Low et al: Machine‐learning prediction of postoperative complications after high tibial osteotomy for canine cranial cruciate ligament disease
Veterinary Surgery 7, 2025

🔍 Key Findings

• Postoperative complications occurred in 20% of stifles, including 7.5% minor, 10.3% surgical, and 3.4% medical complications.
• The PROSPECT machine-learning model achieved high predictive accuracy: 92.3% for surgical complications, 91.9% for minor, and 94.3% for medical.
• Top predictive features included surgical technique, implant type, patient age, and surgeon identity.
• Surgeon-specific variables influenced predictions, indicating operator experience and technique matter.
• Engineered interaction features (e.g., breed × implant) were more predictive than raw clinical data alone.
• Rottweiler, intact male status, and higher bodyweight were associated with increased complication risk; Labradors had decreased risk.
• Model calibration was strong, especially for high and low probability predictions; midrange predictions were less reliable.
• The model supports individualized, probabilistic risk assessment, which could inform client counseling and tailored postoperative care.

PROSPECT = Predicting Risk Of Surgical compli­cations aftEr CCWO and TPLO