Systematic review and evaluation of predictive modeling algorithms in spinal surgeries

Area of Science:

• Neurosurgery
• Orthopedic Surgery
• Medical Informatics

Background:

• Predictive models enhance patient education by stratifying surgical risk.
• Standardizing surgical practices improves patient safety and outcomes.
• Evaluating the efficacy of existing predictive models is crucial for advancing spinal surgery.

Purpose of the Study:

• To systematically review and summarize major predictive models for spinal surgery patient evaluation.
• To identify the most effective models based on performance metrics and applicability.

Main Methods:

• Systematic literature search of PubMed, MEDLINE, and Scopus databases.
• Inclusion of studies reporting Area Under the Receiver Operating Curve (AUROC) scores.
• Exclusion of models not relevant to spinal procedures.

Main Results:

• Several models demonstrated high predictive accuracy, including Scheer et al. (0.89), Spine Sage calculator (0.81-0.85), and Seattle Spine Score (0.712).
• Other promising models include Ratliff et al. (0.70) and Risk Assessment Tool (0.67-0.7).
• Analysis of model inputs and outputs identified key elements for future theoretical models.

Conclusions:

• The Scheer et al. model, Spine Sage calculator, Seattle Spine Score, Risk Assessment Tool, and Ratliff et al. model are ideal for spinal surgery predictive modeling.
• Future models can be optimized using larger prospective databases, longer follow-up, and high-impact variables.