Establishment And Validation Of Risk Prediction Model Of Integrated Traditional Chinese And Western Medicine For Intracerebral Hemorrhage In-Hospital Mortality

Objective:Acute intracerebral hemorrhage has the characteristics of high mortality and disability rate.Combining TCM treatment with Western medicine treatment is beneficial to improve the therapeutic effect.Syndrome differentiation is the basis of treatment.In-depth study of the distribution of syndrome elements in acute intracerebral hemorrhage,especially the distribution and combination of syndrome elements under different factors,is very important for improving the therapeutic effect of traditional Chinese medicine.At the same time,different TCM syndromes and other factors have certain predictive effects on the prognosis of patients,so it is of great benefit to explore the value of TCM syndromes and other factors in the prediction of in-hospital mortality risk of intracerebral hemorrhage,timely identify and actively intervene TCM related death risk prediction factors,and prevent small problems from developing gradually,so as to improve the prognosis of patients.At the same time,it is consistent with the action goal of reducing cardiovascular and cerebrovascular mortality in Healthy China and the thought of "preventing disease from curing existing disease and preventing degeneration" in Traditional Chinese medicine.However,there is still a lack of research on this aspect.Therefore,the distribution of TCM syndromes under different factors and the value of TCM characteristic syndromes,tongue and pulse in the prediction of in-hospital mortality risk need to be further explored.Methods:Study 1:A study on TCM syndromes of acute intracerebral hemorrhage in real world based on cluster analysis and correlation principleBased on intracerebral hemorrhage data warehouse,using cross-sectional study method,patients within 3 days of onset of acute intracerebral hemorrhage were included according to the criteria of intake,and the data extraction and governance,in view of the standardized data,the methods of cluster analysis and correlation principle,study the total population characteristics of TCM syndrome and syndrome factor distribution,And the distribution and combination characteristics of syndrome elements under different gender,age,bleeding site(supratentorial,supratentorial),whether breaking into ventricle or not,and different level of consciousness(GCS 3-8,9-12,13-15 points).Data mining was performed using SPSS26.0 and R software(4.1.1).Study 2:Study on establishing and verifying the risk prediction model of integrated Traditional Chinese and Western medicine for acute cerebral hemorrhage in-hospital mortalityBased on the intracerebral hemorrhage data warehouse,retrospective cohort study method was adopted to include patients within 3 days of onset of acute intracerebral hemorrhage,and data extraction and management were conducted.For standardized data,the occurrence of all-cause death in-hospital was taken as the dependent variable.Twenty-four alternative predictive variables,such as age,gender,GCS score,imaging indicators,TCM syndrome elements,tongue color,moss color,moss and pulse,were used as independent variables.Two modeling methods(traditional Logisitic regression and XGBoost)were used to construct and verify the in-hospital mortality prediction model of integrated Traditional Chinese and Western medicine for acute intracerebral hemorrhage.Results:Study 1:A study on TCM syndromes of acute intracerebral hemorrhage in real world based on cluster analysis and correlation principle1.1 Characteristics of syndrome and syndrome elements distribution in patients with acute intracerebral hemorrhageA total of 84 TCM syndromes were included in 1871 patients with acute intracerebral hemorrhage within 3 days after the onset.Among them,the syndrome of ganyangshangkang was the most(31.11%)).In acute stage,the majority of patients were excessive syndrome(90.18%).After reducing the dimension of TCM syndromes into disease location and disease syndrome elements,a total of 28 syndromes were involved,of which 9 syndromes were disease location syndrome elements,liver(48.82%)was the most common.There were 19 kinds of syndromes,and the most common was yang hyperactivity(47.97%).The association rule analysis found 31 union syndrome combinations,of which 17 were strongly correlated.The combination of "disease location and disease" syndrome elements:{phlegm,heart-shen}→{blockage} has the strongest correlation,and the combination of "disease and disease" syndrome elements:{fire.Heat.}→{wind-agitation},{blockage}→{phlegm}.There was no combination of syndrome elements with association rules.1.2 Distribution characteristics of TCM syndrome elements in different factors① The proportion of "wind-agitation" in males was higher than that in females(P=0.019).② Compared with patients younger than 65 years old,patients older than 65 years old had a higher proportion of syndrome elements of "blood stasis","meridian" and "Yin deficiency"(P<0.05).③Subatentorial hemorrhage accounted for a higher proportion of syndrome element blockage than supratentorial hemorrhage(P=0.031).④The proportion of "liver","hyperyang","wind-agitation" and "blockage"was higher in the group with hematoma breaking into ventricle.The proportion of"wind-agitation","blood stasis","meridians","qi deficiency" and "qi stagnation" in the group without hematoma breaking into ventricle was higher(P<0.05).⑤Syndrome elements "liver","Yang hyperactivity","wind-agitation","blood stasis","meridians","fire(heat)" and "qi deficiency" accounted for a higher proportion in GCS score of 13-15,while syndrome elements "phlegm","heart-shen" and "blockage"accounted for a higher proportion in groups 3-8.Cluster analysis also showed that there were differences in the combination of syndrome elements under different factors.Study 2:Study on establishing and verifying the risk prediction model of integrated Traditional Chinese and Western medicine for acute cerebral hemorrhage in-hospital mortalityThis study included 1403 eligible patients,113 of whom died in hospital.The data set was divided into training set and test set in a random split ratio of 7:3.982 patients in the training set established the model,and 421 patients in the test set verified the model internally.Based on Logistic and XGBoost algorithms,two prediction models were constructed and internally validated to predict in-hospital mortality from acute intracerebral hemorrhage.Seven predictive factors were included in the Logistic regression model,including age,gender,bleeding site,midline,GCS score,qi deficiency,chen mai,and Yang hyperactivity.XGBoost included 16 predictive factors,including GCS,age,chen mai,wind-agitation,bleeding location,hua mai,tongue cover,pulse pressure difference,entering ventricle,midline condition,qi deficiency,blockage syndrome,xi mai,gender,liver and phlegm.The algorithm also showed that chen mai and qi deficiency were the risk factors of in-hospital mortality.According to the principle that the greater the AUC is,the higher the degree of differentiation is,and the Brier score approaches 0,the prediction model of in-hospital mortality risk of acute intracerebral hemorrhage based on Logistic regression was better.In the test set,AUC was 0.844(0.774-0.913)and Brier score was 0.063.Age,bleeding location,male,midline and GCS score in Logistic regression prediction model of in-hospital mortality risk of acute intracerebral hemorrhage were the western medicine predictive factors of in-hospital mortality,which were consistent with internationally published research results.In traditional Chinese medicine,qi-deficiency(OR 4.513,95CI:1.204-14.618),chen mai(OR 2.172,95CI:1.099-4.32)and Yang hyperactivity(OR 1.527,95CI:0.773-3.073)were confirmed characteristic predictors of in-hospital mortality.According to The Youden index(0.68),the optimal truncation value of the total score of the rote chart is 208 points.That is,patients with scores greater than or equal to 208 are more likely to die in hospital,while patients with scores lower than 208 are more likely to survive.In this study,the strong predictive performance of XGBoost is not reflected,which may be related to the small sample size and the number of candidate predictive factors,limiting the advantages of XGBoost algorithm.Conclusion:1 In patients with acute intracerebral hemorrhage within 3 days of onset,and the common pathological factors were hyperactivity of Yang,phlegm,wind-agitation,blood stasis,and heat.And the occurrence of the disease was closely related to liver.Heat is easy to combine with wind-agitation,phlegm plays an important role in the occurrence and development of syndrome blockage.2 The distribution and combination of TCM syndrome elements in patients with acute intracerebral hemorrhage are different in different gender,age,bleeding site,breaking into ventricle and GCS score,suggesting that the above factors can provide some guidance for clinical syndrome differentiation and drug use3 The prediction model of in-hospital mortality of acute intracerebral hemorrhage showed the best prediction performance,risk factors for western medicine were age,male sex,subatentorial hemorrhage,midline deviation,and low GCS score,and the TCM risk factors of in-hospital mortality were qi deficiency,chen mai and Yang hyperactivity.This model provides a new tool to guide the prediction of death risk in hospital,and facilitates active intervention to improve patient prognosis.At the same time,it lays a foundation for the establishment of the early warning system,which is conducive to the promotion and application of the theory of preventing disease and preventing disease from changing,and also has certain value for promoting the prevention and control of major diseases and realizing the goal of healthy China.