Predicting Non-Routine Discharge After Elective Spine Surgery: External Validation of Machine Learning Algorithms Using Institutional Data

Abstract

Abstract INTRODUCTION Non-routine discharge after elective spine surgery negatively impacts patient outcomes and increases healthcare costs. Preoperative prediction in this population could improve discharge planning. We previously developed an algorithm that predicts non-home discharge after elective spine surgery. Here, we validate our algorithm in an institutional population of neurosurgical spine patients from a Transitional Care Program (TCP) at an academic, tertiary care center. METHODS Medical records from elective inpatient surgery for lumbar disc herniation or degeneration in the TCP program (2013–2015) were retrospectively reviewed. Variables included age, sex, body mass index (BMI), American Society of Anesthesiologists (ASA) class, preoperative functional status, number of fusion levels and discharge disposition. The discrimination and calibration of the previous algorithm was assessed in the independent sample, using Python 3.0 software. RESULTS A total of 144 patients underwent elective inpatient surgery for lumbar disc disorders with a non-home discharge rate of 10 (7%). The algorithm generalized well to the institutional data with c-statistic 0.89, calibration slope 1.09 and calibration intercept −0.08. This was comparable to performance in the derivation cohort and substantiates initial use of this algorithm in clinical practice. Prospective validation of this algorithm and comparison to other existing discharge prediction models is ongoing. CONCLUSION This institutional external validation of a previously developed machine learning algorithm suggests a reliable method for identifying lumbar disc disorder patients at risk for non-routine discharge. This tool can be used by multidisciplinary teams of case management and spine surgeons to strategically invest additional time and resources into postoperative plans for this population.