Development Of Prediction Model For Frailty-Prefrailty Elderly In Chengdu Community

Objective1.To investigate the status quo and risk factors of frailty-prefrailty among elderly in Chengdu Community.2.To develop a prediction model for predicting frailty-prefrailty elderly in Chengdu community,which help primary health care providers to screen the community-dwelling elderly at high risk of frailty-prefrailty,and take targeted prophylactic measures early.MethodsThe cross-sectional design was used in this study.From November 2017 to June 2018,a multistage stratified random sampling method was used to select five communities from eastern,western,southern,northern and central Chengdu,the five communities were Huachuan Community,Beijie Community,Shuangyuan Community,Hehuachi Community and Huapaifang Community.According to the method of estimate the required sample size,for each candidate predictor studied require at least 10 events,the prevalence of frailty in Chengdu community is estimated as 10%and the risk factors of the final model are estimated 5-7,in addition consider 10%of the invalid samples,so at least 550 elderly are needed,and each community should collected at least 110 data of elderly.The self-designed questionnaire,Charlson Comorbidity Index(CCI),Pittsburgh Sleep Quality Index(PSQI),Activity of Daily Living Scale(ADL),Mini Nutritional Assessment Short Form(MNA-SF),Faces Pain Scale Revised(FPS-R),Self-Rating Anxiety Scale(SAS),Geriatric Depression Scale-15(GDS-15)and Social Support Rating Scale(SSRS)were used to collected risk factors among 30 known risk factors of frailty,the data was used to develop the prediction model.The significant(P<0.05)predictors in the univariate analysis entered the Logistic regression model,backward LR Logistic regression(P<0.05 for entry and P<0.1 for retention in the model)was performed to select significant independent predictors of frailty-prefrailty.The Hosmer-Lemeshow(H-L)test was used to evaluate the calibration,and the area under the receiver operating characteristic(ROC)was used to evaluate the discrimination.ResultsA total of 550 questionnaires were distributed in this study,545 questionnaires were returned,the return rate was 99.1%,of which 526 were valid questionnaires,and the effective return rate was 96.5%.1.FrailtyThe prevalence of frailty,pre-frailty and robust in 526 elderly people was 9.3%(n=49),45.4%(n=239)and 45.2%(n=238)respectively.The number of frailty phenotype items were weakness(grip strength)154 cases(29.3%),self-reported exhaustion 109cases(20.7%),slow walking speed 101 cases(19.2%),low physical activity 66 cases(12.5%),and unintentional weight loss 31 cases(5.9%).2.Univariate analysisUnivariate analysis showed that 17 candidate predictor variables were significantly associated with frailty-prefrailty in community-dwelling elderly,they were age,education,occupation,CCI,multimorbidity,polypharmacy,stress exposure in prior year,fall in prior year,use of walking devices,self rated health,sleep disorder,ADL decline,malnutrition,pain,anxiety,depression,and social support(P<0.05).3.Logistic regressionA total of 8 significant variables were derived in the final prediction model.They were occupation(OR=0.680,95%CI0.460~1.007,P=0.054),polypharmacy(OR=3.543,95%CI1.442~8.071,P=0.006),use of walking devices(OR=11.434,95%CI1.439~90.874,P=0.021),self rated health(OR=1.974,95%CI1.441~2.704,P<0.001),malnutrition(OR=3.543,95%CI1.442~8.071,P=0.053),sleep disorder(OR=1.508,95%CI1.011~2.248,P=0.044),ADL decline(OR=3.580,95%CI 0.986~13.007,P=0.014)and social support(OR=0.951,95%CI0.923~0.979,P=0.001).The model formula was:Z=1.265×polypharmacy+2.437×use of walking devices+0.680×self rated health+0.410×sleep disorder+1.275×malnutrition+1.059×ADL decline-0.385×occupation-0.05×social support.4.Performance of modelThe Hosmer-Lemeshow test showed that the model demonstrated a good cali-bration(χ~2=6.828,P=0.555>0.05);the analysis showed an area under the receiver operating characteristic curve(AUC)of 0.749(95%CI 0.708~0.790,P<0.001),the model demonstrated a good discrimination.The optimal cut-off for frailty-prefrailty was 0.468with the Yoden index of 0.365,sensitivity and specificity were 72.6%and 63.9%.ConclusionThe study used cross-sectional data of elderly in Chengdu communities and the method of Logistics regression analysis to develop a frailty-prefrailty prediction model for elderly people in Chengdu communities.The model comprised 8 salient risk factors,they were occupation,multimorbidity,use of walking devices,self rated health,malnutrition,ADL decline and social support,the model has a good calibration and discrimination.The prediction model provides a useful tool to identify frailty-prefrailty elderly in the Chengdu community.