Comparison of the safe definitive surgery concept algorithm and IBM Watson trauma pathway explorer for surgical decision-making in real-life polytrauma patients

Introduction

The management of the polytrauma patient still represents a challenge in the trauma bay. Early patient assessment and surgical decision-making are often based on only a few available parameters. The objective of the study is to compare, the treatment algorithm based on the concepts of safe definitive surgery and the IBM WATSON Trauma Pathway Explorer with the outcomes of real trauma patients.

Material & Methods

A retrospective analysis was carried out, using a random selection of severely injured polytrauma patients treated from 2018 to 2022. The Safe Definitive Surgery (SDS) grading and IBM Watson Pathway Explorer were used as reference groups in terms of their proposed treatment strategy (Safe Definitive Surgery/Early total care vs. Damage control orthopedics). Parameters included acidosis, anemia, coagulopathy, hypothermia, injury severity score (ISS) and the presence of severe traumatic brain injury (TBI). A comparison to the chosen treatment and related outcomes was performed. The agreement was calculated using the Cohen"s kappa.

Results

82 patients from this time-period were included into our analysis. 82% were male, the average age was 46 years and the average ISS was 38. The agreement between the IBM WATSON Trauma Pathway, and the performed treatment strategy, using the kappa index was 0.04 (no relationship) while for the SDS algorithm a Kappa index of 0.75, was achieved (substantial agreement).

Conclusions

In the prediction for the initial surgical treatment, SDS thresholds, seem to substantially reflect the surgical reality in our study period, in opposition to the self-learning IBM WATSON, where no agreement could be detected. Being a digital tool in constant evolution created from over 3,500 past trauma cases, the results may be conditioned by the trend of past years, making it evident that it is susceptible to updating and improvement. Algorithms in early decision making process are still a reliable tool in the treatment of trauma patients.

No