Accuracy of diagnosis and health service codes in identifying frailty in Medicare data

BMC Geriatrics

Table 3 Performance of Total Count of Claims-Based Surrogates of Frailty versus a Model-Based Claims-Based Frailty Index Against Clinical Frailty Assessment in the Health and Retirement Study-Medicare Data

Threshold to Define Frailty	Positive for Frailty N (%)	Clinical FI Mean (SD)	Clinical Frailty Phenotype
Threshold to Define Frailty	Positive for Frailty N (%)	Clinical FI Mean (SD)	Sensitivity	Specificity	PPV	NPV
A. Total count of claims-based surrogates of frailty (range: 0–18)^a
≥ 1	1196 (38.6)	0.26 (0.15)	0.62	0.67	0.31	0.88
≥ 2	494 (16.0)	0.32 (0.16)	0.38	0.89	0.45	0.86
≥ 3	284 (9.2)	0.36 (0.15)	0.26	0.95	0.54	0.85
≥ 4	151 (4.9)	0.38 (0.15)	0.16	0.98	0.63	0.83
≥ 5	80 (2.6)	0.40 (0.16)	0.09	0.99	0.66	0.82
B. Model-based claims-based frailty index (range: 0–1)^b
≥ 0.15	1464 (47.3)	0.26 (0.14)	0.76	0.59	0.30	0.91
≥ 0.20	632 (20.4)	0.33 (0.15)	0.47	0.86	0.44	0.87
≥ 0.25	302 (9.8)	0.37 (0.15)	0.30	0.95	0.58	0.85
≥ 0.30	133 (4.3)	0.44 (0.15)	0.16	0.99	0.72	0.83
≥ 0.35	70 (2.3)	0.46 (0.16)	0.09	0.99	0.76	0.82

Abbreviations: FI frailty index, NPV negative predictive value, PPV positive predictive value, SD standard deviation
^aThe optimal cutpoint of the count approach for frailty phenotype was ≥1, which achieved a sensitivity of 0.62 and a specificity of 0.67
^bThe optimal cutpoint of CFI for frailty phenotype was ≥0.17, which achieved a sensitivity of 0.66 and a specificity of 0.72

ISSN: 1471-2318