Discharge to a facility after lower extremity bypass is no small inconvenience—it predicts worse outcomes and sours patient expectations. This study flexes both traditional regression and XGBoost on a modern registry, confirming that factors like age, frailty, and urgent presentation stack the odds against a home discharge. While perioperative details add predictive juice to both models (pushing AUCs up to 0.85), the ML approach showed only marginal gains over regression and, as usual, introduces new interpretability headaches. The real impact might be in moving risk conversations up in the timeline: letting patients and perioperative teams anticipate, rather than react to, a likely need for post-acute placement.
Machine Learning and the Problem of Nonhome Discharge After Lower Extremity Bypass
Do contemporary machine learning models improve prediction of nonhome discharge after lower extremity bypass above standard regression, and how actionable is this for perioperative care?