Temporal and geographical external validation study and extension of the Mayo Clinic prediction model to predict eGFR in the younger population of Swiss ADPKD patients

Table 2 Predictive performance of the validation, the updated models and the sensitivity analysis (in bold)

	R²	Bias (mL/min per 1.73 m²)^a	95% Limits of Agreement (mL/min per 1.73 m2)^b	Correlation	P30 (%)^c	AIC	CRPS
Validation model: risk class (1985 observations)	0.7039	5.29	−16.8,27.3	0.839	88.7	–	–
Validation model: TKV (1985 observations)	0.7853	−2.73	−20.4,15.0	0.871	91.5	14,386.98	73.36
Updated model 1: 2 TKV’s (1867 observations)	0.7704	0.42	−17.4,18.3	0.872	96.6	13,557.91	58.16
Updated model 1 with interaction term	0.7720	0.82	−16.9,18.6	0.879	96.6	1322.3	81.85
Updated model 2: TKV time-varying (1344 observations)	0.7989	0.34	−17.1,17.8	0.889	96.1	9706.15	57.24
Updated model 2 with interaction term	0.8015	0.57	−16.8,17.9	0.895	96.1	9715.05	83.39

^aBias = average (observed eGFR – predicted eGFR)
^b95% Limit of agreement = bias ±1.96*standard deviation of (observed eGFR – predicted eGFR)
^cP30 = percentage of predicted eGFR within 30% of observed GFR (% within 30%)
AIC Akaike information criterion
CRPS continuous ranked probability score

ISSN: 1471-2369