A model predicting operative mortality in the UK has only limited value in Denmark

Purpose

Postoperative mortality from colorectal cancer varies between surgical departments. Several models have been developed to predict the operative risk. This study aims to investigate whether the original and the revised Association of Coloproctology of Great Britain and Ireland (ACPGBI) model can predict 30-day mortality after colorectal cancer surgery in Denmark.

Methods

Data were collected from the Danish Colorectal Cancer Group database which has >?95% completeness. All patients operated on from January 2007 to December 2013 were included. The individual estimated operative risk was calculated with the original and revised ACPGBI models. Discrimination and calibration were evaluated with a Receiver Operating Characteristic (ROC) curve analysis and a Hosmer-Lemeshow test, respectively.

Results

In total, 22,807 patients underwent open or laparoscopic colorectal cancer surgery. After excluding 1437 patients because of missing data, 21,370 patients were left for the analyses. The observed 30-day mortality was 5.0%. The original and revised ACPGBI models estimated an operative risk of 7.0 and 4.0%, respectively, with a significant difference in observed and estimated mortality in both models. However, in patients with an estimated risk of at least 26%, i.e., high-risk, good calibration was found with the original ACPGBI model. Discrimination was good with an AUC of 0.83 (95% CI 0.82–0.84) in both models.

Conclusion

The original and revised ACPGBI models are not suitable prediction models for postoperative mortality in the Danish colorectal cancer population. However, the original model might be applicable in predicting mortality in high-risk patients.