An improved prognostic model for predicting the mortality of critically ill patients: a retrospective cohort study

Figure 1. Study population and protocol flowchart. Flow chart illustrating the major steps in the development and validation of the mionl-MEWS model.

Figure 2. Comparisons of age, body mass index (BMI), red cell distribution width (RDW), neutrophil-to-lymphocyte ratio (NLR), international normalized ratio (INR), lactate (lac) concentration, and osmolarity between survivors and nonsurvivors in the development cohort. (a) Comparison of age by the U test, p < 0.001; (b) comparison of BMI by the U test, p < 0.001; (c) comparison of RDW by the t test, p < 0.001; (d) comparison of NLR by the U test, p < 0.001; (e) comparison of INR by the U test, p < 0.001; (f) comparison of lac concentration by the t test, p < 0.001; and (g) comparison of osmolarity by the chi-squared test, p < 0.001.

Figure 3. Calibration plot of observed versus expected probabilities for assessment of the predictive performance of the mionl-MEWS model in the development cohort.

Figure 4. Predictive performance of the mionl-MEWS, APACHE-II, MEWS, neutrophil-to-lymphocyte ratio (NLR), red cell distribution width (RDW), lactate (lac) concentration, and osmolarity level for 28-day mortality in critically ill patients in the development cohort.

Figure 5. Calibration plot of observed versus expected probabilities for assessment of the predictive performance of the mionl-MEWS model in the development cohort.

Figure 6. Predictive performance of the mionl-MEWS, APACHE-II, MEWS, neutrophil-to-lymphocyte ratio (NLR), red cell distribution width (RDW), lactate (lac) concentration, and osmolarity level for 28-day mortality in critically ill patients in the validation cohort.

Figure 7. Comparison of precision-recall area under the curves (PR-AUCs) between the mionl-MEWS and APACHE-II in the validation cohort.

Figure 8. Comparison of the area under the receiver operating characteristic curve (AUROC) values between the mionl-MEWS and cross validation.

Figure 9. Nomogram for the mionl-MEWS model. On the nomogram, an individual patient’s predicted mortality risk according to the mionl-MEWS model is located on each variable axis, and a line is drawn upward to determine the corresponding score for each variable state. The sum of these numbers indicates the total score, and a line is drawn to the probability axis to determine the likelihood of 28-day mortality (INR international normalized ratio, Lac lactate, MEWS Modified Early Warning Score, NLR neutrophil-to-lymphocyte ratio).