A Data-Driven Method to Discriminate Limb Salvage from Other Combat-Related Extremity Trauma.

ABSTRACT

INTRODUCTION: The aim of this study was to address and enhance our ability to study the clinical outcome of limb salvage (LS), a commonly referenced but ill-defined clinical care pathway, by developing a data-driven approach for the identification of LS cases using existing medical code data to identify characteristic diagnoses and procedures, and to use that information to describe a cohort of US Service members (SMs) for further study. METHODS: Diagnosis code families and inpatient procedure codes were compiled and analyzed to identify medical codes that are disparately associated with a LS surrogate population of SMs who underwent secondary amputation within a broader cohort of 3390 SMs with lower extremity trauma (AIS > 1). Subsequently, the identified codes were used to define a cohort of all SMs who underwent lower extremity LS which was compared with the opinion of a panel of military trauma surgeons. RESULTS: The data-driven approach identified a population of n = 2018 SMs who underwent LS, representing 59.5% of the combat-related lower extremity (LE) trauma population. Validation analysis revealed 70% agreement between the data-driven approach and gold standard SME panel for the test cases studied. The Kappa statistic (κ = 0.55) indicates a moderate agreement between the data-driven approach and the expert opinion of the SME panel. The sensitivity and specificity were identified as 55.6% (expert range of 51.8-66.7%) and 87% (expert range of 73.9-91.3%), respectively. CONCLUSIONS: This approach for identifying LS cases can be utilized to enable future high-throughput retrospective analyses for studying both short- and long-term outcomes of this underserved patient population.