The objective of this systematic review was to establish the accuracy of score systems in the prediction of mortality in patients with suspected infection in hospital settings compared to the ICU.
Three researchers independently performed a systematic search and a review of related articles and their references using the PubMed database. The articles were selected by consensus, based on previously defined inclusion and exclusion criteria.
In total, 21 studies were included, 19 of which were carried out in the ED. The researchers found that the operative characteristics to evaluate the accuracy (calibration and discrimination) of the different scores were insufficiently assessed in most studies. Only two studies evaluated the calibration, using the Hosmer-Lemeshow goodness-of-fit test, and less than half of the studies evaluated the discrimination, using the area under the receiver operator characteristics curve.
The reviewed literature did not provide enough information to assess the accuracy of the prognostic models in patients with suspected infection admitted to the ED and hospital ward. Some reports suggest a better accuracy with new scores like the MEDS (Mortality in Emergency Department Sepsis score), but the results are not consistent.