Table 5.
The five best predictive models for sepsis/septic shock treatment outcome selected by the BMA technique
| Predictors | Probability (%) | Regression coefficient |
||||
|---|---|---|---|---|---|---|
| Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | ||
| Intercept | −18.6 | −1.3 | 19.2 | −18.6 | −2.5 | |
| Age | 0.0 | — | — | — | — | — |
| Gender | 3.6 | |||||
| male | — | — | — | −1.2 | — | |
| eGFR | 0.0 | |||||
| <60 mL/min/1.73 m2 | — | — | — | — | — | |
| Department of treatmenta | 100.0 | |||||
| infectious diseases department | 22.4 | 5.6 | 22.8 | 23.0 | 6.5 | |
| others | 20.3 | 3.3 | 20.4 | 20.4 | 3.8 | |
| Site of infectionb | 0.0 | |||||
| urinary tract | — | — | — | — | — | |
| intra-abdominal | — | — | — | — | — | |
| respiratory tract | — | — | — | — | — | |
| skin and soft tissue | — | — | — | — | — | |
| Presence of septic shock | 83.1 | |||||
| yes | −2.8 | −2.6 | — | — | — | |
| Blood culture | 0.0 | |||||
| positive | — | — | — | — | — | |
| Overall appropriate empirical antibiotic | 66.9 | |||||
| yes | 18.6 | — | 19.2 | 19.2 | — | |
| Number of variables | 3 | 2 | 2 | 3 | 1 | |
| BICc | −446.9 | −445.7 | −443.5 | −441.5 | −441.5 | |
| Posterior probability (%) | 53.6 | 29.6 | 9.8 | 3.6 | 3.5 | |
a
Reference group: ICU.
b
Reference group: unknown source.
c
Bayesian information criterion (BIC) or Schwarz information criterion is a criterion for model selection among a finite set of models; the model with the lowest BIC is preferred.