| Objective:To evaluate the psychometric performances of different frailty screening tools in identifying frailty among institutionalized older adults,using the Comprehensive Geriatric Assessment(CGA)which is based on the concept of frailty as "gold standard".To compare their ability to predict adverse outcomes among institutionalized older adults,it is helpful to provide appropriate frailty screening tool for institutionalized older adults,and then could timely identify frail older adults with higher risk of adverse outcomes.Methods:By convenient sampling,a total of 309 institutionalized older adults were recruited from Jinan City.China.Frailty was assessed by the Physical Frailty Phenotype(PFP).the FRAIL(fatigue.resistance,ambulation,illnesses,and loss of weight).the Study of Osteoporotic Fracture(SOF),Tilburg Frailty Indicator(TFI),Groningen Frailty Indicator(GFI).and Frailty Index(FI),respectively.The Comprehensive Geriatric Assessment(CGA)was used as "gold standard" of frailty.Data on concurrent adverse outcomes including Activities of Daily Living(ADL).fall,and hospitalization were also collected.The Receiver Operating Characteristics(ROC)Curve analysis was conducted to examine the diagnostic accuracy of different frailty tools.The optimal cut-point was determined by the maximum value of the Youden index(YI,calculated as sensitivity+ specificity-1).Diagnostic properties such as sensitivity,specificity,positive predictive value(PPV),negative predictive value(NPV),correctly classified rate(CCR)and kappa value of the six frailty screening tools at the optimal and original cut-points were calculated,respectively.The logistic regression models were used to examine the association between frailty defined by each frailty screening tool and each concurrent outcome with adjustment for age,sex,marital status,and education.Adjusted odds ratios(ORs)and 95%confidence intervals(CIs)were reported.And the ROC analyses were used to compare the ability of six frailty screening tools in predicting the adverse outcomes.Data were analyzed using SPSS 21.0 and Stata 12.0.Rsults:1.Frailty prevalence estimates were 20.1%(SOF)?25.6%(FRAIL).41.7%(PFP)?45.3%(TFI)?56.3%(GFI)?79.6%(FI)2,respectively.The prevalence evaluated by the CGA was 74.1%.2.Areas under ROC curve(AUCs)for SOF,FRAIL,TFI,GFI,Fried,and FI against the gold standard CGA for diagnosis of frailty were 0.77(95%CI 0.71-0.82;P<0.05),0.79(95%CI 0.74-0.85;P<0.05),0.80(95%CI 0.75-0.85;P<0.05),0.80(95%CI 0.75-0.86;P<0.05),0.830(95%CI 0.78-0.88;P<0.05),and 0.89(95%CI 0.85-0.93;P<0.05),respectively.At the original cut-point,the FI presented high sensitivity(92.1%)and low specificity(56.3%)while the other five frailty tools had low sensitivity(26.2%-69.0%)and high specificity(80.0%-97.5%).All the frailty screening tools had high PPVs(85.8%-93.8%)and low NPVs(31.6%-71.4%).The CCR of frailty ranged from 44.7%(SOF)to 82.9%(FI),with varied agreement with the CGA(kappa:0.141-0.520).At the optimal cut-point,the sensitivity and specificity of the PFP,FRAIL,SOF,and FI tended to be balanced,and their CCRs(66.7%-83.2%)and kappa values(0.362-0.597)increased a lot,but their PPVs were still high and their NPVs were still low.On ROC contrasts,the FI had a higher AUC than did other five frailty tools(?2:6.09-18.92,P<.05);the PFP also presented a higher AUC than the SOF(?2:6.36.P<.05);the differences in the AUCs between any other tools were not significant(?2:0.03-1.77.P>.05).3.At the original cut-point,all the frailty screening tools could independently predict ADL disability(OR:3.29-30.23,P<.05)with adjustment for age,sex,marital status,and education.Frailty,as defined by the PFP,FRAIL.SOF,and FI was also associated with fall(OR:1.96-2.76,P<.05).None frailty screening tools were associated with hospitalization(OR:0.72-1.29,P>.05).At the optimal cut-point,the predictive validity for the frailty tools was similar to the predictive validity at the original cut-point,except that the PFP was not associated with fall any more(OR=1.76,95%CI =0.94-3.41).Moreover,using the adverse health outcomes as criterion respectively,the ROC analyses showed that the predictive performances of the six screening frailty tools were moderate for ADL disability(AUC:0.73-0.88).with a higher AUC for the FI than the other tools(?1=:18.15-32.52,P<.001);the predictive performances were poor for fall(AUC:0.57-0.62),with no significant differences(?2:0.01-2.18,P>.05);and they failed to predict hospitalization(AUC:0.46-0.53).Conclusions:The prevalence of frailty varies widely by six frailty screening tools among institutionalized older adults,higher for multidimensional tools than that for purely physical dimensional ones.All frailty screening tools have good diagnostic accuracy in identifying frailty,and the FI performs better in screening frailty against the CGA than do others,as well as the PFP performs better than SOF.There are trade-offs between sensitivity and specificity,and the sensitivity and negative predictive values were lower,but the specificity and positive predictive values were higher.All the six screening frailty tools can independently predict ADL disability,with good predictive ability;only the PFP,FRAIL,SOF and FI can identify older adults with higher risk of fall,and the predictive performances for fall are insufficient;none frailty screening tools can predict hospitalization.The simpler FRAIL may be easier and more practical in an institutionalized setting,for identifying frailty early and accurately,as well as providing effective interventions timely to reduce the risk of adverse outcomes. |
PDF Full Text Request