Predicting rehabilitation discharge after lumbar decompression: Impact of age and preoperative assistive device use.

Background: Elective lumbar decompression, commonly performed via laminectomy or microdiscectomy, is a well-tolerated surgical procedure. Patients requiring discharge to inpatient rehabilitation facilities are infrequent but generate increased healthcare costs and can be burdensome to patients and families. This study aims to develop an accurate predictive model for rehabilitation discharge following elective lumbar decompression, as well as to identify associated factors. Methods: This was a retrospective, single-center, cohort study. Current Procedural Terminology codes for laminectomy and microdiscectomy were used to identify patients from October 2012 to November 2024. Charts were reviewed to obtain additional demographic and clinical characteristics on which an initial univariate analysis was performed. Primary outcomes for this study included measurements of accuracy for predicting rehab discharge. Secondary outcomes included associations of variables with rehab discharge. Those with confounding diagnosis or admitted from a rehabilitation facility were excluded. Machine learning models were trained and evaluated on the data using cross-validation and classification metrics were utilized for model comparison. Results: A total of 486 patients were considered for the analysis. After performing a Benjamini-Hochberg correction, the univariate analysis identified multiple statistically significant variables associated with rehabilitation discharge some of which include age, preoperative leg pain, lower extremity weakness, need for a walking assistive device, and frailty. Of the 5 machine learning models tested, the logistic regression model demonstrated superior performance, with the highest area under curve (0.81), Youden's J statistic (0.55) and balanced accuracy (0.78). The logistic regression model demonstrated that age and the use of an ambulatory assistive device are consistently predictive of rehabilitation discharge (p<.001). Conclusions: When considered alongside frailty, lack of support, and other variables, age and the necessity for a walking assistive device can serve as a reliable predictors of rehabilitation discharge, thereby improving patient workflow, minimizing uncertainty and reducing healthcare costs.