The global demand for end-stage hip and knee osteoarthritis surgical treatment is rising, as is the need of optimal postoperative rehabilitation. Patient stratification is key to provide rehabilitation professionals and policy makers with real-life data in support of early discharge planning and continuous care provision. The aim of this retrospective, observational study was to investigate which factors can predict the burden of care at discharge (BCD) and the inpatient rehabilitation length of stay (LOS) based on a set of demographic, societal, clinical and organizational data collected from a high-volume orthopedic hospital. We included 45.306 variables from 1678 patients. All variables were initially tested individually using a linear regression model for inpatient rehabilitation LOS and a logistic regression model for BCD. Variables that resulted significant (p < 0.05) were subsequently considered in a single, comprehensive linear regression model, or a single, logistic regression model, respectively. Age, living with a family, occupational status, baseline Barthel Index and duration of surgery were predictors of inpatient rehabilitation LOS and BCD. Sex, primary or secondary osteoarthritis, American Society of Anesthesiologists score, body mass index, transfusion, biological risk, type of anesthesia, day of surgery, numeric pain rating scale and baseline cognitive function at baseline were not. Including specific patient comorbidities, surgical access technique and chronic use of pharmacological therapy can improve the predictive power of the model.