Development and Validation of the Shoulder Arthroplasty Risk of Non-Home Discharge (SHORD) Score: A Risk Stratification Model for Non-Home Discharge Following Primary Total Shoulder Arthroplasty.

BACKGROUND: Non-home discharge (NHD) after total shoulder arthroplasty (TSA) increases cost, complications, and readmission, yet most existing risk tools depend on postoperative data or are not built to directly estimate the risk of this outcome. We developed and validated the Shoulder Arthroplasty Risk of Non-home Discharge (SHORD) Score, a purely preoperative bedside tool.

METHODS: We carried out a retrospective prognostic model development and validation study of 29,382 patients undergoing primary TSA in the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP, 2016 to 2021). Patients were split into a derivation (two-thirds) and a held-out validation (one-third) cohort. Preoperative candidate predictors were screened by univariate logistic regression, assessed for collinearity, and entered into multivariable logistic regression; independent predictors were converted to integer points to form the SHORD Score. Discrimination (area under the receiver operating characteristic curve [AUROC]), calibration, and classification were assessed in validation and compared with the 5-item modified frailty index (mFI-5), American Society of Anesthesiologists (ASA) class, age, the Risk Analysis Index (RAI), and the NSQIP morbidity probability using the DeLong test. Sensitivity analyses addressed acute proximal humerus fracture, baseline non-home residence, missing albumin, and dyspnea.

RESULTS: NHD occurred in 2,057 patients (7.0%). 16 independent predictors formed the score, the strongest being age 85 years or older, transfer from acute care, ASA class 4, dependent functional status, and preoperative non-home residence. In validation, the SHORD Score achieved an AUROC of 0.816 (95% confidence interval 0.802 to 0.830) with excellent calibration (calibration slope 0.96; Brier score 0.057). It significantly outperformed every comparator (all DeLong P < .001) and produced a marked gradient exceeding 20-fold, from 0.4% in the lowest risk group to 20.3% in the highest. Discrimination was similar across subgroups and sensitivity analyses, including after excluding the 5.7% of patients with acute proximal humerus fracture (AUROC 0.811).

CONCLUSION: The SHORD Score is an internally validated, preoperative tool whose score is associated with non-home discharge after shoulder arthroplasty and uses only admission-day variables, which may support early discharge planning and resource allocation. External and prospective validation are warranted.