gms | German Medical Science

Objectives: The deep learning-based early warning score(DEWS), an artificial intelligence(AI) assisted system, could be utilized in various clinical practices and for prevention of cardiac arrest. This study investigates the utility of DEWS in predicting intraoperative hemorrhagic risk in spinal surgeries, a critical factor for patient safety and surgical outcomes. Prior research predominantly focused on intraoperative determinants such as fusion levels and blood pressure to predict surgical blood loss. Our research expands this scope by examining the influence of both intraoperative variables and DEWS on blood loss during spinal surgeries. Additionally, we assess the correlation between DEWS scores and the necessity for transfusions in patients undergoing posterior lumbar spine surgery.

Methods: This retrospective study encompasses 48 patients who underwent posterior lumbar spine surgery at Pusan National University Hospital. We collected demographic data, American Society of Anesthesiologists (ASA) scores, pre-operative and post-operative DEWS scores, and complete blood count (CBC) results as primary variables. The DEWS scoring system was employed pre- and postoperatively. Primary outcomes, including the volume of intraoperative blood loss and transfusion status, were meticulously recorded. Supplementary surgical parameters such as duration, fusion levels, estimated blood loss, and total concealed hemorrhage were also incorporated. Patients requiring transfusions (Hb<7.0) were categorized as Group A, while those not requiring transfusions formed Group B.

Results and conclusion: Comparative analysis revealed no significant statistical difference in preoperative hemoglobin (Hb), hematocrit (Hct), and ASA scores between the groups. However, the preoperative DEWS score was notably higher in Group A (p=0.032). Intraoperative fusion levels showed no significant variance (p=0.073), but Group A exhibited longer surgery duration (p=0.039). Postoperative Hb and Hct levels were significantly lower in Group A (p=0.017, 0.031), and postoperative DEWS scores were elevated (p=0.043). The study elucidates a substantial association between elevated pre- and post-operative DEWS scores and increased transfusion requirements. The Receiver Operating Characteristic (ROC) analysis between postoperative DEWS scores and transfusion necessity yielded an area under the curve of 0.752.

Through rigorous retrospective analysis, we discerned the significant prognostic utility of postoperative DEWS scores in forecasting transfusion needs post-spinal fusion surgery. These findings underscore the imperative of integrating postoperative DEWS evaluations in clinical decision-making processes to enhance resource optimization and patient care during the perioperative phase.