Quantifying surgical complexity through textual descriptions of current procedural terminology codes

Models for surgical complications are a requirement for evaluating patients by the bedside or for risk-adjusted quality and outcomes assessment of healthcare providers. Developing such models requires quantifying the complexities of surgical procedures. Existing approaches to quantify procedural complexity rely on coding system generalities or factors designed for reimbursement. These approaches measure complexity of surgical procedures through the time taken for the procedures or their correspondence to rough anatomical ranges. We address this limitation through a novel approach that provides a fine-grained estimate of individual procedural complexity by studying textual descriptions of current procedure terminology (CPT) codes associated with these procedures. We show that such an approach can provide superior assessment of procedural complexity when compared to currently used estimates. This text-based score can improve surgical risk adjustment even after accounting for a large array of patient factors, indicating the potential to improve quality assessment of surgical care providers.