Table 4.

Accuracy of the predicted probabilities in the final model^* including frail or not frail.

Measure	Models
	Final model with frailty	Final model without frailty
Sensitivity	0.813	0.750
Specificity	0.706	0.662
LR+	2.74	2.22
1/LR−	0.26–3.85	0.38–2.63

^*Final model included frailty, age, number of hospitalizations because of exacerbations in the previous year (HOSP) and length of stay (LOS).

The cut-off probability for the full final model was 24%. This probability was calculated with the equation: $1÷1+e^{-(-7.41+0.06\left(LOS\right)+0.07\left(AGE\right)+1.24\left(HOSP\right)+0.40\left(FRAILTYDEGREE\right))}$.

The cut-off probability for the final model without frailty was 27%. This probability was calculated with the equation: $1÷1+e^{-(-7.02+0.05\left(LOS\right)+0.07\left(AGE\right)+1.35\left(HOSP\right))}$.

LR, likelihood ratio.