Predicting length of stay for trauma and emergency general surgery patients

Background: Predicting length of stay (LOS) is difficult for trauma and emergency general surgery (TEGS) patients. Our aim was to determine the accuracy of LOS predictions by TEGS team members and the NSQIP Risk Calculator and the patient factors associated with inaccurate predictions. Methods: LOS for 200 TEGS patients were predicted. Full-model univariate and multivariable linear regressions were used to determine associations between patient characteristics and inaccurate predictions. Results: There were 1,518 predictions of LOS. LOS predictions were rarely correct (TEGS team: 30.7% all patients, 35.6% surgical; NSQIP: 33.0% surgical). No individual group nor NSQIP was significantly better at predicting LOS. Inaccurate predictions were associated with female patients, longer LOS, trauma, frailty, higher comorbidity and injury severity scores, and lesser disposition. Conclusion: Both the TEGS team and NSQIP are poor at predicting LOS for TEGS patients. Further work helping to guide LOS predictions for TEGS patients is warranted. (c) 2020 Elsevier Inc. All rights reserved.