The original SORT model was developed using data from the 2011 NCEPOD study, Knowing The Risk.1,2
Since then, multiple studies have re-evaluated it. Most recently, an international citizen science project led by the Health Services Research Centre at the Royal College of Anaesthetists and the UCL/UCLH Surgical Outcomes Research Centre, studied the validity of the SORT In over 26,000 patients in the UK, Australia and New Zealand, and compared its accuracy with clinical judgment of perioperative teams. This found the discrimination of the SORT to be excellent, but that it over-estimated risk (i.e. was poorly calibrated) particularly in higher risk patients. Clinicians were similarly good at discriminating between high and low risk patients, but were even more pessimistic (worse calibration) than the SORT model. A new model was created which combined clinicians’ assessment with the original SORT and this was both excellent in terms of discrimination and well-calibrated.3
What does all this mean?
It should be noted that no risk prediction tool is completely accurate, and there is always a chance that patients deemed to be low risk will develop complications or die after surgery, or that predicted high risk patients will do well. Further work to externally validate the SORT and other risk stratification tools is required, and the authors of the SORT manuscript would welcome approaches from potential collaborators for future validation work. Please see the Contact tab for further details.