Derivation and validation of a simple calculator to predict home discharge after surgery

Background: Surgical patients and their physicians currently have tools to provide individualized prognostication for morbidity and mortality. For improved shared decision making, formal prediction of patient-centered outcomes is necessary. We derived and validated a simple, interview-based method to predict discharge home after surgery.

Study design: We used the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) Patient User File for 2011. Derivation in general and vascular surgery patients undergoing inpatient surgery was completed using serial multiple logistic regression. Validation was performed within multiple surgical specialties.

Results: The derivation cohort included 88,068 patients, of whom 11,771 (13.4%) were not discharged home. The derived Home Calculator had excellent discrimination (c-statistic = 0.864) using 5 [corrected] variables: age, American Society of Anesthesiologists' performance status, elective surgery, and preadmission residence. Validation cohorts had varying rates of home discharge as follows: general (63,826 of 71,591, 89.2%), vascular (12,319 of 16,102, 76.5%), gynecologic (16,603 of 17,005, 97.6%), urologic (13,662 of 14,435, 94.6%), orthopaedic (12,000 of 19,514, 61.5%), thoracic (4,467 of 5,092, 87.7%). The Home Calculator provided good to excellent discrimination in validation cohorts: general (c = 0.866), vascular (c = 0.800), gynecologic (c = 0.793), urologic (c = 0.814), orthopaedic (c = 0.876), and thoracic (c = 0.800). Comparable discrimination was demonstrated in sensitivity analyses in surgical patients admitted exclusively from home.

Conclusions: We derived and validated a simple Home Calculator that reliably predicts discharge to home after surgery and may be useful when counseling patients about postoperative course. Patient-centered tools such as this may allow physicians to better prepare patients and families for surgery and the recovery process.