A Nomogram Prediction Model For The Risk Of Disability Incidence In The Middle-aged And Older Adults

Aim:With the accelerating rate of population aging,the number of disabled elderly is gradually increasing.The disability process of the elderly is affected by many factors.Preventing risk factors in the early stage of disability can delay or even reverse the decline of functional ability.Therefore,based on two representative middle-aged and elderly community populations,we established a nomogram prediction model for the risk of disability,which aimed to provide a scientific basis for formulating preventive interventions for dysfunction and disability.Methods:The data derived from English Longitudinal Study of Ageing(ELSA,2004-2016)and the Survey of Health,Aging and Retirement in Europe(SHARE,2004-2016).The individuals included in ELSA were randomly divided into a training and a validating set with a ratio of 7:3,and SHARE database was taken as the testing set.Disability was assessed by basic activities of daily living(BADL)scale and instrumental activities of daily living(IADL)scale.18 predictive variables related to disability were included in training group:age,sex,marital status,education level,body mass index(BMI),smoking status,alcohol,physical activity,cognitive function,grip strength,depression symptoms,self-rated health status and chronic diseases(cardiovascular disease,diabetes,chronic lung disease,cancer,memory-related disease and arthritis).The independent risk factors were screened by the Least absolute shrinkage and selection operator(Lasso)and stepwise logistic regression.According to the results of multivariate logistic regression model,the corresponding nomogram model was drawn.The discrimination of the nomogram was evaluated by receiver operating characteristic curve(ROC)area under curve(AUC),the calibration was evaluated by calibration plot and Hosmer-Lemeshow(H-L)test.Results:In the training set,the results of Lasso regression and stepwise Logistic regression analysis found that age(odds ratio(OR)=2.015,95%confidence interval(CI)=1.764-2.301),marital status(OR=1.228,95%CI=1.020-1.477),BMI(OR=1.430,95%CI=1.293-1.582),physical activity(OR=1.291,95%CI=1.104-1.846),cognitive function(OR=0.859,95%CI=0.767-0.963),depression symptom(OR=1.428,95%CI=1.104-1.846),self-rated health status(OR=2.047,95%CI=1.660-2.524),chronic lung disease(OR=1.397,95%CI=1.122-1.740),arthritis(OR=1.711,95%CI=1.443-2.209)and memory-related disease(OR=4.981,95%CI=1.743-14.236)were predictors of BADL disability among community-dwelling middle-aged and older adults.Based on the above predictors,a nomogram prediction model of the risk of BADL disability was established.In ELSA training set,the AUC of the prediction model was 0.734(95%CI:0.715,0.752),and in the validating and testing sets,the AUC of the model were 0.742(95%CI:0.715,0.769)and 0.706(95%CI:0.696,0.715),respectively,indicating good discrimination.The calibration plots showed that prediction and observation data agreed well in both the training set and external validation sets.The~2 of H-L test of training set,validating set and testing set were 10.518(P=0.231),2.077(P=0.979)and 25.804(P=0.001),suggesting that training and validating sets have good goodness of fit.In the training set,the results of Lasso regression and stepwise Logistic regression analysis showed that age(OR=2.460,95%CI=2.092-2.791),BMI(OR=1.312,95%CI=1.185-1.453),physical activity(OR=1.352,95%CI=1.107-1.652),cognitive function(OR=0.850,95%CI=0.759-0.952),grip strength(OR=0.850,95%CI=0.759-0.952),depression symptom(OR=1.652,95%CI=1.274-2.142),self-rated health status(OR=2.143,95%CI=1.729-2.655),chronic lung disease(OR=1.444,95%CI=1.158-1.799)and arthritis(OR=1.317,95%CI=1.107-1.568)were predictors of IADL disability among community-dwelling middle-aged and older adults.Based on the above predictors,a nomogram prediction model of the risk of IADL disability was established.In the training set,the AUC of the prediction model was 0.733(95%CI:0.715,0.751).And in the validating and testing sets,the AUC of the model were 0.769(95%CI:0.744,0.794)and 0.751(95%CI:0.743,0.759),respectively.Good calibrations were demonstrated in the training and validation cohorts,as displayed by calibration curve.The~2 of H-L test of training set,validating set and testing set were 5.075(P=0.749),12.905(P=0.115)and 23.210(P=0.003),suggesting that training and validating sets have good goodness of fit.Conclusion:Based on two large sample cohort,this study successfully developed and validated a nomogram prediction model for predicting the risk of disability in middle-aged and older adults.This model has good performance,included indicators are simple and easy to obtain.The prediction model can accurately and conveniently identify high-risk population,and carry out early intervention to prevent or delay the occurrence of disability,which is suitable for popularizing in the community.