PURPOSE: The goal of this study was to develop a simple, disease-specific multivariable predictive scorecard for mortality to be used in patients with early ALI.
METHODS: The study used two populations: the ARMA trial was used to derive the model and the ALVEOLI trial was used for validation. In both trials (which examined optimal tidal volume and PEEP), subjects were enrolled within 48 hours of onset of ALI. Baseline clinical variables collected within 24 hours of enrollment to the ARMA trial were modeled as predictors of mortality at 28 days using logistic regression and bootstrap resampling to arrive at a parsimonious prognostic model. A point score was constructing based on regression coefficients and discrimination and calibration were assessed. The predictive properties of the index were compared with observed mortality in the ALVEOLI trial for validation. Stratum-specific diagnostic likelihood ratios were generated for each point total.
RESULTS: The following variables had the best predictive properties and comprised the predictive index: hematocrit <26 (1 point), bilirubin > 2 (1 point), fluid balance greater than 3 liters positive (1 point), and age (1 point for age 40-64, 2 points for age > 65). In the derivation set (ARMA), the c-statistic was 0.723, and the model had good fit (goodness of fit test p=0.67). Predicted probabilities, observed mortality, and likelihood ratios for each of the point total strata in the derivation and validation sets are presented in Table 1. The observed mortalities for each of the strata in the validation set (ALVEOLI) were within the confidence bounds of predicted probabilities. Stratum-specific likelihood ratios ranged from 0.31 to 4.47. Discrimination of the model (c=0.706) was comparable to APACHE III score (c=0.701).
CONCLUSION: Mortality in ALI patients can be predicted using a simple point score based on four readily-available baseline clinical variables.
CLINICAL IMPLICATIONS: This index may help inform clinical decisions by allowing rapid prognoses in the ALI population. Further validation of the predictive index in non-clinical trial populations is a logical and necessary next step.
DISCLOSURE: Jason Christie, None.