This literature review examines the use of machine learning to enhance operating room (OR) efficiency by predicting the duration of Total Shoulder Arthroplasty (TSA) procedures more accurately than traditional scheduling methods. Researchers conducted a review of TSA cases from 2013 to 2021, analyzing patient, surgeon, anesthetic, and shoulder-specific factors. Two gradient-boosted decision tree regression models were developed: Model 1 (M1), which considered patient, surgeon, and anesthetic factors, and Model 2 (M2), which incorporated additional shoulder-specific data. The findings of the study showed that human schedulers accurately predicted only 64.1% of cases, while M1 improved accuracy to 79.7% and M2 further increased it to 82.5%. M2 was particularly effective for anatomic TSA (90.6% accuracy) compared to reverse TSA (78.1%). The most influential predictors in M2 included historical median procedure duration, electronic health record (EHR) predictions, surgeon estimates, patient age, and traumatic indications. Factors like younger age, male sex, higher BMI, and B2 glenoid morphology were associated with under-prediction of procedure duration.