Predicting Length of Stay After Shoulder Arthroplasty

Abstract Background/Hypothesis The annual volume of shoulder arthroplasty (SA) in the United States has increased over 400% from 1993 to 2008 and is projected to increase an additional 300% from 2008 to 2030. The purpose of this study was to develop and internally validate a preoperative predictive nomogram for LOS after SA. We hypothesized that patient-specific covariates are responsible for increased LOS following SA. If sufficiently accurate, such a nomogram would be valuable for defining which patients would qualify for outpatient SA. Methods A systematic review of the literature was performed to identify patient variables responsible for increased LOS defined as ≥ 2 days. Patient data was collected from 234 consecutive shoulder arthroplasty cases performed between 2010 and 2016 at a single tertiary health care system by six orthopedic surgeons. The cohort was separated into two groups: short LOS (≤1 day; n=98) and extended LOS (≥2 days; n=136). Logistic regression coefficients were used to construct a LOS nomogram that was internally validated by the bootstrapping method. A receiver operating curve (ROC) plot determined the discriminative ability of the nomogram. Generalizability was assessed using external validation with a new data set of 193 records prospectively collected from 2016 to 2018. Results Univariate analysis demonstrated that female sex, cancer history, anticoagulation and oral hypoglycemic agents, undergoing humeral head replacement for fracture and discharge to a facility were significant predictors for a LOS ≥2 days. The following patient co-variates were found to be significant in the development of the nomogram: age, female sex, procedure type, DMARDs, renal dysfunction, and diabetes drugs. The prediction accuracy of this model was good (area under the curve (AUC) of the ROC plot of 0.793). External validation of the model reported an AUC of 0.664. Conclusion Patients most suitable for expedited postoperative stay are healthy middle-aged males that are not on diabetic medications or DMARDs, have no renal disease, and undergo either resurfacing, total shoulder arthroplasty, or reverse total shoulder arthroplasty. This nomogram may be an additional clinical decision-making tool for effectively preparing for postoperative care and decreasing LOS for an episode of care, subsequently reducing the total cost of care. Future work would focus on undertaking a prospective multicenter study involving institutions that perform a high annual volume of SAs.