Table 4.
Mortality at Day 90 in Phenotypes Identified in the Secondary Model 1 Using the Classifier Models (Probability Cutoff ≥ 0.5)*
| Model | Mortality at Day 90 in the Hypoinflammatory Phenotype |
Mortality at Day 90 in the Hyperinflammatory Phenotype |
P Value | ||||
|---|---|---|---|---|---|---|---|
| Total [n (%)] | Liberal Fluid [n (%)] | Conservative Fluid [n (%)] | Total [n (%)] | Liberal Fluid [n (%)] | Conservative Fluid [n (%)] | ||
| Clinical classifier | 145/678 (21) | 81/321 (25) | 64/357 (18) | 139/322 (43) | 69/176 (39) | 70/146 (48) | 0.0072 |
| Sparse combined | 153/693 (22) | 86/333 (26) | 67/360 (19) | 131/307 (43) | 64/164 (42) | 67/143 (51) | 0.0124 |
| LCA (8) | 161/727 (22) | 93/355 (26) | 68/372 (18) | 123/273 (45) | 57/142 (40) | 66/131 (50) | 0.004 |
Definition of abbreviations: ALVEOLI = Assessment of Low Vt and Elevated End-Expiratory Pressure to Obviate Lung Injury; ARMA = High vs. Low Vt; FACTT = Fluids and Catheter Treatment Trial; LCA = latent class analysis; SAILS = Statins for Acutely Injured Lungs from Sepsis.
Training data sets: ARMA, ALVEOLI, and SAILS; validation data set: FACTT (n = 1,000). P value represents the interaction between phenotype assignment and randomly assigned treatment strategy for mortality at Day 90. LCA-derived classes were extracted from the original LCA study (8) and not from the derivation data set. Outcomes are shown in phenotypes derived by three different models: clinical-classifier model composed of all the predictor variables, the sparse-combined model composed of only laboratory values and vital signs variables, and the original latent class model (8). Outcomes were substratified by randomized treatment strategy in the original trials.