Predictive Model for Estimating the Length of Stay in Hip Arthroplasty Patients by Machine Learning—H.I.P.P.O Score

Abstract

Introduction: Total hip arthroplasty is a major orthopedic intervention, which is increasingly performed, and involves a variable duration of postoperative hospitalization. Estimating this duration is essential for optimizing resources and improving perioperative planning. Materials and Methods: The retrospective study included 85 patients admitted to an orthopedic clinic between 2020 and 2025, undergoing primary hip arthroplasty. Pre- and postoperative clinical, biological, and surgical data were collected. Based on these variables, the H.I.P.P.O. (Hip Intervention Patient Prognostic Outcome) score was developed, using a Random Forest algorithm to predict the length of stay (short, medium, long). Results: The model achieved an overall accuracy of 80%. The most important predictors were as follows: day of surgery, type of prosthesis, preoperative fibrinogen, INR, APTT, preoperative hemoglobin, age, and presence of liver cirrhosis. The H.I.P.P.O. score allowed efficient stratification of patients and showed a high capacity to identify cases at risk of prolonged hospitalization (F1 score = 0.857). Conclusions: The H.I.P.P.O. score is a practical, interpretable, and clinically applicable tool that integrates biological and organizational factors to predict the length of stay after hip arthroplasty. It can support surgical decision making and optimize perioperative management.