Table 2:
Harrell’s C index | Akaike information criterion | Net reclassification index | |
---|---|---|---|
Analysis of risk history | |||
SCORE risk score | |||
Baseline at the risk factor measurement | 0·685 (reference) | 17 255 | ·· |
Baseline measurement + change from the preceding risk factor measurement | 0·690 | 17 200 | ·· |
Change (95% CI)† | 0·004 (0·000 to 0·008) | −57 (−97 to −13)† | ·· |
Optimism in derivation sample‡ | 0·002 | ·· | ·· |
Reclassification in cases (95% CI) | ·· | ·· | −0·006 (−0·060 to 0·063) |
Reclassification in non-cases (95% CI) | ·· | ·· | 0·360 (0·248 to 0·411) |
Continuous net reclassification index (95% CI) | ·· | ·· | 0·353 (0·234 to 0·447) |
ASCVD risk score | ·· | ·· | ·· |
Baseline at the risk factor measurement | 0·699 (reference) | 14 739 (reference) | ·· |
Baseline measurement + change from the preceding risk factor measurement | 0·700 | 14 729 | ·· |
Change (95% CI)† | 0·001 (0·000 to 0·003) | −10 (−28 to 7)† | ·· |
Optimism in derivation sample‡ | 0·002 | ·· | ·· |
Reclassification in cases (95% CI) | ·· | ·· | −0·101 (−0·156 to −0·027) |
Reclassification in non-cases (95% CI) | ·· | ·· | 0·333 (0·110 to 0·442) |
Continuous net reclassification index (95% CI) | ·· | ·· | 0·232 (0·030 to 0·344) |
Analysis of targeted change | ·· | ·· | ·· |
SCORE risk score | ·· | ·· | ·· |
Baseline at the first risk factor measurement | 0·684 (reference) | 17 229 | ·· |
Baseline measurement + change in the following risk factor measurement | 0·690 | 17 200 | ·· |
Change (95% CI)† | 0·006 (0·001 to 0·107) | −29 (−56 to −1)† | ·· |
Optimism in derivation sample‡ | 0·002 | ·· | ·· |
Reclassification in cases (95% CI) | ·· | ·· | −0·085 (−0·137 to −0·010) |
Reclassification in non-cases (95% CI) | ·· | ·· | 0·469 (0·271 to 0·496) |
Continuous net reclassification index (95% CI) | ·· | ·· | 0·384 (0·219 to 0·445) |
ASCVD risk score | ·· | ·· | ·· |
Baseline at the risk factor measurement | 0·690 (reference) | 14 775 (reference) | ·· |
Baseline measurement plus change in the following risk factor measurement | 0·700 | 14 729 | ·· |
Change (95% CI)† | 0·010 (0·004 to 0·015) | −47 (−75 to −19)† | ·· |
Optimism in derivation sample‡ | 0·002 | ·· | ·· |
Reclassification in cases (95% CI) | ·· | ·· | −0·156 (−0·215 to −0·071) |
Reclassification in non-cases (95% CI) | ·· | ·· | 0·534 (0·424 to 0·566) |
Continuous net reclassification index (95% CI) | ·· | ·· | 0·379 (0·226 to 0·475) |
ASCVD=American Heart Association Atherosclerotic Cardiovascular Disease. SCORE=European Society of Cardiology Systematic Coronary Risk Evaluation.
All analyses were adjusted for the baseline risk (measured either in 1991–93 or 1997–99), socioeconomic status, and ethnicity. Follow-up for incident cardiovascular disease starts at later measurement in prospective and retrospective change analysis.
A decrease of 10 units or more in Akaike information criterion is considered strong evidence of better fit.
Optimism index describes overfitting of the final model compared with 200 bootstrap samples with replacement.