Table 2.

Summary of Model Characteristics

	Development only with or without internal validation (n = 40 studies)	Development and external validation (n = 15 studies)	External validation (n = 5 studies)
Modelling method*
Regression analysis, no. of studies (%)	31 (78%)	13 (87%)	3 (60%)
Artificial intelligence, no. of studies (%)	9 (23%)	5 (33%)	0 (0%)
Other†, no. of studies (%)	4 (10%)	0 (0%)	2 (40%)
Internal validation
Split sample, no. of studies (%)	17 (43%)	1 (7%)	–
Bootstrapping or cross-validation, no. of studies (%)	9 (23%)	1 (7 %)	–
None, no. of studies (%)	14 (35%)	13 (87%)	–
Performance measures
Explained variance
R2, no. of studies (%)	21 (53%)	5 (33%)	2 (40%)
Discrimination‡	24 (60%)	10 (67%)	4 (80%)
C-statistic/AUC, no. of studies (%)	22 (55%)	10 (67%)	4 (80%)
Discrimination slope, no. of studies (%)	2 (5%)	0 (0%)	0 (0%)
Other§, no. of studies (%)	2 (5%)	0 (0%)	1 (20%)
Calibration‖	9 (23%)	6 (40%)	4 (80%)
Goodness of fit, no. of studies (%)	4 (10%)	3 (20%)	1 (20%)
Calibration plot, no. of studies (%)	0 (0%)	3 (20%)	4 (80%)
Other¶, no. of studies (%)	5 (13%)	4 (27%)	1 (20%)
Classification
Sensitivity/specificity, no. of studies (%)	14 (35%)	7 (47%)	1 (20%)
Clinical usefulness	1 (3%)	1 (7%)	0 (0%)
Net reclassification index, no. of studies (%)	1 (3%)	0 (0%)	0 (0%)
Decision curve, no. of studies (%)	0 (0%)	1 (7%)	0 (0%)

AUC area under the receiver operating curve

*As some studies use multiple modelling strategies when presenting multiple models, totals may add up to > 100%

†For example, risk stratification with predefined risk tiers

‡As some studies use multiple measures of discrimination, totals may add up to > 100%

^§For example, D-statistic, Brier score, Integrated Discrimination Improvement (IDI)

^‖As some studies use multiple measures of calibration, totals may add up to > 100%

^¶For example, calibration slope, root mean square of approximation (RMSEA), cost capture