Purpose: The Physiologic and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM), Portsmouth revision (p)-POSSUM, and colorectal (Cr)-POSSUM scoring systems were developed as audit tools for comparing outcomes in surgical and colorectal patients on the basis of operative risk assessment. The aim of this study was to evaluate the applicability of these systems to a cohort of colon cancer patients undergoing surgery in the United States.
Methods: POSSUM factors from 890 consecutive patients undergoing major surgical procedures for colon cancer in nine United States hospitals over a two-year period from January 2000 through December 2001 were prospectively collected. The observed over the expected hospital mortality was compared by means of the POSSUM, p-POSSUM, and Cr-POSSUM scoring systems. The effect of missing data on the utility of this process for outcome assessment was assessed with three methods for data imputation.
Results: The number of resections per institution ranged from 13 to 437. The observed mortality rate ranged from 0.8 percent to 15.4 percent among the institutions, with an overall operative mortality of 2.3 percent. The POSSUM, p-POSSUM, and Cr-POSSUM predicted mortality was 10.7 percent, 11.2 percent, and 4.9 percent, respectively. The POSSUM and p-POSSUM models overpredicted mortality in all institutions ( P < 0.01), whereas the Cr-POSSUM demonstrated an observed over expected hospital mortality ratio of >1 in three institutions. The calculations were unaffected by the various methods of inserting missing data.
Conclusion: An apparent overprediction of mortality for colon cancer resection was evident with all three POSSUM variants. This implies that a calibration process is required for use of these variants in the United States health care system. Missing data may be treated as normal values without influencing outcome. The Cr-POSSUM appeared to be the most promising audit tool for colorectal cancer surgery; however, it will require further refinement to provide process control graphs for identification of potential outliers and improvement in the quality of care in the United States.