, improve efficiency and improve services [3]. However, the current diagnostic rules mining research based on electronic medical records often starts from the relationship between disease and symptoms, and the treatment rules mining research starts from the relationship between disease and medication [4]. This will lead to the status quo of the separation of symptoms and medication, which does not accord with the fact that doctors often use medication according to symptoms in clinical practice. Many scholars have insufficient understanding of basic clinical diagnosis and treatment pattern, which will affect the effective development of medical health big data mining research. In addition, in people's common understanding, drugs are often targeted at the disease, symptomatic treatment is not so good. However, in clinical practice, symptomatic treatment is also a very important part, especially in the case of poor treatment of the disease. The drug strategy to alleviate symptoms can greatly improve the quality of patients' life. Bias of symptomatic treatment can reduce patients' compliance with symptomatic treatment. Therefore, it's necessary to use the big data of electronic medical records and the method of association rules to explore basic clinical treatment pattern by analyzing the relationship between diseases, symptoms and drugs. It can help health big data mining by improving the understanding of the basic clinical treatment pattern, and change patients' prejudice towards symptomatic medication to improve treatment compliance.