Evaluation of an AI-Driven Risk Stratification System for Clinical Deterioration in Geriatric Hip Fracture Patients

OBJECTIVES: To determine whether an artificial-intelligence-driven Clinical Deterioration Index (CDI) could identify geriatric hip-fracture patients at risk of early postoperative complications and to establish an orthopaedic-specific cutoff that identified patients at risk of deterioration. METHODS: Design: Retrospective cohort study. SETTING: Single Level I trauma center. PATIENT SELECTION CRITERIA: Patients ≥ 65 years who underwent fixation of OTA/AO 31-A/B/C hip fractures between June 2022 and December 2023 were included using ICD-10 codes S72.0-, S72.1-, and S72.2-. Exclusion criteria were pathologic fracture, revision surgery, or absence of 30-day follow-up. The CDI uses 31 clinical measures to generate a score from 0-100 reflecting risk of rapid deterioration. Patients were separated into two groups based on peak CDI in the immediate 48-hour postoperative period: CDI ≥ 65 and CDI < 65. This cutoff of 65 was established by prior institutional validation. OUTCOME MEASURES AND COMPARISONS: Primary outcomes were in-hospital complications (cognitive changes, cardiac complications, DVT/PE, etc.). Secondary outcomes included length of stay, ambulation distance, and 1-year mortality. Outcomes were compared between CDI groups. RESULTS: Of 197 patients (age range 65-98, 66% female), 15 (7.6%) exceeded the baseline CDI threshold of 65. This group had higher complication rates (93.3% vs 26.4%, P < 0.001), prolonged hospital stays (10.1 vs 5.2 days, P = 0.047), reduced ambulation at discharge (8.8 vs 42.3 feet, P < 0.001), and increased one-year mortality (20.0% vs 3.8%, P = 0.02). CDI ≥ 65 was associated with post-operative complications (OR 37.9). The institutional cutoff of 65 offered high specificity (99.3%) but low sensitivity (22.6%). An optimized threshold (47.7) improved sensitivity (77.4%) with comparable accuracy (75.0%). CONCLUSIONS: An elevated Clinical Deterioration Index (CDI) ≥ 65 correlated with poor outcomes in geriatric hip fracture patients. Specialty-specific cutoffs showed potential to improve postoperative risk stratification and identification of higher-risk patients. LEVEL OF EVIDENCE: III (retrospective cohort study).