The Data Mining Analysis Of Professor Ni Zhuying's Treatment Of Pediatric Cough Disease

Objective In this paper,we used the data mining methods such as complex network,R language,rank sum test,permutation test,etc.to retrospectively study Ni Zhuying's medical treatment for cough in children.This study summarizes Professor Ni Zhuying's medication rules for treating children with cough.And we studied the relationship between the first-diagnosis symptom network of cough and its efficacy.It provides a reference for the clinical use of Chinese medicine for the treatment of cough in children and provides ideas for more accurate and effective treatment.Method This study randomly selected cough cases in children by Professor Ni Zhuying according to the diagnostic criteria for children with cough.We made pediatric cough medical record template.Then recruited medical graduate students to enter case information into the electronic case system.We used Extract-Transformation-Loading support platform to preprocess the data entered,and use structured query language tools to count the frequency of symptoms and medications.Complex network analysis was used to analyze the compatibility of core prescriptions and commonly used drugs.A total of 987 cases were screened for review,and the original data were evaluated for efficacy evaluation.There were 344 medical visits in the valid group and 40 visits in the invalid group.Then we compared the efficiency and chi-square value of the core prescription drugs and commonly used drugs.Then authors used the network's four global indicators,global strength,average shortest path,network diameter and network density,to measure the overall connectivity of the two related networks.The Wilkins rank sum test was used to test the general differences between the two groups of data.A network of correlations and symptoms between two sets of data was constructed using the qgraph package in the R language.The centrality of the symptoms in the network was studied using four local central indicators,which were called strength,closeness,betweenness,eigenvector.Cohen's coefficient was used to calculate the average central effect size of each symptom,and to analyze which of the symptoms had the greatest difference in the two correlation networks.Results A total of 568 children were enrolled in this study,326 male and 242 female.There were 1640 medical visits.The youngest child is 25 days,the oldest child is 14 years old and 6 months old.There are 267 flavors of drugs used for pediatric cough by Professor Ni Zhuying.The highest frequency of the top 12 herbs were tangerine peel,fried bitter almond,honey ephedra,melon cicada skin,peony,bran clam shell,half law,white cicada,skullcap,Houttuynia,Uncaria,bran fried silkworm,bellflower,honeysuckle.Complex network software processing analysis showed that the basic components were honey ephedra,fried bitter almonds,bran fried clam shell,peony,melon cicada skin,dried tangerine peel,Pinellia tuberculata,Angelica.Commonly used drugs were wolfberry,rhubarb,turmeric,mulberry white skin,vinegar,peel,ground bone skin,astragalus,wolfberry,bran fried Atractylodes,Chuanxiong,gardenia,Asarum,Zhejiang Fritillaria,Scrophulariaceae,bran fried Divine Comedy,Fried hawthorn,ghosts,arrows,ground dragons and so on.In the chi-square results for the relationship between drugs and efficacy:P <0.05 of the drugs are Angelica,Gardenia,fried Perilla.There were 7(87.50%)of the core drugs with P>0.5,which were Pinellia ternate,fried bitter almond,honey ephedra,melon peel,bran fried clam shell,prune,orange peel.There were 7 common medicinal herbs with P> 0.5(26.92%),namely Fritillaria cirrhosa,Houttuynia cordata,Fructus acerifolia,Astragalus membranaceus,Astragalus bran,Fry Divine Comedy,and Ape.There were 14 drugs(53.85%)of commonly used drugs with P<0.3.The symptom network of poorly effective group was more closely related and positively related.The global strength,the average shortest path,the network diameter and the network density of the invalid group were significantly greater than in the effective group.In those four local central indicators,the symptom scores of antiadoncus,coughing sounds such as canine warts and expectoration were higher.The maximum effect amount of effect was cough(3.21),constipation(3.03),nasal discharge(2.88),pulmonary breath sounds rough(2.81),anorexia(2.61),antiadoncus(2.51),pharyngeal red(2.30),nasal obstruction(2.20),sneezing(2.03).Conclusion Professor Ni treatment of pediatric cough is mainly from the sputum,from the heat,from the gas on the treatment,pay attention to individual differences in children,syndrome differentiation.The more complex symptoms network predicts a poor prognosis.It is possible to obtain better results by intervening in symptoms with large effects.It can provide more reference to the treatment of clinical pediatric cough disease.