Table 3.
Some methods for external validation and the level of information about the derivation study they require
|
Method |
Aspect |
Information |
|||||
---|---|---|---|---|---|---|---|---|
L1 | L2 | L3 | ||||||
1 |
Regression on PI in validation data |
D, F |
|
|
|
|||
2 |
Check model misspecification/fit |
F |
|
|
|
|||
3 |
Measures of discrimination |
D |
|
|
|
|||
4 |
Kaplan-Meier curves for risk groups |
D, C |
× |
|
|
|||
5 |
Logrank/Cox tests between risk groups |
D |
× |
|
|
|||
6 |
Hazard ratios across risk groups |
D |
× |
|
|
|||
7 | Calibration | C | × | × |
L1, L2 and L3 denote increasing amounts of information (see text). Key to Aspect: D, discrimination; C, calibration; F, model fit.