Performance of prediction models of postoperative mortality in high-risk surgical patients in swedish university hospitals : Predictors, Risk factors and Outcome Following major Surgery study (PROFS study NCT02626546)

• Background: There are several progn ostic prediction models that estimate the probability of postoperative mortality. The role of these models is to support clinical decisions. Before implementation of a prediction model in routine care, it is necessary to analyze its performance in the target population. Our aim was to analyze the performance of four different prediction models of postoperative mortality in a high-risk surgical population.Methods: Data collected from 2015-11-01 until 2016-02-15 in a prospective consecutive observational study (PROFS study) in four university hospitals was used. The inclusion criteria were adult, ASA classification ≥3, and major/complex upper or lower gastrointestinal, urogenital or orthoped ic surgery (UK surgical severity codingA XA PPP). Four prediction models were evaluated: Surgical Outcome Risk Tool (SORT), Surgical APGAR, P-POSSUM and Surgical Risk Scale (SRS). The outcome measure was 90-day mortality. We evaluated the discrimination of the models by area under receiver operator characteristic curve (AUC ROC) before and after recalibration.Results: In total, 1 089 patients were included. Thirteen patients were excluded due to erroneous inclusion, and another three were lost to follow-up, so data from 1 073 was used in this analysis. The mean age was 73 years, the presence of malignancy was 41%, and 90-day mortality was 13% (n = 140). The SORT model had the best discrimination both before and after recalibration. The P-POSSUM model improved after recalibration. The SRS model overestimated, whereas the APGAR model underestimated, the risk of mortality.Conclusions: The original SORT model is promising and could be incorporated as decision support for high-risk surgical patients.

Ämnesord

MEDICIN OCH HÄLSOVETENSKAP  -- Klinisk medicin -- Anestesi och intensivvård (hsv//swe)

MEDICAL AND HEALTH SCIENCES  -- Clinical Medicine -- Anesthesiology and Intensive Care (hsv//eng)

Publikations- och innehållstyp

vet (ämneskategori)

art (ämneskategori)