BACKGROUND: Anatomic total shoulder arthroplasty (TSA) and reverse TSA are the standard of care for end-stage shoulder arthritis. Advancements in implant design, perioperative management, and patient selection have allowed shorter inpatient admissions. Unplanned readmissions remain a significant complication. Identification of risk factors for readmission is prudent as physicians and payers prepare for the adoption of bundled care reimbursement models. The purpose of this study was to identify characteristics and risk factors associated with readmission following shoulder arthroplasty using a large, bi-institutional cohort. METHODS: A total of 2805 anatomic TSAs and 2605 reverse TSAs drawn from 2 geographically diverse, tertiary health systems were examined for unplanned inpatient readmissions within 90 days following the index operation (primary outcome). Forty preoperative patient sociodemographic and comorbidity factors were tested for their significance using both univariable and multivariable logistic regression models, and backward stepwise elimination selected for the most important associations for 90-day readmission. Readmissions were characterized as either medical or surgical, and subgroup analysis was performed. A short length of stay (discharge by postoperative day 1) and discharge to a rehabilitation or skilled nursing facility were also examined as secondary outcomes. Parameters associated with increased readmission risk were included in a predictive model. RESULTS: Within 90 days of surgery, 175 patients (3.2%) experienced an unanticipated readmission, with no significant difference between institutions (P = .447). There were more readmissions for surgical complications than for medical complications (62.9% vs. 37.1%, P < .001). Patients discharged to a rehabilitation or skilled nursing facility were significantly more likely to be readmitted (13.1% vs. 8.8%, P = .049), but a short inpatient length of stay was not associated with an increased rate of 90-day readmission (42.9% vs. 41.3%, P = .684). Parameter selection based on predictive ability resulted in a multivariable logistic regression model composed of 16 preoperative patient factors, including reverse TSA, revision surgery, right-sided surgery, and various comorbidities. The area under the receiver operator characteristic curve for this multivariable logistic regression model was 0.716. CONCLUSION: Risk factors for unplanned 90-day readmission following shoulder arthroplasty include reverse shoulder arthroplasty, surgery for revision and fracture, and right-sided surgery. Additionally, there are several modifiable and nonmodifiable risk factors that can be used to ascertain a patient's readmission probability. A shorter inpatient stay is not associated with an increased risk of readmission, whereas discharge to post-acute care facilities does impose a greater risk of readmission. As scrutiny around health care cost increases, identifying and addressing risk factors for readmission following shoulder arthroplasty will become increasingly important.