Skip to main content
Figure 5 | Critical Care

Figure 5

From: Use of dynamic microsimulation to predict disease progression in patients with pneumonia-related sepsis

Figure 5

Bootstrap validation of the model results. The model was re-evaluated on 100 bootstrapped samples of 50 replications under each of the similarity criteria shown in Figure 3. Only results of the simulations using component SOFA scores are shown. Empiric 95% confidence limits around the predicted number of discharges or deaths each day are constructed from the distribution of simulated discharges or deaths on each day of the simulation. The results indicate that the model results are relatively stable to random fluctuations in the data that were used to calibrate it, and confirm the finding that duration of disease is more important in predicting overall outcome than the instantaneous direction of progression of disease.

Back to article page