Study Of The Prediction Model For Frailty Of Elderly Inpatients With Cardiovascular Disease Based On Decision Tree

Objectives: In this study,the inpatients(age ≥ 60)with cardiovascular diseases in the cardiovascular department of a hospital in Shantou were selected as the objects.The decision tree algorithm in data mining technology is applied to construct the frailty risk prediction model for inpatients,which attempt to refine the decision factors of frailty risk and generate the clinical decision path.The frailty risk prediction model for inpatients aim to provide a new method for clinical nurses to quickly and accurately identify the risk of frailty in inpatients,so that nurses can shorten the assessment time,and take targeted fall prevention measures to reduce the incidence of adverse events in inpatients.Methods: Patients admitted to the Second Affiliated Hospital of Shantou University Medical College in the Department of Cardiovascular from August 1,2019 to August31,2020 were enrolled.Fried Scale was used to evaluate the inpatients.Meanwhile,medical records of inpatients were collected,including general information,medical evaluation,treatment programs,and laboratory biochemical tests and so on.Univariate analysis method was used to statistics significant variables.Combining with the previous frailty risk study,the input variables of the prediction model were established.The incidence of frailty in hospital was the output variable of the prediction model.The frailty risk prediction model was established by the CART decision tree model according to the radio of 7:3 between training set and test set.Finally,the prediction efficacy of prediction model was evaluated.Results:(1)The incidence of frailty in hospitalized patients(age≥60)with cardiovascular disease was 46%.(2)Univariate analysis of the elderly in the frail group and the non-frail group showed statistically significant differences in the factors(P<0.05): age,marital status,residential status,cardiac function grade,fall history,hospitalization times,CCI,type of medication,fall risk assessment,uric acid,albumin and other 11 risk factors;Gender,educational level,medical payment,sleep,history of smoking,drinking,basic life event scale(BADL),hospitalization days,body mass index(BMI),lactate dehydrogenase,hypersensitive c-reactive protein and creatinine,hemoglobin,total cholesterol,triglycerides,high-density lipoprotein cholesterol(hdl-c),low density lipoprotein did not differ between frailty group and non-frailty group.(3)Based on the decision tree algorithm in R language,the frailty risk prediction model of elderly inpatients with cardiovascular diseases was constructed.The results showed that the decision tree had 12 leaf nodes with a depth of 6 layers,and the conditions for the division of this node were indicated at both root node and internal node.The risk factors finally inputted into the tree model were cardiac function grade,age,hospitalization times,albumin,type of medication,uric acid,etc.,and 12 combination rules were generated.Five variables with statistical significance in univariate analysis,including marital status,residential status,fall history,CCI,and fall risk assessment,were not included in the final CART decision tree risk prediction model.Conclusion: The frailty risk prediction model of elderly inpatients with cardiovascular diseases based on CART decision tree algorithm has a good prediction efficiency.Cardiac function grade,age,hospitalization times,albumin,type of medication and uric acid are the decision-making factors for the screening of frailty risk prediction model for elderly inpatients with cardiovascular diseases,among which cardiac function grade is the most important decision-making factor.The prediction modelcan quickly screen out patients with frailty,in order to provide patients with accurate and effective measures to prevent frailty,so can reduce adverse outcomes.