FTP8.13 Validation of a definition of Major Abdominal Surgery (MAS)

Abstract

Abstract Aims To validate a previously derived definition for major abdominal surgery that has potential for widespread adoption and use in patient-driven outcome studies. Methods We codified our narrative definition for MAS using CCSD and OPCS-4 schedules. Data was extracted from electronic health records over a 5-year period to validate this definition by comparing relevant proxy measurements between MAS and non-MAS groups. Unsupervised machine learning, using the Partitioning Around Medoids algorithm, was used to cross-validate the definition by clustering all procedure codes into MAS and non-MAS groups based on these proxy measures. Results 337 CCSD codes and 214 OPCS-4 codes were reviewed. Of these, 84 CCSD and 45 OPCS-4 codes were considered to be MAS, while 195 CCSD and 111 OPCS-4 codes were considered to be non-MAS. The remaining codes were excluded as they involved non-operative procedures. We extracted data on 16,353 procedures, covering 214 CCSD procedure codes over a 5 year period. There was statistically significant difference between MAS and non-MAS operations, when comparing patient age, procedure duration, length of hospital stay, percentage of ICU admissions (p<0.001 for all values). Unsupervised machine learning allocated 127 CCSD codes, that contained data on more than 10 patients, into 2 distinct clusters with percentage agreement being 87.4% and Cohen’s kappa of 0.736 between narrative definition and machine learning analysis. Conclusions We have arrived at a validated definition of major abdominal surgery, which can now be confirmed using a larger data set. We propose that this will prove useful in the development of comparative patient-driven outcome data.