Table 3. Logistic modeling results of LOS by site using cut points defined by classification trees.
| Model | Other covariates+ | Independent Variables | beta | p-value |
|---|---|---|---|---|
| Cincinnati—first sample | none | Actinobacteria > = 0.1% | -2.67 | 0.018 |
| Bacilli < 3.25% | -1.39 | 0.39 | ||
| Cincinnati—last sample | Gravida (p = 0.098) | Pseudomonadales > = 0.1% | -2.65 | 0.029 |
| Clostridiaceae > = 0.35% | 2.43 | 0.030 | ||
| Proteobacteria < 72.3% | -2.59 | 0.011 | ||
| Birmingham—last sample Model 1 | Infant GA (p = 0.087) | Lactobacillales present | -3.6 | 0.017 |
| Birmingham—last sample Model 2 | Chorioamnionitis (p = 0.070) | Lactobacillales present | -3.2 | 0.010 |
Modeling confirmed the statistical significance of cut points identified from classification tree analysis except for the Birmingham first sample tree (not shown) and the unknown Bacillales in the Birmingham last sample tree.
+Covariates that remained in each model are noted along with their p-value.