Skip to main content
. 2012 Jan 11;7(1):e29090. doi: 10.1371/journal.pone.0029090

Table 4. Comparison of the areas under the receiver operating characteristic (ROC) curves of the four frailty instruments compared.* .

Frailty index AUC SE 95% CI Contrast p-value
Follow-up Crude/Adjusted Models 1 vs 2 1 vs 3 1 vs 4
1.MPI 0.7486 0.0209 0.71–0.79
Overall 2. FI-SOF 0.5918 0.0208 0.55–0.63 <0.0001 <0.0001 <0.0001
3.FI-CD 0.7094 0.0208 0.67–0.75
4.FI-CGA 0.6924 0.0218 0.65–0.73
1.MPI 0.7396 0.0332 0.67–0.80
Male 2. FI-SOF 0.6163 0.0313 0.55–0.67 0.0002 0.0017 0.0094
Unadjusted 3.FI-CD 0.7013 0.0333 0.64–0.77
4.FI-CGA 0.6923 0.0340 0.62–0.76
1 month 1.MPI 0.7636 0.0260 0.71–0.81
Female 2. FI-SOF 0.5749 0.0279 0.52–0.63 <0.0001 0.0051 <0.0001
3.FI-CD 0.7216 0.0259 0.67–0.77
4.FI-CGA 0.6994 0.0282 0.64–0.76
1.MPI 0.7655 0.0203 0.72–0.80
Age- and sex- Overall 2. FI-SOF 0.6854 0.0221 0.64–0.73 <0.0001 0.0005 <0.0001
adjusted 3.FI-CD 0.7383 0.0204 0.69–0.78
4.FI-CGA 0.7240 0.0215 0.68–0.77
1.MPI 0.7231 0.0141 0.69–0.75
Overall 2. FI-SOF 0.6034 0.0142 0.57–0.63 <0.0001 <0.0001 <0.0001
3.FI-CD 0.6934 0.0143 0.66–0.72
4.FI-CGA 0.6890 0.0145 0.66–0.72
Unadjusted 1.MPI 0.7164 0.0213 0.67–0.76
1 year Male 2. FI-SOF 0.6196 0.0210 0.58–0.66 <0.0001 0.0010 0.0048
3.FI-CD 0.6836 0.0221 0.64–0.73
4.FI-CGA 0.6855 0.0219 0.64–0.73
1.MPI 0.7416 0.0184 0.70–0.78
Female 2. FI-SOF 0.5974 0.0193 0.56–0.63 <0.0001 0.0005 0.0003
3.FI-CD 0.7081 0.0184 0.67–0.74
4.FI-CGA 0.7042 0.0189 0.67–0.74
1.MPI 0.7507 0.0135 0.72–0.78
Age- and sex- Overall 2. FI-SOF 0.6948 0.0142 0.67–0.72 <0.0001 <0.0001 <0.0001
adjusted 3.FI-CD 0.7297 0.0138 0.70–0.76
4.FI-CGA 0.7278 0.0139 0.70–0.75

AUC: areas under curve; SE: standard error; CI: confidence interval; MPI: Multidimensional Prognostic Index; FI-SOF: Frailty Index derived from the Study of Osteoporotic Fractures; FI-CD: Frailty Index based on cumulative deficits; FI-CGA: Frailty Index based on a Comprehensive Geriatric Assessment.

*AUCs were assessed by crude and adjusted logistic regression models, accounting for clustering due to centre effect.