Analysis Of Influencing Factors Of Frailty In Patients With Rheumatoid Arthritis And Construction And Validation Of Prediction Model

Objective1.Understand the frailty status of RA patients;2.Reveal the influencing factors of frailty in RA patients;3.development and validation the frailty prediction model for RA patients.MethodsIn this study,283 RA patients in the rheumatology and immunology Department of two3 A hospitals in Anhui Province from March 2022 to December 2022 were selected as investigation objects through cross-sectional study and convenience sampling method.The general Information Questionnaire,the Chinese version of the Tilburg Frailty Indicator(TFI),the Self-rating Despression Scale(SDS),social support rate scale(SSRS),Stanford health assessment questionnaire disability(HAQ-DI),Disease activity score in 28 joint counts(DAS28),Visual Analogue Score(VAS),Pittsburgh sleep quality index(PSQI),Charlson comorbidity index(CCI),UCLA Loneliness Scale and36-Item Short Form Health Survey(SF-36)were conducted to investigate RA patients.General demographics,disease-related factors and psychosocial factors were analyzed using R 4.0 software and SPSS 23.0 software.Lasso regression was performed using R4.0 software to screen out the risk prediction model variables of RA patients with frailty.Logistic regression,decision tree and random forest models were used to construct and verify the prediction models respectively,and the best models were selected by comparison.Results(1)Frailty status of RA patients: The average score of frailty scale of 283 RA patients was 6.36±2.889.According to the Chinese version of the TFI scale,among 283 patients with RA,188(66.4%)were frailty and 95(33.6%)were non-frailty.(2)Factors influencing frailty in RA patients: Among the general demographic factors,age,family location,marital status,residence style,education level,occupation,monthly income,payment method,smoking,family smoking,daily intake of fruit,physical activity,mental training and social activities were statistically significant differences between the frailty group and non-frailty group(P<0.05);Among disease-related factors,age at diagnosis,number of medications,walking AIDS,presence of other chronic diseases,number of chronic diseases,number of deformities,joint swelling,tenderness of joints,CRP,functional limitation(HAQ-DI),DAS28 and VAS were significantly different between the two groups(P<0.05);Among the psychosocial factors,depression,sleep disorder and loneliness were significantly different between the two groups(P<0.05).(3)Construction and verification of RA frailty prediction model: All independent variables are included in LASSO regression in this study,and 18 meaningful independent variables are finally selected,which are: Age,residence style,education level,occupation,smoking,frequency of fish intake,physical activity,mental training,social activities,family history of RA disease,number of chronic diseases,morning stiffness,number of malformation,functional limitation,SSRS,SDS,PSQI,UCLA.The variables selected by LASSO regression were used to construct the prediction model in the training set by Logistic regression.The variables finally selected were:physical activity,morning stiffness,social activity,functional limitation,number of deformities,depression and sleep disorders.In the training set,the AUC value was0.935,the accuracy was 0.838,the sensitivity was 0.794,the specificity was 0.925,the positive predictive value was 0.954,and the negative predictive value was 0.697.Among the models in the decision tree,the importance order of the models is as follows:functional limitation,physical activity,age,intelligent training,social activity,depression,sleep disorder,living style,loneliness,education level,frequency of fish intake,social support and occupation.The last variables entering the decision tree model are functional limitation,sleep disorder,loneliness,living style and education level.In the training set,the AUC value was 0.914,the classification accuracy was0.874,the sensitivity was 0.891,the specificity was 0.843,the positive predictive value was 0.912,and the negative predictive value was 0.808.In the random forest model,the relative importance of variables in the process of fitting the training set data was ranked as functional limitation,social activity,physical activity,sleep disorder,depression,intelligent training,loneliness,occupation,social support,morning stiffness,number of deformities,family history of RA disease,education level,frequency of fish intake,number of chronic diseases,age,residence style,smoking.In fact,the variables for random forest model construction are functional limitations,social activities,physical activities,sleep disorders,and intelligent training.In the training set,the AUC value was 0.921,the classification accuracy was 0.843,the sensitivity was0.820,the specificity was 0.886,the positive predictive value was 0.929,and the negative predictive value was 0.729.(4)Validation of RA frailty prediction model: the AUC value of Logistic regression model was 0.909,sensitivity was 0.783,specificity was 0.96,positive predictive value was 0.979,negative predictive value was 0.649,classification accuracy was 0.835;AUC value of decision tree model was 0.785,sensitivity was 0.65,specificity was 0.88,positive predictive value was 0.929,negative predictive value was 0.512,and classification accuracy was 0.718.The AUC value of random forest model was 0.876,the sensitivity was 0.666,the specificity was 1,the positive predictive value was 1,the negative predictive value was 0.556,and the classification accuracy was 0.765.Conclusion(1)The frailty prevalence rate of RA patients was as high as 66.4%.(2)Univariate analysis found age,family location,marital status,residence style,education level,occupation,monthly income,payment method,smoking,family smoking,daily fruit intake,physical activity,mental training and social activities,age at diagnosis,number of medications,walking AIDS,other chronic diseases,number of chronic diseases,number of malformation,tenderized joints,CRP,functional limitation,DAS,VAS,depression,sleep disorders,loneliness are the factors affecting the debilitation of RA patients.(3)Logistic regression model is superior to decision tree and random forest model.Therefore,physical activity,morning stiffness,social activity,functional limitation,number of deformities,depression,and sleep disturbance are predictors of frailty in RA patients.