Table 2.
Random forests models
| Number of features | Accuracy | MCCa | AUCb | Sensitivity | Specificity | |
|---|---|---|---|---|---|---|
| Multi-omics (n = 76) | ||||||
| Untreated | 12 | 95.45 | 0.9129 | 0.9669 | 1 | 0.9167 |
| ZDVc | 7 | 95.65 | 0.9161 | 0.9923 | 1 | 0.9091 |
| PI-ARTd | 20 | 80.65 | 0.6193 | 0.9333 | 0.8571 | 0.7647 |
| Number of features | Accuracy | MCC | AUC | Sensitivity | Specificity | |
|---|---|---|---|---|---|---|
| Plasma (n = 97) | ||||||
| Untreated | 5 | 92.86 | 0.8564 | 0.9641 | 0.9231 | 0.9333 |
| ZDV | 20 | 87.1 | 0.736 | 0.9872 | 0.85 | 0.9091 |
| PI-ART | 20 | 89.47 | 0.7922 | 0.9694 | 0.9375 | 0.8636 |
| Number of features | Accuracy | MCC | AUC | Sensitivity | Specificity | |
|---|---|---|---|---|---|---|
| DBS (n = 76) | ||||||
| Untreated | 20 | 95.45 | 0.9129 | 0.9587 | 1 | 0.9167 |
| ZDV | 4 | 95.65 | 0.9161 | 0.9615 | 1 | 0.9091 |
| PI-ART | 20 | 83.87 | 0.6792 | 0.9208 | 0.8667 | 0.8125 |
Characteristics and performance of random forests models, multi-omic (maternal plasma + maternal DBS), maternal plasma, and maternal DBS for preterm birth in each treatment group
aMatthew’s correlation coefficient
bArea under the receiver operator characteristic curve
cZidovudine monotherapy group
dProtease inhibitor-based antiretroviral therapy group