Anthropometric Measures and Frailty Prediction in the Elderly: An Easy-to-Use Tool

Table 4

Prognostic ability of anthropometric measures as predictors of frailty in older adults who were assisted at primary health care centers via artificial neural models.

Predicted probability of anthropometric measures	Frailty			Se	Sp	PPV	NPV
	Total; prevalence; (%)	Yes; (%)	No; (%)
Neural network (learning sample)
>0.564	268 (67.2)	220 (79.7)	48 (17.9)	0.797	0.610	0.820	0.572
>0.717	217 (54.4)	188 (68.1)	29 (23.6)	0.681	0.764	0.866	0.516
>0.562	271 (68.3)	221 (80.1)	50 (41.3)	0.801	0.587	0.815	0.563
Neural network (testing sample)
>0.564	69 (68.3)	54 (81.8)	15 (42.9)	0.818	0.571	0.782	0.625
>0.717	53 (53.0)	46 (69.7)	7 (20.6)	0.697	0.794	0.867	0.574
>0.562	72 (71.3)	54 (81.8)	18 (51.4)	0.818	0.486	0.750	0.586

Notes. Se: sensitivity; Sp: specificity; PPV: positive predictive value; NPV: negative predictive value; SeC: sensitivity ≈ 80%; SpC: specificity ≈ 60%. YiC: Youden Index.