JMIR Publications

ABSTRACT

Background:

Discharge mobility after osteoporotic vertebral fractures (OVF) is heterogeneous and difficult to predict at admission. Machine learning offers structured approaches for multiclass functional outcome prediction, but existing models predominantly address binary endpoints and report global performance metrics without explicit analysis of error structure.

Objective:

To develop and evaluate a safety-oriented, two-stage machine learning framework for graded discharge mobility prediction in hospitalized patients with osteoporotic thoracolumbar fractures, with explicit minimization of clinically critical misclassification.

Methods:

This prospective, single-center cohort study enrolled consecutive patients aged ≥65 years admitted for acute osteoporotic thoracic or lumbar fractures (n = 158). Discharge mobility was categorized as LOW (<50 m or non-ambulatory), MID (50–500 m), or HIGH (>500 m or stair climbing). Clinical associations were analyzed using proportional odds ordinal regression. A predefined two-stage architecture combined gradient-boosted trees (Stage A: HIGH vs non-HIGH) and a random forest classifier (Stage L: LOW vs non-LOW). Threshold selection was embedded in nested cross-validation with a predefined constraint against HIGH-to-LOW misclassification. All preprocessing steps, including fold-wise multiple imputation, were performed within training partitions only.

Results:

Pre-fracture HIGH mobility was the only significant independent predictor of discharge mobility (OR 3.38, 95% CI 1.68–6.79; P<.001); age, comorbidity burden, fracture grade, and treatment were not. Model performance was moderate (accuracy 0.494; balanced accuracy 0.450; macro-F1 0.444). Binary discrimination yielded AUC 0.693 (HIGH vs non-HIGH) and 0.625 (LOW vs non-LOW). The safety constraint was achieved: no HIGH-to-LOW misclassification occurred. Five LOW-to-HIGH errors (3.2%) were observed, all with elevated predicted HIGH probabilities (0.72–0.83).

Conclusions:

Discharge mobility after osteoporotic thoracolumbar fractures is primarily determined by pre-fracture functional reserve rather than fracture severity or comorbidity. A two-stage machine learning framework achieved moderate multiclass discrimination while preventing clinically critical HIGH-to-LOW misclassification, supporting safety-oriented risk stratification in geriatric OVF care. External validation is required before clinical deployment. Clinical Trial: The study protocol was approved by the institutional ethics committee (approval no. 214/23-ek) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to inclusion in the study.