Disease Comorbidity Analysis Of Large-scale Inpatient Medica Records

The occurrence and development of disease is a complex physiological and pathological process.Because the human body is an organic,there are complex interactions at the levels of molecules,cells,and organs.Therefore,the disease comorbidity(refers to the presence of two or more types of diseases in one patient)is common in clinical practice and have important medical research value.In recent years,the international research on disease comorbidity has caused a wide upsurge,but due to the complexity of disease,people's understanding of disease comorbidity still need to be further explored.Besides,because of the impact of environment,ethnic,social environment and other factors,it is of great significance to the study of large population diseases in China.It can not only provides a deeper understanding of disease,but also prevents the occurrence of comorbid diseases while treating primary diseases.This thesis uses a large-scale data from medical record homepage to construct a larger-scale disease comorbidity network,which integrates the relationship between disease molecules(genes,pathways,and syndromes),conducts macro and micro correlation studies of disease comorbidity.At the same time,based on the patient data of different time points,the prediction of disease is made by the evolution relationship between them.The main findings include the following three aspects:(1)Using the data from the medical records of 453 hospitals collected by the China Academy of Chinese Medical Sciences,a correlation analysis method was used to construct a disease comorbidity network(5702 nodes and 258,535 edges)with a significant comorbidity relationship.Base on,the analysis of the network topology shows that the degree distribution of this network is consistent with the power law distribution and is a scale-free network,indicating a high degree of heterogeneity among different diseases.For example,hypertension has 1926 comorbidities and is at the center of the network,but the comorbidity of choroidal diseases is rare.At the same time,we find that the network is a hierarchical modular network with a significant community structure(module degree of 0.302).(2)Aiming at the macro-and micro-correlation problems of the disease comorbidity relationship,combining disease-related genes,pathways,syndromes,and other data,through a variety of similarity calculation methods and analysis,it is found that the disease intensity has significant positive correlations with molecules between disease share and similarity of syndromes,indicating that the greater the number of shared molecules between diseases or the greater the similarity of disease symptoms,the greater the likelihood that the diseases will form a merging relationship.Combined with literature review,we have focused on the analysis of several disease associations with important clinical value,for example,Alzheimer's disease and arteriosclerotic heart disease(RR=2.585,?=0.0166,shared genes:ACE,APOE,and NOS3).(3)Study on disease progression based on the relationship between the evolution of diseases at the point in time:According to the sequence of the patient's illness,it is predicted that when the patient has certain diseases,it will not trigger a specific disease.Based on the standard datasets of hypertension(20000)and mental illness(7000),we used logistic regression,support vector machine(SVM),random forest and neural network classification models to conduct disease prediction analysis.Initially found the risk factors and protective factors of the two target diseases.And by comparing the experimental analysis,the random forest was better than other models,and it's F1 measure on the two datasets reached 0.6689 and 0.6802,respectively.The results can provide a basis for the early diagnosis of target diseases and optimize the diagnostic process.