Fig 1.
Box summary of PROBAST (Prediction model study Risk Of Bias Assessment Tool).
Fig 2.
Study selection process (PRISMA flow).
Fig 3.
Characteristics of predictive models originally developed from obstetric populations.
*PROBAST does not currently provide a “Maybe” option; however, we have added this term for perceived intermediate/potentially usable models for pregnant and post-partum populations. **for at least one of the WHO criteria.
Fig 4.
Characteristics of predictive models originally developed from non-obstetric population.
* Results were inconsistent across the full number of studies examining this model. ** Sensitivity analysis showed good discrimination and calibration, although results were inconsistent across the full number of studies examining this model. ***Due to concern of a high risk of bias due to a low SMR.
Table 1.
The number and setting of eligible studies in each prediction model, and original population and setting for development of each study.
Table 2.
Summary of 38 eligible articles.
Fig 5.
Box summary of variables of predictive models originally developed from obstetric populations.
Fig 6.
Pooled AUC of APACHE II, SAPS, SOFA and MPM2.