Evaluation of various POSSUM models for predicting mortality in patients undergoing elective oesophagectomy for carcinoma

Br J Surg. 2007 Sep;94(9):1172-8. doi: 10.1002/bjs.5793.


Background: The aim of the study was to validate the use of the Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM), Portsmouth (P) POSSUM and upper gastrointestinal (O) POSSUM models in patients undergoing elective thoracic oesophagectomy for carcinoma.

Methods: The observed in-hospital mortality rates in 545 patients undergoing elective thoracic oesophagectomy for squamous cell carcinoma of the oesophagus in all public hospitals in Hong Kong was compared with rates predicted by POSSUM, P-POSSUM and O-POSSUM. The discriminatory power of these models was assessed using receiver-operator characteristic (ROC) curve analysis.

Results: The observed mortality rate was 5.5 per cent, whereas rates predicted by POSSUM, P-POSSUM and O-POSSUM were 15.0, 4.7 and 10.9 per cent respectively. P-POSSUM showed no lack of fit (P = 0.814), but POSSUM (P < 0.001) and O-POSSUM (P = 0.002) showed lack of fit against observed mortality. POSSUM overpredicted mortality across nearly all risk groups, whereas O-POSSUM overpredicted mortality in patients with low physiological scores and in older patients. POSSUM (area under ROC curve 0.776) and P-POSSUM (0.776) showed equally good discriminatory power but O-POSSUM (0.676) was inferior.

Conclusion: P-POSSUM provided the most accurate prediction of in-hospital mortality in this group of patients who had elective oesophagectomy.

Publication types

  • Evaluation Study
  • Validation Study

MeSH terms

  • Carcinoma, Squamous Cell / mortality*
  • Carcinoma, Squamous Cell / surgery
  • Esophageal Neoplasms / mortality*
  • Esophageal Neoplasms / surgery
  • Esophagectomy / mortality*
  • Female
  • Humans
  • Male
  • Middle Aged
  • ROC Curve
  • Regression Analysis
  • Risk Factors
  • Severity of Illness Index