| Algorithm 1 | Algorithm 2 | ||
---|---|---|---|---|
Features | Hyp and MetSO, DT, NFK, 3-NT, CEL, CMA, G-H1, MG-H1, 3DG-H and pentosidine | anti-CCP antibody positivity and MetSO, DT, 3-NT, FL, CML, CEL, CMA, MG-H1, 3DG-H and pentosidine | ||
Comparators | Disease versus control | eOA | eRA | non-RA |
Training set cross-validation | ||||
 Correct, n | 44/46 | 13/13 | 6/8 | 8/12 |
 Sensitivity | 0.92 (0.64–1.00) | 0.92 (0.64–1.00) | 0.80 (0.44 - 0.97) | 0.70 (0.35 - 0.93) |
 Specificity | 0.91 (0.76–0.98) | 0.90 (0.68–0.99) | 0.78 (0.56–0.93) | 0.65 (0.43–0.84) |
 AUROC | 0.99 (0.97–1.00) | 0.98 (0.96–1.00) | 0.91 (0.82–1.00) | 0.68 (0.49–0.86) |
 Positive predictive value | 1.0 | 1.0 | 0.60 | 0.80 |
 Negative predictive value | 0.85 | 1.0 | 0.84 | 0.83 |
 Positive likelihood ratio | 10.2 | 9.2 | 3.6 | 2.0 |
 Negative likelihood ratio | 0.09 | 0.09 | 0.26 | 0.46 |
Test set cross-validation | ||||
 Correct, n | 119/134 | 30/31 | 22/27 | 27/39 |
 Sensitivity | 0.89 (0.75–0.97) | 0.83 (0.65–0.94) | 0.77 (0.6–0.9) | 0.72 (0.53–0.86) |
 Specificity | 0.90 (0.82–0.95) | 0.84 (0.73–0.92) | 0.76 (0.63–0.86) | 0.71 (0.58–0.81) |
 AUROC | 0.96 (0.93–0.99) | 0.93 (0.88–0.98) | 0.87 (0.8–0.94) | 0.77 (0.68–0.86) |
 Positive predictive value | 0.96 | 0.97 | 0.81 | 0.69 |
 Negative predictive value | 0.75 | 1.0 | 0.81 | 0.91 |
 Positive likelihood ratio | 8.9 | 5.2 | 3.2 | 2.5 |
 Negative likelihood ratio | 0.12 | 0.20 | 0.30 | 0.39 |
Test set validation | ||||
 Correct, n | 63/134 | 18/19 | 8/35 | 31/32 |
 Sensitivity | 0.73 (0.56–0.86) | 0.83 (0.65–0.94) | 0.60 (0.42–0.76) | 0.81 (0.64–0.93) |
 Specificity | 0.72 (0.62–0.81) | 0.84 (0.73–0.92) | 0.61 (0.46–0.72) | 0.80 (0.68–0.89) |
 AUROC | 0.77 (0.69–0.85) | 0.91 (0.84–0.99) | 0.62 (0.5–0.75) | 0.84 (0.77–0.92) |
 Positive predictive value | 0.62 | 0.60 | 0.23 | 0.97 |
 Negative predictive value | 0.05 | 0.99 | 0.97 | 0.43 |
 Positive likelihood ratio | 2.6 | 5.2 | 1.5 | 4.1 |
 Negative likelihood ratio | 0.38 | 0.20 | 0.67 | 0.24 |