Table 3.
Multivariable logistic regression model for predicting probable PTSD 6 Months post-injury (N = 1143).
| OR (95% CI) | p-value | |
|---|---|---|
| Age | 0.995 (0.986, 1.006) | 0.322 |
| Sex Female vs Male | 1.425 (1.019, 1.994) | 0.039 |
| Race Black vs non-Black | 2.816 (1.935, 4.098) | <0.001 |
| Injury cause Violence vs Incident/other | 2.656 (1.489, 4.736) | <0.001 |
| Psychiatric history Yes vs No | 2.408 (1.702, 3.406) | <0.001 |
| Prior TBI Yes vs No | 1.586 (1.138, 2.211) | 0.007 |
| CT + vs − | 0.837 (0.555, 1.262) | 0.395 |
| Sampling Time | ||
| 9–16h vs 0–8h | 1.347 (0.826, 2.196) | 0.232 |
| 17–24h vs 0–8h | 1.800 (1.104, 2.934) | 0.019 |
| GFAP (in log scale) | 0.851 (0.766, 0.946) | 0.003 |
| hsCRP (in log scale) | 1.105 (0.977, 1.251) | 0.113 |
AUC = 0.723, 95% CI: (0.682, 0.76); Nagelkerke’s pseudo-R-squared = 0.168; likelihood ratio test comparing this model to the model without biomarkers and sampling time showed p-value < 0.001.